21
NBER
Edited by James M. Poterba National Bureau of Economic Research
Tax Policy and the Economy
Tax Policy and the Economy
21 2007
Tax Policy and the Economy National Bureau of Economic Research Edited by James M. Poterba
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand: Evidence from the Health and Retirement Survey
Of related interest Tax Policy and the Economy, Volume 20 Edited by James M. Poterba
21
Articles by Alan J. Auerbach; Julie Berry Cullen and Roger Gordon; Nada Eissa and Hilary W. Hoynes; Sondra Beverly, Daniel Schneider, and Peter Tufano; Jeffrey R. Brown and James M. Poterba; Jagadeesh Gokhale and Kent Smetters
Unemployment Insurance Savings Accounts
Tax Policy and the Economy, Volume 19 Edited by James M. Poterba
Evaluating Effects of Tax Preferences on Health Care Spending and Federal Revenues
Articles by Ann Dryden Witte with Marisol Trowbridge; Jonathan Gruber; James R. Hines Jr.; Michelle Hanlon and Terry Shevlin; Randall Morck
978-0-262-16246-3 0-262-16246-6
The MIT Press Massachusetts Institute of Technology Cambridge, Massachusetts 02142 http://mitpress.mit.edu
NBER The MIT Press
Does It Pay, at the Margin, to Work and Save? Measuring Effective Marginal Taxes on Americans’ Labor Supply and Saving Federal Tax Policy towards Energy
Tax Policy and the Economy 21
Tax Policy and the Economy 21
edited by James M. Poterba
National Bureau of Economic Research Cambridge, Massachusetts The MIT Press Cambridge, Massachusetts London, England
NBER/Tax Policy and the Economy, Volume 21, 2007 ISSN: 0892-8649 ISBN-13: 978-0-262-16246-3 (hc.:alk.paper)—978-0-262-66202-4 (pbk.:alk.paper) ISBN-10: 0-262-16246-6 (hc.:alk.paper)—0-262-66202-7 (pbk.:alk.paper) Published annually by The MIT Press, Cambridge, Massachusetts 02142. © 2007 by the National Bureau of Economic Research and the Massachusetts Institute of Technology. All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from the publisher. Standing orders/subscriptions are available. Inquiries, and changes to subscriptions and addresses should be addressed to Triliteral, Attention: Standing Orders, 100 Maple Ridge Drive, Cumberland, RI 02864, phone 1-800-366-6687 ext. 112 (U.S. and Canada), fax 1-800-406-9145 (U.S. and Canada). In the United Kingdom, continental Europe, and the Middle East and Africa, send single copy and back volume orders to: The MIT Press, Ltd., Fitzroy House, 11 Chenies Street, London WC1E 7ET England, phone 44-020-7306-0603, fax 44-020-7306-0604, email
[email protected], website http://mitpress.mit.edu. In the United States and for all other countries, send single copy and back volume orders to: The MIT Press c/o Triliteral, 100 Maple Ridge Drive, Cumberland, RI 02864, phone 1-800-405-1619 (U.S. and Canada) or 401-658-4226, fax 1-800-406-9145 (U.S. and Canada) or 401-531-2801, email
[email protected], website http://mitpress.mit.edu. MIT Press books may be purchased at special quantity discounts for business or sales promotional use. For information, please email
[email protected]. This book was set in Palatino on 3B2 by Sztrecska Publishing, United States. Printed and bound in the United States of America. 10
9
8
7
6
5
4
3
2
1
NBER Board of Directors
Officers Elizabeth E. Bailey, Chairman John S. Clarkeson, Vice Chairman Martin Feldstein, President and Chief Executive Officer Susan Colligan, Vice President for Administration and Budget and Corporate Secretary Robert Mednick, Treasurer Kelly Horak, Controller and Assistant Corporate Secretary Gerardine Johnson, Assistant Corporate Secretary
Directors at Large Peter C. Aldrich Elizabeth E. Bailey John H. Biggs John S. Clarkeson Don R. Conlan Kathleen B. Cooper Charles H. Dallara George C. Eads Jessica P. Einhorn Martin Feldstein Roger W. Ferguson, Jr. Jacob A. Frenkel Judith M. Gueron Robert S. Hamada Karen N. Horn
Judy C. Lewent John Lipsky Laurence H. Meyer Michael H. Moskow Alicia H. Munnell Rudolph A. Oswald Robert T. Parry Marina v. N. Whitman Martin B. Zimmerman
Directors by University Appointment George Akerlof, California, Berkeley Jagdish Bhagwati, Columbia Ray C. Fair, Yale Michael J. Brennan, California, Los Angeles Glen G. Cain, Wisconsin Franklin Fisher, Massachusetts Institute of Technology Saul H. Hymans, Michigan Marjorie B. McElroy, Duke Joel Mokyr, Northwestern Andrew Postlewaite, Pennsylvania Uwe E. Reinhardt, Princeton Nathan Rosenberg, Stanford Craig Swan, Minnesota David B. Yoffie, Harvard Arnold Zellner (Director Emeritus), Chicago
vi
Directors by Appointment of Other Organizations Richard B. Berner, National Association for Business Economics Gail D. Fosler, The Conference Board Martin Gruber, American Finance Association Arthur B. Kennickell, American Statistical Association Thea Lee, American Federation of Labor and Congress of Industrial Organizations William W. Lewis, Committee for Economic Development Robert Mednick, American Institute of Certified Public Accountants Angelo Melino, Canadian Economics Association
NBER Board of Directors
Jeffrey M. Perloff, American Agricultural Economics Association John J. Siegfried, American Economic Association Gavin Wright, Economic History Association
Directors Emeriti Andrew Brimmer Carl F. Christ George Hatsopoulos Lawrence R. Klein Franklin A. Lindsay Paul W. McCracken Peter G. Peterson Richard N. Rosett Eli Shapiro Arnold Zellner
Relation of the Directors to the Work and Publications of the NBER
1. The object of the NBER is to ascertain and present to the economics profession, and to the public more generally, important economic facts and their interpretation in a scientific manner without policy recommendations. The Board of Directors is charged with the responsibility of ensuring that the work of the NBER is carried on in strict conformity with this object. 2. The President shall establish an internal review process to ensure that book manuscripts proposed for publication DO NOT contain policy recommendations. This shall apply both to the proceedings of conferences and to manuscripts by a single author or by one or more co-authors but shall not apply to authors of comments at NBER conferences who are not NBER affiliates. 3. No book manuscript reporting research shall be published by the NBER until the President has sent to each member of the Board a notice that a manuscript is recommended for publication and that in the President’s opinion it is suitable for publication in accordance with the above principles of the NBER. Such notification will include a table of contents and an abstract or summary of the manuscript’s content, a list of contributors if applicable, and a response form for use by Directors who desire a copy of the manuscript for review. Each manuscript shall contain a summary drawing attention to the nature and treatment of the problem studied and the main conclusions reached. 4. No volume shall be published until forty-five days have elapsed from the above notification of intention to publish it. During this period a copy shall be sent to any Director requesting it, and if any Director objects to publication on the grounds that the manuscript contains policy recommendations, the objection will be presented to the author(s) or editor(s). In case of dispute, all members of the Board shall be notified, and the President shall appoint an ad hoc committee of the Board to decide the matter; thirty days additional shall be granted for this purpose. 5. The President shall present annually to the Board a report describing the internal manuscript review process, any objections made by Directors before publication or by anyone after publication, any disputes about such matters, and how they were handled. 6. Publications of the NBER issued for informational purposes concerning the work of the Bureau, or issued to inform the public of the activities at the Bureau, including but not limited to the NBER Digest and Reporter, shall be consistent with the object stated in paragraph 1. They shall contain a specific disclaimer noting that they have not passed through the review procedures required in this resolution. The Executive Committee of the Board is charged with the review of all such publications from time to time.
viii
Relation of the Directors to the Work and Publications of the NBER
7. NBER working papers and manuscripts distributed on the Bureau’s web site are not deemed to be publications for the purpose of this resolution, but they shall be consistent with the object stated in paragraph 1. Working papers shall contain a specific disclaimer noting that they have not passed through the review procedures required in this resolution. The NBER’s web site shall contain a similar disclaimer. The President shall establish an internal review process to ensure that the working papers and the web site do not contain policy recommendations, and shall report annually to the Board on this process and any concerns raised in connection with it. 8. Unless otherwise determined by the Board or exempted by the terms of paragraphs 6 and 7, a copy of this resolution shall be printed in each NBER publication as described in paragraph 2 above.
Contents
Acknowledgments Introduction xiii James M. Poterba
xi
1 Medicaid Crowd-Out of Private Long-Term Care Insurance Demand: Evidence from the Health and Retirement Survey Jeffrey R. Brown, Norma B. Coe, and Amy Finkelstein 2 Unemployment Insurance Savings Accounts
35
Martin Feldstein and Daniel Altman 3 Evaluating Effects of Tax Preferences on Health Care Spending and Federal Revenues 65 John F. Cogan, R. Glenn Hubbard, and Daniel P. Kessler 4 Does It Pay, at the Margin, to Work and Save? Measuring Effective Marginal Taxes on Americans’ Labor Supply and Saving 83 Laurence J. Kotlikoff and David Rapson 5 Federal Tax Policy towards Energy Gilbert E. Metcalf
145
1
Acknowledgments
In planning and organizing this year’s Tax Policy and the Economy meeting, I have incurred debts to many individuals. NBER President Martin Feldstein has been an active supporter of this conference throughout its history. NBER Conference Department Director Carl Beck, Lita Kimble, and especially Rob Shannon have overseen all the logistical details with extraordinary efficiency and with their perennial good spirits. Helena Fitz-Patrick has directed the publication process with outstanding attention to detail and with exceptional speed and efficiency. I am grateful to Dr. Edward Lazear, the Chairman of the Council of Economic Advisers, for delivering a fascinating set of luncheon remarks at this year’s conference. His remarks focused on recent trends in productivity growth, the link between productivity growth and tax revenue, and the broader economic factors that influence the pace of productivity improvement. His remarks provided a broad context for examining many of the specific issues that researchers in public finance study, and underscored the importance of putting in place well-designed economic policies that promote long-term economic growth. I hope that research studies such as the ones included in this volume contribute to that goal. Finally, I wish to thank the authors of this year’s conference papers. They have worked hard to communicate their important research findings in a readable and clear fashion. I appreciate their efforts and their enthusiasm for participating in this interchange between the research and policy communities.
Introduction James M. Poterba, MIT and NBER
The annual NBER Tax Policy and the Economy conference is designed to communicate current academic research findings in the areas of taxation and government spending to policy analysts in government and in the private sector. Research papers for the conference are selected both for their immediate bearing on current policy debates as well as for their insight on broad questions that are of longer-term interest. The papers in this year’s volume focus on a range of topics involving both taxation and social insurance policy. The first two papers are concerned with the implicit taxes associated with Medicaid and unemployment insurance, two important social insurance programs. The first paper is Jeffrey R. Brown, Norma B. Coe, and Amy Finkelstein’s “Medicaid Crowd-Out of Private Long-Term Care Insurance Demand: Evidence from the Health and Retirement Survey.” The authors observe that the Medicaid program imposes a substantial implicit tax on the benefits paid by private long-term care insurance policies. Because Medicaid covers many long-term care costs once a prospective beneficiary’s assets fall below an asset eligibility requirement, those who purchase private long-term care insurance in many cases are simply postponing the date at which they become eligible for Medicaid. The net benefits of purchasing private insurance are therefore smaller than the gross benefits; the difference is the implicit Medicaid tax. The authors demonstrate that while the Medicaid program exerts a negative effect on the demand for private LTC insurance, the effects of changing the asset eligibility thresholds are modest. Even if the Medicaid eligibility thresholds were lowered by about $25,000, LTC insurance demand would rise by less than 3 percentage points. While that represents a 30 percent increase from current levels, it would still imply a very low overall coverage rate for private long-term care insurance in the elderly population.
xiv
Introduction
In the second paper, “Unemployment Insurance Savings Accounts,” Martin Feldstein and Daniel Altman consider an alternative to the current means of providing income support to unemployed workers. They analyze a system of individual accounts that would be funded by employer contributions while the individual was working, and drawn down when the individual was unemployed. Unlike current public unemployment insurance systems, however, which pool all employer contributions and pay benefits to all claimants, “UISAs” would involve dedicated accounts for each worker. A worker who never needed to draw on her account to provide income support while unemployed would be able to cash out the account balance at retirement. This provision changes the beneficiary’s incentives for job search, since finding a job more quickly would provide a greater payout at retirement. A key question about individual-accounts unemployment insurance plans is whether a relatively small number of individuals experience most of the spells of unemployment, and consequently have zero account balances when they need to draw on their individual funds. A high concentration of unemployment spells means that a system of selfinsurance over time will not be able to replicate the current pattern of benefit payments while unemployed. The authors use data from the Panel Survey of Income Dynamics to investigate the performance of a hypothetical UISA system and find that roughly two-thirds of UI benefits are paid to individuals with positive UISA balances at the time of their unemployment spell. This suggests that UISAs offer substantial potential to improve the job search incentives of unemployed individuals, while providing income support to a large fraction of the unemployed population. The third paper addresses the tax treatment of health insurance expenditures, an issue that involves both tax policy and social insurance concerns. In “Evaluating Effects of Tax Preferences on Health Care Spending and Federal Revenues,” John F. Cogan, R. Glenn Hubbard, and Daniel P. Kessler present new estimates of how excluding employer-provided health insurance from taxable income affects the structure of health insurance demand. Many studies have examined what would happen if employer-provided health insurance payments were included in taxable income, thereby placing health outlays financed from such insurance on a similar tax footing as out-of-pocket medical expenditures. This paper focuses on a different approach to reducing the distortion in favor of care financed by employer-provided insurance: allowing out-of-pocket health care expenditures to
Introduction
xv
be deducted from taxable income. This approach would provide a tax subsidy to all health care outlays, regardless of financing method, and it would remove the incentive for financing through employer-provided health insurance. The authors examine how such a change would affect out-of-pocket health care spending, as well as federal tax revenues. They highlight the key behavioral parameters, such as the elasticity of health insurance demand with respect to its price, and the elasticity of participating in employer-provided insurance with respect to its aftertax cost, in determining the effect of expanding the tax deduction. The authors also apply their framework to evaluate the incentive effects of Health Saving Accounts (HSAs), which currently provide a tax-favored mechanism for taxpayers to finance out-of-pocket medical outlays. The next paper, by Laurence J. Kotlikoff and David Rapson, explores one of the most central questions in public finance: what are the effective marginal tax rates on labor supply and saving? In “Does It Pay, at the Margin, to Work and Save? Measuring Effective Marginal Tax Rates on Americans’ Labor Supply and Saving,” the authors model the incentive effects of many different tax structures and means-tested benefit programs to estimate the effective tax rates facing households who earn another dollar of income or decide to save another dollar. The paper focuses on marginal tax rates in a lifecycle setting, and it is more comprehensive than previous work in its detailed analysis of transfer program rules as well as tax schedules. The paper demonstrates that there are a wide range of potentially applicable marginal tax rates, and raises questions about the capacity of many taxpayers to understand the tax rules that they face. The authors note that for some households, particularly those at lower income levels where transfer program phaseouts are in effect, the implied tax rates on labor supply may be very high. Whether the affected households recognize these high marginal tax rates remains an open question. The authors also consider the benefits of contributing to tax-deferred retirement accounts. They find that for high-income taxpayers, the tax benefits associated with contributions are high, and that there are substantial opportunities to increase lifetime consumption levels by participating in these programs. The final paper, “Federal Tax Policy towards Energy” by Gilbert E. Metcalf, is a careful catalogue of the wide range of tax expenditures that affect incentives for energy exploration, production, and consumption. The paper discusses the economic rationale for tax incentives targeted to energy industries and focuses on a variety of externality issues in energy markets. It then examines the distributional effect of current
xvi
Introduction
energy-related tax policies. The paper discusses a wide range of energy policies, including the depreciation provisions that apply to capital equipment used in the production of various forms of energy and the taxes that apply to various end-stage users of energy products. The paper examines policies that bear on electricity production in particular detail. It suggests that relative to the cost of using different generation technologies in a no-tax setting, the current tax structure provides a net subsidy to the use of wind and biomass production technologies, making them after-tax cost competitive with generation of electricity using natural gas. While solar power technologies receive the largest subsidies, the net-of-tax cost of these generation methods is still substantially greater than the cost of other technologies. Each of these papers illustrates the type of policy-relevant research that is carried out by the affiliates of the NBER Public Economics Program. These studies provide important background information for policy analysis, but they do not make recommendations about the merits or demerits of particular policy options. They will hopefully provide a valuable basis for both near-term and long-term policy discussions.
1 Medicaid Crowd-Out of Private Long-Term Care Insurance Demand: Evidence from the Health and Retirement Survey Jeffrey R. Brown, University of Illinois and NBER Norma B. Coe, Tilburg University Amy Finkelstein, MIT and NBER
Executive Summary This paper provides empirical evidence of Medicaid crowd out of demand for private long-term care insurance. Using data on the nearand young-elderly in the Health and Retirement Survey, our central estimate suggests that a $10,000 decrease in the level of assets an individual can keep while qualifying for Medicaid would increase private long-term care insurance coverage by 1.1 percentage points. These estimates imply that if every state in the country moved from their current Medicaid asset eligibility requirements to the most stringent Medicaid eligibility requirements allowed by federal law—a change that would decrease average household assets protected by Medicaid by about $25,000—demand for private long-term care insurance would rise by 2.7 percentage points. While this represents a 30 percent increase in insurance coverage relative to the baseline ownership rate of 9.1 percent, it also indicates that the vast majority of households would still find it unattractive to purchase private insurance. We discuss reasons why, even with extremely stringent eligibility requirements, Medicaid may still exert a large crowd out effect on demand for private insurance. 1.
Introduction
Expenditures on long-term care, such as home health care and nursing homes, accounted for 8.5 percent of all health care spending in the United States in 2004 (Congressional Budget Office 2004). These longterm care expenditures are projected to triple in real terms over the next few decades, in large part due to the aging of the population (Congressional Budget Office 1999). Because over one-third of Medicaid
2
Brown, Coe, and Finkelstein
expenditures are already devoted to long-term care (U.S. Congress 2004), there is rising concern among policy makers about the fiscal pressure that further growth in long-term care expenditures will place on federal and state budgets in the years to come, and growing interest in stimulating the market for private long-term care insurance. For example, in a much-publicized press release issued in October 2004, the National Governors Association announced that states spent nearly as much money on Medicaid in fiscal year 2003 as they did on K-12 education, and expressed concern that Medicaid is putting a “squeeze” on state budgets going forward (National Governors Association 2004). The market for private long-term care insurance is currently quite limited. Only about 10 percent of the elderly have private long-term care insurance (Brown and Finkelstein 2004a). Because these policies tend to be quite limited in scope, only 4 percent of total long-term care expenditures are paid for by private insurance (Congressional Budget Office 2004). By contrast, in the health care sector as a whole, 35 percent of expenditures are covered by private insurance (National Center for Health Statistics 2002). Medicaid provides public long-term care insurance in the form of a payer-of-last resort. It covers long-term care expenditures only after the individual has met asset and income eligibility tests, and after any private insurance policy held by the individual has paid any benefits it owes. In this paper we explore how changes in Medicaid’s meanstested eligibility thresholds might affect demand for private long-term care insurance. We use data from the 1996, 1998, and 2000 waves of the Health and Retirement Survey to study the effect of Medicaid asset protection rules on private long-term care insurance coverage among individuals aged 55 to 69. To investigate Medicaid’s impact, we draw on the substantial variation across individuals in the amount of assets that can be protected from Medicaid based on their state of residence, marital status, and asset holdings. Due to the potential endogeneity of asset holdings to these Medicaid rules, we predict assets based on demographic characteristics of the individual. We find statistically significant evidence that more generous Medicaid asset protection is associated with lower levels of private long-term care insurance coverage. Our central estimate is that a $10,000 increase in the amount of assets an individual can protect from Medicaid is associated with a decrease in private long-term care insurance coverage of
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
3
1.1 percentage points. This implies, for example, that if all states were to adopt the most stringent asset eligibility requirements allowed by federal law in 2000—$16,824 for a married couple and $2,000 for a single individual—and thereby decrease average protected assets by about $25,000. Overall demand for private long-term care insurance would rise by 2.7 percentage points. While such an increase is large relative to the existing ownership rate in our sample of near-elderly and young elderly of 9.1 percent, it suggests that the vast majority of these individuals would remain uninsured. Our empirical findings complement recent simulation-based estimates of the impact of Medicaid on private long-term care insurance demand (Brown and Finkelstein 2004b). Like our empirical estimates, these simulation results also suggest that changes in Medicaid’s asset disregards are unlikely to have a substantial effect on private long-term care insurance demand. At the same time, however, Brown and Finkelstein (2004b) estimate that Medicaid may be able to explain the lack of private insurance purchases for at least two-thirds of the wealth distribution, even if there were no other factors limiting the size of the market. This is because Medicaid imposes a substantial implicit tax on private long-term care insurance; for example, they estimate that about 60 to 75 percent of the expected present discounted value benefits that a median wealth individual would receive from a typical private longterm care insurance policy are redundant of benefits that Medicaid would have provided had the individual not purchased private insurance. Changes in Medicaid’s asset disregards, however, do not have a large effect on this implicit tax. Together, the empirical and simulation results underscore the importance of understanding the mechanism behind the crowd out effect of a particular public program in considering the likely impact of potential reforms to the public program on private demand. The rest of the paper proceeds as follows. Section two provides background information on long-term care expenditure risk and the nature of existing public and private insurance coverage for this risk. It also briefly reviews the insights from simulation estimates of how Medicaid affects private long-term care insurance demand. Section three presents the data and empirical framework. Section four presents our crowd out estimates. Section five uses these crowd out estimates to simulate the likely effects of changes in Medicaid means-testing thresholds. Section six concludes.
4
Brown, Coe, and Finkelstein
2. Background on Long-Term Care Insurance and Medicaid Crowd Out Long-term care represents a significant source of financial uncertainty for elderly households. Although most 65 year olds will never enter a nursing home, of those who do enter a nursing home, 12 percent of men and 22 percent of women will spend more than three years there; one-in-eight women who enter a nursing home will spend more than five years there (Brown and Finkelstein 2004b). These stays are costly. On average, a year in a nursing home cost $50,000 in 2002 for a semiprivate room, and even more for a private room (MetLife 2002). Very little of this expenditure risk is covered by private insurance. According to the 2000 Health and Retirement Survey, among those individuals aged 60 and over, only 10.5 percent own private long-term care insurance. Moreover, Brown and Finkelstein (2004a) estimate that the typical purchased policy covers only about one-third of expected present discounted value (EPDV) long term care expenditures. As a result, only about 4 percent of long-term care expenditures are paid for by private insurance, while about one-third are paid for out of pocket (Congressional Budget Office 2004); by contrast in the health sector as a whole, private insurance pays for 35 percent of expenditures and only 17 percent are paid for out of pocket (National Center for Health Statistics 2002). Medicaid pays for about 35 percent of long-term care expenditures (Congressional Budget Office 2004).1 An extensive theoretical literature has proposed a host of potential explanations for the limited size of the private long-term care insurance market. These explanations include both factors that constrain supply and factors that limit demand. Norton (2000) provides a useful overview of the various potential explanations. On the supply side, market function may be impaired by such problems as high transactions costs, imperfect competition, asymmetric information, or dynamic problems with long-term contracting. There is evidence consistent with the existence of many of these supplyside failures in the private long-term care insurance market. Finkelstein and McGarry (2006) provide evidence of asymmetric information in the market. There is also evidence of dynamic contracting problems arising both from the difficulty of insuring the aggregate risk of rising medical costs (Cutler 1996) and from dynamic adverse selection as individuals who learn that they are better risks than expected drop out of the market (Finkelstein et al. 2005). Brown and Finkelstein (2004a) present evidence that premiums for individuals
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
5
who buy a policy and maintain it until death are marked up about 18 cents per dollar of premium above actuarially fair levels; this markup appears to reflect a combination of transaction costs and imperfect competition. On the demand side, several different factors that may constrain the private insurance market have been suggested. Limited consumer rationality—such as difficulty understanding low-probability high-loss events (Kunreuther 1978) or misconceptions about the extent of public health insurance coverage for long-term care—may play a role. Demand may also be limited by the availability of imperfect but cheaper substitutes, such as financial transfers from children, unpaid care provided directly by family members in lieu of formal paid care, or the public insurance provided by the means-tested Medicaid program (Pauly 1990; Brown and Finkelstein 2004b). There is evidence that these demand side factors are likely to be important in understanding the limited size of the private market. Brown and Finkelstein (2004a) suggest that the loads on policies—and whatever market failures produce them—are unlikely to be sufficient to explain the limited market size. They note that the average load on a typical private policy is about 50 cents on the dollar higher for men than women, yet ownership patterns are extremely similar by gender, a fact that cannot be explained solely by the within-household correlation in ownership patterns. This suggests an important role for demand side factors such as Medicaid. Brown and Finkelstein (2004b) provide more direct evidence of a crowd out effect of Medicaid. They develop and calibrate a utilitybased model of an elderly, life cycle consumer’s demand for private long-term care insurance and compare demand under various counterfactual assumptions regarding the nature of private insurance and of the Medicaid program. Their simulations suggest that given the current structure of Medicaid, even if actuarially fair, comprehensive private insurance policies were to be available, at least two-thirds of the wealth distribution would still not purchase this insurance. They show that the mechanism behind this large estimated Medicaid crowd out effect stems from the fact that a large portion of private insurance benefits are redundant of benefits that Medicaid would have provided in the absence of private insurance, a phenomenon that they label the Medicaid “implicit tax.” For a male (female) at the median of the wealth distribution, they estimate that 60 percent (75 percent) of the benefits from a private policy are redundant of benefits that Medicaid would otherwise have paid.
6
Brown, Coe, and Finkelstein
The Medicaid implicit tax stems from two features of Medicaid’s design that results in private insurance reducing expected Medicaid expenditures. First, by protecting assets against negative expenditure shocks, private insurance reduces the likelihood that an individual will meet Medicaid’s asset-eligibility requirement. Second, Medicaid is a secondary payer when the individual has private insurance. This secondary payer status means that if an individual has private insurance, the private policy pays first, even if the individual’s asset and income levels make him otherwise eligible for Medicaid; Medicaid then covers any expenditures not reimbursed by the private policy.2 Brown and Finkelstein’s (2004b) calibrated life cycle model suggests that changes in Medicaid’s asset disregards would not have a substantial effect on the Medicaid implicit tax, and thus, would not make private long term care insurance desirable for most of the wealth distribution. Specifically, they simulate the likely effect of a policy that has been adopted in several states which makes the Medicaid asset disregards less stringent if the individual purchases private insurance. They estimate that such a policy would not have much effect on the implicit tax or on private insurance demand because, even in the absence of any asset eligibility requirements—i.e., complete asset protection for individuals—Medicaid still imposes a substantial implicit tax on private insurance through its status as a secondary payer. This paper complements the analysis in Brown and Finkelstein (2004b) by examining empirically how the amount of assets that Medicaid allows an individual to keep while receiving Medicaid coverage for long–term care expenses affects demand for private long-term care insurance. Our empirical estimates of the crowd out effect of Medicaid on private long-term care insurance demand are also related to a sizeable empirical literature that has investigated the extent of Medicaid’s crowd out of acute private health insurance among working families. The estimates from this literature range in magnitude, but at the upper end suggest that up to half of the increase in public insurance coverage from increased Medicaid eligibility is offset by reductions in private insurance coverage (see Gruber 2003 for a review of this literature). To our knowledge, only two other empirical papers have examined the impact of Medicaid on private long–term care insurance demand. Sloan and Norton (1997) compare private long-term care insurance holdings in the 1992 and 1994 Health and Retirement Survey (HRS) and the 1993 Aging and Health Dynamics (AHEAD) across individuals in states with different Medicaid income eligibility limits. They find evi-
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
7
dence that higher Medicaid income eligibility limits are associated with lower probability of owning long-term care insurance in the AHEAD data (ages 70+) but not in the HRS data (ages 51–64); they do not examine the effect of asset limits. Kang et al. (2004) use the 1992 through 1998 waves of the HRS to examine the effect of Medicaid asset and income tests on private insurance coverage, using variation in individual financial resources and state Medicaid eligibility limits. They find evidence consistent with a crowd out effect of less stringent Medicaid asset eligibility limits, but not evidence of an effect of Medicaid income limits on long-term care insurance coverage. Our paper builds on this earlier work in two important dimensions. First, we limit our attention to data from 1996 and later waves of the HRS since prior survey waves utilized a confusing question to ascertain longterm care insurance coverage, resulting in substantial under-reporting (coverage rates are about one-fifth of what other surveys from that time suggest) and, more generally, extremely poor data quality; see Finkelstein and McGarry (2006, Appendix A) for more details on these data issues. Second, both previous papers utilize differences in state Medicaid rules to identify the impact of Medicaid on long-term care insurance demand; however, there are other potentially important determinants of the demand for long-term care insurance that vary by state, such as the price and quality of nursing homes. Our empirical approach allows us to surmount this concern, as we discuss in more detail below. 3.
Data and Empirical Approach
3.1 Data and Summary Statistics on Long-Term Care Insurance Coverage We use data from the HRS, a nationally representative sample of the elderly and near-elderly. We use a restricted access version of the HRS that allows us to identify the individual’s state of residence. Our analysis uses data from the 1996, 1998, and 2000 waves of the HRS. The 1996 wave consists exclusively of individuals from the original HRS cohort (individuals born 1931 to 1941). The 1998 and 2000 waves also include individuals from the adjacent, younger cohort (born 1942 to 1947), and the adjacent older cohort (born 1924–1930); these are known respectively as the “War Baby” cohort and the “Children of the Depression” (CODA) cohort. We limit the analysis to individuals aged 55 to 69 in each wave. As discussed, we do not use data from waves prior to 1995 due to data issues with the measurement of long-term care insurance
8
Brown, Coe, and Finkelstein
coverage; we exclude the 1995 AHEAD wave because individuals in this wave are outside our age range. We limit our analysis to individuals aged 55 to 69 to focus on the decisions of individuals who are in the prime buying ages for long-term care insurance (HIAA 2000). Once purchased, the policy is intended to be a “lifetime” policy; indeed, subsequent annual premiums are constant in nominal terms, so that policy payments are quite front-loaded. As a result, it is important to examine the effect of Medicaid rules that were in effect when an individual might be considering the purchase of private long-term care insurance. For this reason, we particularly wish to exclude individuals aged 70 and over from the analysis. Such individuals may well have been making their purchase decisions in the mid– to late 1980s, during which Medicaid eligibility rules were substantially different than they are today. Crucially for our empirical strategy, which relies on the differential treatment of married and single individuals within different states, these rules would not have varied within state by marital status prior to 1989.3 The current structure of Medicaid eligibility rules was adopted with the Medicare Catastrophic Coverage Act of 1988, which was implemented in 1989 (Stone 2002). Because of the panel nature of the data, we observe many individuals multiple times over the waves. Our full sample consists of 28,100 observations on 12,402 unique individuals. We account for the multiple observations of the same individuals in the error structure in our regression analysis. We do not, however, directly exploit the panel nature of the data and the changes in Medicaid eligibility rules for specific individuals over time due to changes in martial status or—more commonly—changes in state rules. We believe the use of such changes provides a questionable form of identification since it is unclear under which set of rules the individual made the (lifetime) purchase of longterm care insurance. Indeed, as we discuss in more detail below, our preferred specification limits the analysis to the sub-sample of individuals who did not change marital status between 1996 and 2000 and who live in one of the 30 states which have not had any real changes to their Medicaid asset allowances between 1991 and 2000 (see Appendix A for details). We refer to this sub-sample as the “Constant Medicaid rules sub-sample” because the individuals faced constant Medicaid rules over our time period. They represent an arguably cleaner sample on which to analyze the crowd-out effects of Medicaid as there is considerably less uncertainty about what rules were in effect when the individuals bought (or considered buying) long-term care insurance.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
9
Table 1.1 presents some summary statistics for both the full sample (column 1) and the constant Medicaid rules sub-sample (column 2). All statistics are based on using household weights. We focus on the summary statistics for the sub-sample in column 2, although the results are generally similar. The long-term care insurance coverage rate is 9.1 percent. This is comparable to the rates found in other surveys for similar age ranges (see e.g., HIAA 2000). Just over 70 percent of the sample is married, just under half is male, and about two-fifths are retired. The average long-term care insurance coverage rate masks important variation across sub-groups in their long-term care insurance holdings. Table 1.2 therefore presents summary statistics on long-term care ownership rates separately by various covariates. Once again, column 1 presents the results for the full sample, and column 2 presents results for the constant Medicaid rules sub-sample. Coverage rates are similar Table 1.1 Summary Statistics (1) Full Sample
(2) Constant Medicaid Rules Sub-sample
Percent with LTC insurance
9.6
9.1
Average age
61.5
61.6
Percent married
70.2
71.4
Percent male
48.1
48.1
Percent retired
40.2
39.8
Average number of children
3.3
3.3
Mean
385
367
25th percentile
50
48
Median
157
153
75th percentile
387
391
28,100
17,623
Household net worth (in thousands)
N
Note: “Full Sample” consists of individuals aged 55–69 in the 1996, 1998, or 2000 HRS who report their marital status and long-term care insurance coverage. “Constant Medicaid Rules” sub-sample is restricted to individuals who did not experience changes in marital status during our data (1996–2000) and who are in one of the 30 states that did not have real Medicaid asset rule changes between 1991–2000. Net worth is the HRS imputed value for net worth. It includes net financing worth, housing wealth, and defined contribution pension values, but does not include DB pension or Social Security wealth. All statistics are calculated using household weights.
Brown, Coe, and Finkelstein
10
Table 1.2 Long-Term Care Insurance Ownership Rates (1) Sample
Full Sample
(2) Constant Medicaid Rules Sub-sample
Entire sample
9.6
9.1
Males
9.7
9.0
Females
9.4
9.3
Singles
7.8
7.1
Marrieds
10.3
10.0
Age 55–61
8.9
8.1
Age 62–69
10.1
10.4
Wave = 1996
10.5
9.9
Wave = 1998
9.3
9.0
Wave = 2000
9.1
8.6
Net worth, bottom quartile
4.1
3.4
Net worth, 2nd quartile
7.7
7.1
Net worth, 3rd quartile
10.2
10.0
Net worth, top quartile
15.1
14.8
Note: Full Sample consists of individuals aged 55–69 in the 1996, 1998, or 2000 HRS who report their marital status and long-term care insurance coverage. “Constant Medicaid Rules” sub-sample is restricted to individuals who did not experience changes in marital status during our data (1996–2000) and who are in one of the 30 states that did not have real Medicaid asset rule changes between 1991–2000. Net worth is the HRS imputed value for net worth. It includes net financing worth, housing wealth, and defined contribution pension values, but does not include DB pension or Social Security wealth. All statistics are calculated using household weights.
by gender, and higher for married individuals than single individuals (10.0 percent vs. 7.1 percent). Coverage rates are higher among 62 to 69 year olds (10.4 percent) than among 55 to 61 year olds (8.1 percent). Coverage rates also vary across states; the inter-quartile range in longterm care insurance coverage rates across states ranges from 0.06 to 0.12 (not shown). The pattern of coverage by net worth is most dramatic. Less than 4 percent of the sample in the bottom quartile owns long-term care insurance, compared to 15 percent in the highest quartile of net worth. In fact, long-term care insurance coverage rates increase monotonically by wealth decile, from 0.03 percent in the bottom decile to 0.17 percent in the top. The wealth profile likely reflects the fact that the means-tested
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
11
eligibility requirements of Medicaid make it a better substitute for private insurance for lower wealth individuals. 3.2 Overview of Medicaid Rules and Our Empirical Approach We focus our analysis on the impact on private long-term care insurance demand of the amount of protected financial assets that an individual can keep while still receiving Medicaid reimbursement for long-term care utilization. Below, we show that other Medicaid rules such as the minimum allowable income retention for the community spouse or the treatment of the community spouse’s house upon its sale or her death do not appear to affect insurance coverage, and do not affect our estimate of the effect of the asset rules on insurance coverage. Medicaid financial asset disregards exhibit substantial variation across individuals based on an individual’s marital status, state of residence, and asset holdings. Our empirical strategy, broadly speaking, is to control for any direct effects of marital status, state, and assets holdings on long-term care insurance demand, and to identify the impact of Medicaid on long-term care insurance demand using the variation in Medicaid generosity that exists across higher interactions of these three variables (i.e., assets by marital status, state by marital status, and assets by state, as well as assets by state by marital status). Thus, for example, we use both differences across states in the amount of assets protected for married individuals relative to single individuals, and differences across states in the amount of assets protected for individuals of different asset levels to identify the impact of Medicaid’s asset protection rules on demand for private long-term care insurance. Throughout the analysis, we use predicted assets to deal with the potential endogeneity of assets to Medicaid spend down rules (Coe 2005). Medicaid asset rules for single individuals are relatively simple and uniform across states and particularly within states: they do not vary with the assets of the individual (as long as the individual has assets of more than the protected amount). The modal rule in 2000 (used by nearly 70 percent of states) allowed single individuals receiving Medicaid coverage for nursing home care to retain no more than $2,000 in financial wealth. The remaining states had asset limits ranging from $1,500 to $6,500. In contrast, the amount of assets a community spouse is allowed to keep when her spouse goes into a nursing home exhibits substantial variation across states at a given household asset level, from a
12
Brown, Coe, and Finkelstein
minimum of $16,824 to a maximum of $84,120 in 2000. Moreover, the amount of assets a community spouse can keep varies with household assets, and this difference across states is highly non-monotonic in the level of household assets. For married households with assets below the minimum amount that federal law requires be kept when one spouse is in a nursing home ($16,824 in 2000), there is no difference across states in Medicaid asset disregards. For most states, there is also no difference in the amount of assets the married couple can keep if their assets are more than double the maximum amount that federal law allows to be kept when one spouse is in a nursing home (which puts the asset amount at $168,240 in 2000). However, for married households within this range—which corresponds to roughly the 20th to 60th percentile of the asset distribution for married households in the relevant age range in the 2000 HRS— there are substantial differences across states in the amount of assets that a married household can keep under Medicaid. By way of illustration, figure 1.1 graphs the difference in the amount of assets a community spouse can keep as a function of total household financial assets in the two most common sets of state rules. Under the most common set of rules—which is used in 26 states—the community spouse is allowed to keep all of their assets up to the federally allowed maximum protected assets ($84,120 in 2000) after which they face a 100 percent marginal tax rate on all further assets. In the second most common set of rules—which is used in another 15 states—the community spouse is allowed to keep all of her assets up to the federally allowed minimum protected assets ($16,824 in 2000), faces a 100 percent marginal tax rate on all assets between this federal minimum and two times the minimum, faces a 50 percent marginal tax rate on all assets between twice the federal minimum and twice the federal maximum, and a 100 percent marginal tax rate on all assets above twice the federal maximum. As seen in figure 1.1, the difference in amount of protected assets that a community spouse with a given amount of assets faces varies non-monotonically with assets. Using the asset distribution for married households in our age range in the 2000 HRS, we estimate that moving from the most common set of state rules to the next most common would on average allow a married household to keep $21,715 more in assets when one spouse entered a nursing home, which represents 29 percent of average financial assets in this range. The maximum difference in the amount of assets that a household would be able to keep
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
13
Figure 1.1 Difference in Protected Assets for Community Spouse (Most common state rules – second most common state rules) Note: Most common state rules apply in 26 states, second most common in 15 states. Dollar amounts are based on rules in 2000. The rules (and affected states) are described in more detail in Appendix A (see “case 1A” and “case 2A” respectively). Percentiles denotes the percentile of the asset distribution for married households in the 2000 HRS.
is $42,060 and occurs for households with assets of $84,120. The minimum difference in protected assets is 0, and occurs for households with assets of less than $16,824 or more than $168,240. If we instead compared these most common state rules (which are also the most generous in terms of the amount of protected assets allowed for married couples) to the least generous state rules (used by three states) the maximum difference in the amount of assets the household would be able to keep would rise to $67,296 (which would occur in households with $84,120 or more in assets). To sum up, we exploit several key sources of variation in the amount of protected assets to identify the impact of Medicaid asset protection on demand for private long term care insurance. These include: differences across states in the average asset disregards for married and single individuals, differences across married individuals of different asset levels in different states, and differences across married and single individuals of different asset amounts, as well as higher order interactions between state of residence, marital status, and assets. In all cases, we control for any direct effects of asset levels, marital status, or state of residence on the probability an individual has private long term care insurance. For interested readers, Appendix A provides considerably
14
Brown, Coe, and Finkelstein
more detailed information on how the Medicaid eligibility rules vary across states by marital status and asset level. 3.3 Econometric Framework Temporarily ignoring several econometric concerns (that we will address below), a natural starting point would be to estimate the following OLS equation: LTCI ist = β1 Protected ist + β 2 Married ist + α s + Xis′ tη + ε ist .
(1.1)
In this estimating equation, the dependent variable LTCIist is a binary indicator for whether individual i in state s and year t owns longterm care insurance, Marriedist is an indicator variable for whether the individual is married and αS represents a full set of state fixed effects.4 The vector of covariates (X) consist of indicator variables for education categorized by highest degree achieved (less than high school, high school, some college, college degree or more), gender, occupation, industry, number of children up to five, Hispanic heritage, race, retired, age, wave, and cohort; in addition, (X) includes interactions of each of the education categories with all of the other control variables. The main covariate of interest is Protected, which we measure in units of $10,000. Protected measures the amount of financial assets that a particular household is allowed to keep and still qualify for Medicaid reimbursement. A higher level of Protected corresponds to a more generous (less means tested) Medicaid program. The mean (median) amount of Protected assets in our sample is $36,345 ($18,152) with a standard deviation of $36,135. Protected varies across households depending on state of residence (s), marital status (m), and household assets according to the following formula: Protectedims = Assets ims ≤ Minimum ms ⎧ Assets ims if ⎪ Minimum + . 5 * ( − ) if Minimum Assets Minimum ⎨ ms < Assets ims <Maximum ms . ms ims ms ⎪ Assets Maximum if ims ≥ Maximum ms ms ⎩
(1.2)
The state sets the level for the minimum and maximum amount of assets protected by the Medicaid program, –Minimumms and Maximumms respectively, within the constraints imposed by Federal
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
15
law. We calculate Protectedims for each individual in our sample, based on their assets, marital status, and the specific state Medicaid rules detailed in Appendix A. By including a dummy for whether the individual is married and a full set of state fixed effects we control, respectively, for any fixed differences across married and single individuals or across individuals in different states in their demand for private long–term care insurance. The covariates (X) are designed to control for demographics that may directly affect insurance demand, perhaps through their effect on asset levels or perhaps through other means. (We do not directly control for assets in equation (1.1) because of its potential endogeneity, although we have verified that controlling flexibly for net worth decile does not in fact affect the results). Protected is therefore identified off of two-way and three-way interactions between state, marital status, and assets. We note that the state fixed effects allow us to control flexibly for a number of other potentially important determinants of demand for private long-term care insurance. They condition out any differences across states in the price and quality of nursing homes, which may affect demand for long-term care insurance. They also condition out any differences across states in the Medicaid program that may influence insurance coverage but are the same for married and single individuals within a state or individuals of different asset levels within a state. These include, for example, the Medicaid rules regarding the nature and extent of coverage provided for home health care, and the Medicaid reimbursement rates relative to private payer rates in the state. Our estimates therefore focus precisely on the impact of Medicaid eligibility rules for nursing home coverage on long-term care insurance demand. A potential concern with estimating equation (1.1) is that—as equation (1.2) makes clear—Protected is a function of assets, and therefore savings decisions, which may themselves be affected by Medicaid rules. Thus assets may be endogenous to insurance purchase decisions. Indeed, there is empirical evidence that the savings of the elderly appear to respond to the incentives embodied in Medicaid’s rules for eligibility for coverage for long-term care expenditures (Coe 2005). This is consistent, more generally, with the evidence that savings decisions are affected by the incentives provided by means tested public insurance programs (see e.g., Hubbard, Skinner, and Zeldes 1995; and Gruber and Yelowitz 1999). To address the potential endogeneity of assets to Medicaid rules, we calculate predicted assets for each household based on a reduced form
16
Brown, Coe, and Finkelstein
prediction model that uses only plausibly exogenous demographic characteristics to predict asset accumulation. Specifically, we estimate: Log( Assets)ist = Xist′ δ + υ ist .
(1.3)
We estimate the asset equation in logs because the highly skewed nature of the asset distribution results in a much better fit in predicting log assets than assets. We define assets to be Medicaid-taxable assets; these are the same as net worth for single individuals, but exclude housing wealth from net worth for married individuals, since housing wealth is not treated as a Medicaid-taxable asset for married individuals. As covariates we include the same set of covariates used in X in equation (1.1) that we described above. We also include a marital status dummy since savings behavior may well differ across single and married individuals. Note that we do not use state dummies—or state Medicaid rules—in predicting wealth. The goal of equation (1.3) is not to develop the best prediction model of assets but to isolate the portion of assets that can be explained by plausibly exogenous demographic characteristics rather than asset accumulation decisions that are themselves endogenous to the state Medicaid rules. We estimate equation (1.3) using the full data sample, and household weights. Estimation of the prediction equation (1.3) yields an R-squared of 0.24. Using the results of equation (1.3), we generate predicted assets for each individual in the sample. We then use predicted assets—instead of actual assets—as well as the individual’s state of resident and marital status to calculate the amount of assets that would be protected by the Medicaid program. We refer to these protected assets calculated using predicted rather than actual assets as Protected_Hat. Thus, Protected_Hat represents the amount of assets the Medicaid program would disregard if the household’s actual assets were as predicted by their characteristics. By contrast, Protected denotes the amount of assets the Medicaid program would protect based on their actual (potentially endogenous) assets. Like Protected, Protected_Hat is measured in units of $10,000. The mean (median) value of Protected_Hat in our sample is $43,121 ($39,929), with a standard deviation of $34,348. In the results reported below, we estimate equation (1.1) by instrumental variables, instrumenting for Protected with Protected_Hat. In all of our regression estimates we use the HRS household weights. We adjust the standard errors to allow for an arbitrary variance-covariance matrix in the error term within each state. To take account of the sam-
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
17
pling variation in the predicted variable Protected_Hat (Murphy and Topel 1985) we also report standard errors from a non-parametric bootstrap. Specifically, we bootstrap the prediction equation (equation 1.2) and for each iteration of the bootstrap, calculate predicted assets, use these to calculate Protected_Hat, and then estimate equation (1.1) using Protected_Hat as an instrument for Protected on the drawn sample; we run 200 iterations of the bootstrap. In practice, the standard errors are not affected much by this procedure; we report both sets of standard errors in the results below. Because we are using multiple waves of the HRS, we calculate Protected using the state rules and individual demographics in effect in the year in which the interview takes place. As mentioned above, 21 states, affecting 32 percent of the sample, experience real changes in the community spouse asset disregards between 1991 and 2000. In addition, about 5 percent of our sample changes marital status over the waves 1996–2000 In principle, these changes in state rules and marital status over time provide us with a fourth source of variation in Medicaid asset protection rules faced by an individual. We do not, however, believe that such changes in Medicaid asset protection are a particularly clean or useful source of variation, as it is unclear for these individuals which Medicaid asset protection rules were in effect—and thus the relevant rules—when the individual was considering whether to purchase longterm care insurance. Although we report estimation results for the full sample, our preferred specification limits the sample to the approximately three-fifths of the original sample (17,623 observations consisting of 7,923 unique individuals) who did not change martial status between 1996 and 2000 and are from states whose Medicaid rules did not change in real terms since 1991. Our estimates of crowd out become larger and more precise in this sub-sample, which is consistent with greater measurement error in the full sample in the relevant Medicaid rules in effect when an individual is making his long–term care insurance coverage decision. Finally, it is worth noting a potential limitation to our approach is that we are using current predicted assets, while what matters for the Medicaid asset tax is the assets an individual has at the time of nursing home entry. This will bias against finding an effect of Medicaid. In practice, however, the relatively low rates at which the elderly appear to spend down their assets over their retirements suggest that this may not be too great of a problem (see e.g., Hurd 1989; Hurd 2002; and Mitchell and Moore 1997).
Brown, Coe, and Finkelstein
18
4.
Crowd Out Estimates
Table 1.3 reports the main results from estimating equation (1.1) by instrumental variables, using Protected_Hat to instrument for Protected. The first column shows the results for the whole sample. The coefficient on Protected is –0.0056, and is statistically significant at the 10 percent level. The point estimate suggests that a $10,000 increase in the amount of assets an individual can retain while qualifying for Medicaid is associated with a 0.56 percentage point decline in long-term care insurance coverage. The remaining columns report analysis when the sample is limited to individuals who face constant Medicaid rules. While we lose almost two-fifths of our observations due to these data cuts, we believe this sub-sample will provide a cleaner estimate of the impact of Medicaid on long-term care insurance coverage. Consistent with this view, column 2 indicates that the estimated effect of Medicaid on long-term care insurance demand is larger (and more statistically significant) in the constant Medicaid rules sub-sample than in the full sample. The point estimate on Protected rises to –0.109, and is statistically significant at the 5 percent level. This suggests that a $10,000 increase in the amount of assets a household can hold and still be eligible for Medicaid is associated with a 1.1 percentage point decline in the probability of holding long-term care insurance. The results in column 2 constitute our preferred specification, and we use these results for our central estimate. The remaining columns of table 1.3 explore the sensitivity of our central estimate to using different sources of variation to identify the effect of Medicaid protected asset rules on long-term care insurance demand. As discussed above, variation in Protected_hat comes from the two-way interaction of predicted assets with state, the two-way interaction of predicted assets with marital status, the two-way interaction of marital status with state, and the three way interaction of marital status, predicted assets, and state. To investigate whether each of these sources of variation yields similar results, columns (3) through (6) show the results in which we control one by one for various sources of variation, and therefore identify only off of the others. Specifically, in column (3) we add a control for predicted assets interacted with marital status, in column (4) we add controls for predicted assets interacted with state dummies, and in column (5) we add controls for marital status interacted with state dummies. Finally, in column (6) we include controls for all two-way interactions (predicted
Protected
Full Sample
Constant Medicaid Rules Sub-sample
(1)
(2)
(3)
(4)
(5)
(6)
–0.0056*
–0.0109**
–0.0093
–0.0103*
–0.0172**
–0.0153
(0.0031)
(0.0048)
(0.0060)
(0.0050)
(0.0064)
[0.0045]
[0.0048]
[0.0052]
[0.0055]
[0.0053]
[0.0079]
Asset HAT * Married
Asset HAT * State
Married * State
All two-way Interactions
15,576
15,576
15,576
15,576
Additional controls Observations
24,841
15,576
(0.0112)
Notes: Table reports the results of estimating equation (1) by instrumental variables (IV); Protected_Hat is used as an instrument for Protected. All regressions use household weights. Protected and Protected_Hat are measured in units of $10,000. “Constant Medicaid rules sub-sample” restricts the sample to individuals whose marital status did not change between 1996 and 2000 and who are in states whose real Medicaid asset disregards did not change between 1991 and 2000. Columns (3) through (6) add additional controls to the specification in column (2), as indicated in the row labeled “additional controls.” Specifically, column (3) adds an interaction between predicted assets and marital status, Column (4) adds interactions between predicted assets and state dummies, Column (5) adds interactions between marital status and state dummies, and column (6) adds all three of these sets of two-way interactions in the previous three columns. Standard errors allowing for an arbitrary variance covariance matrix within each state are (in parentheses); standard errors calculated using a non-parametric bootstrap are [in square brackets] (see text for more details). ***, **, * denotes statistical significance at the 1 percent, 5 percent, and 10 percent level respectively, based on the standard errors (in parentheses). All regressions include indicator variables for marital status, state fixed effects, education categorical dummies, gender, occupation, industry, number of children up to five, age, retired, HRS cohort, race, Hispanic, and HRS wave, as well as education categorical dummies interacted with the indicators for marital status, gender, occupation, industry, number of children up to five, age, retired, HRS cohort, race, Hispanic, and HRS wave.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
Table 1.3 Main Crowd Out Estimates
19
20
Brown, Coe, and Finkelstein
assets by marital status, predicted assets by state, and married by state) so that the only variation used to identify Protected_Hat is the three-way interaction of state by marital status by predicted assets. Although the analysis often loses power when various sources of identifying variation are eliminated, the results indicate that the coefficient on Protected always remains negative and roughly of the same magnitude as the –0.011 in the baseline specification; it various from –0.0093 to –0.017 depending on the specification. The fact that all the sources of variation yield similar estimates increases our confidence in the empirical strategy and our baseline estimates. Table 1.4 reports results from a number of additional sensitivity analyses. Column 1 replicates the IV estimates from our preferred specification (table 1.3, column 2). One potential concern is that two other aspects of Medicaid vary across state by marital status and may also affect insurance demand: the treatment of income and estate recovery practices.5 Multi-collinearity in various Medicaid program rules’ generosity could produce a misleading estimate of the impact of Medicaid asset rules. Moreover, the impact of these other features of Medicaid on long-term care insurance demand are of independent interest. Column 2 therefore adds two variables to control for these two features. The variable “Income” measures the amount of income (in units of $10,000) the household is allowed to keep and still qualify for Medicaid; this varies across states and within state by marital status. “Liens” is an indicator variable for whether a state will put a lien on a house when one spouse is in the nursing home in order to recoup expenses upon the death of the community spouse. This practice means that the house is no longer a bequeathable asset for married couples and the house is only a temporarily protected asset; there is no change for single households since the house is not a protected asset for them in any state. Appendix A describes the state income and housing (“liens”) rules in more detail. The results in column 2 of table 1.4 show the expected positive coefficient on “Liens,” but the positive coefficient on “Income” is the opposite of what was expected. Neither coefficient is statistically significant, and an F-test indicates that they are not jointly significant (not shown). Perhaps most importantly, inclusion of these variables does little to change the parameter of interest, the coefficient on Protected. As discussed previously, the variation in our variable of interest Protected occurs mostly in the range of 20th to 60th percentile of the asset distribution of married individuals (see figure 1.1). Therefore, column 3 shows the results limiting the sample to this (albeit endogenous) range;
Protected
Baseline (1)
Other Medicaid Parameters (2)
20th to 60th Pctile of Asset Distribution (3)
Age 55–61 (4)
Age 62–69 (5)
High School Education or Less (6)
Some College or More (7)
–0.0109**
–0.0114**
–0.0223*
–0.0136**
–0.0043
–0.011**
–0.012*
(0.0048)
(0.0055)
(0.0130)
(0.0053)
[0.0048]
[0.0049]
[0.0127]
6,029
Income
(0.0085)
(0.0052)
(0.006)
[0.0067]
[0.0068]
[0.0059]
[0.010]
8,552
7,024
9,452
6,124
0.0363 (0.4285)
Liens
0.0087 (0.0175)
Observations
15,576
15,576
21
Note: Table reports the results of estimating equation (1) by instrumental variables (IV) on the constant Medicaid rules sub-sample; Protected_Hat is used as an instrument for Protected. All regressions use household weights. Protected and Protected_Hat are measured in units of $10,000. Column (1) replicates the baseline results (from table 1.3, column (2)). Column (2) adds controls for other Medicaid rules, specifically the amount of income (in units of $10,000) that a household can keep and still qualify for Medicaid (“Income”), and an indicator variable for whether a state will put a lien on a house when one spouse is in a nursing home in order to recoup expenses upon the death of the community spouse (“Liens”). Column (3) limits the sample to individuals in the 20th to 60th percentile of the asset distribution of married individuals (which is the range in which the variation in Protected is largest). Columns (4) and (5) look separately at individuals aged 55–61 and 62–69. Columns (6) and (7) look separately at individuals with a high school education or less and individuals with some college education or more. Standard errors allowing for an arbitrary variance covariance matrix within each state are (in parentheses); standard errors calculated using a non-parametric bootstrap are [in square brackets] (see text for more details). ***, **, * denotes statistical significance at the 1 percent, 5 percent, and 10 percent level respectively, based on the standard errors (in parentheses). All regressions include indicator variables for marital status, state fixed effects, education categorical dummies, gender, occupation, industry, number of children up to five, age, retired, HRS cohort, race, Hispanic, and HRS wave, as well as education categorical dummies interacted with the indicators for marital status, gender, occupation, industry, number of children up to five, age, retired, HRS cohort, race, Hispanic, and HRS wave.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
Table 1.4 Sensitivity Analysis: Alternative Specifications and Sub-samples
22
Brown, Coe, and Finkelstein
as expected, the point estimate increases in absolute value. However, even with a doubling of the point estimate—to –0.0223 (standard error = 0.0130), the results still imply that even if all of the states decreased the amount of protected assets to the minimum allowable under federal law in 2000, the vast majority of individuals in our sample would remain without private insurance. Columns (4) and (5) report the results from doing the analysis separately for younger ages (55–61) and older ages (62–69), respectively. The sample specification suggests the effect is stronger on younger ages— which may be because these individuals are more likely to be buying during the time of the analysis and thus the state rules in effect at that time are more likely to be the relevant ones. Columns (6) and (7) report results for, respectively, those with a high school education or less and those with some college or more; the results are substantively and statistically indistinguishable. Finally, we have verified (in results not reported) that estimation of the reduced form OLS—in which Protected is replaced by Protected_Hat on the right hand side of equation (1)—yields qualitatively similar results to the instrumental variables estimation of equation (1.1), in which Protected is instrumented for with Protected_Hat. The coefficients on this reduced form estimation tend to be somewhat smaller (although still statistically significant) than the instrumental variables version; for example, a reduced form estimation of our preferred specification (shown in table 1.3, column 2) yields a coefficient on Protected_Hat of –0.0052 (statistically significant at the 5 percent level) compared to the IV estimate of –0.109. This is consistent with the introduction of measurement error in using Protected_Hat instead of Protected to measure the Medicaid rules faced by a given household. By contrast, estimating equation (1.1) with Protected rather than Protected_Hat on the right hand side results in a positive coefficient; this suggests that the issue of the potential endogeneity of assets to the Medicaid rules is in fact quantitatively important for our estimates. 5.
Simulated Effects of Potential Medicaid Reforms
The preceding analysis suggests a statistically significant crowd out effect of Medicaid on demand for private long-term care insurance. Our central estimate suggests that a $10,000 increase in the amount of assets a household can hold and still be eligible for Medicaid is associated with a 1.1 percentage point decline in the probability of hold-
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
23
ing long-term care insurance. Relatedly, these findings suggest that increasing the stringency of Medicaid’s means testing—i.e., decreasing the amount of protected assets—would induce greater demand for private insurance. In this section we provide a gauge of the magnitude of these crowd out estimates by exploring their implications for the effect of potential Medicaid reforms in the size of the private long-term care insurance market. In particular, we consider what our estimates imply for how much long-term care insurance holdings would increase if all of the states decreased the amount of protected assets to the minimum allowable under federal law in 2000. These minimum federally allowable asset protection laws were $16,824 for a married couple, and $2,000 for a single individual. Currently, only three states (Arkansas, D.C., and Oregon) have rules this stringent, while 26 states have the most generous asset protection rules allowed by federal law. Given current state rules and the distribution of assets in the data, we estimate that this change would decrease the average protected assets for an individual in our sample by a little under $25,000. Our point estimates in column 4 therefore imply that this decrease in asset protection would be associated with an increase in private long–term care insurance coverage by about 2.7 percentage points. This represents about a 30 percent increase in coverage relative to the current coverage rate of 9.1 percent. However, it suggests that the vast majority of the individuals in our sample would remain without private insurance. We also considered what our results imply for private long-term care insurance coverage if the minimum federally allowed asset protect laws were reduced by half (to $8,421 for a married couple and $1,000 for a single individual) and all states were to set their asset protection laws at their new minimum. Given the current state rules and the distribution of assets in the data, we estimate that this (out-of-sample) change would decrease the average protected assets for an individual in our sample by almost $30,000, or by $5,000 more on average than the previous reform we considered. Our crowd out estimates imply that a $30,000 decline in the amount of protected assets would be associated with a 3.3 percentage point increase in private long-term care insurance coverage rates. While this represents a more than one-third increase over current insurance coverage rates, it would still leave over 85 percent of individuals in the sample without private insurance. These empirical findings are broadly consistent with the life-cycle simulation-based evidence in Brown and Finkelstein (2004b). They find
Brown, Coe, and Finkelstein
24
that recent policy reforms adopted in several states to allow individuals who purchase private insurance to qualify for Medicaid coverage while retaining substantially more assets would have relatively little effect on the implicit tax that Medicaid imposes on private insurance, and hence little effect on demand for private insurance. Importantly, however, Brown and Finkelstein (2004b) estimate that Medicaid’s implicit tax has a large crowd out effect on private insurance demand. Changes in asset protection by themselves, however, do not affect this implicit tax much because as long as Medicaid remains a secondary payer, even without any asset limits to Medicaid eligibility a large portion of private insurance benefits are redundant of what Medicaid would otherwise have paid. (By the same token, removing the secondary payer status without changing the Medicaid asset limits similarly leaves a large Medicaid implicit tax.) Our empirical findings, coupled with the simulationbased evidence in Brown and Finkelstein (2004b) thus underscore the importance of understanding not just the size of the crowd out effect, but also the mechanism behind it in considering the likely impact of potential reforms to the public program on private demand. 6.
Conclusion
Long-term care is a large, and largely uninsured, potential expense facing the elderly. Medicaid serves as the insurer of last resort. As the baby boomers age, long-term care expenditures are expected to rise substantially, and with them Medicaid expenditures. This will put increasing pressure on state and federal budgets. As a result, increasing attention is focused on how public policy can stimulate the private long-term care insurance market. This paper looks empirically at the effects of the Medicaid program on private long-term care insurance demand. We draw on the substantial variation in the level of assets that an individual can protect from Medicaid based on an individual’s state of residence, marital status, and asset holdings to identify the impact of Medicaid on private longterm care insurance demand. Our estimates suggest that more generous Medicaid asset protection is associated with less private long-term care insurance coverage. Our central estimate implies that a $10,000 increase in the amount of assets a household can retain while qualifying for Medicaid coverage of long-term care expenditures is associated with a 1.1 percentage point reduction in long-term care insurance coverage.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
25
Although our findings point to a crowd out effect of Medicaid asset protection on long-term care insurance demand, they also suggest that even large scale reductions in Medicaid asset protection are unlikely to stimulate private insurance coverage among most of the elderly population. We estimate that, if all states were to adopt the most stringent asset eligibility requirements allowed by federal law in 2000— $16,824 for a married couple and $2,000 for a single individual—and thereby decrease average protected assets by about $25,000, overall demand for private long-term care insurance would rise by 2.7 percentage points, leaving almost 90 percent of the elderly still without private insurance. These empirical findings complement recent simulation research (Brown and Finkelstein 2004b) which also suggest that changes in Medicaid’s asset protection rules would do little to address the lack of private long-term care insurance among most of the elderly. At the same time, Brown and Finkelstein (2004b) find that Medicaid may have a substantial crowd out effect on long-term care insurance demand through the large implicit tax it places on the benefits from private long-term care insurance policies. Changes in Medicaid asset rules appear to not have much affect on this implicit tax, which may explain the simulation and empirical evidence that changes in Medicaid asset rules do not appear to have much effect on demand for private insurance. Together, these findings raise the important question of whether it is feasible to design Medicaid in a way that reduces the implicit tax it places on private insurance, and thus the constraints it appears to place on private insurance demand. Notes We are grateful to the Robert Wood Johnson Foundation, the National Institute of Aging, and the Campus Research Board at the University of Illinois at Urban-Champaign for financial support, and to Jim Poterba for helpful comments. 1. This leaves a remaining one-quarter of expenses that are covered by Medicare. However, this apparently large Medicare share is somewhat misleading. About half of Medicare long-term care spending consists of Medicare’s home health care benefit, which is a genuine long-term care service. However, the other half comes from Medicare’s coverage of short-term, skilled nursing home facilities following an acute hospital stay; this is not the custodial nursing home care that accounts for the vast majority of nursing home days and is covered by private long-term care insurance and by Medicaid, and is somewhat misleadingly included in long term care spending estimates (Congressional Budget Office 2004; US Congress, 2000).
26
Brown, Coe, and Finkelstein
2. Understanding Medicaid’s implicit tax also helps explain the ostensibly puzzling finding that men and women purchase private insurance in very similar proportions, despite substantially higher loads on male policies. Since women have much higher expected lifetime long-term care utilization, the expected proportion of long-term care expenditures paid for by Medicaid is higher for women than men of the same asset levels, and thus the Medicaid implicit tax on private insurance is higher for women than for men. Indeed, Brown and Finkelstein (2004b) show that the net loads on polices—i.e., the load on the net benefits from the private policy, which omits any benefits paid by the private policy that Medicaid would otherwise have paid—are quite similar for men and women. 3. Consistent with this, using our empirical strategy we find statistically insignificant effects of current (1996–2000) Medicaid rules on long-term care insurance coverage for individuals who are 70 and older (mean age of 79) and who therefore may have been at the prime buying age under a very different set of rules (results not reported). We also show in the sensitivity analysis below that the crowd out effects we estimate in our 55–69 year old sample are stronger at younger ages within this range. 4. We specify equation (1.1) as a linear probability model because it allows us to handle instrumental variables most flexibly; as we discuss in more detail below, we are concerned about endogeneity of the right hand side variable Protected and therefore estimate equation (1.1) by instrumental variables. We have confirmed, however, that the marginal effects from Probit specifications evaluated at the mean yield nearly identical results to the linear probability model specified in equation (1.1). 5. Of course, many other aspects of Medicaid vary across state—such as reimbursement rates for nursing homes and whether and how much coverage is provided for home care. An advantage of our strategy is that because we do not use cross-state differences in Medicaid to identify its effects, we purge these differences and are able to focus on the effect of one particular Medicaid parameter of interest. 6. For the less common case when both spouses need nursing home care, they are essentially treated as two single individuals in terms of the treatment of assets, income and housing. The one exception is that some states set a lower threshold for the amount of assets the couple can keep ($3,000 combined instead of $2,000 each). 7. In 2000, the FEDMIN and FEDMAX were, respectively, $16,824 and $84,120 (Stone 2002). They are indexed to the CPI but have otherwise remained unchanged between the 1991 and 2000 period. 8. The five state exceptions are AR, DC, NY, OR, and SC. 9. These are summary stats on the 2000 HRS, using financial assets. The sample is limited to married households aged 60–70. The cut points in the distribution are identical for the age 75+ sample (20th–58th percentile). 10. The 26 states in this category are: AK, AZ, CA, CO, FL, GA, HI, IL, KY, LA, MA, MD, ME, MI, MS, MO, ND, NE, NV, OK, SD, VT, WA, WI, WV, and WY. 11. The 15 states in this category are: CT, ID, IN, KS, MT, NC, NH, NJ, OH, PA, RI, TN, TX, UT, and VA. 12. Our specific sources for state Medicaid rules from 1991–2000 are: Bruen et al (1998); Congressional Research Service (1993); Horvath (1997); Kassner and Shirley (2000); the National Association of Medicaid Directors; Price (1996); Sabatino and Wood (1996); Schwab (1998); Stone (2002), and telephone calls to particular states.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
27
References Brown, Jeffrey, and Amy Finkelstein (2004a). “Supply or Demand: Why is the Market for Long-Term Care Insurance So Small?” NBER working paper no. w10782, September. Brown, Jeffrey, and Amy Finkelstein (2004b). “The Interaction of Public and Private Insurance: Medicaid and the Long-Term Care Insurance Market,” NBER working paper no. w10989, December. Bruen, Brian K., Joshua M. Wiener, Johnny Kim, and Ossai Miazad (1999). “State Usage of Medicaid Coverage Options for Aged, Blind, and Disabled People.” Urban Institute Press, August, Washington, D.C. Cutler, David (1996). “Why Markets Don’t Insure Long-Term Risk?” Unpublished mimeo. Available at http://econweb.fas.harvard.edu/faculty/dcutler/papers/ltc_rev.pdf. Coe, Norma B. (2005). “Financing Nursing Home Care: New Evidence on Spend-Down Behavior,” MIT dissertation, September. Congressional Budget Office (2004). “Financing Long-Term Care for the Elderly,” April. Congressional Budget Office (1999). “Projections on Expenditures for Long-term Care Services for the Elderly,” March. Congressional Research Service (1993). “Medicaid Source Book: Background Data and Analysis,” (The Yellow Book). Finkelstein, Amy, and Kathleen McGarry (2006). “Private Information and its Effect on Market Equilibrium: New Evidence from Long-Term Care Insurance,” American Economic Review, forthcoming. Finkelstein, Amy, Kathleen McGarry, and Amir Sufi (2005). “Dynamic Inefficiencies in Insurance Markets: Evidence from Long-term Care Insurance,” American Economic Review Papers and Proceedings, May. Gruber, Jonathan (2003). “Medicaid” in Moffitt, Robert (ed.), Means Tested Transfer Programs in the United States. Chicago: University of Chicago Press, 15–77. Gruber, Jonathan, and Aaron Yelowitz (1999). “Public Health Insurance and Private Savings,” The Journal of Political Economy, 107(6):1249–1274. Health Insurance Association of America (HIAA) (2000). “Who Buys LTC Insurance in 2000?” Hurd, Michael (2002). “Are Bequests Accidental or Desired?” RAND Labor and Population Program working paper series DRU-3010. Hurd. Michael (1989). “Mortality Risks and Bequests,” Econometrica, 57(4):779–813. Horvath, Jane (1997). “Medicaid Financial Eligibility for Aged, Blind and Disabled: Survey of State Use of Selected Options,” Portland, ME: National Academy for State Healthy Policy. Hubbard, R. Glenn, Jonathan Skinner, and Stephen Zeldes (1995). “Precautionary Saving and Social Insurance,” The Journal of Political Economy, 103(2):360–399. Kang, Hyojin, Alan Mathios, and Sharon Tennyson (2004). “Medicaid Crowd-Out of Private Insurance: The Case of Long-Term Care,” unpublished mimeo.
28
Brown, Coe, and Finkelstein
Kassner, Enid, and Lee Shirley (2000). “Medicaid Financial Eligibility for Older People: State Variations in Access to Home and Community-Based Waiver and Nursing Home Services,” The Public Policy Institute, AARP. Kunreuther, Howard (1978). Disaster Insurance Protection: Public Policy Lessons. New York: Wiley. MetLife Mature Market Institute (2002). “MetLife Market Survey of Nursing Home and Home Care Costs 2002.” Mitchell, Olivia, and James F. Moore (1997). “Retirement Wealth Accumulation and Decumulation: New Developments and Outstanding Opportunities,” NBER working paper no. w6178, September. Murphy, Kevin, and Robert Topel (1985). “Estimation and Inference in Two-Step Econometric Models.” Journal of Business and Economic Statistics, 3(4):370–379. National Association of State Medicaid Directors (NASMD). “Aged, Blind & Disabled State Summaries.” Available at http://www.nasmd.org/research/ABD/abd.htm. National Governors Association (2004). “Medicaid Squeezes State Budgets,” available at http://www.nga.org/portal/site/nga/menuitem.6c9a8a9ebc6ae07eee28aca9501010a0/ ?vgnextoid=04ee137945da2010VgnVCM1000001a01010aRCRD&vgnextchannel=759b8f2 005361010VgnVCM1000001a01010aRCRD. National Center for Health Statistics (2002). Health, United States, 2002 with Chartbook on Trends in the Health of Americans. Hyattsville, MD. Norton, Edward (2000). “Long-term Care,” in A.J. Culyer and J.P. Newhouse (eds.), Handbook of Health Economics, volume 1, chapter 17. Amsterdam: Elsevier Science. Pauly, Mark (1990). “The Rational Nonpurchase of Long-Term-Care Insurance,” Journal of Political Economy, 98(1):153–168. Price, Richard (1996). “Medicaid: Spousal Impoverishment Protections.” CRS Report for Congress 95–1053 EPW, March 4. Sabatino, C.P., and Wood, E. (1996). Medicaid Estate Recovery: A Survey of State Programs and Practices. Washington, D.C.: AARP Public Policy Institute. Schwab, Carol A. (1998). “Medicaid Eligibility for Nursing Home Benefits: A Guide for North Carolinians.” North Carolina Cooperative Extension Service, North Carolina State University, Raleigh, North Carolina. Sloan, Frank A., and Edward C. Norton (1997). “Adverse Selection, Bequests, Crowding Out, and Private Demand for Insurance: Evidence from the Long-term Care Insurance Market,” Journal of Risk and Uncertainty, (15):201–219. Stone, Julie Lynn (2002). “Medicaid: Eligibility for the Aged and Disabled,” CRS Report for Congress RL31413. Washington, D.C.: Penny Hill Press. United States Congress, Committee on Ways and Means (2004). “2004 Green Book.” Washington, D.C.: Government Printing Office. United States Congress, Committee on Ways and Means (2000). Green Book 2000: Background Material on Programs Under the Jurisdiction of the Committee on Ways and Means. Washington, D.C.: Government Printing Office.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
29
Appendix A: Overview of Medicaid Rules This appendix discusses in some detail the rules that govern financial eligibility for Medicaid. We focus primarily on the rules regarding the amount of financial assets that an individual or couple is permitted to keep and still receive Medicaid reimbursement for nursing homes. We label this value Protected in our empirical work in the main body of the paper; it is our key variable of interest. We also briefly discuss the rules regarding Medicaid income disregards and Medicaid treatment of housing wealth; we briefly explore the impact of these variables in the sensitivity analysis presented in table 1.4. Medicaid rules vary considerably across states. Within each state, the rules pertaining to a single individual who goes into a nursing home differ from those pertaining to married individuals who go into a nursing home whose spouse remains in the community. Since the differential rules within state by marital status form the main source of our empirical identification strategy, we discuss these differences in some detail. Appendix table 1.A1 provides a summary of the various Medicaid rules for nursing home expenses for single and for married people in each state as of 2000. We now discuss each briefly in turn. At the end of this section, we turn to a discussion of state rule changes over the 1991 to 2000 period.
Rules for Single Individuals in 2000 Medicaid rules for single individuals are relatively simple and uniform across states. Maximum amount of retainable assets: They can keep no more financial wealth than the Medicaid-specified asset limit. Anything above this must go toward paying for the care. In 2000, the modal asset limit was $2,000 (which nearly 70 percent of states used). The remaining states have an asset limit that ranges from $1,500 to $6,500. Maximum amount of retainable monthly income: They can keep no more monthly income than the Personal Needs Allowance (PNA). In 2000, this ranged from $30 to $77 per month. Rules regarding housing wealth: They must sell their house (and use the proceeds to pay for that care), unless there is a chance of recovery or a dependent child living in the house.
Rules for Married Individuals (One Spouse in NH, One in Community)6 Treatment of financial assets: When one spouse enters a nursing home, total household financial (non-housing) assets (A) are attributed evenly between the
Brown, Coe, and Finkelstein
30
Table 1.A1 Medicaid Eligibility Parameters by State, 2000 Rules for Married Individuals with Community Spouse
State
Max. Amount of Retainable Assets
Max. Amount of Retainable Monthly Income (PNA)
Min . Amount of Retainable Assets for Community Spouse (STATEMIN)
Min. Amount of Retainable Monthly Income for Community Spouse (CSIL)
Is Lien Put on Community Spouse’s House?
No Change Asset Rules for Married Individuals (1991–2000)
AK
2,000
75
84,120
2,103
0
1
AL
2,000
30
25,000
1,407
1
0
AR*
2,000
40
16,824
1,407
0
1
AZ
2,000
76.80
84,120
2,103
0
1
CA
2,000
35
84,120
2,103
1
1
CO
2,000
50
84,120
1,407
1
0
CT
1,600
52
16,824
1,407
1
1
DC*
2,600
42
16,824
2,103
0
1
DE
2,000
70
25,000
1,407
1
0
FL
2,000
35
84,120
2,103
0
1
GA
3,000
30
84,120
2,103
0
1
HI
2,000
30
84,120
2,103
1
1
IA
2,000
30
24,000
2,103
0
0
ID
2,000
30
16,824
1,407
1
1
IL
2,000
30
84,120
2,103
1
1
IN
1,500
50
16,824
1,407
0
1
KS
2,000
30
16,824
1,407
0
1
KY
2,000
40
84,120
2,103
0
1
LA
2,000
38
84,120
2,103
0
1
MA
6,500
60
84,120
1,407
1
0
MD
2,500
40
84,120
2,049
1
0
ME
3,000
40
84,120
2,103
0
0
MI
2,000
60
84,120
2,103
0
0
MN
3,000
67
23,774
1,407
1
1
MO
2,000
30
16,824
1,407
0
1
MS
2,000
44
84,120
2,103
0
1
MT
2,000
40
16,824
2,103
1
1
NC
2,000
30
16,824
2,103
0
1
ND
3,000
40
84,120
2,103
0
1
NE
4,000
50
84,120
1,407
0
0
NH
2,500
50
16,824
2,103
0
1
NJ
4,000
35
16,824
1,407
0
1
NM
2,000
45
31,290
1,407
0
1
Rules for Single Individuals
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
31
Table 1.A1 (Continued) Medicaid Eligibility Parameters by State, 2000 Rules for Married Individuals with Community Spouse
State
Max. Amount of Retainable Assets
Max. Amount of Retainable Monthly Income (PNA)
Min . Amount of Retainable Assets for Community Spouse (STATEMIN)
Min. Amount of Retainable Monthly Income for Community Spouse (CSIL)
Is Lien Put on Community Spouse’s House?
No Change Asset Rules for Married Individuals (1991–2000)
NV
2,000
35
84,120
2,103
0
0
NY*
3,600
50
74,820*
2,103
1
0
OH
1,500
40
16,824
1,407
0
1
OK
2,000
50
84,120
2,103
0
0
OR*
2,000
30
16,824
1,407
0
1
PA
2,400
30
16,824
1,407
0
1
RI
4,000
50
16,824
1,407
SC*
2,000
30
66,480
1,662
0
0
SD
2,000
30
84,120
1,407
0
0
TN
2,000
30
16,824
1,407
0
1
TX
2,000
45
16,824
2,103
0
1
UT
2,000
45
16,824
1,407
0
1
VA
2,000
30
16,824
1,407
0
1
VT
2,000
47.66
84,120
1,407
0
1
WA
2,000
41.62
84,120
1,407
0
1
WI
2,000
45
84,120
1,875
1
1
WV
2,000
50
84,120
1,407
0
0
WY
2,000
30
84,120
2,103
0
1
Rules for Single Individuals
1
Notes: For all states, the maximum amount of assets the community spouse is allowed to keep (STATEMAX) is the same (and equal to FEDMAX) unless the state is denoted with a * in which case STATEMAX=STATEMIN. Treatment of housing for single individuals is not described in the table since it is the same in all states (see text). PNA stands for Personal Needs Allowance CSIL stands for Community Spouse Income Limit. In 2000, FEDMIN and FEDMAX were $16,824 and $84,120, respectively. Source: Stone (2002), Sabatino and Wood (1996), and authors’ corrections based on telephone conversations with particular states where other sources disagreed with those listed here.
32
Brown, Coe, and Finkelstein
two spouses. From this even attribution, the spouse that goes into a nursing home is allowed to keep only the Medicaid-specified income and asset limits for single individuals (i.e., $30–$77 a month of income and approximately $2,000 of assets). The main source of variation used in our empirical work is the amount of assets that the community spouse is allowed to keep. This amount depends on the total amount of household financial assets (A) and the state rules regarding the minimum and maximum assets that the community spouse is allowed to keep (STATEMIN and STATEMAX respectively). A community spouse whose share of the assets is below the state minimum allowable (STATEMIN) is allowed to take assets from the nursing home spouse to top-up their asset level up to STATEMIN. In setting the minimum and maximum amount of assets the community spouse can retain, the states are constrained to set a minimum (STATEMIN) that is at least as high as the federal minimum (FEDMIN) and to set the maximum (STATEMAX) no higher than the Federal maximum (FEDMAX).7 In all states therefore, married couples with combined assets of less than the Federal Minimum ($16,824 in 2000) face the same treatment of their assets (they are allowed to keep all of them). Furthermore, in all but five states, married couples with combined assets of more than twice the Federal maximum (i.e., $168,240 in 2000) are allowed to keep the same amount ($84,120).8 The differential treatment across states of married couples’ assets therefore occurs mainly for couples with assets between the Federal Minimum and twice the Federal Maximum. In 2000, these limits were $16,824 and $168,240 respectively, and correspond to the 20th and 58th percentile of financial assets for married households in the 2000 HRS. Married households in this “affected range” have average assets of $74,266.9 For married couples with assets between the Federal Minimum and twice the Federal Maximum, there is substantial variation across states in the amount of assets they are allowed to retain. This variation arises from where states choose to set their State Minimum and State Maximum allowable retainable assets. Figure 1.A1 compares the amount of assets that the community spouse can keep under the two most common state rules. In the first, which is used in 26 states, the state sets both the minimum and maximum allowable assets (STATEMIN and STATEMAX) to the Federal Maximum. Under these rules, the household faces a 0 percent marginal tax rate on its assets until it reaches the state minimum amount of allowed retainable assets, at which point it faces a 100 percent marginal tax rate on all further assets. We refer to this as “Case 1A” and illustrate it with the dark line in figure 1.A1.10 The second most common set of state rules (which apply in 15 states) set the state minimum equal to the federal minimum allowable retainable assets, and the state maximum equal to the federal maximum allowable retainable assets.
Medicaid Crowd-Out of Private Long-Term Care Insurance Demand
33
Figure 1.A1 Medicaid Marginal Tax Rates Married Households
This is shown by the dashed line in figure 1.A1. In this case, the marginal tax rate faced on assets goes from 0 percent, to 100 percent, to 50 percent and then back to 100 percent, as shown in the figure.11 Figure 1.1 graphs the difference in the amount of assets a community spouse can keep (as a function of total household financial assets) if they are in a Case 1A state relative to a Case 2A As is readily apparent, this difference is nonmonotonic in the couples’ assets. The difference in the amount of allowable retained assets is quite substantial. For example, if all married couples were to move from Case 1A states to Case 2A states, the average difference in the amount a household is allowed to keep would be $8,277 in the whole sample, which is approximately 2 percent of average assets. For those in the “affected range” (20th to 58th percentile of assets i.e., between $16,824 and $168,240), this move would on average allow them to keep $21,715 more in assets, which represents 29 percent of average financial assets in this range. The maximum change in the amount of assets that the household would be able to keep is $67,296 and occurs for households with assets of $84,120. Finally, we note that while we have focused on the two major types of state rules, there are ten other states whose rules differ from those in Cases 1A and 2A. For the sake of brevity we do not discuss these rules in detail; they are summarized in table 1.A1. Like the two more common cases discussed above, the difference across states in the treatment of assets is non-monotonic in the couples’ assets across these other cases as well (relative to each other or the two more common cases). It is also non-trivial in magnitude.
34
Brown, Coe, and Finkelstein
Treatment of Income: Income is split based on the “name on the check” rule, rather than evenly between the two spouses as is the case for assets, in all but two states. The institutionalized individual is allowed to keep the same amount of income as a single household (defined above as the PNA). The community spouse is allowed to keep an unlimited amount of income if it is in her name. If the community spouse’s income is below a minimum amount known as the community spouse income limit (CSIL), then she is eligible to keep enough of her institutionalized spouse’s income to bring her total income up to that limit. This minimum income amount varies across states from approximately $1,407 to $2,133 per month as shown in table 1.A1. Treatment of Housing: The house of a community spouse is left out of the calculation of assets or income and it is completely protected for the community spouse during her lifetime. It may be of unlimited value. However, about onequarter of states will put a lien on this house, which allows the state to collect money from the sale of the house to reimburse them for their Medicaid outlays upon the sale of the house or the death of the community spouse. We refer to such states as LIEN states. Enforcement of estate recovery practices varies across states (Sabatino and Wood 1996).
Changes in Rules All of the preceding discussion applies to the state rules in 2000. For purposes of the empirical work, it is important to know whether states changed their rules at some point, as individuals might have purchased or considered purchasing long-term care insurance under a different set of rules. As discussed in the text, there was a major change in rules in 1988 (effective in 1989), which motivated our focus on an age group who was likely to be of buying age after 1988. We also tracked down information on rule changes between 1991 and 2000. There is no central database for state-specific Medicaid eligibility rules. We compiled a timeline for these state-specific rule changes by collecting a variety of different sources that covered the different years; where sources disagreed, we telephoned the relevant agency in the state to ascertain the correct information. We were unable to obtain state-specific information between 1989 and 1991.12 There were no major changes in the Medicaid rules for single individuals during this time period. However, for married individuals, 21 states changed their assets protection rules for the community spouse between 1991 and 2000 (see appendix table 1.A1). In addition, 27 states changed the allowable income limit for institutionalized individuals (PNA). Finally, over our period 13 states introduced estate recovery plans (LIENS).
2 Unemployment Insurance Savings Accounts Martin Feldstein, Harvard University and NBER Daniel Altman
Executive Summary We examine a system of Unemployment Insurance Saving Accounts (UISAs) as an alternative to the traditional unemployment insurance system. Individuals are required to save up to 4 percent of wages in special accounts and to draw unemployment compensation from these accounts instead of taking state unemployment insurance benefits. If the accounts are exhausted, the government lends money to the account. Positive accounts earn the return on commercial paper and negative accounts are charged that rate. Positive UISA balances are converted into retirement income or bequeathed if the individual dies before retirement age. Negative account balances are forgiven at retirement age. Money taken by an unemployed individual from a UISA with a positive balance reduces the individual’s personal wealth by an equal amount. In this case, individuals fully internalize the cost of unemployment compensation. UISAs provide the same protection to the unemployed as the current UI system but with less of the adverse incentives. The key empirical question is whether accounts based on a moderate saving rate can finance a significant share of unemployment payments or whether the concentration of unemployment among a relatively small number of individuals implies that the UISA balances would typically be negative, forcing individuals to rely on government benefits with the same adverse effects that characterize the current UI system. To resolve this issue we use the Panel Study on Income Dynamics to simulate the UISA system over a 25 year historic period. Our analysis indicates that almost all individuals have positive UISA balances and therefore remain sensitive to the cost of unemployment compensation. Even among individuals who experience unemployment, most have positive account balances at the end of their unemployment
Feldstein and Altman
36
spell. Although about half of the benefit dollars would go to individuals whose accounts are negative at the end of their working life, less than one-third of the benefits go to individuals who also have negative account balances when unemployed. These facts suggest a substantial potential improvement in the incentives of the unemployed. The cost to taxpayers of forgiving the negative balances is substantially less than half of the taxpayer cost of the current UI system. Our analysis of the distribution of lifetime UISA payments and taxes of household heads shows the top quintile gaining a small cumulative amount while those in the bottom quintile lose a very small cumulative amount. Other quintiles are small net gainers. 1.
Introduction
Unemployment insurance (UI) exists to provide protection against the hardship that would otherwise be caused by unemployment. Unfortunately, it also distorts incentives in ways that cause inefficient increases in total unemployment. In this paper we analyze empirically a modification of the traditional unemployment insurance system. We show that this alternative, based on individual savings accounts, can substantially reduce the adverse incentive effects of the existing unemployment insurance system without any decrease in the protection of those who become unemployed. Since the working paper version of this paper was distributed in December 1998 there have been several papers that have supported the basic premise of our study that the primary problem of the uninsured is the lack of liquidity rather than the need for insurance as such (see Chetty 2005; Card, Chetty, and Weber 2006; Shimer and Warning 2006; and the papers cited in those studies.) In our approach, unemployed individuals receive immediate access to liquidity through a combination of pre-unemployment asset accumulation and access to an additional government line of credit when they are unemployed. There is also a pure insurance feature in the form of forgiveness at the time of retirement of the accumulated unemployment loan balance, if any, that results from the government line of credit. Stiglitz and Yun (2005), Brown, Orszag, and Snower (2006), Orszag and Snower (1997), and Coloma (1996) provide a more explicit analysis of the type of unemployment insurance saving account developed in this study. Our analysis is therefore fundamentally different from previous studies investigating how the adverse incentive effects of the current
Unemployment Insurance Savings Accounts
37
tax-financed unemployment insurance system could be reduced by changes in basic program parameters such as the level and duration of benefits, the experience rating rules, and the provision of bonuses for hiring the unemployed (Baily 1978; Meyer 1995; Mortenson 1994; Blanchard and Tirole 2006). The basic system that we examine requires each individual to save a fraction of his or her wage takings in a special Unemployment Insurance Saving Account (UISA). If the individual loses his job and would be eligible for unemployment benefits under the current UI rules, he withdraws an amount equal to the regular UI benefits from his personal UISA. If the funds in the account are not sufficient to pay the benefit, the government lends the necessary amount to the account. Accounts earn a market rate of return on existing balances and pay the government the same return on borrowed amounts. At retirement age the funds in the UISA are merged into the individual’s IRA or other investmentbased retirement saving plans. An individual who dies with a positive account balance bequeaths that amount to his spouse or other heirs. The government cancels the debt of those who reach retirement age (or die before then) with negative account balances. More details of the plan are described in the third section. All unemployed individuals would therefore receive the same cash amounts during spells of unemployment from their UISAs as they would under the existing unemployment insurance rules. Their full current protection is thus maintained. Any individual whose UISA always has a positive balance (and who expects that it would remain positive) would completely internalize the cost of unemployment benefits and therefore would not have any incentive to increase in an inefficient way the frequency or duration of his unemployment spells because of the availability of those benefits. The adverse incentive problem would arise only for individuals who expect that they will retire or die with negative balances in their UISAs. For such individuals, the benefits received because of additional unemployment have no personal cost. They face the same incentives to excess unemployment that they would under the existing unemployment system, but without the discipline that comes from employer experience rating.1 The feasibility of this savings account approach to unemployment insurance depends on the extent to which insured unemployment is concentrated in a subgroup of the population. Some individuals experience a disproportionate share of the total unemployment days and this concentration applies to insured unemployment as well as to
38
Feldstein and Altman
unemployment in general (see, e.g., Meyer and Rosenbaum 1996). If the insured unemployment is sufficiently concentrated, individuals may not be able to finance their own unemployment benefits by saving moderate shares of their earnings in the UISAs.2 The use of individual savings accounts to finance unemployment benefits would be irrelevant if those who collect benefits would typically have negative balance accounts and therefore be drawing on the government guarantee. Before carrying out the current research, we regarded this as a potentially serious problem that could make the savings account approach unworkable. It is important therefore to assess the proportion of individuals who develop negative account balances and the extent to which UI benefits are now paid to individuals who would have negative accounts. To do so we examine the extensive experience represented by individuals in the Panel Study of Income Dynamics (PSID). Our analysis of these data implies that approximately 5 percent of employees would retire or die with negative account balances and that only about half of all benefits from the UISAs would be paid to such individuals. The cost to the government of the unrecovered loans in the negative accounts is substantially less than the cost of the current unemployment insurance system, permitting a reduction in the current distortionary payroll tax as well as in the distortionary effects of the existing benefit system. These findings tell us that the savings account approach to unemployment insurance, combined with a government guarantee, can be an economically viable policy option. The second section of this paper summarizes the existing unemployment insurance system and discusses the various ways in which it causes a rise in the frequency and duration of unemployment. In the third section we describe the operation of the Unemployment Insurance Savings Accounts and the several alternative funding options that we will analyze in the remainder of the paper. The fourth section then uses the Panel Study of Income Dynamics to show how a large group of individuals would have been affected by these alternatives to the existing unemployment insurance system over periods of up to 25 years. The fifth section analyzes the distributional effects of the UISA system and of the associated reduction in the UI payroll tax. The sixth section briefly discusses the potential mutually reinforcing effect of a behavioral response of unemployment to the improved incentives implied by the UISA system. There is a brief concluding section.
Unemployment Insurance Savings Accounts
2.
39
The Current Unemployment Insurance System and Its Problems
To discuss the distorting effects of the current UI system on the frequency and duration of unemployment, it is useful to begin by reviewing the current system’s basic rules. Although unemployment insurance rules differ among the individual states, the basic structure is quite similar throughout the country. An individual who has worked a sufficient amount or earned a sufficient amount of wage income during the past year is eligible to receive benefits if he or she is laid off.3 Benefits are approximately 50 percent of the unemployed individual’s previous gross wage, subject to a minimum weekly benefit floor that raises the percentage for low wage workers and a maximum weekly benefit ceiling that lowers the percentage for high wage workers. Some states also provide supplementary benefits if the unemployed individual has a dependent spouse or children. The average weekly benefit in 1997 was $193 (an amount that rose to $262 in 2003). Benefits are generally payable for a maximum of 26 weeks. Benefits are subject to federal personal income tax but not to the Social Security payroll tax (or the equivalent tax for self-employed individuals). Some states include unemployment benefits in taxable income for assessing the state income tax. Unemployment benefits are financed by taxes levied on firms by the state governments. Each firm pays a percentage of the earnings of each employee up to a relatively low maximum level that varies among the states; the maximum taxable wage for the UI tax was only $7,000 in most states in 1997; by 2006, some states still had a maximum of $7,000 but Massachusetts had increased to $14,000. The percentage that each firm pays depends on the past experience of that firm as a UI taxpayer and of its employees as UI benefit recipients. This “experience rating” system is intended to cause the firms to internalize the cost of the unemployment benefits of its employees. However, because there are both lower and upper limits on these state UI tax rates, many firms are not effectively experience rated, i.e., an additional layoff or an additional week of unemployment by a former employee would have no effect on the firm’s UI tax bill.4 The most obvious and most thoroughly researched effect of the existing UI system on unemployment is the increase in the duration of the unemployment spells. By reducing the cost of remaining unemployed, UI benefits induce individuals to have longer spells in order to search
40
Feldstein and Altman
for a better job or simply to enjoy some leisure or the opportunity to work at home. There is substantial evidence that the level and maximum duration of UI benefits affects the level of reservation wages and the duration of unemployment spells (Feldstein and Poterba 1984; Katz and Meyer 1990; Moffitt 1985). This evidence and the underlying search theory (e.g., Baily 1977) would seem to provide a clear case that UI induces excessive search. Calculations for a typical employee imply that the combination of UI benefits and personal taxes reduces the net cost of search to about one-fourth of the unemployed individual’s potential marginal product.5 However, against this presumption that UI benefits cause excessive search it is sometimes argued that in the absence of unemployment benefits individuals would not search long enough because they lack access to the capital market and therefore could not finance the optimal amount of search. The Unemployment Insurance Savings Accounts provide the access to funds to finance the optimal search with a reinsurance mechanism provided by the government in case the individual’s fund is exhausted. Individuals with positive UISA balances are motivated to take the costs and benefits of search into account correctly6 while those with permanently negative account balances are in the same situation as today’s UI recipients. The current system of UI benefits not only increases the duration of unemployment of those who are unemployed but also increases the frequency of temporary layoffs. Because of the subsidy inherent in the current system of benefits, individuals will prefer to be unemployed rather than to work at a time when the marginal revenue product of their labor is depressed (Feldstein 1976). Empirical research (Card and Levine 1994; Feldstein 1978) shows that this is true for seasonal unemployment and other forms of temporary layoffs. If individuals were instead to finance such spells of unemployment by drawing from their own UISAs they would have no incentive to choose excessive spells of temporary unemployment. The current payment of benefits to workers who become unemployed reduces the wage premium required to compensate employees for the risk of being laid off. The unemployment insurance system thus lowers the cost of production in firms that have above-average layoff rates, whether for cyclical, seasonal or other reasons. The reduced cost of production in such unemployment-intensive firms lowers the price of the associated product and therefore raises its share in GDP relative to what it would be without the UI subsidy. This shift in the mix of
Unemployment Insurance Savings Accounts
41
products raises the overall unemployment rate. Once again, the UISAs would eliminate the subsidy for those with positive balances, leading to an adjustment in wages that raises the cost of those products and of those firms that contribute most to overall unemployment. 3. Unemployment Insurance Savings Accounts: Five Alternative Options In a UISA plan each individual (or that individual’s employer7) would be required to contribute a fraction of wage income to a UISA. The magnitude of this mandatory saving is limited in different ways in the alterative options described in this section. The options specify different limits on the maximum annual income to which the saving fraction applies. Some options permit deposits to stop when the accumulated balance reaches a specified fraction of the individual’s annual earnings. The funds deposited to the UISA would come from pretax income, just as current UI tax payments do. They would accumulate tax-free. If the funds are withdrawn in lieu of UI benefits, they would be considered taxable income just as UI benefits are today. It would be natural to apply the tax to the funds withdrawn in retirement or by heirs, just as 401k and traditional IRA funds are taxed. Alternatively, the funds deposited in UISAs could come from after-tax income and subsequent withdrawals would be untaxed (as they are in Roth IRAs). The funds in the UISAs might be invested by the individuals in a variety of ways similar to IRA or 401k investments. Since the government augments the funds in those accounts that have insufficient funds to meet benefits during spells of unemployment, the nature of the investments might be more tightly regulated than the funds in IRAs or 401k accounts. We shall not explore this issue here but will discuss calculations based on two alternative investment strategies. In the more conservative strategy, the UISAs are invested in money market mutual funds that earn the six-month commercial paper rate of interest. In an alternative investment strategy, the accounts are invested in a continuously rebalanced mixture consisting of 60 percent corporate stock (represented by the Standard and Poors 500 index) and 40 percent corporate bonds (based on the Salomon Brothers bond index). This portfolio produced a real rate of return of 5.9 percent for the period from 1946 to 1992 (Feldstein and Ranguelova 1998; extending the sample period to the present time would have very little
42
Feldstein and Altman
effect on this rate of return.) We reduce this yield by 0.4 percentage points (to 5.5 percent) to allow for administrative costs of the portfolio management. We assume that individuals must choose permanently between the commercial paper strategy and the 5.5 percent strategy and may not change their rates of return at any time in the analysis. The individuals whose account balances are insufficient to pay the benefits to which they are entitled can borrow from the government at the same rate as they earn in their account. The results that we present below show that our conclusions are not sensitive to the choice between these two rates of return. The amounts that individuals would withdraw from their UISAs when they are eligible for benefits under current UI rules are the same as the benefits that they would receive under the current UI system. Each of the five options that we study requires individuals to contribute 4 percent of their wages up to the maximum amount specified by that particular option. We assume a five year start-up period during which individuals contribute to their UISAs but during which the unemployed continue to receive government UI benefits under the current system. After describing these five alternatives, we use the Panel Study of Income Dynamics data to assess how the choice among these options affects the performance of the system. Option 1 High Saving Base Individuals contribute 4 percent of earnings up to a maximum of about three times the average weekly wage. For the first year of the PSID data (1967), the annual wage ceiling for our UISA contributions is $15,000. This ceiling then grows in proportion to the growth of the average weekly wage, reaching about $52,000 in 1991, the last year of the PSID sample which we studied. Option 2 Low Saving Base The saving base in option 1 permits a rapid accumulation of UISA balances by high wage earners but is arguably unnecessarily high for two reasons. First, the dollar limit on the level of weekly benefits is equivalent to providing a 50 percent replacement rate only up to a level that is approximately equal to the average weekly wage. Second, the frequency of unemployment declines as wages rise and is substantially less among individuals with above average wages.
Unemployment Insurance Savings Accounts
43
Option 2 requires individuals to contribute 4 percent of all earnings up to a level only slightly above the median wage, a level that is also roughly equal to the level of wages on which UI benefits are currently based. For the first year of the PSID data (1967), the wage ceiling is taken to be $6,000. This ceiling then grows in proportion to the growth of the average weekly wage, reaching $21,000 in 1991. Option 3 Target Account Fund In the first two options, individuals are required to continue contributing to their UISAs regardless of their unemployment experience and of the amounts accumulated in their accounts. Since benefits are 50 percent of wages (up to the ceiling) and last for no more than six months in a spell, the maximum benefit that can be drawn in a single spell is only one-fourth of a year’s earnings. Most spells of unemployment are substantially shorter than six months, the median spell being less than ten weeks in almost all years. Spells would be even shorter with the change in incentives provided by the savings account approach. Option 3 therefore provides that the individual stops contributing to the UISA when the accumulated balance reaches 50 percent of the individual’s wage income in the previous year or 50 percent of the ceiling amount in option 2 if that is smaller. Option 4 Experience-Based Target Account Fund Individuals with substantial risk of unemployment should have larger account balances than those who are less likely to be unemployed. Option 3 can be modified to reduce the target level of the account fund for those with low unemployment experience and to increase it for those with substantial unemployment experience. Option 4 provides one such modification. Individuals save until the fund reaches the sum of (1) 30 percent of the individual’s annual wage (or of the wage ceiling specified in options 2 and 3 if that is lower) plus (2) twice the individual’s total UI withdrawals during the past two years. Consider, for example, an individual with $30,000 of base period annual wage income who has had two eight-week spells of compensated unemployment during the past two years. The UISA withdrawals during those 16 weeks would be $3,200.8 Option 3 would require that individual to save 4 percent of wages until the fund balance reached $10,500. In contrast, option four would change this to the sum of $6,300 (30 percent of the specified earnings “ceiling”) plus $6,400 (twice the benefits withdrawn in the past two years), a total of $12,700. The
Feldstein and Altman
44
accumulation would still be at a rate of 4 percent of the first $20,000 of wages. Accumulating more in this way should not be seen as a penalty since individuals own the funds in their UISAs and can eventually consume or bequeath them. The funds are there as a buffer to reduce the government’s risk in guaranteeing that benefits will be paid even if the UISAs have insufficient funds. Option 5 An Experience Rating Component Although a system of UISAs can substantially reduce many individuals’ incentives for longer or more frequent spells of unemployment, it does eliminate the effect of experience rating.9 While experience rating is not needed to correct incentives when individuals have positive UISA balances, it would improve incentives when individuals have negative balances and are therefore motivated to act as if the government provides their unemployment benefits. Option 5 combines the basic contribution requirement of Option 2 with a requirement that employers pay the first five weeks of benefits in each spell.10 This has two potentially favorable effects on incentives. First, by reducing the amount that individuals withdraw from their accounts during any given spell, they are more likely to have a positive balance and therefore to be sensitive to the cost of providing benefits. Second, even for those individuals with negative accounts, the employer has an incentive not to create excess unemployment, the traditional role of experience rating. 11 4.
Analyzing the Options with the PSID Data
The Panel Study of Income Dynamics (PSID) provides longitudinal data on individuals that are well-suited to analyzing the feasibility of substituting UISAs for the existing unemployment insurance system.12 The PSID contains linked interview data on a national probability sample of households and subsequent split-offs for the period from 1967 through 1991. The data for each year and each head of household include the total UI benefits received as well as demographic and labor market information. We focus our analysis on the individuals who were heads of households in 1967. The head of household can be either a single individual or the individual in a household who is designated as the head of the unit. We include only those individuals who were still in the sample and under age 65 in 1972, i.e., in the first year after the five-year period
Unemployment Insurance Savings Accounts
45
in which individuals make deposits to the savings accounts but draw benefits only from the regular state UI program. We then follow these 2,773 individuals until the end of the data sample in 1991 or until the year in which they die, retire or otherwise leave the sample. Separate tabulations are presented in the appendix for the sub-sample of 1,990 individuals who by 1991 are no longer employed or in the PSID sample. We impute retirements at age 65 for all workers. Our procedure is very straightforward. For each of the options, we accumulate funds according to the rules of that option. The accounts earn the commercial paper rate in one simulation and a 5.5 percent real return in the alternative simulation. Those are also the rates charged on negative balances. In each year, starting with the sixth year of the simulation, we subtract from each account the UI benefits that the individual received in that year. The key results for the full sample based on the commercial paper rate of return are shown in table 2.1 and for the 5.5 percent rate of return in Appendix table 2.A1. The corresponding results for the sub-sample of individuals who had died, retired or otherwise left the sample are presented in Appendix tables 2.A2 and 2.A3. All of the results in these tables assume no behavioral response to the change in unemployment incentives. We return to this issue in the sixth section and present some results that suggest the sensitivity of our calculations to possible behavioral responses. Consider first the results in table 2.1 for Option 1. Row one shows that only 5.2 percent of all the individuals in the PSID sample of household heads had negative balances at the end of their time in the PSID. If employees correctly anticipated their final condition, almost 95 percent of employees would be fully sensitive to the cost of unemployment benefits. The figure is essentially the same (5.3 percent, from table 2.A1) if we look only at “finishers,” i.e., those who had died, retired or left the sample by 1991. A stricter measure of sensitivity is the fraction of individuals whose accounts were ever negative. Row 2 shows that only 6.8 percent of individuals ever had negative accounts. For the remaining 93 percent, receiving unemployment benefits would mean drawing from their own funds. Although many of these did not have any spells of unemployment, their positive UISA balance gave them a stronger inventive to avoid unemployment than they had in the existing UI system. Most individuals who become unemployed have positive accounts at the beginning and at the end of their spell. Row 3 shows that in only
46
Table 2.1 Analysis of Alternative Options with PSID Data Percentages Option 1
Option 2
Option 3
Option 4
Option 5
All employees (1) Negative terminal balance (2) Negative balance ever
5.2 6.8
6.6 8.8
6.7 8.9
7.0 9.4
5.6 7.1
24.3 19.9
30.5 25.2
30.9 25.5
33.1 26.8
27.3 21.6
23.4
24.5
24.3
25.3
21.0
31.1 44.1
38.7 54.8
39.3 55.5
42.2 58.3
34.5 48.9
27.4
36.1
36.7
39.3
28.7
Eligible unemployment spells (3) Negative balance at end of spell (4) Negative balance & negative terminal balance Employees with negative balances ever (5) Return to positive terminal balance Unemployment compensation dollars (6) Negative balance & negative terminal balance (7) Negative terminal balance Net government payments
The analysis is based on the full sample of 2,773 original heads of households in the PSID sample from 1967 to 1991, including those still working in 1991. The calculations use the commercial paper rate of return on UISA balances, both positive and negative. See text for definitions. Option 1: High wage base Option 2: Low wage base Option 3: Target account fund Option 4: Experience-based target account fund Option 5: Experience rating: Employer pays five weeks
Feldstein and Altman
(8) Percent of total UISA payments
Unemployment Insurance Savings Accounts
47
one-quarter of the unemployment spells in which benefits are received is the account negative at the end of the spell. This reflects the fact that most spells are short and come when the individuals have accumulated enough in their UISAs to finance the spell. The result is similar when we look at the terminal UISA balances at the end of the PSID experience, taking into account future deposits to the UISA and future spells of unemployment; 20 percent of accounts in which individuals receive UI benefits are negative at the end of the spell and at the end of the PSID sample (shown in row 4 of the table). Even individuals whose accounts are negative at some point in time (those shown in row 2) need not assume that they will remain negative. About one-quarter of those individuals (23.4 percent, shown in row 5) have positive terminal balances when they retire or leave the sample, or in 1991 when the sample ends. When we turn from the numbers of individuals to the amount of UISA payments, we find that only 31.1 percent of UISA benefits are paid in spells that end with negative balances for individuals that also have negative terminal balances (row 6). This is the group most likely to assume that the costs of the UISA benefits will be borne by the government. A somewhat higher percentage of UISA payments go to individuals in spells that are not necessarily negative but that lead to a negative terminal balance (44.1 percent, shown in row 7). Members of this group may be sensitive to the cost of UISA payments during those spells (and years) when their balances are positive and they have not yet concluded that the terminal balance will be negative. Row 8 shows the dollars paid by the government and not subsequently repaid by the individuals as a percentage of the total UISA payments received by all individuals. Because the benefits in the UISA system are the same as the UI benefits in the current UI system, this ratio is also the ratio of the tax-financed UISA benefits to the total taxfinanced benefits under the existing UI system. The estimate of 27.4 percent shown in row 8 means that the cost of the UISA to taxpayers with option 1 and no behavioral response would be only 27.4 percent of the cost to taxpayers of the existing UI system.13 The distorting effects of the existing UI payroll tax are separate from the distorting effects of the benefits conditioned on unemployment. In thinking about the incentive effects implied by these results it is important to consider the effects on both the duration and the frequency of unemployment. Individuals who have positive balances or who believe that they will end their careers with positive balances will not
48
Feldstein and Altman
want to become unemployed and, if they do became unemployed, will have no incentive to remain unemployed. The evidence that most individuals have positive balance accounts and that they end their careers with positive balance accounts shows that (assuming they understand this likelihood) they generally face the cost of unemployment and, in contrast to the situation with the existing UI rules, would have little incentive for behavior that would increase either the frequency or duration of unemployment. The percentage of funds withdrawn by individuals who end their working life (or are working when the PSID ends) with accounts that have negative balances appears to suggest that in about 44 percent of the unemployed weeks the individuals that currently receive unemployment insurance benefits would face the same adverse incentives under the UISA system as they do under the current UI system. While eliminating the distortion for the other 56 percent of the weeks would be a substantial achievement of the UISA approach, this 56 percent figure understates for several reasons the extent of the improvement in incentives that would occur. First, some of the spells of unemployment that now end with permanently negative balances would never occur if the individuals’ incentives were different. Second, since the duration of the spells would be shortened by the change in incentives, fewer of the spells would actually lead to negative balance accounts. While we do not have an estimate of the effect of these behavioral responses, we believe that the evidence here indicates that the incentives would be improved for substantially more than half of the spells and weeks of current insured unemployment. We return below (in the sixth section) to consider the implications of a behavioral response to the improved incentives. The results with the other options are generally similar to the results with option 1, although options 2, 3, and 4 involve a smaller saving base and therefore more frequent negative balances. More specifically, shifting to a lower wage base for the saving requirement (option 2 requires a 4 percent saving rate on wage income up to about the median wage) raises the percentages that become negative or end negative by about two percentage points. With the lower amount of saving, the percentage of spells that end with negative account balances becomes 30.5 percent instead of the 24.3 percent with option 1 (row 3). The number of spells that end with negative balances and that also go on to negative terminal balances rises from 20 percent with option 1 to 25 percent (row 4). Capping the saving requirement for workers with
Unemployment Insurance Savings Accounts
49
above average incomes thus reduces the sensitivity but still leaves most individuals who experience unemployment with a positive account balance. Option 3 allows workers to stop contributing to their accounts when the balance reaches 50 percent of their savings wage base under option 2. This has essentially no effect on any of the performance measures. For those who experience no unemployment, it reduces substantially the amount of lifetime saving that is required in the UISAs without changing the likely sensitivity of this group or others. Option 4 makes the target level of accumulation for the UISA depend on the recent unemployment experience, lowering the basic target to only 30 percent of the savings wage base under option 2 but then adding the benefits drawn in the past two years to this amount. The positive and negative effects are reasonably balanced, causing little affect on the various performance measures shown in table 2.1. For those with little or no unemployment, this option permits a substantially lower rate of saving. Option 5 requires the employer to pay the first five weeks of unemployment benefits in every spell and is otherwise similar to option 2. Only after the five weeks does the individual draw benefits from the UISA. This makes the employer directly sensitive to the cost of unemployment for all employees, including those with negative balances. It also reduces substantially the probability that employees who experience unemployment will develop a negative balance or end their career with a negative balance. Thus row 4 shows that among unemployment spells resulting in negative balances, the percentage of spells from individuals who end their careers (or the time in the PSID sample) with a negative balance falls from 25 percent with option 2 to 22 percent with option 5. 5.
Distributional Effects of Switching to UISA System
The effect on each individual’s disposable income of shifting from the current UI rules to a UISA system depends on the individual’s unemployment experience and the level of the individual’s income. There are three components of the effect: (1) the required saving contribution to the individual’s UISA account; (2) the net balance in the UISA account at retirement age; and (3) the change in the payroll tax payment. The benefits paid during unemployment can be ignored because they are always the same in the two systems.
50
Feldstein and Altman
Any analysis of the distributional effect of shifting from one system to another involves the usual incidence issues about the effect of induced behavioral changes on wages and other pretax factor incomes. These incidence issues are particularly difficult in the current case because the program change involves not only taxes but also transfers conditioned on unemployment experience. We limit our analysis therefore to the nominal analysis, i.e., to the estimated distribution of individual payments with no changes in gross wages or other factor incomes. We assume moreover that all payments are borne by the individuals, regardless of whether they are made by the individual or the firm. A second caveat is necessary about interpreting the distributional effects by income class tabulated in this section. This analysis refers only to heads of households and makes no attempt to incorporate the distributional effects of a shift to a UISA system on others in the same household. The sample is further restricted to those individuals who were less than 45 years old in 1967 in order to study a relatively long working period. Before looking at the calculations, it is useful to note the way that the shift from the existing UI rules to a UISA system affects individuals at two extremes: those with no unemployment during their working life and those whose unemployment is so substantial that they end their working life with a negative UISA balance. For individuals who experience no unemployment, the net present value of UISA saving deposits and the balances withdrawn at retirement is zero (discounting at whatever rate is used to accumulate those balances). Such individuals are net gainers from the switch to the UISA system since the taxes required to fund the benefits of those who have negative final balances are less than the taxes required by the current system to fund all UI benefits. Because the tax is levied on earnings up to a relatively low level ($7,000 in many states in 1997), the favorable tax reduction effect is the same for all individuals above that low level and declines with income below that level. For individuals who experience substantial unemployment and retire with negative accounts, the net discounted present value of the required UISA savings represents a net tax. Against this must be balanced the reduction in the regular UI payroll tax (which is the same reduction as that enjoyed by those with no unemployment). Since the current payroll tax is less than the UISA saving requirement, the reduction in the payroll tax is clearly less than the UISA saving requirement, implying
Unemployment Insurance Savings Accounts
51
that individuals who have negative balances incur a net reduction in the present value of their disposable income. These are of course the two extreme cases. To assess the overall distributional effect of the switch, we divide our sample into lifetime income quintiles based on real mean annual income during the individual’s working years between 1967 and 1991. For each quintile, we calculate the discounted present values as of 1967 of (1) the required savings deposited to UI saving accounts; (2) the funds available at retirement age; and (3) the reduced payroll tax in each year. We estimate the payroll tax reduction of individual I in year t as TAXCUTIt = (1 − reltax) θ t TI It where: reltax is the ratio of the payroll tax with the UISA system to the payroll tax with the existing UI system, as shown in row 8 of table 2.1 for each UISA option; TI It is the taxable wage income for individual I in year t (up to the payroll tax ceiling in that year); and θ t is the national average UI payroll tax rate in year t under the current UI rules. To estimate these values, we assume that the maximum taxable earnings for the UI payroll tax (T tmax ) is $7,000 in 1997 and scale it down in earlier years in proportion to the average weekly earnings in the total private U.S. economy. For each individual, the value of TI It is the lesser of (T tmax ) and that individual’s wage in year t. The national average UI payroll tax rate in year t is estimated as θ t = BEN t /[0.9 T tmax N t ] where BENt is the aggregate national UI benefits paid in year t, 0.9 T tmax is the estimated average taxable earnings for the UI tax (we scale by 0.9 since not all workers will earn $7,000 in 1997 dollars), and N t is the number of individuals in covered employment.14 The present values, calculated using the six-month commercial paper interest rate, are shown in table 2.2; a separate calculation based on the 5.5 percent real rate of return is presented in Appendix table 2.A4. We present estimates for options 1, 2, and 3. The lowest quintile of households corresponds to those in which the head earned an average lifetime income of only $12,293 a year in 1991 dollars during the years that the individual worked between 1967 and 1991. The second and third quintiles had average lifetime earnings in 1991 dollars of $23,976 and $31,948 while the top group had average earnings of $71,561. The first three rows of table 2.2 show the present discounted value in 1967 of the positive terminal UISA balances (the refunded amounts) minus the UISA saving deposits, discounting at the commercial paper discount rate in each year. Thus individuals in the lowest lifetime income quintile paid on average $591 more in UISA saving
52
Table 2.2 Distributional Effects of Shifting from Current UI Rules to a UISA System Income Quintile Lifetime mean annual income (1991$)
1st 12293
2nd 23976
3rd 31948
4th 40977
5th 71561
PDV of positive UISA terminal balances minus UISA saving deposits Option 1
–591
–653
– 768
– 539
–314
Option 2
–573
–622
– 666
– 492
–244
Option 3
–565
–615
– 663
– 488
–243
Option 1
496
675
701
733
782
Option 2
438
595
617
645
688
Option 3
433
588
611
639
681
PDV of payroll tax reductions
PDV of positive UISA terminal balances plus payroll tax reductions minus saving deposits –95
22
–67
94
468
Option 2
–135
–27
–49
153
444
Option 3
–132
–27
–52
151
438
Option 1: High wage base; Option 2: Low wage base; Option 3: Target account fund. See text for other definitions. Present discounted values are calculated using the commercial paper rate.
Feldstein and Altman
Option 1
Unemployment Insurance Savings Accounts
53
deposits (discounted to 1967) than the amount that they had in their UISA accounts (treating negative accounts as zero, since such debts are forgiven) at the time of retirement or death (also discounted to 1967). The $591 amount excludes the value of the benefits received since that does not change as we go from current UI rules to the UISA system. In the aggregate, this bottom quintile receives a disproportionately large share of the benefits relative to the amount that it provides in UISA deposits, causing it to receive a substantial transfer from the UISA system just as it does from the current UI system. But because we are interested in the distributional effects of shifting from the current UI rules to a UISA system, rather than the distributional effects of the UISA system itself, and since the benefits are exactly the same in the two systems, we ignore the benefits and focus on the difference between the amount that the individuals get in refunds at the time of retirement or death (i.e., the positive UISA balances at those times) and the amounts contributed as saving deposits, all discounted to the beginning of the sample. This negative effect is balanced by the positive effect of lower payroll taxes. With option 1, the tax saving associated with replacing the current UI system with a UISA system (as noted in row 8 of table 2.1) is 73 percent of the UI payroll taxes that would otherwise be paid under the current rules. For individuals in quintile 1, the present value of these payroll tax reductions (shown in the first row of the second part of table 2.2) is $496. These tax savings offset most of the PDV excess of UISA savings over balance refunds for this group, leaving a net negative present value cost of $95 for the shift from current UI rules to UISA rules. Since this is the present value of the net effects over the 25 year period, it is essentially too small to be of significant concern. This “loss” is of course before considering any of the potential gains—both financially and in terms of economic efficiency—that would result from the behavioral effects of the shift to the UISA system. The second quintile has a net positive gain of $22, again essentially close to zero when compared to the 25 year present value of the earnings of individuals with average annual earnings of nearly $24,000. The largest effect is the positive gain of the top quintile, a lifetime present value gain of $468, which is also quite small relative to the average annual earnings of more than $70,000 in this group. The results for the other two options are similar, with relatively small lifetime present value losses in the bottom half of the distribution of lifetime income and relatively small lifetime gains in the top half of the distribution.
54
Feldstein and Altman
6. Effects of Unemployment Responses to the Unemployment Compensation System The first section of this paper discussed the various ways in which the existing unemployment insurance system increases the frequency and duration of unemployment. We have also considered how shifting to the UISA system would change these incentives in ways that reduce unemployment. An explicit model of the effect of shifting to a UISA system on the frequency and duration of unemployment would have to deal with individuals’ expectations about the probability that they will shift from an existing positive account balance to a negative account balance at the time of retirement (and therefore should not currently be sensitive to the effect of unemployment on their UI account) or from an existing negative UISA account balance to a positive terminal balance (implying that they should be concerned about the cost of their current unemployment benefits). Although such an analysis lies beyond the scope of this paper, it is clear that the shift to the UISA system would initiate a mutually reinforcing process in which reduced subsidies to unemployment would reduce the frequency and duration of unemployment which would in turn imply that the a larger fraction of UISA payments were from individual account balances rather than from the government. This virtuous spiral would converge to lower probabilities of unemployment and lower durations of unemployment spells than are observed in the historic data. To indicate how such a virtuous spiral might improve the performance and reduce the taxpayer cost of the UISA system, we present simulations of the UISA option 1 on the assumption that all UI spells are reduced by either 10 percent or 30 percent in duration. These simulations are shown in table 2.3. Although a reduction in the frequency of unemployment is not specifically included, the analysis can be regarded as a way of observing the effect of 10 percent or 30 percent fewer unemployed days, regardless of whether this is from changes in frequency or duration. As a rough generalization, the results show that a 30 percent reduction in the amount of eligible unemployment reduces the proportion of individuals who have negative balances or who end with a negative account balance by at least one-quarter. The percentage of spells that end with negative balances and the percentage of spells by individuals who eventually have negative balances
Unemployment Insurance Savings Accounts
55
Table 2.3 Effects of 10% and 30% Reductions in Unemployment Days on the Implications of UISA Option 1 All Employees (1) Negative terminal balance (2) Negative balance ever
No Change
10% Reduction
30% Reduction
5.2 6.8
4.8 5.8
3.7 5.1
24.3 19.9
21.6 18.3
16.7 13.6
23.4
17.4
27.6
31.1 44.1
28.5 42.3
20.5 32.5
27.4
22.0
13.7
Eligible unemployment spells (3) Negative balance at end of spell (4) Negative balance & negative terminal balance Employees with negative balances ever (5) Return to positive terminal balance Unemployment compensation dollars (6) Negative balance & negative terminal balance (7) Negative terminal balance Net government payments (8) Percent of total UISA payments
The analysis is based on the full sample of 2,773 original heads of households in the PSID sample from 1967 to 1991, including those still working in 1991. The calculations use the commercial paper rate of return on UISA balances, both positive and negative. See text for definitions.
at the end of their careers also fall, this time by about one-third. The same is also true of the aggregate dollar value of benefits. This does not show that a 30 percent reduction in aggregate unemployment days is likely but only that, if it does occur, it will cause a large reinforcing decline in the number who face or can expect to face negative balances. The 30 percent reduction in the amount of insured unemployment also has the effect of cutting the tax-financed benefits in half, from 27.5 percent of the current UI benefits with no behavioral response to 13.7 percent with a 30 percent reduction in days with compensated unemployment. 7.
Summary and Conclusion
In this paper we have examined a system of Unemployment Insurance Saving Accounts as an alternative to the traditional unemployment
56
Feldstein and Altman
insurance system. The system requires individuals to save a modest share of wages in special accounts and to draw unemployment compensation from these accounts instead of taking state unemployment insurance benefits. If the accounts are exhausted, the government lends money to the account. Negative account balances are forgiven at retirement age. Positive UISA balances are converted into retirement income or bequeathed if the individual dies before retirement age. Any dollar taken from a UISA with a positive balance reduces the individual’s personal wealth by a dollar. As such, the UISAs cause individuals to internalize the cost of unemployment compensation. The UISAs can therefore in principle provide the same level of protection to the unemployed with less of the adverse incentives that now increase the frequency and duration of unemployment. The key operational question about the feasibility of UISAs is whether accounts based on a moderate saving rate can finance a significant share of unemployment payments or whether the concentration of unemployment among a relatively small number of individuals implies that the UISA balances would typically be exhausted, forcing individuals to rely on government benefits with the same adverse effects that characterize the current UI system. To resolve this issue we use the Panel Study on Income Dynamics to simulate the UISA system over a 25-year historic period. Our analysis indicates that almost all individuals have positive UISA balances and therefore remain sensitive to the cost of unemployment compensation. Even among individuals who experience unemployment, most would still have positive account balances at the end of their unemployment spell. Although about half of the benefit dollars would go to individuals whose accounts are negative at the end of their working life, less than one-third of the benefits go to individuals who currently have negative account balances or who will have negative account balances at the end of their current unemployment spell. All of this suggests a substantial improvement in the incentives of the unemployed. The reduction in the cost to taxpayers of more than 60 percent of the current taxpayer burden represents a substantial further potential improvement in the efficiency of the labor market. Our analysis of the distribution of lifetime UISA payments and taxes shows that the household heads in the top quintile gain a small cumulative amount while those in the bottom quintile lose a very small amount. Other quintiles are small net gainers.
Unemployment Insurance Savings Accounts
57
Notes Martin Feldstein is Professor of Economics at Harvard University and President of the National Bureau of Economic Research. This is a slightly revised version of NBER Working Paper 6860, distributed in December 1998. At that time, Daniel Altman was a graduate student at Harvard University and a NBER-National Institutes on Aging Pre-Doctoral Fellow. The authors are grateful to Richard Freeman, Ed Glaeser, John Gruber, Caroline Hoxby, Larry Katz, Bruce Meyer, Jim Poterba, and members of the Harvard Seminar on Labor Economics back in 1998 and to Raj Chetty, Ivan Werning, and Jim Poterba for more recent comments. Although we planned to do additional work on this problem, Dr. Altman immediately began a career in economic journalism, writing for the Economist, the New York Times, and others. Since it is clear that we will not get back to this, we decided to publish it essentially as originally written. 1. Experience rating can affect firms’ decisions to lay off employees and the duration of unemployment among those on temporary layoff. Although the U.S. unemployment insurance rules provide for experience rating, many firms do not face effective experience rating. We return to these issues below, including an option that provides some of the incentive effects of experience rating. 2. To the extent that the identities of those who will experience large amounts of lifetime unemployment are unknown at the start of their working lives, the social provision and subsidy of unemployment benefits would therefore be a kind of optimal catastrophic insurance. 3. Individuals who quit a job may be eligible for benefits if their quitting is found to be “for just cause.” In some states quitters are eligible after an extensive waiting period. Unemployed individuals who are new entrants to the labor force or reentrants without recent work experience are not eligible for unemployment insurance benefits. 4. On experience rating and its potential effects, see Feldstein (1976). 5. Consider an individual who can earn $100 a day and faces a federal marginal income tax rate of 28 percent, a state marginal income tax rate of 5 percent and a payroll tax rate of 7.65 percent. Taxes reduce the net take-home pay of that individual from the $100 gross pay to $59.35. If the individual is unemployed, he or she receives gross unemployment benefits of $50, subject to a 28 percent federal income tax; the resulting net benefit is therefore $36. The net cost to the individual of remaining unemployed for the day is the difference between these two net amounts or $23.35. By contrast the individual’s marginal product of labor is the sum of the $100 gross pay and the additional $7.65 employer payroll tax. The net cost of remaining unemployed is thus only 22 percent of the marginal product of labor. Bringing this up to date by substituting a 25 percent marginal tax rate for the earlier 28 percent rate would only change these net amounts very little. 6. The opportunity cost of search is still substantially less than the marginal product of labor but the gain from search is also reduced by the same set of marginal tax rates. With the capital market problem solved, the amount of search done by a rational and risk-neutral individual will be optimal. 7. The current UI taxes are paid by employers but the incidence of the tax would presumably be the same if the tax were paid by employees. Similarly, gross wages would adjust down if UISA deposits were made by employers rather than employees, since these deposits are the property of the individual workers and are similar to a form of taxpreferred cash compensation.
58
Feldstein and Altman
8. The maximum weekly benefits are assumed here to be based on income up to $20,000 or a maximum benefit of $200 per week. 9. On the nature and limits of experience rating, see above, page 5. 10. This idea was previously suggested in Feldstein (1975). 11. The experience rating could be strengthened under any of the options by using employers’ experience with unemployment as the basis for the tax used to finance the cost of benefits paid to individuals with negative UISA balances. 12. See Katz (1986) for an earlier use of the PSID data to study unemployment insurance. 13. This ratio is calculated as follows. The denominator is the total UISA payments to all participants over the period from 1972 to 1991. To calculate the numerator, we focus on those individuals who had negative terminal balances (at death, retirement, departure from the sample, or upon reaching 1991). We then identify the last year in which the balance of each of these “negative terminal balance” individuals was positive and ignore government payments in all prior years. We then calculate the sum of (1) the negative balance in that year (the difference between benefits in that year and the sum of the prior positive balance and the savings deposited that year in the account) and (2) any UISA benefits taken in subsequent years by the individual. We subtract from this cumulative total (3) the amounts that the individual paid to the account (actually directly to the government as repayment for past credit) in all subsequent years. The combination of these three terms is the net amount that the government pays to negative accounts in excess of the amounts repaid. 14. The annual values of BENt and Nt are presented in the statistical appendix of each year’s Economic Report of the President.
References Baily, M. (1977). “Unemployment Insurance as Insurance for Workers,” Industrial and Labor Relations Review, 30(4):495–504. Baily, M. (1978). “Some Aspects of Optimal Unemployment Insurance,” Journal of Public Economics, 10(3):379–402. Blanchard, O., and J. Tirole (2006). “The Joint Design of Unemployment Insurance and Employment Protection,” Paper presented at the NBER Summer Institute, July 24. Brown, Alessio J.G., J. Michael Orszag, and Dennis J. Snower (2006). “Unemployment Accounts and Employment Incentives,” Kiel working paper no. 1274. Card, David, Raj Chetty, and Andrea Weber (2006). “Cash-on-Hand and Competing Models of Intertemporal Behavior: New Evidence from the Labor Market.” NBER working paper no. 12639 (Cambridge, MA). Card, D., and P.B. Levine (1994). “Unemployment Insurance Taxes and the Cyclical and Seasonal Properties of Unemployment,” Journal of Public Economics, 53(1):1–29. Chetty, R. (2005). “Why do Unemployment Benefits Lengthen Unemployment Durations?” NBER working paper no. 11760.
Unemployment Insurance Savings Accounts
59
Coloma, C.F. (1996). “Seguro do Desempleo: Teoria, Evidencia y Una Propuesta (Unemployment Insurance: Theory, Evidence and a Proposal),” Cuadernos de Economia, 33(99):256–320. Feldstein, M. (1975). “Unemployment Insurance: Time for Reform,” Harvard Business Review, (March–April). Feldstein, M. (1976). “Temporary Layoffs in the Theory of Unemployment,” Journal of Political Economy, 84(5):937–957. Feldstein, M. (1978). “The Effect of Unemployment Insurance on Temporary Layoff Unemployment,” American Economic Review, 68(5):834–846. Feldstein, M., and J. Poterba (1984). “Unemployment Insurance and Reservation Wages,” Journal of Public Economics, 23(1-2):141–167. Feldstein, M., and E. Ranguelova (1998). “Individual Risk and Intergenerational Risk Sharing in an Investment-Based Social Security System,” NBER working paper no. 6839, December. Katz, L. (1986). “Layoffs, Recall and the Duration of Unemployment,” NBER working paper no. 1825, 1986. Katz, L., and B.D. Meyer (1990). “The Impact of the Potential Duration of Unemployment Benefits on the Duration of Unemployment,” Journal of Public Economics, 41(1):45–72. Meyer, B. (1995). “Lessons from the U.S. Unemployment Insurance Experiments,” Journal of Economic Literature, 33(1):91–131. Meyer, B.D., and D.T. Rosenbaum (1996). “Repeat Use of Unemployment Insurance,” NBER working paper no. 5423, January. Moffit, R. (1985). “Unemployment Insurance and the Distribution of Unemployment Spells,” Journal of Econometrics, 28(1):85–101. Mortenson, D. (1994). “Reducing Supply-Side Disincentives to Job Creation,” in Reducing Unemployment: Current Issues and Policy Options. Kansas City: Federal Reserve Bank. Orszag, J.M., and D. Snower (1997). “From Unemployment Benefits to Unemployment Accounts,” mimeo, Birkbeck College, London. Shimer, R., and I. Werning (2006). “On the Optimal Timing of Benefits with Heterogeneous Workers and Human Capital Depreciation,” NBER working paper no. 12230, May. Stiglitz, J., and Jungyoll Yun (2005). “Integration of Unemployment Insurance with Retirement Insurance,” Journal of Public Economics, 89(11–12):2037–2067.
60
Table 2.A1 Analysis of Alternative Options with PSID Data: Finishers Only Percentages Option 1
Option 2
Option 3
Option 4
Option 5
All employees (1) Negative terminal balance (2) Negative balance ever
5.3 6.4
6.6 8.2
6.6 8.2
6.9 8.5
5.5 6.5
23.9 20.0
30.9 25.8
31.5 26.1
33.4 27.7
27.6 21.4
17.6
19.5
19.4
19.0
15.0
31.9 47.6
39.2 58.6
40.0 58.6
42.7 61.4
34.6 50.6
31.5
40.7
41.2
43.8
31.4
Eligible unemployment spells (3) Negative balance at end of spell (4) Negative balance & negative terminal balance Employees with negative balances ever (5) Return to positive terminal balance Unemployment compensation dollars (6) Negative balance & negative terminal balance (7) Negative terminal balance Net government payments
The analysis is based on the full sample of 1,990 original heads of households in the PSID sample from 1967 to 1991 who were retired, dead or missing by the end of 1991. The calculations use the commercial paper rate of return on UISA balances, both positive and negative. See text for definitions. Option 1: High wage base Option 2: Low wage base Option 3: Target account fund Option 4: Experience-based target account fund Option 5: Employer pays first five weeks of benefits
Feldstein and Altman
(8) Percent of total UISA payments
Percentages Option 1
Option 2
Option 3
Option 4
Option 5
All employees (1) Negative terminal balance (2) Negative balance ever
4.9 5.7
5.8 7.4
6.0 7.8
6.4 8.2
4.8 6.0
20.6 18.2
26.1 22.4
27.1 23.0
28.2 24.1
23.8 18.4
13.6
21.2
22.7
22.4
19.7
27.6 42.4
34.0 49.6
35.2 51.4
37.6 54.5
29.5 42.9
24.6
31.5
33.7
35.4
24.9
Eligible unemployment spells (3) Negative balance at end of spell (4) Negative balance & negative terminal balance Employees with negative balances ever (5) Return to positive terminal balance Unemployment compensation dollars (6) Negative balance & negative terminal balance (7) Negative terminal balance
Unemployment Insurance Savings Accounts
Table 2.A2 Analysis of Alternative Options with PSID Data: 5.5% Return
Net government payments (8) Percent of total UISA payments
The analysis is based on the full sample of 2,773 original heads of households in the PSID sample from 1967 to 1991. The calculations use a 5.5 percent real rate of return on UISA balances, both positive and negative. See text for definitions.
61
Option 1: High wage base Option 2: Low wage base Option 3: Target account fund Option 4: Experience-based target account fund Option 5: Experience rating: Employer pays five weeks
62
Table 2.A3 Analysis of Alternative Options with PSID Data: Finishers Only: 5.5% Return Percentages Option 1
Option 2
Option 3
Option 4
Option 5
All employees (1) Negative terminal balance (2) Negative balance ever
4.9 5.3
5.7 6.9
5.8 7.1
6.2 7.4
4.6 5.4
19.9 18.0
26.8 23.2
27.5 23.8
28.5 25.0
23.7 18.0
8.5
16.5
17.3
16.0
14.7
27.8 45.0
33.9 53.2
35.1 54.3
37.7 58.1
29.4 44.1
28.9
35.6
36.9
39.5
28.0
Eligible unemployment spells (3) Negative balance at end of spell (4) Negative balance & negative terminal balance Employees with negative balances ever (5) Return to positive terminal balance Unemployment compensation dollars (6) Negative balance & negative terminal balance (7) Negative terminal balance Net government payments
The analysis is based on the full sample of 1,990 original heads of households in the PSID sample from 1967 to 1991 who were retired, dead or missing by the end of 1991. The calculations use a 5.5 percent real rate of return on UISA balances, both positive and negative. See text for definitions. Option 1: High wage base Option 2: Low wage base Option 3: Target account fund Option 4: Experience-based target account fund Option 5: Experience rating: Employer pays five weeks
Feldstein and Altman
(8) Percent of total UISA payments
Income Quintile Lifetime mean annual income (1991$)
1st 12293
2nd 23976
3rd 31948
4th 40977
5th 71561
PDV of UISA positive terminal balances minus UISA saving deposits Option 1
–410
–428
–521
–341
–202
Option 2
–397
–410
–454
–325
–161
Option 3
–390
–407
–453
–319
–159
Option 1
331
448
463
482
509
Option 2
314
408
420
437
464
Option 3
292
394
407
423
448
PDV of payroll tax reductions
Unemployment Insurance Savings Accounts
Table 2.A4 Distributional Effects of Shifting from Current UI Rules to a UISA System With 5.5 % Rate of Return
PDV of UISA positive terminal balances plus payroll tax reductions minus saving deposits Option 1
–79
–58
141
307
Option 2
–96
–2
20
–34
112
303
Option 3
–98
–13
–46
104
289
Option 1: High wage base; Option 2: Low wage base; Option 3: Target account fund. See text for other definitions. Present discounted values are calculated using the commercial paper rate.
63
3 Evaluating Effects of Tax Preferences on Health Care Spending and Federal Revenues John F. Cogan, Stanford University R. Glenn Hubbard, Columbia University and NBER Daniel P. Kessler, Stanford University and NBER
Executive Summary In this paper, we calculate the consequences for health spending and the federal budget of an above-the-line deduction for out-of-pocket health spending. We show how the response of spending to this expansion in the tax preference can be specified as a function of a small number of behavioral parameters that have been estimated in the existing literature. We compare our estimates to those from other researchers. And, we use our analysis to derive some implications for tax policy toward HSAs. 1.
Introduction
As Pauly’s (1986) classic review shows, virtually all observers of health policy since Feldstein’s (1973) seminal article have agreed that the tax preference for employer-provided health insurance—under which employer contributions to employee health insurance are deductible to the employer and non-taxable to the employee—encourages overconsumption of health services in the United States. By making health spending in general, and insured health spending in particular, appear less costly than they are, the tax preference gives employees the incentive to take compensation as health insurance rather than cash, even if they would otherwise prefer not to. Both the budget cost of the tax preference, and its potential implications for efficiency in markets for health services, are large. Table 3.1 provides three estimates of the federal revenue loss from the tax preference in 2004. Shiels and Haught (2004) estimate the revenue loss to be $188.5 billion. According to them, the loss from the exclusion from the personal income tax base of employer contributions to employee
Cogan, Hubbard, and Kessler
66
Table 3.1 Cost to the Federal Budget of Existing Tax Preferences for Health Spending 2004 (in billions of dollars) Study Shiels and Haught (2003) Total
JCT (2003)
US Treasury in OMB (2003)
$188.5
Exclusion of employer contribution to HI premiums from… Social security payroll tax base
$52.2
Medicare payroll tax base
$14.2
Personal income tax base
$114.7
$101.0
$123.9
$7.4
$5.9
$6.3
Deduction for out-of-pocket expenses in excess of 7.5% of adjusted gross income
Note: Three studies’ estimation methods differ slightly; see Shiels and Haught (2003) for discussion.
health insurance alone was $114.7 billion. The Joint Committee on Taxation (2003) estimates this loss to be slightly less ($101.0 billion); the Department of the Treasury in the Office of Management and Budget (OMB 2003) estimates it to be slightly more ($123.9 billion). Everyone agrees, however, that it dwarfs the revenue loss from all other health tax preferences (such as the deduction for health spending in excess of 7.5 percent of adjusted gross income), and indeed is the single largest tax preference in the federal budget. According to the Department of the Treasury in OMB (2003), the loss from the employer exclusion surpasses the loss from the deductibility of mortgage interest, state and local property taxes, and all capital gains tax preferences. Indeed, the only personal tax preferences that come close are the various exclusions for retirement savings contributions. Table 3.2 presents trends in health spending by payor and form of spending for 1993–2003. The table documents the well-known growth in the magnitude of real spending over this period. Particularly noteworthy is the change in the form of health spending, from (largely taxable) out-of-pocket to tax-preferred insured spending. According to the table, real employer and employee payments for insured health spending rose about 50 percent over the period, while out-of-pocket spending rose less than half as much. If the tax preference contributed to this, and
Tax Preferences on Health Care Spending and Federal Revenues
67
Table 3.2 Health Spending, by Payor and Form of Spending 1993–2003 (in billions of 2003 dollars) % change 1993–2003
Type of Payor and Form of Spending
1993
Year 1998
2003
TOTAL
$1,087.5
$1,257.2
$1,614.2
48.4%
Private
$694.4
$806.8
$992.2
42.9%
Private Business
$280.9
$325.1
$423.0
50.6%
$205.1
$237.2
$320.6
56.3%
Employer payments of Medicare payroll taxes
$45.5
$60.8
$64.3
41.4%
Workers’ compensation and other
$30.4
$27.1
$38.1
25.5%
$367.5
$423.5
$512.6
39.5%
$110.4
$129.5
$174.1
57.8%
Employee payments of Medicare payroll taxes
$55.5
$78.2
$86.0
55.0%
Individual payments of Medicare SMI premiums
$15.1
$17.5
$22.0
45.6%
$186.6
$198.4
$230.5
23.6%
$46.0
$58.2
$56.6
23.1%
Public
$393.1
$450.5
$622.0
58.2%
Federal Government
$223.0
$243.0
$344.0
54.3%
$14.6
$12.9
$19.7
34.9%
Employer contribution to HI premiums
Household Employee contribution to HI premiums
Out-of-pocket spending Other Private
Employer contributions to HI premiums Medicaid
$99.2
$115.0
$160.9
62.2%
$106.2
$112.3
$160.2
50.9%
$170.1
$207.5
$278.1
63.5%
Employer contributions to HI premiums
$48.5
$55.1
$86.2
77.7%
Medicaid
$58.2
$83.1
$111.8
92.2%
Other
$57.2
$61.9
$71.5
25.1%
Medicare (net of payroll tax receipts) and other State and Local Government
Source: Health United States (2005), table 127, deflated with the CPI.
insured spending is subject to a greater degree of moral hazard, then the impact of the preference on efficiency could be substantial. Taken together, these factors have led academic researchers to focus on the consequences of revoking the tax preference. Yet, policymakers over the past 30 years have taken an alternative approach; they have sought to level the tax playing field by expanding the tax preference rather than eliminating it.1 In 1978, changes to section 125 of the Internal Revenue Code allowed health expenditures made through an employer-provided Flexible Spending Account (FSA) to be deductible to the employer but nontaxable to the employee.2 In 1996, the Health
68
Cogan, Hubbard, and Kessler
Insurance Portability and Accountability Act allowed employees of small businesses who were covered by certain high-deductible health plans (HDHPs) to make tax-free contributions to a Medical Savings Account (MSA). Funds from an MSA can be withdrawn, tax free, to pay for medical expenses in the present or the future; if used for other purposes after age 65, MSA distributions are taxed as ordinary income. Under Treasury regulations issued in 2002, sections 105 and 106 of the Internal Revenue Code allow health reimbursement accounts (HRAs) to reimburse employees for medical expenses with before-tax dollars, without the use-it-or-lose-it provision of section 125 cafeteria plans.3 In 2003, the Medicare Prescription Drug, Improvement, and Modernization Act allowed employers and individuals with any HDHP to make tax-free contributions to a health savings account (HSA). President Bush has proposed expanding the use of HSAs by liberalizing their contribution limits. Conditional on the tax preference for insurance remaining in place, the consequences of these expansions for health spending, and economic efficiency, are theoretically indeterminate. Expanding the tax preference has two opposing effects. First, expansion lowers the overall price of health care relative to other goods and services, which increases distortionary spending. Second, expansion raises the price of purchasing health care through insurance relative to out-of-pocket. The second effect induces people to shift to health plans with higher deductibles and coinsurance rates, which, in turn, lowers distortionary spending. Thus, assessing the effects of expanding the tax preference to out-ofpocket spending is important for evaluating existing and proposed tax policies toward health care. Yet, very little work has sought to estimate these effects and to understand their sensitivity to assumptions about the demand for health services and insurance. In this paper, we present a simplified version of the approach in John Cogan, R. Glenn Hubbard, and Daniel Kessler (CHK 2005) to calculating the effects of an abovethe-line deduction for out-of-pocket health spending, which we term “full deductibility.” In that paper, we show how the response of total health spending to an expansion in the tax preference can be specified as a function of a small number of behavioral parameters that have been estimated in the existing literature. This paper expands on that work in three ways. First, it calculates the effects of full deductibility on out-of-pocket spending, total spending, and the government budget under a range of parameter values. Second, it compares our estimates
Tax Preferences on Health Care Spending and Federal Revenues
69
to those from other researchers. Third, it uses our analysis to derive some implications for tax policy toward HSAs. 2.
Assessing the Effects of Tax Deductibility on Health Spending
As reviews by Pauly (1986) and, more recently, Selden and Moeller (2000) show, a substantial body of research has sought to assess the effects of revoking the tax preference for employer-provided health insurance. Considerably less work has focused on the effects of extending the preference to out-of-pocket spending. Jack and Sheiner (1997) simulate the effects on insurance policy choice, health spending, and efficiency of both revoking and extending the tax preference. Those authors show that extending deductibility might actually reduce health spending and improve efficiency, by leading to such a large increase in the effective coinsurance rate that the gain from the reduction in moral hazard swamps the loss from the reduction in the overall price of health care. A recent working paper by Jack, Levinson, and Rahardja (2005) provides empirical support for this hypothesis. They show that, correcting for selection effects, FSAs are associated with effective coinsurance rates that are about 7 percentage points higher, relative to a sample average coinsurance rate of 17 percent. This finding suggests that making out-of-pocket health spending deductible, which an FSA effectively does, would significantly change the form of the average health insurance contract. In CHK (2005), we derive the relationship between the impact on health spending of making out-of-pocket expenses tax deductible and two parameters from economic studies: the price elasticity of health spending, and the elasticity of the coinsurance rate with respect to the tax preference for insured spending. We specify health spending E as a function of the after-tax price of health services relative to all other goods p and the tax preference for out-of-pocket spending relative to insured spending to/ti, E(p, to /ti). In a world without taxes, p is the price of health services p*. In a world with tax preferences, p is p* multiplied by the weighted average of the tax preferences for out-of-pocket spending to and insured spending ti, p = p* × [cto + (1 – c) ti], where to and ti are weighted by the quantity shares of out-of-pocket and insured spending c and (1 – c), respectively. The share c can also be thought of as the coinsurance rate—that is, the share of spending out-of-pocket in the absence of tax preferences.
Cogan, Hubbard, and Kessler
70
So ∂(t / t ) ∂E ∂p dE ∂E ∂p ∂E ∂E 1 = × × + × o i = + × dto ∂p ∂to ∂(to / ti ) ∂to ∂p ∂to ∂(to / ti ) ti and t ∂(t / t ) ∂E ∂p dE ∂E ∂p ∂E ∂E = × × + × o i = − × o dti ∂p ∂ti ∂(to / ti ) ∂ti ∂p ∂ti ∂(to / ti ) ti2 Then the sum of these two equations, in elasticity terms, is: dE / E dE / E ∂E / E ∂p / p ∂E E / E to ∂E / E + = × + × + dto / to dti / ti ∂p / p ∂to / to ∂(to / ti ) ti ∂p / p ×
∂p / p ∂E / E to − × ∂ti / ti ∂(to / ti ) ti
or dE / E dE / E ∂E / E ⎡ ∂p / p ∂pp / p ⎤ + = ×⎢ + ⎥ dto / to dti / ti ∂p / p ⎣ ∂to / to ∂ti / ti ⎦ =
⎤ ∂E / E ⎡ p * ×[cto + (1 − c)ti ] ×⎢ + θ (to , ti , p *) ⎥ ∂p / p ⎣ p ⎦
or e(to ) + e(ti ) = e( p) × [1 + θ (to , ti , p *)], where e(to) is the elasticity of spending with respect to the tax preference for out-of-pocket spending; e(ti) is the elasticity of spending with respect to the preference for insured spending; and e( p) is the price elasticity of spending. In CHK (2005), we show that under reasonable assumptions and current market conditions and tax preferences,4 θ(.) is small, so e(to) + e(ti) ≈ e( p). Finally, we translate the results from previous studies into these terms. For example, the equation above can be rewritten as e(to) + (e(c) × e(c,ti)) ≈ e( p) where e(c) is the elasticity of spending with respect to the coinsurance rate and e(c,ti) is the elasticity of the coinsurance rate with respect to
Tax Preferences on Health Care Spending and Federal Revenues
71
the tax preference for insured spending. If demand curves are locally linear, then e( p) = e(ap) for any positive constant a, so e( p) = e(c), which implies: e(to) ≈ e( p) × (1 – e(c,ti)). Assessing the effects of extending deductibility thus requires estimates of e( p) and e(c,ti). There is a range of estimates of e( p). Based on the RAND Health Insurance Experiment, Manning and colleagues (1987) estimate e( p) = –0.2 in arc elasticity terms.5 In more recent work, Eichner (1998, table 1) estimates e( p) = –0.7 (average for all employees 1990–92, also in arc elasticity terms).6 In addition, even the high end of this range may understate the impact of a market-wide change in incentives such as extending deductibility. All of the estimates of e( p) are based on responses to individual-level changes in copayments, which may be smaller than the responses to more widespread changes in insurance contracts that fundamentally alter how doctors practice medicine. Finkelstein (2005), for example, shows that the change in hospital spending associated with the introduction of Medicare was far greater than the elasticities from the RAND Experiment would have predicted. Less disagreement surrounds the magnitude of e(c, ti). Several studies have assessed the effect of the tax preference on coinsurance rates. These can be used to compute e(c, ti). Early simulations by Feldstein and Friedman (1977) suggest that revoking the tax preference for employerprovided insurance would lead to a doubling in the coinsurance rate (from approximately 25 to 50 percent). This finding is consistent with an unpublished estimate by Phelps (1986). More recent work leads to virtually the same conclusions. At conservative levels of consumer risk aversion and e( p), simulations by Jack and Sheiner (1997, table 2) find that the tax preference for insurance has led optimal coinsurance rates to shrink from 33–67 percent to 20–30 percent. Assuming an average marginal (payroll plus income) tax rate of 30 percent,7 revocation leading to doubling of coinsurance rates from c to 2c implies an e(c, ti) in arc elasticity terms of 1.9, because: 2c − c [2 c + c ]
2 = 1.9. 1 − (1 − .3) [1 + (1 − .3)] 2
Cogan, Hubbard, and Kessler
72
Table 3.3 presents calculations of the effects of revoking and extending the income and payroll tax preference for e(p) = –0.2, –0.45, and –0.7 at e(c, ti) = 1.9. The first two rows of table 3.3 present the effects of revoking the income and income-and-payroll tax preference which is simply e( p) × e(c , ti ) ×
dti
ti .
The third and fourth rows of table 3.3 present the effects of extending the income and income-and-payroll tax preference which is simply e( p) × (1 − e(c , ti )) ×
dto
to .
Our estimates of the effects of tax policy on spending are within the range of those from other research. For example, according to Gruber (2002, table 5), removing the income tax subsidy for health insurance would result in a 32.8 percent decline in health spending, expressed as a percentage point change from its initial value. Expressed as a percentage-point change at the average (in order to make his estimate comparable to those in table 3.1), this amounts to a 39.2 percent decline—larger than the 31.8 percent decline in spending that we would predict even assuming an elasticity of spending with respect to the coinsurance rate of 0.7. Gruber’s estimate, when combined with the consensus estimate Table 3.3 Effect on Health Spending of Changing the Tax Preference Elasticity of Spending with Respect to After-Tax Price of Health Care –0.2
–0.45
–0.7
Effect of revoking: Income tax preference
–9.1%
–20.4%
–31.8%
–13.4%
–30.2%
–46.9%
Income tax preference
–2.7%
–6.1%
–9.5%
Income plus payroll tax preference
–5.6%
–12.6%
–19.7%
Income plus payroll tax preference Effect of extending to out-of-pocket:
Note: Assumes a health-spending-weighted average marginal income tax rate of .19, out-of-pocket-spending-weighted average marginal income tax rate of .14, and an average payroll tax rate of .13. Average marginal tax rates were calculated using MEPS, and include both households with and without income tax liabilities. See CHK (2005) for details.
Tax Preferences on Health Care Spending and Federal Revenues
73
of e(c, ti) = 1.9, implies that extending the income tax preference to outof-pocket spending would result in a decline in overall spending of 11.8 percent.8 According to Jack and Sheiner (1997, table 1), extending the income and payroll tax preference to out-of-pocket spending would lead to a decline in spending of 4.9 percent (=0.3/6.15), slightly lower than the decline in spending that we would predict assuming an elasticity of spending of 0.2. 3. Assessing the Budget Implications of Extending Deductibility to Out-of-Pocket Spending In addition to reducing inefficient health spending, full deductibility reduces federal tax revenues. Deductibility has two effects on revenues—a loss from making previously taxable spending deductible, and a gain from the shift away from previously deductible health spending. We do not account for any possible spillover effects from privately-purchased health care to the Medicare or Medicaid programs. The revenue loss consists of two components—the loss from allowing the above-the-line deduction of out-of-pocket spending, and the loss from purchases on health insurance being deducted above-the-line that are currently not deducted or deductible. The revenue gain also consists of two components. Tax revenues rise because higher policy deductibles will translate into a shift in employees’ compensation away from excludable health spending to taxable wages.9 The government picks up both payroll and income taxes on the portion of the wage increase directed to non-health spending (first component), and payroll taxes on the portion directed to out-of-pocket health spending (second component). Table 3.4 presents our calculations of these losses and gains on an annual basis, in 2004 dollars. The top panel of the table shows the two components of the gross losses from full deductibility; the middle panel shows the two components of the gross gains; and the bottom panel shows the intermediate steps underlying the calculation of each of the components of the gross gains. As the top panel shows, the gross losses are a mechanical consequence of the policy; they do not depend on behavior. We calculate that full deductibility would have a gross revenue cost of $26.8 (= $16.4 + $10.4) billion per year.
74
Cogan, Hubbard, and Kessler
Table 3.4 Effect on Tax Revenues of Full Deductibility Elasticity of Spending with Respect to After-Tax Price of Health Care –0.2
–0.45
–0.7
Loss from deduction of taxable out-of-pocket spending
–$16.4
–$16.4
–$16.4
Loss from deduction of taxed or taxable insurance payments
–$10.4
–$10.4
–$10.4
Pickup of payroll taxes
$2.4
$5.3
$8.1
Pickup of income taxes
$3.5
$7.7
$11.8
$6.3
$5.3
$4.3
–$14.5
–$8.4
–$2.5
Reduction in health spending
–$18.4
–$40.7
–$62.3
New coinsurance rate
33.3%
33.3%
33.3%
Reduction in employer-insured spending
–$67.2
–$81.6
–$95.6
$48.8
$40.9
$33.3
Revenue Losses
Revenue Gains Shift away from health spending
Shift away from employer-insured spending toward out-of-pocket Pickup of payroll taxes Total Intermediate calculations underlying estimates of revenue gains
Increase in out-of-pocket spending (difference)
Note: Assumes $117b of currently taxable out-of-pocket spending, $74b of current taxed or taxable insurance payments, $500b of current spending on employer health insurance, $688b of total private health spending, an average coinsurance rate of 25%, a health-spending weighted average marginal income tax rate of .19, out-of-pocket spending-weighted average marginal income tax rate of .14, and an average payroll tax rate of .13. See note to table 3.1 and CHK (2005) for details on calculation of average marginal tax rates.
The first two rows of the middle panel present estimates of the first component of gross gains, the increase in tax revenues from the shift from insured spending to wages. Given the percentage effects on spending from table 3.3, full deductibility leads to a decline in spending of $18.4 to $62.3 billion (bottom panel, first row),10 and in turn to an increase in payroll (income) tax revenues of $2.4 to $8.1 billion ($3.5 to $11.8 billion). The third row of the middle panel presents estimates of the second component of gross gains, the increase in tax revenues from the shift
Tax Preferences on Health Care Spending and Federal Revenues
75
from insured to out-of-pocket spending. To calculate this shift, we make the (conservative) assumption that spending for individuals with employer-provided health insurance responds to full deductibility the same as does all private health spending. Under this assumption, the coinsurance rate that obtains under full deductibility c’ (bottom panel, second row) is equal to ⎡ 2(η / e( p)) ⎤ c × ⎢1 + ⎥ ⎣ 2 − (η / e( p)) ⎦
1−τ ,
where τ is the average marginal tax rate, and 2(η/e( p))/[2 – (η/e( p))] is the implied rise in the after-tax coinsurance rate necessary to induce the spending decline from table 3.3. The coinsurance rate under full deductibility would be 33.3 percent, regardless of the elasticity of demand for health services, up from a pre-deductibility average of 25 percent. (That the coinsurance rate does not vary with the elasticity of demand is, of course, a product of our model’s assumption of a constant e(c, ti).) This translates into a decline in spending on employer-provided health insurance, in percentage terms, of ⎛ 1 − c′ ⎞ ⎜⎝ ⎟ × (1 + η *), 1− c ⎠ or in dollar terms, of $67.2 to $95.6 billion (bottom panel, third row). The increase in out-of-pocket spending subject to the payroll tax (bottom panel, fourth row) is thus the difference between this decline and the decline in total spending. According to this simple model, then full deductibility would lead to a significant but plausible increase in out-of-pocket spending, from $149 billion in 2004 dollars (=$117 billion in currently non-deductible out-of-pocket spending plus $32 billion in deductible spending, see CHK (2005, Appendix E)) to $190 billion (=$149 billion + $41 billion, see column 2), or 27.5 percent (=($190 billion – $149 billion)/$149 billion). The calculations presented in table 3.4 make the important point that much of the gross revenue losses from full deductibility will be made up by revenue gains from the reduction in the inefficiency due to the highly-distortionary existing tax preference. Indeed, even assuming an elasticity of demand for health services of –0.2, $12.2 billion of the $26.8 billion (or 46 percent) of the losses will be undone; if the elasticity of
Cogan, Hubbard, and Kessler
76
demand is –0.45, fully $18.3 billion of the $26.8 billion (or 68 percent) of losses are undone. 4.
Implications for Policy toward HSAs
Like full deductibility, allowing HSA contributions to be tax-deductible gives a tax preference to out-of-pocket spending. Under current law, a holder of an HDHP (i.e., a health plan with a deductible of at least $1,050/$2,100 in 2006 (individual/family)) can contribute to an HSA the amount of the deductible, but not more than $2,700 (individual) or $5,450 (family).11 The contribution is deductible from federal income taxes and from income taxes in 44 states.12 If the contribution is made by a person’s employer, it is also excludable from the Social Security tax base. The contribution accumulates interest taxfree and is non-taxable on distribution, if spent on health services; it is taxable as ordinary income if distributed for any other purpose after age 65. HSAs differ from full deductibility in three key ways. First, an individual can only have an HSA if they are enrolled in an HDHP. Second, an individual can deduct HSA contributions from his or her taxable income up to the amount of their HDHP’s deductible, whether or not they incur any health expenses, but can not deduct more, even if they have coinsurance payments in excess of the deductible. Third, an HSA allows an individual to save tax-free for future health expenses or retirement, whereas full deductibility only allows deduction of current health expenses.13 For consumers who use HSAs only as a vehicle to deduct current health expenses, the most important difference between HSAs and full deductibility is the minimum deductible requirement of HDHPs. If, for these individuals, all of the expenditure-reducing incentive effects of full deductibility were channeled through insurance policy deductibles (rather than coinsurance rates), then deductibles would have to rise from a typical value of $221 (2004 dollars)14 to approximately $290,15 far less than the 2006 mandated HDHP minimum of $1,050. Because HSAs are indistinguishable from full deductibility for these consumers, this implies that they would prefer a lower deductible than the mandated minimum. Because most taxpayers do not exhaust their existing retirement savings incentives (CBO 2003), and therefore are likely to treat HSAs
Tax Preferences on Health Care Spending and Federal Revenues
77
primarily as a vehicle to deduct current expenses, HSAs as currently formulated will likely be taken up by fewer people than would full deductibility. However, among HSA enrollees, HSAs bring health spending much closer to the efficient level (the level that would be preferred in the absence of any tax preference) than full deductibility. If all of the expenditure-reducing incentive effects of revoking the income and payroll tax preferences were channeled through insurance policy deductibles, then deductibles would rise to approximately $1,680 in 2004 dollars.16 These back-of-the-envelope comparisons between full deductibility and HSAs are consistent with empirical studies of HSAs (see, for example, Melinda Buntin et al. 2005; and Roger Feldman et al. 2005) and MSAs (see, for example, Larry Ozanne 1996; and Emmett Keeler et al. 1996). The most efficient way to expand HSAs would be to allow deductibility of all out-of-pocket payments for people with insurance (not just those toward the policy deductible), but limit the budget consequences of HSAs by capping deductible contributions at a fixed dollar amount (such as $1,000/$2,000 for an individual/family, indexed to inflation) in excess of current health expenses. Two considerations support this shift. First, recall that the back-of-the-envelope calculations above suggest that lowering the minimum deductible requirement is an important policy for increasing the take-up rate. For most consumers, the minimum deductible requirement is simply too high, given the magnitude of the existing tax preference for employer-provided insurance. Allowing people to choose their policy deductible will solve this problem. Just as the take-up of managed care had beneficial spillovers to fee-for-service insurance (for example, Laurence Baker 1997), the take-up of HSAs and insurance plans with more cost sharing will as well. Second, evidence from the RAND Experiment suggests that most of the expenditure-reducing effects of policy deductibles occur at low levels of deductibles (for example, Emmett Keeler et al. 1988). Extending deductibility to out-of-pocket expenses above the policy deductible will provide an important incentive for individuals to increase coinsurance rates as well. The results from the RAND experiment suggest that a mix of higher deductibles and coinsurance rates would achieve greater efficiency in health spending than mandating that all of the savings be channeled through the deductible.
Cogan, Hubbard, and Kessler
78
5.
Conclusion
The U.S. health care system, the envy of the world in innovation, faces criticisms from policymakers about the cost of care. From an economic perspective, an alternative approach is to ask whether private consumers of health care—and taxpayers who fund public programs—are obtaining the highest “value” for the resources devoted to health care. Healthy, competitive markets generally offer the greatest opportunity to maximize value. As academic researchers have long observed, limiting or revoking altogether the tax preference for health insurance would improve the performance of markets for health services on this dimension. Current policy generally allows individuals to receive employer-provided health insurance expenditures tax-free, but requires direct out-of-pocket medical spending to be financed from after-tax income. This tax preference has given consumers the incentive to purchase health care through low-deductible, low-copayment insurance instead of out-of-pocket. However, likely because the vast majority of voters benefit from this preference, policymakers over the past 30 years have instead sought to level the tax playing field by expanding the tax preference rather than eliminating it. In this paper, we show that extending deductibility to out-of-pocket spending, while a second-best policy change, is nonetheless likely to lead to significant improvements in efficiency under a range of assumptions about demand for health care and health insurance. Although we are not the first to recognize this fact, we quantify the actual health spending and revenue effects of such a policy using a transparent accounting model and a small number of behavioral parameters from existing studies. Providing additional evidence on the sensitivity of health insurance contracts to tax changes is a subject for future research. Also, while not emphasized here, expanding deductibility may also significantly reduce rates of uninsurance by lowering the cost of health insurance. Finally, we view as an important topic for future work more analysis of the relationship between tax deductibility and Health Savings Accounts. Notes Cogan is the Leonard and Shirley Ely Senior Fellow at the Hoover Institution at Stanford University. Hubbard is Dean and Russell L. Carson Professor of Finance and Economics at the Graduate School of Business and Professor of Economics at Columbia University;
Tax Preferences on Health Care Spending and Federal Revenues
79
and Research Associate of the National Bureau of Economic Research. Kessler is Professor of Business and Law at Stanford University, Senior Fellow at the Hoover Institution at Stanford University, and Research Associate of the National Bureau of Economic Research. This paper was prepared for the NBER Tax Policy and the Economy Conference, to be held in Washington, D.C., on September 14, 2006. We are grateful to Joe Antos, Doug Holtz-Eakin, Jim Poterba, and conference participants for helpful comments. 1. Such a pattern likely reflects politics as much as economics: the vast majority of voters benefit from the excludability of health insurance. When President Reagan expressed interest in eliminating or even limiting the exclusion, his proposal was soundly rejected in Congress. Indeed, the Clinton health reform plan explicitly rejected any such limitations (Cutler 1994). 2. An FSA allows employees to allocate a portion of their compensation to nontaxable fringe benefits instead of taxable wages. Currently, once the amount of the FSA contribution has been designated, the employee is not allowed to change it or drop the plan during the year unless he or she experiences a change of family status. By law, the employee forfeits any unspent funds in the account at the end of the year. 3. HRAs, however, are owned by the employer and contributions to them are subject to nondiscrimination rules; that is, they can not be at the employee’s discretion. See U.S. Department of Labor (2003). 4. For example, to = 1 and ti = 0.7. 5. Because the tax preference leads to large changes in effective prices for health services, point elasticities expressed in (current) after-tax terms will be very different from those expressed in (counterfactual) pre-tax terms. For example, the effect of a 1 percent increase in the effective coinsurance rate from its current (lower) base is much smaller than the effect of a 1 percent decrease in the effective coinsurance rate from its (higher) base in the absence of the tax preference. Some of the studies we review provide estimates in the former terms; some provide estimates in the latter. We follow the convention used in the RAND study and convert all elasticities into arc terms, expressed at the average between pre- and post-tax prices. 6. The published estimates in Eichner (1997, table 1) are based on models that assume that consumers make marginal health spending decisions throughout the year based on the coinsurance rate that they face at the end of the year. This assumption is important because many plans’ coinsurance rates vary with a consumer’s level of cumulative spending over a calendar year. For example, a plan may have a $500 deductible (i.e., a coinsurance rate of 100 percent on the first $500 of spending), a coinsurance rate of 25 percent, and a $2,000 out-of-pocket maximum (i.e., a coinsurance rate of 0 percent after $1,500 in coinsurance payments or $6,500 in total spending). If consumers have rational expectations, then this assumption is correct. Regardless of when in the year the choice to make a (marginal) health expenditure arises, the effective coinsurance rate for any marginal expenditure would be the rate in effect after all of the year’s expenditures had occurred. In the text of the article, Eichner points out that estimates of e(p) from models that do not assume rational expectations are generally lower. However, in our view, the rational expectations assumption is more justifiable than the alternatives, so we use the estimates from the table. 7. We discuss how we assess the magnitude of the average marginal tax rate in more detail below. 8. Gruber’s estimate implies e(p) = –0.87 = 0.392 / (1.9* 0.19*2/(0.7 + 0.89)), which implies an effect of –0.118 = –0.87*0.9*2*0.14/(1 + 0.86).
Cogan, Hubbard, and Kessler
80
9. As Gruber (2000) points out, empirical evidence supports the hypothesis that the costs of health insurance premiums are fully shifted out of wages. 10. Table 3.3 presents percentage effects measured at the average η, so the percentage effect measured at current levels 2η η* = , 2 −η because
η=
η* [1 + (1 + η *)]
2 11. This summary is taken from the detailed explanation of the tax treatment of HSAs in Internal Revenue Service (2004). 12. See National Conference of State Legislatures (2006). 13. Of course, adoption of full deductibility does not preclude HSAs. Indeed, full deductibility enhances the incentive to finance current health spending out of pocket, while HSAs (when used as a savings vehicle) enhance the incentive to accumulate assets to finance future health expenses out of pocket. 14. See Gabel and Rice (2003). 15. We reach this conclusion by using parameters from the RAND Health Insurance experiment, the approach suggested by Phelps (2003), and the increase in health spending reported by CMS (2006). In 1984 dollars, the current deductible of $221 would be equivalent to $44. According to CMS (2006), spending per private health insurance enrollee rose from $675 in 1984 to $3,379 in 2004, a factor of five. According to Phelps (2003, table 5.6), to achieve the spending reduction of 2.7 percent from full deductibility predicted by the RAND Experiment, deductibles would have to have risen to $58 in 1984, or $290 (=58*5) in 2004. 16. To achieve the spending reduction of 13.4 percent from revoking the income and payroll tax preferences predicted by the RAND Experiment, deductibles would have to have risen to $336 in 1984, or $1,680 (=336*5) in 2004.
References Baker, Laurence (1997). “The Effect of HMOs on Fee-for-Service Health Care Expenditures: Evidence from Medicare,” Journal of Health Economics, 16(4):453–81. Buntin, Melinda et al. (2005). “Consumer Directed Health Plans: Implications for Health Care Quality and Cost,” available at http://www.chcf.org/documents/insurance/ConsumerDirHealthPlansQualityCost.pdf, 2005. Centers for Medicare and Medicaid Services, Office of the Actuary, National Health Statistics (2006). Available at http://www.cms.hhs.gov/NationalHealthExpendData/ downloads/tables.pdf. Cogan, John F., R. Glenn Hubbard, and Daniel P. Kessler (2005). Healthy, Wealthy, and Wise. Hoover Insitution/AEI Press, Washington, D.C. Congressional Budget Office (2003). “Utilization of Tax Incentives for Retirement Saving,” August.
Tax Preferences on Health Care Spending and Federal Revenues
81
Cutler, David (1994). “A Guide to Health Care Reform,” Journal of Economic Perspectives, 8:13–30. Eichner, Matthew (1998). “The Demand for Medical Care: What People Pay Does Matter,” American Economic Review, 88:117–121. Feldman, Roger et al. (2005). “Health Savings Accounts: Early Estimates of National Take-up,” Health Affairs, 24:1582–1591. Feldstein, Martin (1973). “The Welfare Loss from Excess Health Insurance,” Journal of Political Economy, 81:251–280. Feldstein, Martin, and Bernard Friedman (1977). “Tax Subsidies, the Rational Demand for Insurance, and the Health Care Crisis,” Journal of Public Economics, 7:155–178. Finkelstein, Amy (2005). “The Aggregate Effects of Health Insurance: Evidence from the Introduction of Medicare,” NBER working paper no. 11619. Gabel, Jon, and Tom Rice (2003). “Insurance Markets: Understanding Consumer Directed Health Care in California,” California Health Care Foundation, available at www.chcf. org/documents/insurance/ConsumerDirectedHealthCare.pdf. Gruber, Jonathan (2000). “Health Insurance and the Labor Market,” in A.J. Culyer and J.P. Newhouse, (eds.), Handbook of Health Economics. Amsterdam: North-Holland. Gruber, Jonathan (2002). “Taxes and Health Insurance,” in James Poterba, (ed.), Tax Policy and the Economy, Volume 16. Cambridge, MA: MIT Press. Health United States, US Centers for Disease Control (2005). Internal Revenue Service, Notice 2004–2 (2004). “Health Savings Accounts,” available at http://www.irs.gov/irb/2004-02_IRB/ar09.html. Jack, William, Arik Levinson, and Sjamsu Rahardja (2005). “Employee Cost-Sharing and the Welfare Effects of Flexible Spending Accounts,” forthcoming, Journal of Public Economics. Jack, William, and Louise Sheiner (1997). “Welfare-Improving Health Expenditure Subsidies,” American Economic Review, 87:206–221. Joint Committee on Taxation (2003). “Estimates of Federal Tax Expenditures for Fiscal Years 2004-08,” Pub. No. JCS–8–03. Keeler, Emmett et al. (1996). “Can Medical Savings Accounts for the Nonelderly Reduce Health Care Costs?” JAMA, 275(21):1666–1671. Keeler, Emmett et al. (1988). “The Demand for Episodes of Medical Treatment in the Health Insurance Experiment,” RAND R3454-HHS, March. Manning, Willard G., et al. (1987). “Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment,” American Economic Review, 77(3):251–277. National Conference of State Legislatures, 2004–2006 (2006). State Legislation on Health Savings Accounts and Consumer-Directed Health Plans, available at http://www.ncsl. org/programs/health/hsa.htm#2005. Office of Management and Budget (2003). “Budget of the US Government: Analytical Perspectives, Fiscal Year 2004,” Chapter 6: Tax Expenditures.
82
Cogan, Hubbard, and Kessler
Ozanne, Larry (1996). “How Will Medical Savings Accounts Affect Medical Spending?” Inquiry, 33:225–236. Pauly, Mark V. (1986). “Taxation, Health Insurance, and Market Failure in the Medical Economy,” Journal of Economic Literature, 24:629–675. Phelps, Charles E. (1986). “Large-Scale Tax Reform: The Example of Employer Paid Health Insurance Premiums,” University of Rochester working paper. Phelps, Charles E. (2003). Health Economics, 3d. Edition. Boston, MA: Addison-Wesley. Selden, Thomas F., and John Moeller (2000). “Estimates of the Tax Subsidy for Employment-Related Health Insurance,” National Tax Journal, 53:877–888. Shiels, John, and Randall Haught (2004). “The Cost of Tax-Exempt Health Benefits in 2004,” Health Affairs Web Exclusive W4106–112, February. U.S. Department of Labor, Bureau of Labor Statistics (2003). “Health Spending Accounts,” available at www.bls.gov/opub/cwc/cm20031022ar01p1.htm.
4 Does It Pay, at the Margin, to Work and Save? Measuring Effective Marginal Taxes on Americans’ Labor Supply and Saving Laurence J. Kotlikoff, Boston University and NBER David Rapson, Boston University
Executive Summary Building on Gokhale, Kotlikoff and Sluchynsky’s (2002) study of Americans’ incentives to work full- or part-time, this paper uses ESPlanner, a life-cycle financial planning program, in conjunction with detailed modeling of transfer programs to determine (1) total marginal net tax rates on current labor supply, (2) total net marginal tax rates on lifecycle labor supply, (3) total net marginal tax rates on saving and (4) the tax-arbitrage opportunities available from contributing to retirement accounts. In seeking to provide the most comprehensive analysis to date of fiscal incentives, the paper incorporates federal and state personal income taxes, the FICA payroll tax, federal and state corporate income taxes, federal and state sales and excise taxes, Social Security benefits, Medicare benefits, Medicaid benefits, Foods Stamps, welfare (TAFCD) benefits, and other transfer program benefits. The paper offers four main takeaways. First, thanks to the incredible complexity of the U.S. fiscal system, it’s impossible for anyone to understand her incentive to work, save, or contribute to retirement accounts absent highly advanced computer technology and software. Second, the U.S. fiscal system provides most households with very strong reasons to limit their labor supply and saving. Third, the system offers very high-income young and middle-aged households as well as most older households tremendous opportunities to arbitrage the tax system by contributing to retirement accounts. Fourth, the patterns by age and income of marginal net tax rates on earnings, marginal net tax rates on saving, and tax-arbitrage opportunities can be summarized with one word—bizarre.
Kotlikoff and Rapson
84
1.
Introduction
Households both want and need to understand the incentives they face at the margin for working and saving. Yet any American seeking to understand her total effective net marginal tax on either choice faces a daunting challenge. First, she needs to consider a host of taxes and transfers including federal personal income taxes, federal corporate income taxes, federal payroll taxes, federal excise taxes, state personal income taxes, state corporate income taxes, state sales taxes, state excise taxes, Social Security benefits, welfare benefits (TAFDC), Supplemental Security Income benefits (SSI), Medicaid benefits, Medicare benefit, food stamps, nutrition benefits (WIC), and energy assistance benefits (LIHEAP). Second, she needs to understand in very fine detail how each of these taxes and transfers is calculated. Third, she needs to understand the interactions of the different tax and transfer programs. Fourth, she needs to consider the fact that these taxes and transfers are paid and received over time. And fifth, she needs to have a method for translating all of these interconnected time-dated tax payments and benefit receipts into a simple and comprehensible statement of her marginal reward for working and saving. This paper uses ESPlannerTM (Economic Security PlannerTM) in conjunction with detailed modeling of non-Social Security transfer programs (ESPlanner incorporates Social Security) to generate total effective (net) marginal taxes on labor supply and saving for stylized American households. It also examines the tax arbitrage opportunity available to households from saving in either (1) 401(k), traditional IRA, or other tax-deferred retirement accounts or (2) Roth IRAs, Roth 401(k)s, or other Roth accounts. The paper builds and draws on Gokhale, Kotlikoff, and Sluchynsky (2002), which studied the incentives of Americans to work full- or part-time. That study showed that the overall tax/transfer system is progressive, particularly at the very low end of the earnings distribution, that all households face very high marginal taxes on the choice of working full- or part-time, that many low- and moderate-income households face substantially higher marginal taxes on working full- or part-time than do high-income households, and that many low-income households face confiscatory taxes on switching from part- to full-time work or switching from full-time work by one spouse to full-time work by both spouses.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
85
The value added of this paper relative to Gokhale, Kotlikoff, and Sluchynsky (2002) is that we consider the marginal net taxes on working extra hours in the current year, working extra hours throughout one’s career, and increasing one’s current saving. We also examine the tax arbitrage opportunity available to different households from contributing to (1) 401(k), traditional IRA, or similar tax-deferred accounts or (2) Roth IRAs, Roth 401(k)s, or other Roth accounts. With the exception of certain very low-earning households, we find high to very high marginal net tax rates—ranging from 24 to 45 percent—on current and life-cycle labor supply. These calculations are made at particular levels of pre-tax and pre-transfer earnings and are based on discrete increments in earnings. As we also demonstrate, marginal net tax rates on current and life-cycle labor supply are astronomical over much smaller increments in gross earnings at particular levels of earnings at which income and asset eligibility tests of particular tax and transfer programs become relevant. The Congressional Budget Office’s (2005) recent study of effective tax rates on labor supply reports much lower marginal rates, particularly for low-income households, than those we report. The reason is that the CBO ignores transfer payments and federal and state sales and excise taxes. At low incomes (when transfer benefits are often linked directly to income) our estimates of marginal effective rates are 80 to 100 percentage points higher than the CBO in some cases. For example, 60 year old couples earning $10,000/yr are within the EITC phase-in region, which results in a CBO estimated marginal rate of –40 percent. However, at this income they also face a one-for-one reduction in food stamps. After accounting for all of the relevant transfer programs, the resulting effective marginal rate is 50 percent, or 90 percentage points higher than the CBO estimate. Aside from these few extreme cases, the differences are smaller, but still substantial. Our estimates for low- to mid-income households are 30 to 50 points higher than the CBO, and 10 to 25 points higher for mid- to high-income households. In addition to finding high to very high marginal net taxes on labor supply for virtually all American households, we also find high to very high marginal net tax rates on saving for most households. For some low-income households, we find astronomical net tax rates on saving; for these households higher saving means higher future assets and higher asset income, which can reduce eligibility for transfer payments
86
Kotlikoff and Rapson
via asset and income tests. Finally, we find huge arbitrage opportunities for particular households of particular ages and earnings levels from contributing to either tax-deferred retirement accounts or Roth IRAs, Roth 401(k)s, or other Roth accounts. The paper provides four main takeaways. First, thanks to the incredible complexity of the U.S. fiscal system, it’s essentially impossible for anyone to understand her incentive to work, save, or contribute to retirement accounts absent highly advanced computer technology and software. Second, the U.S. fiscal system provides most households with very strong reasons to limit their labor supply and saving. Third, the system offers very high-income young and middle-aged households as well as most older households tremendous opportunities to arbitrage the tax system by contributing to retirement accounts. Fourth, the patterns by age and income of marginal net tax rates on earnings, marginal net tax rates on saving, and tax-arbitrage opportunities can be summarized with one word—bizarre. We proceed in second section by laying out our methods for measuring total marginal net taxes on working additional hours and on saving. The third section describes ESPlanner and its use in this paper. The fourth section presents our stylized households. The fifth section presents results and the sixth section concludes. 2. Measuring Total Effective Marginal Tax Rates and the Tax Arbitrage Opportunities Afforded by Retirement Accounts Economists measure the gain from extra work or saving in terms of its potential impact on consumption. The gain from extra current work is typically measured in terms of its maximum impact on current consumption. Thus, if a worker earns an extra $100 this year, permitting this year’s consumption to rise, at most, by $50, we say the worker faces a 50 percent effective marginal tax on her labor supply. The term “effective” references marginal taxes paid net of marginal transfer payments received. Since a large component of some households’ incomes, particular those of low income households, comes from government transfer programs, including such payments in the analysis of earnings and saving incentives is essential. Of course working and earning more in the current year is just one potential margin of choice when it comes to expanding labor supply. We say “potential” because some workers may be in jobs in which the hours they work are pre-set by their employer and can’t be changed.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
87
For such workers, the only way to adjust their annual hours worked is to switch jobs. In this paper we calculate net marginal tax rates on working additional hours in just the current year. But we also determine the net marginal incentives associated with permanently adjusting annual hours worked by switching from a job with a low fixed-level of annual hours to one with a high fixed-level of annual hours. We refer to such a job change as an increase in life-cycle labor supply. To measure this net tax rate we compare the change in the present value lifetime income before any taxes and transfer payments arising from a uniform increase in annual hours (and earnings, since we consider fixed real wages per hour) to the change in the present value of lifetime spending permitted by this additional labor supply. Our third marginal tax of interest is that on extra saving. The gain from extra saving can be measured in terms of the impact on future consumption of forgoing a fixed amount of current consumption. Consider, for example, a two-period (youth and old age) framework. In the absence of any effective marginal tax on saving, reducing current consumption when young by $100 would lead to an increase in consumption when old, measured in present value, of exactly $100. If consumption when old, measured in present value, rises by only $50, the saver faces a 50 percent marginal net tax on saving.1 Our analysis involves, of course, households that live for many years, not just two periods. When there is more than one period (more than one future year) in which to consume, there is no standard definition of the effective tax rate on saving. One could, for example, consider how much reducing this year’s consumption by, say, $100 will increase the present value of future consumption spending assuming the additional future spending power is all allocated to next year’s consumption. Alternatively, one could allocate all the future spending power to consumption ten years out, or 20 years out, or in any future year one chooses. One could also spread the extra spending power uniformly over all future years. Each such choice will generate a different measure of the effective tax rate. The reason is that the longer one pushes out the allocation of the extra spending power, the higher will be the effective tax rate thanks to the nature of compounding. Our response to this surfeit of computable saving tax rates is to present the saving rate associated with reducing current consumption and raising all future consumption levels by the same percentage. More precisely, we compare the present value increase in future spending that
88
Kotlikoff and Rapson
can be financed by a given reduction in current spending assuming that spending in each future year rises by the same percentage. Our final goal is to illustrate the arbitrage opportunities available to households for saving in either (1) 401(k), traditional IRA, or taxdeferred accounts or (2) Roth IRAs, Roth 401(k)s, or other Roth accounts. As described below, we arrange this analysis such that one can directly compare the arbitrage opportunities from contribution to tax-deferred accounts with those from contributing to Roth IRAs, Roth 401(k)s, or other Roth accounts. 2.1 Accounting for Transfer Payments Both marginal earnings and marginal saving can alter the amount of transfers received, which will, in turn, affect the calculation of effective tax rates. As is well known, marginal-transfer schedules are highly nonlinear. For example, in Massachusetts—the state in which we assume our stylized households reside—a household is eligible to receive welfare (TAFDC) if its assets are below $2,500. If this household currently receives welfare and holds $2,499 in assets, an additional dollar saved will render it TAFDC-ineligible. As another example, consider a twoparent family that earns $25,736 per year in labor income and has two dependent children. In Massachusetts, this family is eligible to receive nearly $14,000 in transfers, most of which come from Medicaid.2 Earning an additional dollar or, indeed, an extra penny, causes the family to lose Medicaid eligibility. Accounting for government transfer programs in the estimation of tax rates raises three issues. One is simply their accurate measurement, which requires taking into account each program’s eligibility, income and asset tests. This is a significant undertaking given that ESPlanner does not compute transfer payments apart from Social Security benefits. As described in the Appendix, our transfer benefit calculator assesses household eligibility for each of the transfer programs and applies all applicable income and asset taxes in determining benefit levels. The second issue is the fungibility of transfer payments. Certain benefits, like Medicare and Medicaid, are in-kind and must be consumed in the year received. Others, like TAFDC and, potentially, Food Stamps are fungible. Ideally, one would want to enter fungible benefits as special receipts in ESPlanner and treat non-fungible benefits as consumption in the year they are received. But given the time involved in entering a large number of fungible special receipts in a large number of ESPlanner profiles, we opted to treat all transfer payments as non-fungible, i.e., as consumed in the year they are received.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
89
A third challenge in incorporating transfer payments is identifying the precise point at which marginal net tax rates spike. As is well known, marginal net tax rates can be extremely high at certain levels of earnings and saving because of the discontinuous nature of tax and transfer schedules.3 The examples just sighted in which earning an extra penny of income triggers major losses in TAFCD and Medicaid benefits are cases in point. Identifying these spikes requires considering very small increments in earnings and saving in the range of earnings and saving where such spikes are known to occur. Our initial analysis uses discreet increments equal to the maximum of $100 or 1 percent of earnings to determine the general pattern of labor supply incentives. We then consider much smaller increments to determine precisely where marginal net tax rates spike. 2.2 Calculating Marginal Net Taxes on Current Labor Supply To calculate marginal net tax rates on current labor supply we simply calculate the marginal income net of taxes and gross of transfer payments that would be generated from earning additional income in the current year and then assume this additional net income is spent in the current year.4 To determine how much current net income rises for a given increment in current earnings, we run each of our stylized households through ESPlanner as well as through our annual transfer benefit calculator twice— first, based on their initial levels of earnings and then based on an incremented level of earnings. Equation (4.1) provides a formula for the our net tax rate, τc, on current labor supply. In the formula, ΔE stands for the change in currentyear labor earnings, ΔT for the change in current-year taxes, ΔX for the change in current-year transfer payments received, θs for the state sales tax, and θe for the rate of federal excise taxation.5
τc = 1 −
ΔE − ΔT + ΔX . (1 + θ s + θ e )ΔE
(4.1)
Note that the standard formula for the net tax rate on labor supply is
τc =
ΔT − ΔX . ΔE
But the standard formula ignores sales and excise taxes; i.e., it treats both θs and θe as equaling zero. This is clearly inappropriate since sales and excise taxes, like income and payroll taxes, limit the amount of actual consumption (not consumption expenditure) a worker can enjoy by working more and earning more income.6 Dividing the change in
Kotlikoff and Rapson
90
expenditure associated with additional earnings (ΔE – ΔT – ΔX) by the sales- and excise-tax inclusive consumer price of a dollar of expenditure, (1 + θs + θe), determines how much actual consumption a worker ends up with if she increases her earnings by ΔE.7 2.3 Calculating Marginal Net Taxes on Life-Cycle Labor Supply We define the net marginal tax on life-cycle labor supply, τl, in (4.2).
τl = 1 −
PV ΔC , (1 + θ s + θ e )PV ΔE
(4.2)
where PVΔC denotes the change in the present value of total consumption and other “off-the-top” spending (on housing, insurance premiums, and special expenditures) and PVΔE denotes the change in the present value of lifetime earnings arising from a uniform increase in annual earnings. As discussed in more detail shortly, the discount rate used to form these present values is the return before both corporate and individual taxes. To calculate PVΔC we (1) use ESPlanner to calculate the present value of total spending (consumption spending, housing spending, special expenditures, and insurance premiums) given base-case annual earnings and (2) add to this present value of total spending the present value of transfer payments accruing to the household given ESPlanner’s calculated annual time path of annual total income and assets. Next we increase annual household earnings by a fixed amount each year (specifically, 1 percent of each household’s assumed fixed annual real earnings) through retirement and use ESPlanner plus our transfer calculator to obtain new present values of remaining lifetime earnings and total spending. Differencing the new and previously derived present values of total spending provides the numerator in (4.2). The denominator is determined by simply forming the present value of annual increases in pre-tax and pre-transfer payments earnings. Since ESPlanner smooths households’ living standards subject to borrowing constraints, it will spend extra earnings in a given year on consumption in all years provided doing so does not violate the userspecified limit on borrowing. For purposes of calculating τl we specify this limit at zero. To the extent that borrowing constraints permit, ESPlanner will freely spend in one year earnings generated in another. In so doing, the program will alter the time path of regular asset, regular asset income, and taxes levied on regular asset income. Hence, our tax rate τl on life-cycle earnings will pick up more than simply taxes levied
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
91
on earnings. It will also capture marginal taxation of saving. Thus, we don’t claim τl to represent solely a marginal net tax on life-cycle earnings, but rather a marginal net tax on increased annual earnings that is then subject to as much consumption smoothing as possible.8 2.4 Calculating Effective Marginal Taxes on Regular Saving As indicated, we measure the effective tax rate on saving assuming that the reduction in 2005 spending is allocated uniformly to all future periods such that the living standard in all future periods rises by the same percentage. To effect this outcome in ESPlanner we do two things. First, we permitted all our stylized households to borrow as much as they needed in order to fully smooth their living standards as well as to use additional current saving to effect a uniform rise in their future living standards.9 Second, we raised the program’s living standard index for all years from 2006 onward by 10 percent and compared the increase in the present value of consumption spending from 2006 onward with the associated reduction in consumption spending in 2005. This second step leads the program to lower current consumption spending, while increasing future consumption spending each year by the same percentage, thus effecting a uniform rise in living standard in all future years. The discount rate used to determine the present value change in future consumption, all measured in 2005 dollars, is 7.0 percent, which is our assumed pre- all taxes real rate of return. This pre-tax return is the return one would receive before the application of any federal and state personal or corporate income taxes. In using this return, we are, in effect, incorporating marginal effective corporate capital income taxes as well as marginal effective personal capital income taxes. To see why one needs to discount at the pre- all taxes return, consider a two-period framework with lifetime household budget constraint given by c y + co / (1 + r ) = e y + eo / (1 + r ) − Ty − To / (1 + r ) .
(4.3)
The return r is pre all taxes. The terms cy and co stand for consumption when young and old. The terms ey, eo, Ty, and Tc stand, respectively, for pre-tax earnings when young, pre-tax earnings when old, net taxes paid when young, and net taxes paid when old. Net taxes here are comprehensive; for example, taxes when old include, in the U.S. context, corporate income taxes, personal capital income taxes, personal labor income taxes, state income taxes, payroll taxes, sales taxes, and excise taxes net of all manner of available transfer payments.
Kotlikoff and Rapson
92
Consumption, earnings, and taxes when old are discounted at rate r. For a given reduction in current consumption equal, say, to Δcy, the marginal net tax rate on saving, τs, is given by
τs =
ΔTy + ΔTo / (1 + r ) Δc y
.
(4.4)
The formula for τs tells us the percentage degree to which the present value of future consumption, Δco /(1 + r), fails to rise by the same amount (in absolute value) that current consumption falls; i.e., were τs to equal zero, Δco /(1 + r) would equal –Δcy according to (4.3) under our assumption that ey + eo/(1 + r) don’t change. Note that if one knows r and the value of Δco, one can compute ΔTy + ΔTo / (1 + r ) Δc y by calculating Δco / (1 + r ) − Δc y and subtracting 1 from the resulting ratio. Now we know r, but how do we determine Δco? For purposes of this study, the answer is that we use ESPlanner to determine Δco (actually, the change in each future year’s consumption). To be clear, ESPlanner is operating not off the budget constraint (4.3), but off the following budget constraint, c y + co / (1 + r n ) = e y + eo / (1 + r n ) − T y −Ton / (1 + r ),
(4.5)
where rn is the return households earn pre-individual capital income taxes, but post corporate income taxes and T no are individual income taxes paid when old (i.e., T no does not include corporate income tax payments). Given the assumed linearity of the corporate income tax, the two budget constraints (4.3) and (4.5) are mutually consistent, so there is no problem using (4.5) to determine Δco and then plugging this amount into the formula 1 – Δco /(1 + r)/ – Δcy to form the desired marginal net tax rate on saving. To see this, write rn = r(1 – τc), where τc is the corporate income tax rate. If one substitutes this expression for rn in (4.5) and notes that To – T no = (ey – Ty – cy) r τc (i.e., the two variables differ by the amount of the corporate tax revenue), one arrives at (4.3).
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
93
2.5 Return Assumptions Used in Running ESPlanner In running ESPlanner we enter an 8.33 percent nominal rate of return. Given our 3 percent inflation rate assumption, this translates into a 5.17 percent post-corporate tax real return.10 We use a 7.0 percent real precorporate tax rate of rate (the r in equation (4.3)) to do the discounting needed to form tax rates on life-cycle labor supply and saving. We arrived at these values based on consultations with Jane Gravelle. 2.6 Assessing the Tax-Arbitrage Opportunities in Contributing to Retirement Accounts So far we’ve considered only marginal net taxation of regular saving. But much of household saving is currently being done within either 401(k) and other tax-deferred retirement accounts or within Roth IRAs, Roth 401(k)s, or other Roth accounts. Contributing to these accounts does not, however, necessarily entail any reduction in current consumption. Indeed, contributing to these accounts represents a tax arbitrage opportunity if, as we’ve been assuming, households are not liquidity constrained. To assess these tax-arbitrage opportunities we measure the increase in the present value of all consumption—current as well as future—per net dollar contributed to either type of retirement account. The “net” in “per net dollar” refers to the contribution net of current taxes saved. Thus, if we have a household contribute X to a 401(k) account and it saves the household Y in current taxes, we define the net dollar contribution to be X – Y. This is the amount by which the household’s liquid assets are reduced by the transactions. Since Roth contributions are made after tax and do not affect current taxable income, we consider contributions of size X – Y in order to maintain comparability with respect to our analysis of contributions to tax-deferred accounts. Our analysis here does not include any marginal employer matching contribution. The reason is that we want to understand the pure tax arbitrage incentives presented by retirement “saving” as opposed to the incentive to “save” in retirement accounts presented by employers. 3.
Using ESPlanner to Measure Total Effective Marginal Tax Rates
The methods discussed above to calculate marginal net taxes on lifecycle labor supply and on saving require the use of a dynamic life-cycle model that jointly calculates all future taxes and transfer payments.
94
Kotlikoff and Rapson
ESPlanner is clearly one such model. It determines a household’s highest sustainable living standard within each non-liquidity constrained interval of its life and the consumption, saving, and term life insurance holdings needed to smooth the household’s living standard within each non-constrained interval. The program uses dynamic programming in forming its recommendations. Dynamic programming is needed to deal both with potential borrowing constraints and with non-negativity constraints on life insurance holdings. The program takes into account the following user-specified inputs: the household’s state of residence, current and future planned children and their years of birth, current and future regular and self-employment earnings, current and future special expenditures and receipts (as well as their tax status), current and future levels of a reserve fund, current regular and retirement account balances, current and future own and employer contributions to retirement accounts (with Roth account contributions treated separately), current and future primary and vacation home values, mortgages, rental expenses, and other housing expenditures, current and future states of residence, ages of retirement account withdrawals, ages of initial Social Security benefit receipt, past and future covered Social Security earnings, desired funeral expenses and bequests, current regular saving and life insurance holdings, the economies of shared living, the relative cost of children, the extent of future changes in Social Security benefits, the extent of future changes in federal income taxes, FICA taxes, and state income taxes, current and future pension and annuities (including lump sum and survivor benefits), the degree to which the household will annuitize its retirement account assets and values of future earnings, special expenditures, receipts, and other variables in survivor states in which either the head or her spouse/partner is deceased. The living standard of members of a household is defined by ESPlanner as the amount of consumption expenditure an adult would need to make to enjoy as a single person with no children the same living standard she enjoys in the household. The equation relating a household’s living standard per member to its total consumption expenditure takes into account economies in shared living and the relative cost of children.11 Consumption expenditure is defined by ESPlanner as all expenditures apart from special expenditures, such as college tuition for children, housing expenditures, taxes, life insurance premiums, regular saving, and contributions to retirement accounts.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
95
3.1 ESPlanner’s Tax Calculations ESPlanner makes highly detailed federal income, FICA, and state-specific income tax as well as Social Security benefit calculations. These tax and benefit levels are the only non-user specified variables influencing the program’s consumption smoothing calculations. The program’s federal and state income-tax calculators determine whether the household should itemize its deductions, compute deductions and exemptions, deduct from taxable income contributions to taxdeferred retirement accounts, include in taxable income withdrawals from such accounts as well as the taxable component of Social Security benefits, check, in the case of federal income taxes, for Alternative Minimum Tax liability, and calculate total tax liabilities after all applicable refundable and non-refundable tax credits including the Earned Income Tax Credit, the Child Credit, and the Saver’s credit. These federal and state tax calculations are made separately for each year that the couple is alive as well as for each year a survivor may be alive. Given the non-linearity of tax functions, one can’t determine a household’s tax rates in future years without knowing its regular asset and other taxable income in those years. But one can’t determine how much a household will consume and save and thus have in asset income in future years without knowing the household’s future taxes. Hence, there is a chicken and egg problem—a simultaneity problem—that needs to be resolved to make sure that consumption and saving decisions are consistent with the future tax payments they help engender. 3.2 ESPlanner’s Social Security Benefit Calculations In determining Social Security benefits the program takes full account of the earnings test, early retirement reduction factors, the delayed retirement credit, the re-computation of benefits, the family benefit maximum, the phase-in to the system’s ultimate age-67 normal retirement age, as well as offset and windfall elimination provisions. ESPlanner’s survivor tax and benefit calculations for surviving wives (husbands) are made separately for each possible date of death of the husband (wife). That is, ESPlanner considers separately each date the husband (wife) might die and calculates the taxes and benefits a surviving wife (husband) and her (his) children would receive each year thereafter. Moreover, in calculating survivor-state specific retirement, survivor, mother, father, and child dependent and survivor Social Security benefits, ESPlanner takes account of all the just-mentioned benefit adjustment factors.
96
Kotlikoff and Rapson
3.3 Checking the Calculations Each component of ESPlanner’s tax code and transfer calculator, whether it be the basics of the 1040 form, the provisions of the Earned Income Tax Credit, the details of the Alternative Minimum Tax, the tax treatment of housing capital gains, the taxation of Social Security benefits, the TAFDC earnings test, the payment in the case of low-income households of Medicare premiums by Medicaid, etc.—has been rigorously checked on a component by component basis. This is not to say that no bugs were found. On the contrary, a goodly number were found thanks to independent checking over the years by three software engineers and four economists as well as a large number of ESPlanner users, including professional financial planners, who have examined the tax and Social Security benefit calculations with extremely sharp eyes.12 3.4 ESPlanner’s Algorithm ESPlanner generates recommended levels of annual consumption expenditure, saving, and term life insurance holdings. All recommendations are presented in today’s dollars. Consumption in this context is everything the household gets to spend after paying for its “off-thetop” expenditures—its housing expenses, special expenditures, life insurance premiums, special bequests, taxes, and net contributions to tax-favored accounts. Given the household’s demographic information, preferences, borrowing constraints, and non-negativity constraints on life insurance, ESPlanner calculates the highest sustainable and smoothest possible living standard over time, leaving the household with zero terminal assets (apart from the equity in homes that the user has chosen not to sell) if either the household head, her spouse/partner, or both live to their maximum ages of life. The amount of recommended consumption expenditures needed to achieve a given living standard varies from year to year in response to changes in the household’s composition. Moreover, the relationship between consumption and living standard in a given year is nonlinear for two reasons. First, a non-linear function governs the program’s assumed economies of shared living, with the function depending on the number of equivalent adults. Second, the program permits users to specify that children are less or more expensive than adults in terms of delivering a given living standard. The default setting is that a child is 70 percent as expensive as an adult. Hence a household with two adults and two children is specified, under the default assumptions, to entail 3.4 equivalent adults.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
97
The program’s recommended consumption also rises when the household moves from a situation of being liquidity constrained to one of being unconstrained. Finally, recommended household consumption will change over time if users intentionally specify, via the program’s standard of living index, that they want their living standard to change. Dealing with the simultaneity issues as well as the borrowing and non-negative life insurance constraints all within a single dynamic program appears impossible given the large number of state variables such an approach entails.13 To overcome this problem, ESPlanner uses an iterative method of dynamic programming. Specifically, the program has two dynamic programs that pass data to one another on an iterative basis until they both converge to a single mutually consistent solution to many decimal points of accuracy. One program takes age-specific life insurance premium payments as given and calculates the household’s consumption smoothing conditional on these payments. The other program takes the output of this consumption smoothing program—the living standard in each year that needs to be protected—as given. This second program calculates how much life insurance is needed by both potential decedents and their surviving spouses/partners. This iterative procedure also deals with our two simultaneity issues. The trick here is to form initial guesses of future taxes and survivor life insurance holdings and update these guesses across successive iterations based on values of these variables endogenously generated by the program in the previous iteration. When the program concludes its calculations, current spending is fully consistent with future taxes and vice versa, and the recommended life insurance holdings of heads and spouses/partners are fully consistent with the recommended life insurance holdings of survivors. 3.5 Accounting for Employer-Paid FICA Taxes and Corporate Income Taxes Since users enter their earnings net of employer-paid FICA taxes ESPlanner does not explicitly calculate these taxes. Nor does it explicitly calculate corporate income taxes since users enter their expected returns net of such taxes. From an economics perspective, employer-paid payroll taxes are no less of a burden or a work or saving disincentive than are those paid directly by employees. Indeed, there is only one economic difference between employer-paid and employee-paid payroll taxes; employer-paid payroll taxes are excludable from the calculation
Kotlikoff and Rapson
98
of adjusted gross income in determining federal personal income tax liability, whereas employee-paid payroll taxes are not. Our procedure for including the employer FICA tax is to input into ESPlanner a given increase in earnings, say X (where X is either an increase in current earnings or an increase in the present value of future earnings), and compare the associated increase in spending not with X, but with X plus the additional FICA tax paid on X. This sum represents the full pre-tax compensation being paid to the household. Like employer-paid payroll taxes, corporate income taxes, both federal and state, also reduce the return to input suppliers. But unlike payroll taxes, where the input supply is labor, the input supply relevant to the corporate income tax is household saving. This saving helps finance corporations, and when corporations have to pay taxes, they can’t pay as high a return to their investors. To capture this discrepancy between the pre- and post-corporate tax rates of return, we use the pre-corporate tax discussed above in all the discounting used to form present values. However, in actually running ESPlanner, we enter the post-corporate return as an input in the program since, to repeat, ESPlanner doesn’t calculate corporate taxes. 3.6 Non-Social Security Transfers As indicated, our transfer calculator determines the level of benefits of seven government programs available to residents of Massachusetts: Transitional Aid to Families with Dependent Children (TAFDC), Supplemental Security Income (SSI), Food Stamps, Special Supplemental Nutrition Program for Women with Infants and Children (WIC), Medicare, Medicaid, and Low Income Home Energy Assistance Program (LIHEAP). For each year of potential life of our stylized households, we consider whether the household is eligible for the transfer based on it demographics, income, and assets and, if eligible, compute the appropriate benefit level taking into account any relevant earnings and asset tests. These provisions can include earnings deductions, net income adjustments (such as non-reimbursed out-of-pocket medical expenses), child deductions, and housing deductions. Often the earnings tests are tied explicitly to the federal poverty lines, which vary by the number of household members. 4.
Our Stylized Households
Our stylized households consist of either single individuals or married couples, whose spouses are the same age. We consider households
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
99
age 30, 45, and 60. Both the single-headed households and the married households have two children to whom they gave birth at ages 27 and 29. Table 4.1 lists key assumptions about the seven single and seven married households we consider. The single households have initial labor earnings ranging from $0 to $250,000. For the married couples, the spread is double that of the singles, i.e., it ranges from $0 to $500,000. All household heads and spouses retire and start collecting Social Security benefits at age 65. Earnings between the household’s current (2005) age and retirement at the beginning of age 65 are assumed to remain fixed in real terms. Each household is assumed to have a home, a mortgage, and nonmortgage housing expenses. The 30 year-old households have initial assets equal to a quarter of a year’s earnings. The older households are assumed to have the same assets that ESPlanner determines the 30 year-olds to have at the age at which we consider the older households. The households are also assumed to incur non-housing expenses, the most significant component of which is annual college tuition. For ease of implementation, and to avoid unrealistic profiles, tuition is assumed to be a quarter of a year’s earnings, subject to a ceiling of $50,000 per child. The households pay these amounts each year for four years for each child when the child is age 19 to 22. The final assumption to discuss concerns longevity. The default assumption in ESPlanner is that users have maximum ages of life of 100. Since the program is focused on economic security, this seems appropriate; users may live this long and need to plan for this eventuality. But for purposes of understanding the marginal net taxes households pay, on average, the appropriate longevity assumption is expected, rather than maximum lifespan. Hence, for this analysis, we run the stylized households through ESPlanner under the assumption that household heads and their spouses or partners live to age 85. This is greater than current life expectancy at birth, but seems appropriate given that we are considering households age 30, 45, and 60. 5.
Results
Tables 4.2 and 4.3 present our calculated marginal net tax rates on current labor supply for couples and singles, respectively. The increment we consider in current earnings is the maximum of $100 or 1 percent of current earnings. Consequently, the marginal net tax rates we compute are relative to this increment. We discuss below marginal net tax rates over 1 penny increments in earnings.
100
Table 4.1 Characteristics of Stylized Households Single Households Total Annual Household Earnings
Assets at Age 30
Annual College Expense
House Value
Mortgage
Monthly Mortgage Payment
Annual Property Taxes
Annual Home Maintenance
$10,000
$2,500
$2,500
$30,000
$24,000
$300
$300
$150
$15,000
$3,750
$3,750
$45,000
$36,000
$450
$450
$225
$25,000
$6,250
$6,250
$75,000
$60,000
$750
$750
$375
$35,000
$8,750
$8,750
$105,000
$84,000
$1,050
$1,050
$525
$50,000
$12,500
$12,500
$150,000
$120,000
$1,500
$1,500
$750
$100,000
$25,000
$25,000
$300,000
$240,000
$3,000
$3,000
$1,500
$250,000
$62,500
$50,000
$750,000
$600,000
$7,500
$7,500
$3,750
$20,000
$5,000
$5,000
$60,000
$48,000
$600
$600
$300
$30,000
$7,500
$7,500
$90,000
$72,000
$900
$900
$450
$50,000
$12,500
$12,500
$150,000
$120,000
$1,500
$1,500
$750
$70,000
$17,500
$17,500
$210,000
$168,000
$2,100
$2,100
$1,050
$100,000
$25,000
$25,000
$300,000
$240,000
$3,000
$3,000
$1,500
$200,000
$50,000
$50,000
$600,000
$480,000
$6,000
$6,000
$3,000
$500,000
$125,000
$50,000
$1,500,000
$1,200,000
$15,000
$15,000
$7,500
Married Households
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
101
Table 4.2 Marginal Net Tax Rates on Current-Year Labor Supply (Couples) Total Annual Household Earnings ($000s) Age
10
20
30
50
75
100
150
200
300
30
–14.20%
42.50%
42.30%
24.40%
36.90%
37.00%
45.90%
36.80%
43.90% 44.00%
45
–11.40%
41.70%
41.80%
35.80%
36.10%
36.10%
45.10%
35.90%
40.90% 43.20%
60
50.90%
32.00%
36.30%
36.50%
45.50%
45.50%
47.70%
43.20%
45.80% 45.00%
500
Table 4.3 Marginal Net Tax Rates on Current-Year Labor Supply (Singles) Total Annual Household Earnings ($000s) Age
10
20
30
50
75
100
125
150
200
250
42.90%
42.90%
37.00%
37.00%
36.10%
36.20%
36.90%
42.00% 41.50%
45
–9.80%
42.90%
42.60%
37.00%
36.90%
36.10%
36.10%
36.90%
42.00% 41.50%
60
39.50
37.30%
37.70%
46.40%
45.50%
38.80%
38.80%
44.00%
45.00% 44.00%
30
72.30%
The first impression one gets from glancing at these tables is that marginal rates calculated with respect to the aforementioned discrete earnings increments are either moderate or high for essentially all households except for very low-earning young and middle–age couples as well as middle–aged singles. For all households with $20,000 or more in annual earnings, marginal net tax rates range from 24 percent to 45 percent. The relationship of marginal rates to income is anything but monotonic in earnings. Nor does it take on the U-shaped pattern suggested by optimal income tax theory (see Diamond 1998). Take couples age 30. The marginal rate is –14 percent at $10,000 in earnings, 42 percent at $20,000, 24 percent at $50,000, 37 percent at $75,000, 46 percent at $150,000, 37 percent at $200,000, and 44 percent at $500,000. In addition to anomalous patterns of marginal rates with income, holding age constant, there are also unusual patterns with respect to age, holding income fixed. Take singles earning $10,000. Thirty-year old members of this group face a marginal net tax rate of 72 percent. Were they age 45, their marginal rate would be –10 percent. And were they 60, their marginal rate would be 39 percent. As another example of the surprising relationships between age and marginal rates, note that rates
102
Kotlikoff and Rapson
fall with age for couples with $30,000 in earnings, but rise with age for couples with $75,000 in earnings. 5.1 Explaining Patterns of Work Incentives by Age and Earnings How does one make sense of these findings? Well, the size of each marginal net tax rate is easily traced to underlying marginal changes in particular taxes or transfer payments. Take, for example, married households age 30 that earn $10,000 per year. Their –14 percent net tax rate reflects the major marginal subsidy being provided to them by the Earned Income Tax Credit; this subsidy significantly exceeds the marginal payroll and sales and excise taxes they pay on additional earnings.14 If this same household were to earn $20,000, rather than $10,000, its marginal net tax rate would be 42 percent rather than –14 percent. The reason is that at this higher earnings level, the EITC is being clawed back at a rate of more than 20 cents on the dollar. In addition, the household pays, at the margin, FICA and state income taxes and also gets hit by sales and excise taxes. Next consider the $10,000 couple, but at age 60. Unlike their younger counterparts, this couple is no longer eligible for the EITC because it no longer has young children and its earnings exceed the income cutoff. On the other hand, the couple does receive Food Stamps. But because it has no young children, the couple is in the Food Stamps claw back range, where it loses 24 cents in Food Stamps per dollar earned. This marginal tax in conjunction with the 15.3 percent employer and employee FICA, the Massachusetts 5.3 percent income tax, the Massachusetts 5.0 percent sales tax, and the .9 percent assumed federal excise tax rate delivers a net marginal rate of 51 percent.15 As a third example of one’s ability to precisely trace the anomalous nature of these marginal net taxes, consider 30 year old singles who earn only $10,000 per year. Unlike their married counterparts who face a 14 percent subsidy on additional current earnings, these single households face a 72 percent marginal net tax. The major difference between the two cases involves the claw back of TADFC. Because the single household’s family size is smaller, it faces the TADFC claw back of 100 cents on the dollar when it earns $10,000, whereas the married household faces this effective marginal tax only at a higher earnings level. Surprisingly, if the $10,000 single household is age 45 rather than age 30, the marginal net tax is –10 percent rather than 72 percent. What explains this huge difference? The answer has to do with the TAFDC benefit. Because the 45 year old single household has older children, it
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
103
no longer qualifies for the TAFDC daycare allowance or, consequently, any TAFDC benefits. At the margin it therefore faces no TAFDC claw back tax. On the other hand, its earnings are so low that it’s in the Earned Income Tax Credit’s positive subsidy range. This subsidy is sufficiently high to produce a negative net marginal tax on labor supply notwithstanding the state, FICA, sales, and excise taxes this household must pay on marginal earnings. If we advance this household’s age by another 15 years and consider it at age 60, we find it again faces a very high, positive marginal net tax, in this case 39 percent. Because this household’s children are now grown, it finds itself in the EITC claw back range, which contributes significantly to the total net marginal tax it faces. Tracing each household’s marginal net tax on supplying more current earnings is one thing. Understanding why anyone would intentionally design a fiscal system with such a bizarre pattern of work incentives by age and earnings is another. The explanation is that these patterns are unintended. Indeed, for federal and state government officials to have intentionally designed these incentives would have required them to know what they were doing. But, to our understanding, this is the very first study to have incorporated all of the major federal and state taxtransfer programs.16 Thus, those who designed this sausage could literally not have known what they were doing. But why didn’t they try to find out? The answer is that no single government body is responsible for the overall structure of our fiscal incentives. Instead, the 20 or so major tax-transfer programs/provisions that combine to produce these bizarre incentives are being set by various federal and state governmental committees/bodies each of whom ignore, for the most part, the workings of the others and focus only on the details of the program/provision over which they have responsibility. 5.2 Marginal Net Tax Rates on Life-Cycle Labor Supply Table 4.4 presents marginal life-cycle net tax rates for our 30 year old households. In these calculations, the increment in annual earnings is the maximum of $100 or 1 percent of each year’s earnings. First consider couples. Their net tax rates are generally similar to the current marginal tax rates reported in Table 4.2 for 30 year old couples. The main differences occur at $10,000, $50,000 and $500,000 in income. At these income rates the life-cycle net tax rates are significantly higher than the current-year rates. This is not to suggest that life-cycle rates are
104
Table 4.4 Marginal Net Tax Rates on Life-Cycle Labor Supply Couples—Total Annual Household Earnings ($000s) 10
20
30
50
75
100
150
200
300
500
2.10%
40.20%
40.10%
32.30%
36.60%
33.30%
42.20%
41.60%
42.80%
49.60%
Singles—Total Annual Household Earnings ($000s) 10
20
30
50
75
100
125
150
200
250
0.80%
34.70%
36.70%
32.60%
34.60%
39.50%
37.30%
37.70%
40.30%
41.30%
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
105
always higher for given income levels. There are several income levels in tables 4.2 and 4.4 at which the life-cycle rates are lower. For single households age 30, life-cycle and current-year marginal rates are very different for earnings below $125,000, but quite similar at that level of earnings and above. Take the $10,000 earnings case. The current-year marginal net tax rate is 72 percent, whereas the life-cycle rate is only 2 percent. At $75,000 in earnings, the life-cycle rate is 76 percent, whereas the current-year rate is 37 percent. Part of what is going on here is that low-income households that are eligible for Medicaid, TAFDC and other welfare benefits in the current year will not be receiving these benefits throughout their lives because of changes in their household demographics and levels of non-labor income. 5.3 Budget Constraints Now that we’ve provided a broad brush overview of marginal net tax rates measured over discrete intervals, we turn to a more detailed analysis of the highly non-linear and complex budget constraints facing typical earners. The figures at the end of the paper show current year and lifetime budget constraints. The current year budget constraints relate current year net income to current year gross income. The slope of this constraint determines the current year marginal net tax rate. The lifetime budget constraints show how the present value of lifetime spending varies with annual real earnings, where we’re assuming the same annual earnings in all years of work.17 The slope of this budget constraint determines what we’ve been referring to as the life-cycle marginal net tax rate. We also present figures indicating marginal net tax rates on current labor supply as well as the marginal net tax rates on life-cycle labor supply confronting 30 year old households. Take, as an example, the figure relating current net income to current gross income for 45 year old couples. And consider a $25,000 initial total household earnings level, which is close to what the head and spouse would collectively earn were they to work full–time at the minimum wage. This income places the couple about 30 percent above the federal poverty line, but is low enough that the whole family is eligible for Medicaid benefits in Massachusetts. Recall that this household has two dependent children, both of whom are college bound. It also has a $75,000 house with a 15 year remaining mortgage whose balance is just over $30,000. Because of the Earned Income Tax Credit (EITC), Medicaid and other benefits provided by federal and state transfer programs, this house-
106
Kotlikoff and Rapson
hold has net income of just over $40,000 per year. If the couple earns additional wage income, several things will happen. First, every additional dollar earned will generate a claw back of the EITC at the rate of 21 cents per dollar earned. More importantly, if the couple earns enough additional income, it will lose eligibility for roughly $15,000 in Medicaid benefits. The figure showing marginal net tax rates levied on this household’s current labor supply identify where these rates become extremely high. This occurs at points where the households’ incomes exceed income-test thresholds for the various transfer programs. One way to appreciate the size of work disincentives facing this household is to ask how much more it must earn, after losing all its benefits, to achieve the same living standard it enjoys when earning $25,000 and receiving all its benefits. The answer is roughly $50,000. That is, the couple has to double its earnings simply to break even with respect to maintaining its living standard. Such high net taxes apply to all low-income households, regardless of age or marital status. The life-cycle labor supply budget figures as well as their associate marginal net tax-rate diagrams also indicate kinks and high rates of marginal net taxes but these kinks and high rates don’t necessarily line up with those associated with current labor supply. These life-cycle figures tell us not just about the incentives to work more each year, but also about the incentives to take costly steps, such as enhancing one’s education or switching to a more demanding job, that will raise one’s annual earnings for a given level of labor supply by raising one’s hourly wage rate. To further appreciate the nature of life-cycle labor supply disincentives, consider our 60 year old couple earning only $10,000. For this couple earning $55,000 a year for the duration of its working life is only marginally better than earning $10,000. The $10,000/yr household has remaining lifetime spending of $473,000 whereas the $55,000/yr household will spend $480,000. As can be seen in the figure below, all households with incomes between $10,000 and $55,000 will have lower remaining lifetime spending than the $10,000 household. The reason is simple: between $12,000 and $13,000/yr in income, the couple loses its Medicaid benefits in retirement, thanks to the Medicaid asset test, and between $17,000 and $18,000/yr in income it loses Medicaid benefits from age 60–65. These losses (which occur every year between age 60 and death) amount to hundreds of thousands of dollars in present value.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
107
Younger households face similar life-cycle budget constraints, but the life-cycle labor supply disincentives are considerably smaller. This is because Medicaid expenditures comprise a larger fraction of remaining lifetime consumption at age 60 than age 30 or 45. Discounting these future losses to present value and recognizing that younger couples have far more years of working over which to make up the transfer losses makes clear why younger households are not as adversely affected. For 30 year old couples and singles, they must earn $10,000 to $15,000/yr more to overcome their loss of Medicaid when it occurs; 45 year olds must earn $15,000–$25,000/yr more; and, to repeat, 60 year olds must earn $25,000–$45,000/yr more. 5.4 Measuring Marginal Net Taxes on Saving Tables 4.5 and 4.6 present our marginal net tax rates on regular and retirement account saving by age and earnings levels. The increment in current saving we consider ranges from $500 to $5,500 depending on the household’s earnings level. Consider first the regular saving findings for couples. Most of the net tax rates fall in the range of 20 to 40 percent. The highest rate is 52 percent, which applies to 30 year old households making $500,000 per year. This is part of a pattern for young and middle-aged households in which the net tax rate on regular saving rises with income. But for 60 year old couples, the rate is 39 percent at the lowest earnings level, then falls to 22 percent and then climbs to 36 percent for households with $500,000 in earnings. The regular saving net tax rates for singles are far afield from those for couples. For very low earning, young and middle–aged singles, the rates are astronomical reflecting the impact of asset tests on various transfer benefits. At higher incomes and at older ages, the rates range from around 20 percent to around 40 percent. Above $34,000 in annual earnings these rates generally rise. 5.5 Measuring Retirement Account Tax Arbitrage Opportunities Tables 4.5 and 4.6 present our findings on tax arbitrage via contributions to tax-deferred retirement accounts, which we reference as “401(k)”-type accounts and Roth accounts. As indicated above, the results are presented in terms of cents of arbitrage gain per dollar of net contribution. Take, as an example, the 401(k) results for our 45 year old couples with $70,000 in total annual household earnings. At the margin, these households can increase the present value of their lifetime consump-
Kotlikoff and Rapson
108
Table 4.5 The Marginal Net Taxation of Saving (Couples) Marginal Effective Tax on Regular Saving Total Household Annual Income ($000s) Age
20
30
50
70
100
200
500
30
20.50%
20.10%
20.50%
23.30%
24.90%
32.00%
51.50%
45
20.10%
21.40%
22.00%
22.60%
25.90%
30.30%
43.40%
60
38.60%
22.10%
22.00%
27.90%
34.10%
34.30%
36.50%
100
200
500
401(k) Arbitrage Opportunity Total Household Annual Income ($000s) Age
20
30
50
70
30
5.7¢
5.6¢
5.9¢
8.6¢
20.4¢
53.9¢
154.7¢
45
6.2¢
7.5¢
24.1¢
23.3¢
21.4¢
44.1¢
79.9¢
60
171.1¢
183.9¢
46.4¢
28.0¢
36.1¢
47.7¢
49.2¢%
Roth Arbitrage Opportunity Total Household Annual Income ($000s) Age
20
30
50
70
100
200
500
30
1.1¢
0.9¢
1.2¢
3.9¢
19.1¢
33.4¢
121.9¢
45
1.1¢
2.9¢
4.0¢
4.4¢
17.6¢
30.8¢
57.0¢
60
47.5¢
48.0¢
16.2¢
15.6¢
25.3¢
23.9¢
27.8¢
tion by 23.3 cents for every dollar they contribute on net (net of their immediate tax savings) to a tax-deferred retirement account. This is a significant money machine. But it’s de minimis compared with the 154.7 cent money machine available to 30 year old couples with $500,000 in annual earnings. On the other hand, it’s huge compared with the .7 cent money machine available to 30 year old single households with earnings of $15,000. As the two tables indicate, the arbitrage opportunities are greatest for high-earning young and middle-aged households and for older households. That said, the pattern of arbitrage opportunities by age and earnings is far from monotonic with respect to either age or by earnings. Take singles households with $35,000 in annual earnings. The size of their 401(k) money machine is 16.3 cents at age 30, 64.9 cents at age
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
109
Table 4.6 The Marginal Net Taxation of Saving (Singles) Marginal Effective Tax on Regular Saving Total Household Annual Income ($000s) Age
10
15
25
35
50
100
250
30
82.70%
260.40%
18.80%
18.70%
20.40%
25.50%
30.60%
45
109.40%
19.60%
19.70%
20.10%
20.20%
30.70%
39.20%
60
20.50%
41.40%
22.00%
23.40%
30.30%
37.60%
35.80%
100
250
401(k) Arbitrage Opportunity Total Household Annual Income ($000s) Age
10
15
25
35
50
30
1.0¢
0.7¢
5.5¢
16.4¢
5.4¢
31.0¢
73.4¢
45
5.8¢
5.9¢
6.6¢
64.9¢
18.0¢
33.8¢
69.4¢
60
47.7¢
76.2¢
64.1¢
32.0¢
42.0¢
33.6¢
55.4¢
100
250
Roth Arbitrage Opportunity Total Household Annual Income ($000s) 15
25
35
50
Age
10
30
1.0¢
0.7¢
0.6¢
0.6¢
2.2¢
28.6¢
53.3¢
45
1.3¢
0.9¢
1.7¢
9.6¢
1.4¢
32.1¢
50.6¢
60
7.1¢
23.9¢
35.0¢
9.6¢
18.2¢
28.0¢
26.5¢
45, and 32.0 cents at age 60. Or consider couples age 60. If they earn $20,000 per year in total, their 401(k) money machine generates 171.1 cent per net dollar contributed. With $70,000 in annual earnings, their 401(k) machine produces only 28.0 cents per net dollar contributed. But at $500,000 in annual earnings, the machine has improved. It now produces 49.2 cents per net dollar contributed. The Roth arbitrage opportunities are uniformly smaller than the 401(k)-type arbitrage opportunities.18 Nonetheless, they can be quite substantial. For example, 45-year old singles earning $100,000 per year stand to receive 32.1 cents per dollar placed in a Roth account.19 The top Roth arbitrage opportunity is that of couples age 30 with $500,000 in annual earnings. Their money machine generates 121.9 cents for free for each dollar they place in a Roth account.
110
Kotlikoff and Rapson
As in the case of marginal net tax rates on labor supply and saving, one can decipher the reason a particular arbitrage opportunity is of a given size. In this regard, the 5.7 cent and 171.1 cent respective arbitrage opportunities of 30 and 60 year old couples earning $20,000 are worth comparing. The 30 year olds have zero (or very small positive) federal tax obligations at age 30, before considering the EITC. To take advantage of the federal Saver’s Credit, they must be paying positive federal taxes. The Saver’s Credit, enacted in 2001, matches low-income households’ retirement account contributions by as much as dollar for dollar, but it does so by reducing their tax payments to the extent these payments are positive; i.e., the Saver’s Credit is not refundable, making many lowincome households ineligible for it. Our 60 year old couple with $20,000 is low-income, but is eligible for the Saver’s Credit. The reason is that the couple no longer has dependent children. With fewer deductions, its adjusted gross income is higher than that of its 30 year old analogue, resulting in a higher (positive) federal tax liability. So when these households contribute to a 401(k) vehicle, they not only reduce their current taxes by exempting their contribution to the 401(k) from their taxable income; they also reduce them because of the Saver’s Credit. These factors, in combination with the fact that these households will be in very low tax brackets in the future, explain the fantastic size of this arbitrage opportunity. Interestingly, the same age 60 couple has a much smaller arbitrage potential if it contributes not to a 401(k)-type vehicle, but to a Roth account. In this case, the money machine spews forth only 47.5 cents per dollar contributed. There are two reasons this machine does so poorly compared to the 401(k) machine. First, the Roth contributions generate no immediate reduction in taxes. Hence, there is no ability, as there is with the 401(k) contribution, to arbitrage between current high and future low marginal tax brackets. Second, each dollar of net contribution to a 401(k) entails a larger gross contribution than in the case of a contribution to a Roth account. Since the Saver’s Credit is paid on the basis of the gross contribution, not the net contribution, a given net contribution to a 401(k)-type account generates a much larger Saver’s Credit than does the same size net contribution made to a Roth account. Another comparison between arbitrage incentives that’s worth making is that between 45 year old 401(k) contributing couples who earn
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
111
$25,000 per year and those who earn $35,000. The lower-earning couple is again not eligible to receive the Saver’s Credit because of its negligible federal tax obligations, whereas the higher earning couple is eligible. A final arbitrage opportunity worth highlighting is that of 30 year old couples with $500 in total annual earnings. These couples can earn 154.7 cents for free per net dollar placed in a 401(k)-type account. This reflects the value of their current tax saving, the fact that they are in much lower tax brackets in the future and their ability to benefit from tax-deferral (the ability to earn capital income on a tax-free basis). As the size of the corresponding Roth arbitrage opportunity makes clear, the deferral advantage for this household is significant. 6.
Conclusion
The study of effective marginal tax rates is hardly new.20 Nor is the observation that transfer programs can dramatically affect effective marginal tax rate calculations, and that marginal rates depend critically and sensitively on household demographic and economic circumstances. But what is new here is the inclusion in one study of all the major tax and transfer programs/elements that materially affect incentives to work and save. On the tax side, this list includes federal and state personal income, corporate income, sales and excise, and payroll taxes. On the transfer side, the list includes Social Security, Medicare, Medicaid, Food Stamps and TAFDC benefits. America’s tax-transfer system confronts the vast majority of American households with either high, very high, or astronomically high total effective marginal tax rates on labor supply and saving. It also provides very substantial tax arbitrage opportunities to a subset of households, particularly those with high incomes or advanced ages. The pattern of net marginal tax rates and arbitrage opportunities with respect to age, marital status, and earnings is quite simply all over the map. But this is what one would expect given the amazing complexity of the fiscal system, the fact that the various components of the system are being developed with little or no thought to their interaction, and that the various governmental bodies responsible for the different elements of our tax-transfer system appear to make little or no attempt to understand the overall work and saving disincentives as well as arbitrage opportunities they are producing.
112
Kotlikoff and Rapson
Notes We thank Jane Gravelle, Alexi Sluchynsky, and Adam Looney for critical advice and comments. This paper builds and draws on Gokhale, Kotlikoff, and Sluchynsky (2002). 1. Alternatively, we can say that the tax on future consumption is 100 percent since the price, measured in present value, of consuming $50 when old has risen from $50 to $100. 2. In assuming that eligible households receive average benefits from transfer programs like Medicaid to particular households we are ignoring the insurance value of these programs. 3. If one could earn infinitesimal amounts, effective marginal net tax rates in these cases would be infinite. But since the smallest increment one can earn is a penny, effective marginal net tax rates, while potentially extremely high, are finite. 4. In maintaining fixed current saving, we’re ensuring no change in future incomes and transfer payments with one exception—future Social Security benefits. These benefits are potentially changed due to the presence of higher current earnings in the worker’s ultimate earnings record. Including the impact of these Social Security benefit changes on current consumption is a goal of our future research. However, it’s important to bear in mind that Social Security benefit changes, to the extent they arise, can only influence current spending insofar as the worker (or household to which the worker belongs) is not liquidity constrained. Many of our stylized households are so constrained. 5. The sales tax in Massachusetts is 5 percent, and the federal excise tax accounts for approximately 0.9 percent of aggregate consumption in the U.S. Hence, we set θs = 0.05 and θe = 0.009. 6. Sales and excise taxes also represent taxes on wealth since, like earnings, when wealth is spent, the spender pays these taxes and ends up getting less actual consumption than would otherwise be the case. 7. In a static setting a worker’s budget constraint is (1 + θs + θe)C = w(1 – τ), where τ is the sum of income and payroll tax rates and w is the pre-tax wage. But one can rewrite this constraint as C = w(1 – τ)/(1 + θs + θe). Letting τe stand for the effective tax rate on labor supply, we have C = w(1 – τe), where τe = 1 – (1 – τ)/(1 + θs + θe), which is the same as equation (4.1). 8. Roughly two-thirds of young American households appear to be liquidity constrained (see Kotlikoff, Marx, and Rizza 2006). This doesn’t necessarily mean that they have zero current fungible assets. Instead it means that their living standard per person in the future will be higher than it is in the present and that whatever saving they are doing is for purposes of smoothing their living standards in the short or medium runs. Like typical young households, all but the highest earning of our stylized young households are liquidity constrained. 9. In assuming that all of our stylized households are able to borrow, we don’t mean to suggest that such borrowing is feasible. Instead, we seek to understand how our taxtransfer system affects the incentive to save were households actually able to do so. 10. The formula for the real return is actually (1 + i)/(1 + π) – 1.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
113
11. Let C stand for a household’s total consumption expenditure, s for its living standard per equivalent adult, ki for the number of children age i, θi for relative cost of a child age i, N for the number of adults, and υ for the degree of economies of shared living. The relationship between C and s in a given year is ⎛ C = s⎜ N + ⎜⎝
∑ i
ν
⎞ θ i ki ⎟ . ⎟⎠
12. Indeed, in the case of Social Security benefit calculations, a number of individual users and financial planners have double checked ESPlanner’s Social Security’s benefit calculations with those produced by Social Security Administration’s detailed ANYPIA calculator. A number have complained that ESPlanner’s calculated benefits were too high. As they were told, ESPlanner’s benefit projections accord precisely with those of the ANYPIA calculator in the case of users whose covered earnings all lie in the past. But in the case of users with projected future covered earnings, ESPlanner’s projection of future benefits differ from the ANYPIA’s projection for a simple reason. The ANYPIA calculator assumes no future rise in the U.S. price level and no future real wage growth. This seems remarkable until one realizes that the government doesn’t want to be in a position of implicitly promising higher benefits than it knows for sure it will pay. 13. The simultaneity issue with respect to taxes mentioned above is just one of two such issues that need to be considered. The second is the joint determination of life insurance holdings of potential decedents and survivors. ESPlanner recognizes that widows and widowers may need to hold life insurance in order to protect their children’s living standard through adulthood and to cover bequests, funeral expenses, and debts (including mortgages) that exceed the survivor’s net worth inclusive of the equity on her/his house. Accordingly, the software calculates these life insurance requirements and reports them in its survivor reports. However, the more life insurance is purchased by the potential decedent, the less life insurance survivors will need to purchase, assuming they have such a need. But this means survivors will pay less in life insurance premiums and have less need for insurance protection from their decedent spouse/partner. Hence, one can’t determine the potential decedent’s life insurance holdings until one determines the survivor’s holdings. But one can’t determine the survivor’s holdings until one determines the decedent’s holdings. 14. This household pays no state income tax at the margin. 15. To be clear, there are interactions in the separate marginal net tax provisions, so these rates are not simply additive for this or any other household. 16. To its credit, the Congressional Budget Office has been providing Congress with detailed studies of marginal effective federal income tax rates. But Congressional Budget Office (2005) and prior studies do not include state income taxes, sales or excise taxes, or any of the seven major transfer programs included here. Moreover, these studies do not use a dynamic/intertemporal model and, consequently, cannot address saving or lifecycle labor supply incentives. 17. The present value of lifetime spending includes here the present value of non-Social Security transfer payments, which, to recall, we are treating as being consumed/spent in the year received. 18. This analysis abstracts from potential future tax hikes that could significantly limit the marginal arbitrage gain available from contributing to tax-deferred retirement accounts.
114
Kotlikoff and Rapson
19. Note that all contributions to Roth accounts are on a net basis because there is no reduction in current taxes associated with adding to one’s Roth account. 20. Recent contributions to the literature on marginal net tax rates include CBO (2005) and Feenberg and Poterba (2003). 21. Access to these data was generously provided to us by Professor Jonathan Skinner of Dartmouth College.
References Congressional Budget Office (2005). “Effective Marginal Taxes on Labor Income,” November. Diamond, Peter (1998). “Optimal Income Taxation: An Example with a U-Shaped Pattern for the Optimal Marginal Rates,” American Economic Review, 88(1):83–95. Feenberg, Daniel, and James Poterba (2003). “The Alternative Minimum Tax and Effective Marginal Tax Rates,” NBER working paper no. 10072, April. Gokhale, Jagadeesh, Laurence J. Kotlikoff, and Alexi Sluchynsky (2002). “Does It Pay to Work,” NBER working paper no. 9095, August. Kotlikoff, Laurence J., Ben Marx, and Pietro Rizza (2006). “Americans’ Dependency on Social Security,” Boston University. Available at http://people.bu.edu/kotlikof/Americ ans%27%20Dependency%20on%20Social%20Security1.pdf.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.1a Gross Income vs. Net Income (1 Year) 30 Year Old Couples Earning $0–$50,000 Per Year
Figure 4.1b Gross Income vs. Net Income (1 Year) 30 Year Old Couples Earning $0–$500,000 Per Year
115
116
Figure 4.2a Marginal Effective Tax Rate 30 Year Old Couples Earning $0–$50,000 Per Year
Figure 4.2b Marginal Effective Tax Rate 30 Year Old Couples $50,000–$500,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.3a PV Lifetime Spending Including Transfers ($’000s) 30 Year Old Couples
Figure 4.3b PV Lifetime Spending Including Transfers ($’000s) 30 Year Old Couples
117
118
Figure 4.4a Marginal Lifetime Effective Tax Rate 30 Year Old Couples $0–$50,000 Per Year
Figure 4.4b Marginal Lifetime Effective Tax Rate 30 Year Old Couples $50,000–$500,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.5a Gross Income vs. Net Income (1 Year) 45 Year Old Couples Earning $0–$50,000 Per Year
Figure 4.5b Gross Income vs. Net Income (1 Year) 45 Year Old Couples Earning $0–$500,000 Per Year
119
120
Figure 4.6a Marginal Effective Tax Rate 45 Year Old Couples Earning $0–$50,000 Per Year
Figure 4.6b Marginal Effective Tax Rate 45 Year Old Couples $50,000–$500,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.7a PV Lifetime Spending Including Transfers ($’000s) 45 Year Old Couples
Figure 4.7b PV Lifetime Spending Including Transfers ($’000s) 45 Year Old Couples
121
122
Figure 4.8a Gross Income vs. Net Income (1 Year) 60 Year Old Couples Earning $0–$50,000 Per Year
Figure 4.8b Gross Income vs. Net Income (1 Year) 60 Year Old Couples Earning $0–$500,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.9a Marginal Effective Tax Rate 60 Year Old Couples Earning $0–$50,000 Per Year
Figure 4.9b Marginal Effective Tax Rate 60 Year Old Couples $50,000–$500,000 Per Year
123
124
Figure 4.10a PV Lifetime Spending Including Transfers ($’000s) 60 Year Old Couples
Figure 4.10b PV Lifetime Spending Including Transfers ($’000s) 60 Year Old Couples
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.11a Gross Income vs. Net Income (1 Year) 30 Year Old Singles Earning $0–$25,000 Per Year
Figure 4.11b Gross Income vs. Net Income (1 Year) 30 Year Old Singles Earning $0–$250,000 Per Year
125
126
Figure 4.12a Marginal Effective Tax Rate 30 Year Old Singles Earning $0–$25,000 Per Year
Figure 4.12b Marginal Effective Tax Rate 30 Year Old Singles $25,000–$250,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.13a PV Lifetime Spending Including Transfers ($’000s) 30 Year Old Singles
Figure 4.13b PV Lifetime Spending Including Transfers ($’000s) 30 Year Old Singles
127
128
Figure 4.14a Marginal Lifetime Effective Tax Rate 30 Year Old Singles $0–$25,000 Per Year
Figure 4.14b Marginal Lifetime Effective Tax Rate 30 Year Old Singles $25,000-$250,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.15a Gross Income vs. Net Income (1 Year) 45 Year Old Singles Earning $0–$25,000 Per Year
Figure 4.15b Gross Income vs. Net Income (1 Year) 45 Year Old Singles Earning $0–$250,000 Per Year
129
130
Figure 4.16a Marginal Effective Tax Rate 45 Year Old Singles Earning $0–$25,000 Per Year
Figure 4.16b Marginal Effective Tax Rate 45 Year Old Singles $25,000–$250,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.17a PV Lifetime Spending Including Transfers ($’000s) 45 Year Old Singles
Figure 4.17b PV Lifetime Spending Including Transfers ($’000s) 45 Year Old Singles
131
132
Figure 4.18a Gross Income vs. Net Income (1 Year) 60 Year Old Singles Earning $0–$25,000 Per Year
Figure 4.18b Gross Income vs. Net Income (1 Year) 60 Year Old Singles Earning $0–$250,000 Per Year
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
Figure 4.19a Marginal Effective Tax Rate 60 Year Old Singles Earning $0–$25,000 Per Year
Figure 4.19b Marginal Effective Tax Rate 60 Year Old Singles $25,000–$250,000 Per Year
133
134
Figure 4.20a PV Lifetime Spending Including Transfers ($’000s) 60 Year Old Singles
Figure 4.20b PV Lifetime Spending Including Transfers ($’000s) 60 Year Old Singles
Kotlikoff and Rapson
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
135
Appendix Our Transfer Calculator The following is a list of the non-Social Security transfer benefit calculated by our transfer calculator. –Transitional Assistance to Families with Dependent Children (TAFDC) –Food Stamps (FS) –Medicaid –Medicare –Supplementary Security Income (SSI) –Low-Income Home Energy Assistance Program (LIHEAP) –Special Supplemental Nutrition Program For Women, Infants and Children (WIC) The annual levels of each transfer benefit are determined taking into account all eligibility criteria, which often include demographics (e.g., number and ages of children), as well as applicable income and asset tests. Each program, however, has eligibility rules and benefit formulae that deal with special cases. For this study, we consider the rules and benefit formulae that apply to the standard cases.
Modeling Specific Benefit Programs Transitional Aid to Families with Dependent Children—TAFDC Transitional Aid to Families with Dependent Children (TAFDC) is a cash assistance program designed to assist needy families with dependent child or pregnant women. TAFDC is the formal name in Massachusetts of the program formerly known as AFDC (Aid to Families with Dependent Children). Most states have adopted the name Temporary Assistance to Needy Families (TANF). The terms “transitional” and “temporary” reflect the new objective of the programs, namely to provide short-term assistance to needy families and to encourage such families to return to the labor force. Under the current rules of the TAFDC, eligible households may generally receive assistance for no more than 24 months within any five year period. There are several steps in defining eligibility for benefits. The calculations needed to determine eligibility, both non-financial and financial, and benefit levels can be complicated even for the standard cases we consider. Non-Financial Eligibility requires that the child must be deprived of the care or support of at least one parent. Deprivation factors include: death, continued absence, physical or mental incapacity, unemployment or underemployment of (a) parent(s). A dependent child may be under age 19 or, if a full–time school student, age 19. We assume that our family units meet these program-specific requirements. To meet requirements for Financial Eligibility a household must pass two income tests. First, family unit gross income cannot exceed 185 percent of the
Kotlikoff and Rapson
136
Table 4.A1
Household Size
Eligibility Standard (185% of the Need Standard)
Need Standard/ Payment Standard
2
982
531
3
1,171
633
4
1,352
731
Need Standard that applies given family size. Second, gross income minus certain applicable deductions cannot exceed the Need Standard itself (see table 4.A1). Standard monthly deductions include –a $90 deduction for each employed family member. –an extra $30 plus one-half of gross income above $120 deduction for the employed TAFDC benefit recipients or applicants who received benefits in the previous 4 months. –dependent-care deductions that range between $50 to $200 for a child under two and $44–$175 for a child two or over, depending on the hours worked by a recipient. We applied the $90 deduction per working individual for all 12 months of each year of eligibility and the maximum deduction levels for childcare for children between ages one and five. However, we did not implement the extra deduction because of its complex dynamic nature. If the family unit passes both income tests it gets financial assistance defined as the difference between the maximum payment standard and net income after deductions. In accordance with standard program restrictions on the length of benefit receipt, we limited the receipt of benefits to no more than 24 months within any five year period. Hence, for those of our stylized households who are eligible for assistance, benefits follow a cyclical pattern: two years on followed by three years off, provided the asset test criterion is met. TAFDC regulation in Massachusetts assumes that families receiving benefits may also receive $40 of monthly housing allowance, which we add to the monthly TAFDC benefit.
Source 1. Mass Resources. Transitional Aid to Families with Dependent Children (TAFDC). http://www.massresources.org/pages.cfm?contentID=17&pageID =4&Subpages=yes
Food Stamps The purpose of the Food Stamp Program is to improve the diet of low-income families by increasing their food purchasing power. Households must satisfy both state and federal requirements to qualify for food stamps. There are sev-
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
137
Table 4.A2
Household Size
Gross Monthly Income Limit
Net Monthly Income Limit
Maximum Monthly Benefit
1
1,009
776
149
2
1,354
1,041
274
3
1,698
1,306
393
4
2,043
1,571
499
eral steps in determining program eligibility and calculating the value of the stamp benefits. First, gross monthly (earned and unearned) income cannot exceed the limits specified in table 4.A2 for households of different sizes. Unearned income includes Social Security and private pension benefits, SSI benefits, unemployment insurance benefits and TAFDC payments. In our study we include SSI and TAFDC payments as part of the income used to calculate the value of food stamps. The following monthly deductions apply: –$134 per household. –20 percent of gross income. –Dependent day care: under two years of age, up to $200 per month; over two years of age, up to $175 per month. We apply here the TAFDC program dependent care deduction for every child between the ages of one and five. –Medical expenses of individuals over 60 years old are deductible beyond the first $35. These expenses are calculated as the sum of payments for prescription drugs, Medicare premiums, deductibles and coinsurance payments. –Excess housing costs, which are defined as housing expenses in excess of half of the household’s income after other deductions. Prior to age 60 there is a maximum level of $388 for deductible excess housing costs. Net monthly income (monthly income after deductions) cannot exceed the family-size specific limits given in table 4.A2. The value of the stamps is the maximum monthly allotment less 30 percent of net income. The 30 percent figure reflects the expectation that recipient households will spend about 30 percent of their resources on food.
Source 1. Mass Resources. Food Stamps Program. http://www.massresources.org/ pages.cfm?contentID=12&pageID=3&Subpages=yes
Medicare Medicare is a federal health insurance program for the aged and disabled (we ignore disability benefits and focus on the benefits for the aged only). It incorporates two parts: Hospital Insurance (HI), also known as “Part A,” and
Kotlikoff and Rapson
138
Supplementary Medical insurance (SMI), also known as “Part B.” Hospital Insurance is generally provided automatically to individuals aged 65 and over who are entitled to Social Security benefits. Part A helps pay for care in hospitals, skilled nursing facilities, hospice and some home health care. Enrolling in SMI is optional; part B helps pay for: doctors, outpatient hospital care, clinical laboratory tests, durable medical equipment, most supplies and some other services not covered by Part A. Medicare Part A is primarily financed through a mandatory 2.9 percent payroll tax. Part B is financed in part by participant premium payments of $78.20 per month regardless of benefits received. In addition, there are specific costsharing arrangements. In particular, under Part A in each benefit period a recipient of benefits pays: $776 for a hospital stay of 1–60 days; an additional $194 per day for days 61–90; an additional $338 per day for days 91–150; and all costs for each day beyond 150 days. We assume that at age 65 both husband and wife enroll in both Part A and Part B. It is typical for individual to enroll in both plans. We assumed that in each year an individual, if s/he receives benefits, stays in the hospital less than 60 days and so pays the fixed fee of $776. Under Part B, participants receiving benefits must first meet an annual $110 deductible and, in most cases, cover 20 percent of the approved amount after the deductible. In our calculations, we impute to each age-eligible spouse at a particular age their expected net Medicare benefits at that age. Any actual out-of-pocket cost sharing and premium payments were deducted from the gross income in calculations of the Food Stamps benefits for eligible individuals. Our data on Medicare benefits for aged come from the Dartmouth Atlas of Healthcare Database.21 This database provides average Medicare benefits under Part A and under Part B classified by age and sex in 2003. We found that, in the recent past, average benefits per person enrolled were 26 percent and 5 percent greater, respectively, under Plan A and Plan B, in Massachusetts compared to the national averages. We incorporated that adjustment for all age cohorts and both sexes. We converted all 2003 amounts to 2005 dollars using CPI for medical expenditures, provided by the Bureau of Labor Statistics (see table 4.A3). Table 4.A3 Medicare Reimbursement per Eligible Enrollee (2005) Part A
Part B
Age
Men
Women
Men
Women
65–69
2,987
2,504
2,104
2,218
70–74
3,923
3,368
2,731
2,640
75–79
5,005
4,376
3,249
2,912
80–84
6,004
5,274
3,498
2,877
85+
7,072
6,400
3,413
2,581
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
139
Sources 1. Dartmouth Atlas of Healthcare Database (September 2005). 2. Centers for Medicare and Medicaid Services. Internet: http://www.cms. hhs.gov/
Medicaid Medicaid is a joint federal-state program that provides medical care to the poor. In 2002 Medicaid recipients constituted 17 percent of the U.S. population. Over 50 percent of all Medicaid income-eligible infants, children, and adults had no access to any other form of private or public health insurance. However, not all eligible individuals apply for Medicaid. For purposes of this study we assume that our households, when eligible, do apply and receive all Medicaid benefits to which they are entitled. Medicaid covers most, but not all, medically necessary medical care and services provided to eligible individuals. Each state establishes its eligibility standards and general rules. The policies are complex and vary considerably from state to state. In Massachusetts, Medicaid is officially known as MassHealth. In addition to serving the poor in general, MassHealth incorporates special programs to assist poor pregnant women and children, the disabled, and immigrants who are in need of emergency care. MassHealth provides the following services: –Inpatient hospital services. –Outpatient services: hospitals, clinics, doctors, dentists (limited dental coverage for adults), family planning, and home-health care. –Medical services: lab tests, X rays, therapies, pharmacy services, dental services, eyeglasses, hearing aids, medical equipment and supplies, adult day health, and adult foster care. –Mental health and substance abuse services: inpatient and outpatient. –Living in nursing homes. –Payment of the Medicare premium, coinsurance, and deductibles for certain groups of elderly. Like Medicare, Medicaid operates as a vendor payment program; recipients receive benefits directly in the form of medical services provided by qualified vendors. Benefits are provided as long as the individual meets general and financial eligibility criteria. Financial eligibility criteria include income eligibility requirements, which may be different for different family members, and assets eligibility requirements. MassHealth Standard Program specifies that the family monthly income before taxes and deductions cannot exceed: –200 percent of the FPL (Federal Poverty Level) for pregnant women and infants. –150 percent of the FPL for children under age 19. –133 percent of the FPL for parents with children under age 19.
Kotlikoff and Rapson
140
Under MassHealth the income limit for an eligible individual (couple) aged 65 and over is 100 percent of the FPL. In addition, in Massachusetts if an individual is eligible for SSI, s/he would also be eligible for Medicaid. Table 4.A4 presents the respective monthly income limits. Medicaid eligibility may be extended to individuals with incomes greater than the above income limits if they are deemed “medically needy.” States provide residual financing of such individuals’ medical treatment costs, provided they spend their excess resources (income and assets) down to the eligibility limits. This is particularly the case for individuals moving into nursing homes with insufficient resources to fully finance their stays. For simplicity, we do not consider coverage of the medical needy in this analysis. In each year we determine for each family member of a particular age and sex if s/he meets appropriate income standards of eligibility and then allocate to that individual the Medicaid age- and sex-specific benefit projected to prevail in that year. Fortunately, statistics on Medicaid eligibles, recipients, and total vendor payments are available by sex and age. When the beneficiary in our stylized case is a child under 19, we ignore gender difference in benefits. If a person over age 65 is eligible for Medicaid, his/her Medicare cost-sharing will be partially or fully financed by Medicaid. There are two broad groups of dual-eligibles: those for whom Medicaid pays only Medicare part B premiums (so-called, SLMB eligibles), and those who get extensive coverage from Medicaid (see the discussion on Medicaid-Medicare interactions below). Our calculated average benefit values for aged eligibles reflect Medicaid payments made for both these groups. However, we impute full Medicaid benefits only to the elderly with incomes less than 100 percent of the federal poverty line; and we treat SLMB eligibles separately. Specifically, for those over 65, who are eligible for the full coverage, we adjust the average Medicaid benefits by excluding payments for SLMB eligibles, using data on the fraction (4.6 percent) of those receiving benefits from both Medicare and Medicaid who are SLMB recipients, the size of the SLMB Medicaid benefit (equal to the annual Part B premium), and the overall average Medicaid benefit net of Nursing Home financing. Our final calculated adjusted age- and sex-specific Medicaid benefits for 2005 are presented in table 4.A5. We used the BLS index of medical expenditure growth to measure 2002 benefit levels in 2005 dollars. Table 4.A4 Federal Poverty Lines (2005) Household Size
100%
133%
150%
200%
1
798
1,061
1,196
1,595
2
1,069
1,422
1,604
2,138
3
1,341
1,783
2,011
2,682
4
1,613
2,145
2,419
3,225
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
141
Table 4.A5 Estimated 2005 Medicaid Benefits in Massachusetts, Net of SLMB Program Financing Average Net Benefit per Eligible Age
Female
Male
All Ages
$6,145
$5,711
Under 1
$3,468
$3,747
1–5
$1,839
$2,148
6–12
$1,660
$2,094
13–14
$2,134
$2,777
15–18
$2,807
$3,018
19–20
$2,814
$2,590
21–44
$4,503
$6,653
45–64
$10,216
$11,424
65–74
$9,353
$11,021
75–84
$15,914
$15,300
85 and over
$26,960
$23,243
In each year we determine for each family member of a particular age and sex if s/he meets appropriate income standards for eligibility and then allocate to that individual the Medicaid age- and sex-specific benefit projected to prevail in that year. When the beneficiary in our stylized case is a child under 19, we ignore gender difference in benefits.
Sources 1. 2005 Health and Human Services poverty guidelines. Internet: http://aspe. hhs.gov/poverty/05poverty.shtml. 2. MassHealth. Internet: www.mass.gov. 3. Medicaid. Center for Medicare and Medicaid Services. Internet: http://www. cms.hhs.gov.
Supplementary Security Income (SSI) Supplementary Security Income is a federal program that makes monthly payments to people who have limited income and resources if they are 65 or older or are disabled. In our study we ignore payments to the disabled. If individuals meet the program’s income limits, after deductions, they receive monthly benefits. Payments up to the Federal income limits are paid by the federal government, while states provide supplements that are calculated as the difference between state and federal income limits. Standard deductions are $20 per month plus the sum of a) an additional $65 per month if labor income exceeds $65 per month and b) one-half of wages over $65. In Massachusetts, an SSI-eligible person is automatically enrolled in Medicaid. See table 4.A6.
Kotlikoff and Rapson
142
Table 4.A6 Household Size
Income Limit
1
708
2
1,071
For every year we first determine age eligibility for each spouse, and then income eligibility for the household. When both spouses are eligible, their combined benefit equals the difference between the income limit for a two-person household and the spouses’ combined income after deductions. When only one spouse is age eligible, the eligible spouse’s benefit is calculated according to the regulations using either an individual- or couple-income limit depending on the level of the income of the ineligible spouse.
Source 1. Mass Resources. Supplemental Security Income (SSI). http://www.massresources.org/pages.cfm?contentID=18&pageID=4%20&Subpages=yes.
Low-Income Home Energy Assistance Program (LIHEAP) LIHEAP is a block-grant program of the Federal Government that allocates funds between states to operate various home energy assistance programs for needy households. The funds may be used for the purposes of home heating and cooling assistance, energy-crisis intervention, and low-cost weatherization or other energy-related home repairs. LIHEAP assists eligible low-income households in meeting the heating or cooling portion of their residential energy needs. Low-income households are defined as households with incomes that cannot exceed the greater of 150 percent of the poverty level or 60 percent of state median income ($31,952, $39,469, and $46,987 for 2-, 3-, and 4- person families respectively in Massachusetts in 2005). The states have flexibility in setting their income eligibility at or below this maximum standard. LIHEAP payments can be made to households where one or more persons are receiving Supplemental Security Income (SSI), Aid to Families with Dependent Children (AFDC/TANF), or food stamps. Priority may be granted to those households with the greatest energy cost in relation to income, taking into consideration the presence of children and elderly. In Massachusetts in 2004, 134 thousand households received LIHEAP benefits. However, this represents only 15.5 percent of LIHEAP-eligible households. As such, while the average benefit per recipient is $480, the amount received per eligible household is only a fraction thereof. In our calculations, we assume that each eligible household received 15.5 percent of the maximum possible LIHEAP benefit according to their income test relative to the poverty line.
Measuring Marginal Taxes on Americans’ Labor Supply and Saving
143
Sources 1. Massachusetts Department of Housing and Community Development. Internet: http://www.mass.gov/dhcd/components/cs/1PrgApps/LIHEAP/ chart.pdf. 2. Massachusetts Department of Housing and Community Development. Internet: http://www.mass.gov/dhcd/components/cs/Fuel/default.htm# income%20chart. 3. U.S. Department of Health and Human Services. Internet: http://www. liheap.ncat.org/tables/FY2004/heatbenefit04.htm.
Special Supplemental Nutrition Program for Women, Infants and Children (WIC) WIC is a program designed to improve the health of pregnant women, new mothers and their infants. WIC targets population groups that have low income and are at risk nutritionally, specifically: –pregnant women through pregnancy and up to 6 weeks after birth or after pregnancy ends. –breastfeeding women through their infant’s first birthday. –infants through their first birthday. –children up to age 5. WIC benefits include: supplemental nutrition, nutrition counseling and screening services. In most WIC State agencies, WIC participants receive either actual food items or food vouchers to purchase specific foods to supplement their diets. Different food packages are provided for different categories of participants. Although federally funded, WIC is administrated by state agencies and managed by local agencies. The WIC Program has certain eligibility requirements that are based on income and nutritional risk. In order to qualify, WIC applicants must show medically verified evidence of health or nutrition risk. In addition, their family income generally must be below 185 percent of the federal poverty level (FPL). Certain applicants can be judged income-eligible for WIC based on their participation in Food Stamps, Medicaid and AFDC/TANF programs. WIC does not serve all eligible individuals—participation is limited by the availability of Federal funding. Usually, program applicants are ranked by need. The estimated 2004 average monthly benefit for WIC recipients (be they women, infants, or children) in Massachusetts is $33.80. For our calculations, we assume that all eligible households receive this average benefit times the probability of receipt, which was 81 percent in 2004. The average monthly benefit of the $33.80 multiplied by 0.81 is $27.38, which implies probability-adjusted annual benefits of $328.52 to all of our eligible households.
Sources 1. WIC Program. Food And Nutrition Service. Internet: http://www.fns.usda. gov/wic/ and http://www.fns.usda.gov/pd/wisummary.htm. 2. Massachusetts state government. Internet: http://www.mass.gov.
5 Federal Tax Policy towards Energy Gilbert E. Metcalf, Tufts University and NBER
Executive Summary On Aug. 8, 2005, President Bush signed the Energy Policy Act of 2005 (PL 109–58). This was the first major piece of energy legislation enacted since 1992 following five years of Congressional efforts to pass energy legislation. Among other things, the law contains tax incentives worth over $14 billion between 2005 and 2015. These incentives represent both pre-existing initiatives that the law extends as well as new initiatives. In this paper I survey federal tax energy policy focusing both on programs that affect energy supply and demand. I briefly discuss the distributional and incentive impacts of many of these incentives. In particular, I make a rough calculation of the impact of tax incentives for domestic oil production on world oil supply and prices and find that the incentives for domestic production have negligible impact on world supply or prices despite the United States being the third largest oil producing country in the world. Finally, I present results from a model of electricity pricing to assess the impact of the federal tax incentives directed at electricity generation. I find that nuclear power and renewable electricity sources benefit substantially from accelerated depreciation and that the production and investment tax credits make clean coal technologies cost competitive with pulverized coal and wind and biomass cost competitive with natural gas. 1.
Introduction
On Aug. 8, 2005, President Bush signed the Energy Policy Act of 2005 (PL 109-58). This was the first major piece of energy legislation since 1992 and culminated five years of efforts to pass energy legislation by
146
Metcalf
the Bush Administration. Among other things, the law contains tax incentives worth over $14 billion between 2005 and 2015. These incentives represent both pre-existing initiatives that the law extends as well as new initiatives. In this paper I review federal tax energy policy focusing both on programs that affect energy supply and demand. In the next section, I discuss an economic rationale for energy tax incentives. Next, I review current energy taxes in the third section. In the following section, I summarize the various energy incentives in the tax code.1 These include accelerated depreciation of various types as well as production and investment tax credits. In addition, special incentives are targeted towards electric utilities and the transportation sector. In the fifth section, I briefly discuss the distributional and incentive impacts of many of these tax incentives. I also conduct a levelized cost analysis of various electricity generation technologies to assess the impact of the production and investment tax incentives directed at electricity generation. In summary, the energy taxes or tax incentives currently in effect are difficult to justify on the basis of economic theory. Energy taxes totaled $36.1 billion in fiscal year 2004 with the vast bulk of the revenues coming from motor vehicle fuel taxes. The most pressing case for taxation— externalities—suggests direct pollution or driving charges rather than a gasoline tax. The other motor vehicle related tax, the Gas Guzzler Tax, suffers from the defect of excluding light trucks and Sport Utility Vehicles (SUVs) from the tax. These make up the majority of motor vehicles currently sold. With the passage of the Energy Policy Act of 2005 (EPACT), energy tax preferences are worth roughly $6.7 billion in fiscal year 2006. The production and investment tax credits have been effective at making certain renewable energy sources (mainly wind and biomass) competitive with natural gas in electricity generation. I note, however, that tax credits are a socially costly way of making these renewable sources competitive with fossil fuel sources. Finally, while fossil fuel and nuclear power continue to receive the majority of benefits from tax incentives, the tilt towards these fuels is not as large as it once was. Percentage depletion and expensing of intangible drilling costs, for example, have been scaled back relative to the situation in the 1950s and 1960s. And the investment tax credits for solar generated electricity combined with generous depreciation tax treatment contribute to negative effective tax rates on solar generated electricity.
Federal Tax Policy towards Energy
2.
147
Rationale for Government Energy Tax Incentives
Why should the federal government have an energy policy? More particularly, why should the tax code be used as an instrument of an energy policy? To help evaluate the various provisions of the tax code that affect energy supply and demand, I briefly review four major arguments for government intervention in energy markets: energy externalities, national security, market failures and barriers in energy conservation markets, and rent expropriation. For a more in-depth review, see Newbery (2005) or Lazzari (2005). A broad array of externalities are associated with our consumption of energy. Burning fossil fuels contributes to air pollution (sulfur dioxides, nitrogen oxides, particulates) and generates greenhouse gases. In addition, our use of petroleum in transportation contributes to roadway congestion, accident externalities, and other traffic related market failures (see Parry and Small 2005 for a fuller discussion of driving related externalities). Economic theory suggests that we should tax externalities directly rather than a proxy for the externality (here, motor vehicle fuels). Road congestion suggests the use of congestion or time-of-day pricing on highways. Tailpipe emissions from vehicles call for emissions pricing if technologically feasible.2 Accident externalities call for changes in automobile insurance pricing. None of these externalities suggest a policy of taxing motor vehicle fuels directly. The one externality that might suggest a motor fuels tax is global warming arising from burning fossil fuels given the tight relationship between petroleum consumed and carbon emitted.3 But even here a stronger case could be made for a comprehensive tax on the carbon content of all energy sources rather than a specific tax on motor vehicle fuels.4 Positive spill-overs from research and development are frequently cited as an argument for tax incentives for particular technologies. Supporters of production tax credits for renewable fuels, for example, argue that experience in the marketplace and learning by doing will bring about cost savings that support the initial subsidies. The difficulty with such an argument, of course, is that all research and development spending has elements of non-appropriability leading to a policy prescription of support for general R&D rather than sector or technology specific R&D support. A second broad rationale for government intervention in energy markets is national security concerns. Here the argument is that our dependence on imported energy, oil in particular, makes us vulnerable
148
Metcalf
to economic coercion from foreign owners of energy resources. In 2004, the United States imported over 60 percent of the 20.5 million barrels per day of petroleum that it consumed (Energy Information Administration 2005). The need to protect a stable source of energy imports, it is argued, requires increased spending on defense and national security and has made the country more vulnerable to unstable governments in the Middle East and other oil rich regions. Oil import tariffs are a proposed solution to this problem. By reducing our dependence on foreign oil, it is argued, the United States reduces its vulnerability to political and economic instability elsewhere. The difficulty with this argument is that oil is a commodity priced on world markets. Even if the United States were to produce all the oil it consumes, it would still be vulnerable to oil price fluctuations. A supply reduction in the Middle East would raise prices of domestic oil just as readily as it raises prices of imported oil.5 A third argument for government intervention in energy markets is the existence of market barriers to energy efficient capital investment. A long-standing “energy paradox” claims that consumers need very high rates of return on energy efficient capital (appliances, housing improvements, lighting, etc.) and a variety of market barriers have been proposed to explain this paradox and to motivate market interventions. I critique the market barriers literature elsewhere (Metcalf 2006) and simply note two relevant issues here that support possible market interventions. First, many have argued that consumers are poorly informed about the potential for energy savings (as well as the value of the savings) associated with new more expensive technologies. This is a reasonable point given the public good nature of information acquisition and suggests the value of government information programs. Programs such as energy efficiency labeling on new appliances can help overcome information failures at low cost. Second, principal-agent problems may deter energy efficient investments. A good example is the provision of energy efficient appliances and housing in rental housing. Tenants may desire more energy efficient housing and appliances but landlords may be reluctant to make the investments out of concern that they may be unable to recoup their incremental investment through higher rents. In addition, many apartment buildings are not easily converted to allow for tenant control over and payment for energy consumption (especially heating services) in individual units. This removes incentives for tenants to conserve energy. Landlords, meanwhile, may be reluctant to invest in energy conservation capital if the effectiveness of the invest-
Federal Tax Policy towards Energy
149
ment depends on tenant use characteristics (installing additional insulation is likely not cost-effective if tenants open windows during the winter when apartments become overheated). The appropriate policy response in this situation is to provide a subsidy to tenants (or landlords) for investments in energy conservation capital. Finally, a number of authors (Newbery 1976; Bergstrom 1982; Karp and Newbery 1991) have noted that an oil import tariff can expropriate some portion of the Hotelling rents associated with oil. The intuition is straightforward if all consuming countries could act in collusion. Since potential oil supply from known oil fields is fixed, a tax that doesn’t alter the relative scarcity rents of oil over time will not affect the time profile of extraction. Thus, we can collapse the analysis to that of a tax on an inelastically supplied product. Since the entire burden of such a tax is on the supplier, the result follows.6 Newbery (2005) estimates that the optimal oil import tariff for the EU and the United States ranges between $3.10 and $15.60 per barrel in 2002. Summing up, we shall see that the arguments for using the tax code to affect energy supply and demand are poorly related to existing energy tax policy. The most compelling case can be made for energy taxes related to carbon emissions and for an oil import fee to transfer some of the Hotelling rents from oil suppliers to the United States. I turn next to a discussion of current energy tax provisions at the federal level. 3.
Federal Energy Taxes
Table 5.1 lists federal taxes that are specifically linked to energy production or consumption. By far the largest are the excise taxes on gasoline and diesel fuels that are dedicated to the Highway Trust Fund accounting for over 95 percent of federal energy excise tax collections in FY 2004. The federal excise tax rate on gasoline is 18.4¢ per gallon. Of that, 0.1¢ is dedicated to the Leaking Underground Storage Tank Trust Fund and the remaining 18.3¢ to the Highway Trust Fund.7 Of that 18.3¢ per gallon, 2.86¢ is dedicated to the Mass Transit Account and the remaining 15.44¢ to the Highway Account. In fiscal year 2004, this tax raised $35.1 billion. In comparison, total outlays for grants to state and local governments from the highway and urban mass transit programs in fiscal year 2004 were $30.0 billion. Non-trust fund aid to sub-federal governments for highways and urban mass transit totaled an additional $7.8 billion with nearly all of that designated to urban mass transit.
Metcalf
150
Table 5.1 Federal Energy Excise Taxes
Tax
Tax Rate
Revenue Projection for FY2004 ($millions)
Highway trust fund revenues
18.3¢ per gallon of gasoline*
$34,711
Gas guzzler tax
$1,000–$7,700 per vehicle depending on mileage
Oil spill liability trust fund
5¢ per barrel
Leaking underground storage tank tax
0.1¢ per gallon of motor fuels
189
Coal excise tax
Lower of 4.4 percent of sale price and $1.10 per ton ($.55 per ton for surface mined coal)
566
Aquatic resources trust fund tax on motorboat gasoline and other fuels
Motorboat gasoline proceeds from highway trust fund revenues
416
Inland waterway fuels tax
$.224 per gallon for commercial vessels Total
141 —
91 $36,114
*Diesel fuel is taxed at the federal level at 24.3¢ per gallon. State excise taxes on gasoline and diesel fuel averaged 18.1¢ as of April 2006. According to the American Petroleum Institute (2006), taking into account all taxes on gasoline (diesel) including the Leaking Underground Storage Tank Tax, the average tax rate is 46.5¢ (53.02¢) per gallon. Source: Budget of the United States, Historic Tables, Table 2.4. Gas Guzzler tax revenue from SOI Historic Tables, Table 21.
Because the federal motor fuels gas tax is an excise tax, its ad valorem equivalent rate fluctuates with gas prices. Figure 5.1 shows how the rate has changed between 1978 and 2005. With the decline of gasoline prices in the late 1980s, the tax peaked at 27 percent of the refiner price of finished motor gasoline to end users in 1998 and has fallen since then to 10 percent in 2005. The United States has the lowest tax rate on unleaded gasoline among all the OECD countries (see figure 5.2). Its tax rate per litre ($.104) in the fourth quarter of 2005 was less than half that of the next closest country and compares to an OECD average rate of $.789 per litre. Returning to the importance of motor fuels taxes in total energy tax collections, consider the United Kingdom. Its tax on gasoline is the third highest among OECD countries (see figure 5.2). Yet the UK tax on hydrocarbons is only 90 percent of its total energy tax collections.8
Federal Tax Policy towards Energy
151
Figure 5.1 Federal Excise Tax Rate as Share of Before-Tax Refiner Price of Gasoline Sources: EIA Monthly Energy Review and Jackson (2006).
The Gas Guzzler Tax was enacted as part of the Energy Tax Act of 1978. It levies a tax on automobiles that obtain fuel mileage below 22.5 miles per gallon.9 Tax rates range from $1,000 to $7,700 per vehicle. In 2004 the tax collected $141 million. The gas guzzler tax explicitly excludes sport utility vehicles, minivans, and pickup trucks, which represent 54 percent of the new vehicle sales in 2004 (U.S. Census Bureau 2006, table 1027). The light truck category (comprising SUVs, minivans, and pickup trucks) is the fastest growing segment of the new vehicle market, growing at an annual rate of 5.5 percent between 1990 and 2004. In contrast, new car sales are falling at an annual rate of 1.6 percent. Unofficial Congressional estimates suggest that phasing out the SUV loophole over four years would raise roughly $700 million annually once the phase-out was complete. The Energy Policy Act of 2005 (EPACT) resurrected the Oil Spill Liability Trust Fund tax at the original rate of 5¢ per barrel. This tax had
152
Sources: IEA, Energy Prices and Taxes, Fourth Quarter 2005.
Metcalf
Figure 5.2 Tax Rate on Unleaded Gasoline (Fourth Quarter 2005)
Federal Tax Policy towards Energy
153
previously been in effect from 1990 through 1994. The Joint Committee on Taxation estimates that this tax will raise $1.25 billion between 2005 and 2010. The tax is imposed on crude oil received at U.S. refineries as well as imported petroleum products. Domestic crude oil for export is also subject to the tax if the tax had not been previously paid. The coal excise tax funds the Black Lung Disability Fund. It is levied on coal mined in the United States at a rate of 4.4 percent of the sales price up to a limit of $1.10 per ton of underground coal and $.55 per ton of surface mined coal. This tax raised $566 million in 2004. Gasoline sold for sport motorboats is taxed at the same rate as highway gasoline and diesel fuel and the funds allocated to the Aquatic Resources Trust Fund (subject to an annual cap on transfers that effectively reduces the share of tax on motorboat fuels shifted to this trust fund). Finally the Inland Waterways Fuels Tax levies a tax of 22.4¢ per gallon of fuel sold to commercial vessels using the Inland Waterway System (barges for the most part). 4.
Energy Incentives in the Corporate and Personal Income Tax
The President’s FY2007 Budget Submission lists over $20 billion of tax expenditures (for the fiscal years 2007–2011) associated with energy.10 Table 5.2 lists these tax expenditures sorted from highest to lowest cost (over the five year budget window). Both the number of tax provisions and the revenue cost have increased as a result of 2005’s energy legislation. The General Accounting Office (2005) listed nine income tax preferences totaling $4.2 billion in revenue loss in fiscal year 2003.11 In fiscal year 2006, the Administration’s budget lists 29 preferences totaling $6.7 billion for that year. The single largest tax expenditure is associated with alcohol fuels. Most of this revenue loss arises from the reduction in motor vehicle fuel tax revenues ($12,500 million) with the remainder arising from the $.51 per gallon income tax credit for this fuel.12 After alcohol fuels is the tax expenditure for investment and production tax credits for new energy technologies. Investment tax credits range from 20 to 30 percent depending on the technology and production tax credits exist, primarily for electricity produced from renewable energy sources. Before turning to a discussion of the current code, it may be useful to provide some historical perspective on the federal tax treatment of energy.13 Prior to the first oil embargo in 1973 the federal government’s tax policy was designed to encourage fossil fuel exploration and
Metcalf
154
Table 5.2 Energy Tax Expenditures
Tax Provision
Revenue Cost: FY 2007–2011 ($millions)
Alcohol fuel credits
12,730
New technology credit (secs. 45 & 48)
4,060
Alternative fuel production credit (sec. 29)
3,450
Expensing of exploration and development costs
3,230
Excess of percentage over cost depletion
3,230
Temporary 50% expensing for equipment used in the refining of liquid fuels
830
Credit for investment in clean coal facilities
780
Amortize all geological and geophysical expenditures over 2 years
630
Natural gas distribution pipelines treated as 15–year property
560
Credit for energy efficiency improvements to existing homes
530
Exclusion of interest on energy facility bonds
510
Tax credit and deduction for clean-fuel burning vehicles
420
Capital gains treatment of royalties on coal
400
Exclusion of utility conservation subsidies
380
Allowance of deduction for certain energy efficient commercial building property
340
Exception from passive loss limitation for working interests in oil and gas properties
200
Credit for holding clean renewable energy bonds
180
Credit for business installation of qualified fuel cells and stationary microturbine power plants
150
Credit for energy efficient appliances
80
Credit for construction of new energy efficient homes
40
Enhanced oil recovery credit
20
30% credit for residential purchases/installations of solar and fuel cells
20
Pass through low sulfur diesel expensing to cooperative owners Deferral of gain from dispositions of transmission property to implement FERC restructuring policy
–30 –210
Credit for production from advanced nuclear power facilities
—
Alternative fuel and fuel mixture tax credit
—
Source: FY2007 Budget Submission of the President, Analytical Perspectives.
Federal Tax Policy towards Energy
155
production. Expensing of intangible drilling costs was introduced in 1916 and percentage depletion in 1926. Percentage depletion for oil and gas was particularly generous with a rate of 27.5 percent (relative to the current rate of 15 percent) for oil and gas and was available to all companies, not simply independent producers as at present. During the 1970s the sharp increase in the price of oil combined with growing environmental concerns associated with oil and gas drilling as well as a rising federal budget deficit led to a curtailment of the preferential treatment for fossil fuels. The percentage depletion rate, for example, was reduced to 15 percent for oil and gas and restricted to independent producers (i.e., producers without refining or retailing operations). The Energy Tax Act of 1978 introduced the Gas Guzzler Tax, tax subsidies for gasohol, and investment tax credits for conservation and renewable energy production. This was followed by the Windfall Profits Tax which, in addition to its efforts to tax profits on old oil, enacted the section 29 production tax credits for non-conventional oil. The election of Ronald Reagan in 1980 ushered in a new era in the federal government’s tax treatment of energy. According to Lazzari (2006), Reagan brought a free-market approach to energy policy. As such, he worked to eliminate the Windfall Profits Tax (largely repealed in 1988) and to end federal tax credits for energy production or investment. By 1988 all that remained of the federal tax credits for energy were the section 48 investment tax credits for solar and geothermal power. The post-Reagan era saw a number of changes to the tax code with the most significant being the Energy Policy Act of 1992 (PL 102–486). This law enacted the section 45 production tax credits for wind and closed loop biomass generated electricity. As discussed below, this credit was gradually expanded to cover other renewable sources and remains in effect today.14 Other laws passed during the post-Reagan era generally extended existing production and investment tax credits and raised the gasoline tax.15 The most recent change is the Energy Policy Act of 2005, which extended and expanded coverage of the section 45 production and section 48 investment tax credits among a variety of other provisions. I discuss the current tax code’s various provisions in the next section.16 4.1 Depreciation Depreciation is the wearing out of an asset over time and is properly recognized as a cost of doing business. Economic depreciation refers to the actual wearing out of the asset as reflected in changes in the asset’s
156
Metcalf
value over time. A pure income tax would allow a deduction for economic depreciation. Because of the difficulties involved in measuring economic depreciation, the tax code groups assets into broad categories and mandates depreciation schedules for assets in each category. Tax depreciation may bear some resemblance to economic depreciation but it should be stressed that tax depreciation is a policy tool that may be used to encourage or discourage certain types of investment at the expense of other types. Accelerated depreciation refers to a depreciation schedule that allows for more rapid tax depreciation than economic depreciation. The limit of accelerated depreciation is expensing, an immediate deduction for the entire cost of the asset. Expensing an asset reduces the effective tax rate on this asset to zero. To see this consider an asset worth $100 that generates additional net profits of $20 per year. In the absence of taxation, this asset produces a net return of 20 percent. Now impose a 35 percent tax with expensing. In the first year, the firm takes a deduction for the cost of the machine and enjoys a reduction in taxes of $35 (35 percent times $100). The aftertax cost of the machine has been reduced to $65. In subsequent years, the firm obtains additional after-tax profits of $13. The net return on this investment is still 20 percent ($13/$65). Under the current tax code, capital assets are depreciated according to the Modified Accelerated Cost Recovery System (MACRS) with recovery periods ranging from 3 to 39 years under the General Depreciation System (GDS).17 Most capital is depreciated using a declining balance method at either 200 percent (3, 5, 7, and 10 year property) or 150 percent (15 and 20 year property). Table 5.3 shows the recovery period for various types of energy related capital. Most electric generating capital is depreciated over 20 years with the major exception being nuclear power plants (15 years) and renewable energy generating capital (five years). High voltage electricity transmission lines received a 15 year recovery period in the Energy Policy Act of 2005. That act also clarified the depreciation of natural gas gathering pipelines (seven years) and reduced the recovery period of distribution pipelines from 20 to 15 years. In addition, the new law also contains a provision allowing partial expensing for new refinery capacity placed in service before 2012. The provision allows for 50 percent expensing with the remainder deducted as under current law. Below, I provide some analysis of the impact of accelerated depreciation (as well as other tax provisions) on the cost of capital for various types of electricity generating property and show that nuclear power
Federal Tax Policy towards Energy
157
Table 5.3 Recovery Periods for Energy Capital
Property
Recovery Period (Years)
Electric utility generation and distribution property
20
Electric transmission property (below 69 kV)
20
69 kV and higher electric transmission property
15
Electric utility nuclear power generator
15
Industrial electric generation
15
Liquefied natural gas plant
15
Natural gas distribution pipelines
15
Pipeline transportation (including storage of integrated producers)
15
Coal gasification production property
10
Refineries
10
Natural gas gathering pipelines
7
Natural gas production property
7
Electric utility nuclear fuel assemblies
5
Oil and gas drilling rigs
5
Section 48 alternative energy property
5
Source: U.S. Internal Revenue Service (2006).
and electricity generated from renewable sources receive particularly generous tax treatment from accelerated depreciation. Oil drilling receives an additional depreciation benefit from the ability to expense dry holes. One can view dry holes as part of the cost of drilling a successful well. This tax provision raises the effective value of the depreciation deductions for oil rigs. Technology, however, has reduced the percentage of dry holes. In 1960, 40 percent of all wells drilled were dry holes. By 2004, that percentage had fallen to 12 percent reducing the tax advantage of dry hole expensing.18 While not energy capital explicitly, motor vehicles have a significant impact on energy consumption and depreciation rules treat different types of motor vehicles very differently. Clean fuel vehicles may be expensed up to limits (ranging from $50,000 for trucks or vans with gross vehicle weight exceeding 26,000 pounds to $2,000 for motor vehicles weighing less than 10,000 pounds). Clean fuel vehicles include vehicles that burn natural gas, LNG, LPG, hydrogen, electricity, gasohol (if at least 85 percent alcohol) and certain hybrid electric vehicles.
Metcalf
158
Passenger cars, SUVs and pickup trucks used in a small business can have very different depreciation treatment. Small businesses are allowed a section 179 deduction of up to $100,000 per year in capital expenses (subject to phase-out rules). The section 179 deduction is limited for motor vehicles in certain ways. First, passenger vehicles and light trucks with a gross vehicle weight of 6,000 pounds or less are treated as listed property and subject to annual depreciation deduction limits arising from luxury passenger car rules written in the Deficit Reduction Act of 1984 (PL 98–369). These vehicles must be depreciated over a five year period with specified annual depreciation caps.19 The luxury vehicle limits are such that any passenger car costing more than $13,860 and any light truck costing more than $15,360 will have a recovery period longer than the standard five year recovery period for motor vehicles. Second, small businesses purchasing SUVs weighing more than 6,000 pounds (but not more than 14,000 pounds) can expense $25,000 and depreciate the balance over five years using double-declining balance rules. Table 5.4 illustrates how the various depreciation rules affect the after-tax price for a small business owner in the top personal income tax bracket choosing among three new 2005 vehicles each costing $30,000. The passenger car must be written off over 21 years as opposed to 19 years for the light SUV and six years for the heavy SUV. The differences in depreciation treatment raise the price of the passenger car and light SUV by 13 and 10 percent respectively relative to the heavy SUV. 4.2 Tax Treatment Specific to Fossil Fuel Production Traditionally, fossil fuels have received preferential tax treatment. Percentage depletion and the expensing of intangible drilling costs are the Table 5.4 Depreciation Treatment for Motor Vehicles Toyota Avalon Gross vehicle weight rating Years to depreciate
3,490
Ford Escape 5,520
Ford Expedition 7,300
21
19
6
6,632
7,135
9,308
After-tax vehicle cost
23,368
22,865
20,692
Percentage mark-up over heavy SUV
13%
10%
—
PV of tax shield
Curb weight reported for Toyota Avalon. Table assumes purchase price of $30,000 for all vehicles, tax rate of 35 percent and a discount rate of 10 percent.
Federal Tax Policy towards Energy
159
most well known. While not as generously treated as in the past, the tax preferences for fossil fuel production are still important. As natural resources are extracted from booked reserves, the value of those reserves is diminished. This is a legitimate cost of business and a Haig-Simons income tax would allow a deduction for the value of the resource extracted. Rather than take deductions for the value of the extracted resource, oil, gas, and coal producers have historically been allowed a deduction based on percentage depletion. Percentage depletion allows the firm to deduct a fraction of the revenue arising from sale of the resource. Historic percentage depletion rates have been as high as 27.5 percent. Currently percentage depletion is allowed for independent producers at a 15 percent rate for oil and gas and 10 percent for coal.20 Percentage depletion is allowed on production up to 1,000 barrels of average daily production of oil (or its equivalent for natural gas). In addition, the depletion allowance cannot exceed 100 percent of taxable income from the property (50 percent for coal) and 65 percent of taxable income from all sources.21 Despite the curtailed availability of percentage depletion, it continues to be a significant energy tax expenditure, costing $3.2 billion over five years in the federal budget (see table 5.2). The following example illustrates the tax benefits of percentage depletion over cost depletion. John Doe purchases an interest in a property for $300,000 that contains reserves of 50,000 barrels of oil. He produces 10,000 barrels of oil in the first year which he sells for $630,000. Under cost depletion, he would be allowed to deduct $60,000 for the reduction in reserves ⎛ 10, 000 ⎞ ⎜⎝ 50, 000 × $300, 000⎟⎠ . Percentage depletion allows him to deduct $94,500 (.15 × $630,000). Note that percentage depletion can exceed the basis in the asset. Continuing to assume a first purchase price of $63 per barrel, the total value of oil extracted would be $3.15 million and the percentage depletion deduction would be $472,500.22 Obviously the benefit of percentage depletion would be considerably higher at the historic depletion rate of 27.5 percent. The second major tax benefit available to oil and gas producers is the ability to expense intangible drilling expenses (labor and material costs associated with drilling wells). Normally the non-capital expenses associated with oil exploration and drilling would be capitalized and the costs allocated as income is earned from the well over its useful
160
Metcalf
life. Instead firms may deduct these expenses in the first year. Corporations may only deduct 70 percent of the costs and must depreciate the remaining 30 percent over five years. Additionally, geological and geophysical costs associated with exploration can be amortized over a two year period.23 As noted in table 5.2, this is the third largest energy tax expenditure in the federal budget totaling $3.2 billion over five years. In addition to the tax preferences described above, I note two additional significant tax preferences. First, owners of coal mining rights who lease their land for mining receive royalties for coal extracted from their property. Owners who are individuals may elect to treat those royalty payments as capital income in lieu of taking percentage depletion on the property. Second, owners of working interests in oil and gas properties are exempt from passive loss limitations for income from these properties. 4.3 Production and Investment Tax Credits The federal tax code includes a number of production and investment tax credits on fossil, alternative, nuclear, and renewable fuels. Those credits include the following: Section 29 Non-Conventional Oil Production Credit24 The 1980 Windfall Profits Tax (PL 96–223) was a failed effort to simultaneously capture profits on old oil as a result of oil price increases in the 1970s and encourage exploration for and production of new oil. The law was repealed in 1988.25 One part of that law that was retained was the section 29 Alternative Fuel Production Credit for production of non-conventional oil (e.g., shale oil, synthetic fuel oils from coal). The section provides for a $3.00 per barrel of oil-equivalent production tax credit (indexed in 1979 dollars and worth $6.79 in 2005). The 2005 energy act adds coke and coke gas to the list of qualified fuels and makes the credit part of the general business credit.26 The credit phases out for oil prices above $23.50 in 1979 dollars ($53.20 in 2005).27 The tax expenditure for this credit is estimated to be $3.4 billion between FY 2007 and 2011 and is the second largest energy tax expenditure in the federal budget. The main beneficiary of the credit is coalbed methane, natural gas that is extracted from tight seams in coal mines. Traditionally this gas was vented to reduce safety problems in mines. But with higher gas prices, the credit and advances in technology, it has become economic to recover this gas. This is not a non-conventional fuel per se but its extraction method might be viewed as non-conventional.
1.
Federal Tax Policy towards Energy
161
2.
Section 45 Production Tax Credits Section 45 of the IRS code, enacted in the Energy Policy Act of 1992, provided for a production tax credit of 1.5¢ per kWh (indexed) of electricity generated from wind and closed-loop biomass systems.28 The tax credit has been extended and expanded over time and currently is available for wind, closed-loop biomass, poultry waste, solar, geothermal and other renewable sources at a current rate of 1.9¢ per kWh.29 Firms may take the credit for ten years. Refined coal is also eligible for a section 45 production credit at the current rate of $5.481 per ton.30 EPACT added new hydropower and Indian coal with the latter receiving a credit of $1.50 per ton for the first four years and $2.00 per ton for three additional years. While EPACT extended the section 45 tax credits for two additional years (through 2007), it did not extend the credit for solar generated electricity beyond 2005. Finally, EPACT allowed for the issuance of $800 million in Clean Renewable Energy Bonds (CREBs) to finance projects eligible for section 45 production tax credits (with the exception of Indian coal). CREBs do not pay interest. Rather the holder of a CREB on its credit allowance date is entitled to a tax credit to be determined by the Treasury Department so that the bond may be sold at par.31 3. Other Production Tax Credits EPACT provided a production tax credit for electricity produced at nuclear power plants (section 45J). Qualifying plants are eligible for a 1.8¢ per kWh production tax credit up to an annual limit of $125 million per 1,000 megawatts of installed capacity. This limit will be binding for a nuclear power plant with a capacity factor of 80 percent or higher. The law places an aggregate limit of 6,000 megawatts of capacity eligible for this credit. The American Jobs Creation Act of 2004 (PL 108–357) created a production credit (section 45I) for marginal oil and gas producers of $3.00 per barrel of oil ($.50 per mcf of natural gas) in year 2005 dollars. The full credit is available when oil (gas) prices fall below $15 per barrel ($1.67 per mcf) and phases out when prices reach $18 per barrel ($2.00 per mcf).32 Marginal wells produce on average 15 or less barrels of oil (or oil equivalent) per day. This same law provided for small refinery expensing of 75 percent of capital costs associated with low sulfur diesel fuel production and a 5¢ per gallon small refiner’s credit for the remaining 25 percent of qualified capital costs for the production of low sulfur diesel fuel. The
162
Metcalf
2005 Energy Policy Act allowed a pass through of this credit to owners of cooperatives. The Omnibus Budget Reconciliation Act of 1990 contained a provision for a 15 percent credit (section 43) for expenditures on enhanced oil recovery tangible property and intangible drilling and development costs and other related capital expenditures. The credit is phased out as the section 29 reference oil price exceeds $28 in 1990 dollars ($37.44 for 2005). At current prices, producers cannot take this credit. 4. Section 48 Investment Tax Credits Nonrefundable investment tax credits for alternative energy were initially put in place in the Energy Tax Act of 1978 (PL 95–618) at a rate of 10 percent for solar and geothermal property. That law provided a number of investment tax credits including a credit for residential energy conservation investments. This latter credit expired in 1982. EPACT increased the investment tax credit for solar to 30 percent. In addition the 30 percent tax credit applies to fuel cells used to produce electricity while the 10 percent credit is available for qualifying microturbine power plants. The section 48 investment tax credits for renewable energy production were extended by EPACT to apply to investments in certain clean coal facilities. Integrated gasification combined cycle (IGCC) plants are eligible for a 20 percent credit (up to a maximum of $800 million in credits); other advanced coal-based projects are eligible for a 15 percent credit (up to a maximum of $500 million in credits); and certified gasification projects are also eligible for a 20 percent credit (maximum of $350 million in credits). The section 45 and 48 credits combined are the single largest energy tax expenditure in the federal budget worth $4.1 billion over a five year period. 5. Section 40 Alcohol and Biodiesel Fuels Credit The Energy Policy Act of 1978 included an exemption from the motor fuels excise tax for alcohol and alcohol blended fuels, generically known as gasohol.33 The Windfall Profits Tax allowed an immediate tax credit in lieu of the exemption.34 The credit was set at a rate to be equivalent to the tax exemption. The alcohol fuel mixture credit is currently $.51 per gallon of ethanol in gasohol and $.60 for other alcohol based fuels (excluding petroleum based alcohol fuels). In addition small producers may take a credit of $.10 per gallon. The 2005 Energy Policy Act increased the small producer production capacity limit from 30 million to 60 million gallons per year.
Federal Tax Policy towards Energy
163
The American Jobs Creation Act also added section 40A of the code to provide an income tax credit for biodiesel fuels at a rate of $.50 per gallon of bio-diesel (other than agri-biodiesel) and $1.00 for agri-biodiesel. Like the alcohol fuel tax credit, it is first applied to motor fuel excise tax payments with the excess added to the general business credit. 6. Other Issues Bearing on Production and Investment Tax Credits Firms or individuals may not receive the full value of the production and investment tax credits (along with other energy related tax incentives) described above depending on the taxpayer’s alternative minimum tax (AMT) status. All of these credits are part of the general business credit. Credits included in the general business credit may be used to the extent that they do not reduce the taxpayer’s after-credit liability below the tentative minimum tax. Note that this limitation may occur even if the taxpayer pays no alternative minimum tax.35 According to Carlson (2005), 70 percent of firms in the mining industry in 2002 were either in a loss or an AMT status and so unable to avail themselves of many if not all of their tax credits.36 It is unclear how the limitation on the use of credits under the general business credit affects investment. 4.4 Tax Incentives for Electric Utilities Many of the tax incentives that affect the electric utility industry have to do with accelerated depreciation and are discussed above. EPACT provided for several additional incentives. First, electric utilities are allowed to carry back net operating losses (NOLs) for five years (as opposed to the standard two year carry back) for an NOL occurring in tax years 2003–2005. The NOL must be used before January 1, 2009 and the tax refund arising from the use of the NOL is limited in any year to 20 percent of the utility’s prior year investment in electric transmission equipment rated at 69 kV or higher and specified pollution control equipment.37 Second, owners of nuclear power plants are required to make contributions to a decommissioning fund for the plant. The Deficit Reduction Act of 1984 allowed those contributions to be tax deductible at the time of contribution if those contributions were funded as part of the cost of service to ratepayers of regulated utilities. The cost of service rules were repealed in EPACT so that all taxpayers, including unregulated utilities, could deduct their contributions to decommissioning funds. Finally, EPACT extends a current provision allowing electric utilities who dispose of certain transmission property to implement FERC restructuring policy to recognize the gain over an eight year period
Metcalf
164
rather than in the current year. Proceeds from the sale must be reinvested in other utility property. 4.5 Transportation The Energy Policy Act of 1992 allowed a 10 percent credit (up to $4,000) for the purchase of an electric vehicle. After 2005, the maximum credit falls to $1,000 and the credit terminates as of the end of 2006. Hybrid clean-fuel vehicles and other clean-fuel vehicles were allowed a $2,000 deduction. These deductions and credits were replaced by the Alternative Motor Vehicle Credit (sec. 30B) as enacted in EPACT. Section 30B of the tax code provides a credit for fuel cell vehicles, alternate fuel vehicles (natural gas, LNG, LPG, hydrogen, and 85 percent methanol fuel vehicles), and hybrids. The credit depends on different vehicle attributes depending on the type of vehicle. Table 5.5 lists the credit information for different fuel types. EPACT also replaced a deduction for installing clean-fuel vehicle refueling property with a 30 percent tax credit for property placed in service before Jan. 1, 2008. 4.6 Energy Efficiency Prior to the passage of last year’s energy legislation, the only remaining tax incentive pertaining to energy conservation was section 136 of the tax code enacted in the Energy Policy Act of 1992. Section 136 provides an exclusion from gross income of any subsidies provided by an electric utility for the purchase or installation of any energy conservation measure. EPACT provided a number of new incentives. First, the law allows a 30 percent personal income tax credit not to exceed $2,000 for photovoltaic and solar water heating property (excluding equipment for heating swimming pools and hot tubs) installed at residential Table 5.5 Clean Vehicle Tax Credits
Fuel Type
Credit Determining Characteristics
Maximum Credit with Gross Vehicle Rating Less Than 8,500 Pounds
Fuel cell
Gross vehicle weight, fuel economy
$12,000
Hybrids
Fuel economy, lifetime fuel savings
$3,400
Alternative fuels
Gross vehicle weight
$4,000
Advanced lean-burn diesel
Fuel economy, lifetime fuel savings
$3,400
Federal Tax Policy towards Energy
165
properties. Fuel cell power plants are also eligible for the 30 percent credit not to exceed $500 per 0.5 kW of capacity. Second, the law provides a 10 percent personal income tax credit for insulation materials, energy saving windows and doors, and energy conserving metal roofs. In addition, taxpayers may take a credit for specific energy efficiency appliances such as advanced main air circulating fans ($50), furnace and hot water boilers ($150), and qualifying energy efficient property (e.g., designated heat pumps and air conditioners) ($300) with a maximum credit per home of $500 no more than $200 of which may be for windows. Third, contractors may take a tax credit of $1,000 ($2,000) for new home construction that is certified to obtain a 30 percent (50 percent) reduction in energy usage. Fourth, commercial property energy conservation expenditures are eligible for a deduction of costs up to $1.80 per square foot if the spending effects a 50 percent or more reduction in energy usage. For buildings that do not achieve a 50 percent reduction, a partial allowance is allowed based on guidelines for specific technologies to be established by the Secretary of the Treasury. Finally, appliance manufacturers are provided a production credit for energy-efficient dishwashers, clothes washers and refrigerators. The maximum credit is $100 for dishwashers, $200 for clothes washers and $175 for refrigerators. 5.
Incentive and Distributional Effects of Energy Tax Incentives
Who benefits from the various taxes and tax incentives described in sections 3 and 4? And what are the impacts on energy demand and supply? In this section, I discuss the incidence and behavioral impacts of various tax provisions. 5.1 Motor Fuels Excise and Gas-Guzzler Taxes Consider first the incidence impact of the federal excise tax on motor fuels. Doyle and Samphantharak (2006) find that roughly 80 percent of increases in state sales taxes on gasoline are passed forward to consumers. This is consistent with other studies and likely underestimates the shifting of federal excise taxes to consumers given state-level competition.38 Gasoline taxes are generally viewed as regressive. This view is confirmed by Poterba (1991) when households are ranked by annual income. But when ranked by current expenditures as a proxy for lifetime income, Poterba finds that gasoline taxes are much less regressive.
Metcalf
166
Assessing the gas guzzler tax is more difficult. Surprisingly few studies of the gas guzzler tax have been carried out that take the light truck loophole into account. Greene et al. (2005) undertake simulations of a gas-guzzler tax and find that removing the exemption for light trucks increases mileage for these vehicles as manufacturers cluster vehicles (both passenger cars and light trucks) just above the miles per gallon cutoff for the tax. Their study holds all characteristics other than fuel economy and price constant. Thus, the recent phenomenon of using engine improvements to obtain higher power and performance at the expense of fuel efficiency cannot be modeled in their analysis. 5.2 Oil and Natural Gas Production Incentives Turning to production and investment tax incentives, consider first oil and natural gas production. The favorable treatment accorded oil producers and refiners lowers the cost of oil and could affect prices of final petroleum products. But since oil is priced in world markets and to a great extent is a homogenous product, it is not clear that the domestic tax incentives would have a large impact on the price of gasoline or other petroleum products.39 In this case, the benefits largely accrue to producers in the form of higher wages for specialized workers in oil production and refining and higher dividends to shareholders. It may be, however, that the U.S. supply incentives have an impact on worldwide supply and price given the magnitude of U.S. oil production. As noted in endnote 5, the United States is the third largest producer of oil with 8.5 percent of the world’s production in 2004. A simple rough calculation suggests however that the U.S. supply incentives are unlikely to have a large impact on world oil prices or supply. Let Q *S be the world supply of oil, and QS the U.S. supply (with analogous variables defined for oil demand). Also let ps and pˆs be the price received by U.S. oil suppliers and oil suppliers in the rest of the world, respectively. Finally, let pD = pˆs = ps – s be the worldwide demand price and s the subsidy arising from tax incentives provided to domestic oil suppliers. Setting world oil supply equal to demand and differentiating, we obtain the relationship between world oil prices and the domestic subsidy: ⎛ QS ⎞ ⎜⎝ Q * ⎟⎠ ηD
dpD S = ds ηS − ηD
(5.1)
Federal Tax Policy towards Energy
167
where ηD and ηS are the demand and supply elasticities for oil. Longrun estimates of these elasticities are in the neighborhood of –0.5 and 0.5 respectively.40 The percentage change in worldwide oil supply is then ⎛ QS ⎞ ⎜⎝ Q * ⎟⎠ ηD ds dQS* S = − ηD . ηS − ηD p QS*
(5.2)
The tax incentives for oil (percentage depletion and expensing of IDCs) are most valuable for small producers. Taking the oil industry as a group, let us say that the value of the subsidies is worth 10 percent of the current price of oil.41 Table 5.6 shows the impact on world oil production for various demand and supply elasticities: Table 5.6 Percentage Change in World Oil Supply Demand Elasticity
Supply elasticity
–0.1
–0.3
–0.5
–0.7
0.1
0.0%
0.1%
0.1%
0.1%
0.3
0.1%
0.1%
0.2%
0.2%
0.5
0.1%
0.2%
0.2%
0.2%
0.7
0.1%
0.2%
0.2%
0.3%
Change arising from a subsidy to domestic production equal to 10 percent of oil price.
The supply response ranges from zero to 0.3 percent with 0.2 percent the response associated with the central parameter value assumptions. Table 5.7 shows that the price response is also small: Table 5.7 Percentage Change in World Oil Price Demand Elasticity
Supply elasticity
–0.1
–0.3
–0.5
–0.7
0.1
–0.4%
–0.2%
–0.1%
–0.1%
0.3
–0.6%
–0.4%
–0.3%
–0.3%
0.5
–0.7%
–0.5%
–0.4%
–0.4%
0.7
–0.7%
–0.6%
–0.5%
–0.4%
Change arising from a subsidy to domestic production equal to 10 percent of oil price.
168
Metcalf
The price decline ranges from 0.1 percent to 0.7 percent with a central parameter response of 0.4 percent. To some extent, a similar story can be told for natural gas. Natural gas is not as easily transportable as oil and regional price differences can persist over time. Improvements in transportation and the increase in LNG shipping, however, are breaking down these regional barriers.42 This analysis is consistent with a recent analysis of a precursor bill to the Energy Policy Act of 2005 done by the U.S. Energy Information Administration (2004). This report reviewed section 29 and 45 tax credits along with other production incentive tax provisions and concluded that with the exception of the section 29 credits, the provisions did little to increase domestic production of gas or oil. Section 29 credits would increase domestic natural gas from non-conventional sources (coalbed methane for the most part). Ultimately, domestic consumption would be unaffected by these provisions. Recall the discussion of national security as a rationale for an energy tax policy. The analysis in this section suggests that the production and investment tax credits embodied in current law will have little effect on world production or on efforts to stabilize domestic energy prices. Where a policy to encourage domestic production of energy may be effective is to increase the proportion of energy that is not subject to supply disruptions due to political upheaval. But here the rationale is a bit murky. Many of the tax incentives encourage the production of electricity from nuclear or renewable sources. But currently only 3 percent of oil is used for electricity production. It may well be that concerns about natural gas disruptions motivate these policies (natural gas accounts for 24 percent of electricity production). Natural gas however is more subject to price spikes arising from bottlenecks in production and distribution than from political shocks. A concern with oil supply disruptions would suggest a focus on reducing petroleum use in transportation, currently responsible for two-thirds of all petroleum consumption. 5.3 Electric Generation from Alternative and Renewable Fuels The production and investment credits for renewable and alternative fuels can have a large impact on whether various electric generation technologies are cost competitive in the marketplace. With the shift from regulated utilities to an environment in which electricity generation is increasingly unregulated, cost considerations become increasingly important for firms contemplating constructing merchant power
Federal Tax Policy towards Energy
169
plants. In this section, I present estimates of the levelized cost for different sources of electricity under varying assumptions about the availability of federal tax incentives. The levelized cost analysis is similar in spirit to the Hall-Jorgenson cost of capital framework. It asks what price must be received for electricity sold by a generator to cover fixed and variable costs of providing the electricity including the required return for equity owners.43 This approach has been used in a variety of studies of electric power generation (e.g., Deutch and Moniz 2003; Tolley and Jones 2004; and Sekar et al. 2005). The steps to constructing an estimate of levelized cost are: • Compute the present discounted value of costs in each year over life of a project. This includes all capital and operating costs net of tax deductions. • Sum all costs over life of project. This is the present discounted value of the project’s overall costs. • Compute the amount of constant real before-tax revenue required each year that will equal the total present discounted value of costs over the life of the project. • Divide this required revenue value by total kilowatt-hours produced by plant to obtain a cost per kWh. I estimate the levelized cost for the following eight electricity generation sources: nuclear, conventional (pulverized) coal, clean coal using an integrated gasification combined cycle (IGCC) process, natural gas combined cycle, biomass, wind, solar thermal and photovoltaics. The first three technologies are generally used as baseload generators and the latter are either shoulder or peaking generators. Table 5.8 provides key parameter value choices for the eight different technologies.44 The capacity factor describes what fraction of the time a plant is operating. Nuclear power plants are designed to operate continuously but are shut down for routine and unexpected maintenance. The capacity factor for nuclear power is taken from Deutch and Moniz (2003). I’ve chosen capacity factors for coal and natural gas to be comparable to nuclear. The capacity factors for the renewable resources are for the most part taken from the Energy Information Administration’s National Energy Modeling System (NEMS).45 The overnight cost is the total capital construction cost of the plant in year 2004 dollars. Construction costs are covered with short-term debt financing until the plant goes into service. At that point, ten year
170
Table 5.8 Selected Parameter Values for Levelized Cost Analysis Technology
Nuclear
Coal PC
Coal IGCC
Gas-CC
Biomass
Wind
Solar Thermal
PV
Capacity factor
85%
85%
85%
85%
83%
35%*
31%*
21%*
Construction time
6
4
4
3
4
3
3
2
Overnight cost ($/kW)
2,014
1,249
1,443
584
1,809
1,167
3,047
4,598
% Debt finance
50%
60%
60%
60%
60%
60%
60%
60%
Economic life
40
30
25
25
20
20
20
20
MACRS life
15
15
15
15
5
5
5
5
Production tax credit ($/kWh)
125
0
0
0
0.019
0.019
0
0
Section 48 ITC
0
0
20%
0
0
0
30%
30%
*I have not assumed any additional costs for capacity to account for the intermittency of these power sources. Source: See Appendix.
Metcalf
Federal Tax Policy towards Energy
171
bonds are issued and equity financing raised to cover those costs. I’ve assumed 60 percent debt financing on all projects except nuclear. I assume a lower debt financing rate of 50 percent to acknowledge the greater perceived risk of nuclear financing in the marketplace. The economic life of these assets varies and they have a MACRS recovery period of five to 15 years. Finally, I assume that the Section 45J production tax credit for nuclear power hits the $125 million cap per 1,000 MW of capacity. See the Appendix for more details on the computation of levelized costs. Table 5.9 reports levelized costs of electricity in cents per kWh (year 2004 dollars). I assume that the plant will be placed in service after Jan. 1, 2006 so that solar power is not eligible for a production tax credit but does obtain the more generous 30 percent section 48 investment tax credit.46 The first column provides the levelized cost under current law. Coal has the lowest levelized cost with the cost of IGCC comparable to that of a conventional pulverized coal plant given the new investment tax incentive for IGCC enacted in EPACT. Nuclear and natural gas are the next most expensive followed by biomass and wind.47 Either of the solar generating plants are considerably more expensive than other electricity sources with photovoltaics (PV) over four times the cost of natural gas. Note that wind and solar are intermittent power sources and so require stand-by generation. A recent study by The Royal Academy of Engineering (2004) found that the requirement for stand-by power raised the cost of onshore wind power by nearly 50 percent. I have not factored such costs into this analysis. Comparing the first two columns, eliminating the section 45 production tax credit only modestly raises the cost of biomass and wind (4 percent cost increase) but raises the cost of nuclear by nearly 30 percent. Next, eliminating the section 48 investment tax credits raises the cost of the IGCC plant by 15 percent and the cost of solar by over 35 percent. Column 4 reports levelized costs assuming the various production and investment tax credits but replacing the accelerated depreciation with economic depreciation (modeled as straight-line depreciation) over the asset’s life.48 Accelerated depreciation is most generous to wind and solar generated electricity. Replacing accelerated depreciation with economic depreciation would raise the cost of wind and solar thermal by 13 percent and PV by 15 percent. For the other fuel sources, replacing accelerated depreciation with economic depreciation would raise the cost of nuclear by 9 percent, biomass by 8 percent, coal by 7 percent and natural gas by 2 percent.
172
Table 5.9 Real Levelized Costs of Electricity Technology
Current Law (1)
No PTC (2)
No ITC (3)
Economic Depreciation (4)
Level Playing Field (5)
No Tax (6)
Nuclear
4.31
5.55
4.31
4.70
5.94
4.57
Conventional coal
3.53
3.53
3.53
3.79
3.79
3.10
Clean coal (IGCC)
3.55
3.55
4.06
3.80
4.37
3.53
Natural gas
5.47
5.47
5.47
5.61
5.61
5.29
Biomass
5.34
5.56
5.34
5.74
5.95
4.96
Wind
5.70
5.91
5.70
6.42
Solar thermal
12.25
12.25
16.68
13.74
18.82
13.84
PV
22.99
22.99
32.60
26.34
37.39
26.64
6.64
4.95
Author’s calculations. See Appendix for further detail. Cost are reported in cents per kWh at year 2004 prices.
Metcalf
Federal Tax Policy towards Energy
173
Column 5 reports levelized costs assuming a tax system that provides a level playing field. This scenario assumes economic depreciation and no production or investment tax credits. In terms of the impact on levelized cost, conventional coal and natural gas receive the fewest tax preferences. Leveling the playing field raises the cost of biomass by 11 percent with the bulk of the benefit arising from accelerated depreciation (based on a comparison of columns 2 through 4 with 5). The cost of wind is higher by 16 percent with the majority of the benefit arising from accelerated depreciation. The cost of IGCC is higher by 23 percent with roughly two-thirds of the benefit arising from the investment tax credit. The cost of nuclear is higher by 38 percent with the production tax credit providing the bulk of the benefits. Finally, the cost of solar is over 50 to 60 percent higher with about two-thirds of the benefit arising from the production tax credits. In the final column, I compute levelized costs assuming zero taxes. While the levelized cost of most technologies falls, the cost of nuclear and solar rises indicating that these technologies face a negative effective average tax rate. Eliminating taxes raises the cost of nuclear by 6 percent and solar by 13 to 16 percent. From a social welfare perspective, the production and investment tax credits are costly ways to encourage renewable electricity generation since the subsidies must be financed by raising distortionary taxes. An alternative approach to encouraging renewable electricity generation would be to place a tax on traditional fuels.49 As a final calculation, I computed the levelized cost of biomass and wind assuming no investment or production tax credits. In this case, the levelized costs of biomass and wind are 5.56¢ and 5.91¢ per kWh respectively. A tax on carbon dioxide of $12 per metric ton would raise the price of natural gas sufficiently to make biomass and wind cost-competitive with natural gas. Unlike the subsidies, however, the tax would raise revenue which could finance reductions in other distortionary taxes.50 In units perhaps more familiar to most readers, a carbon tax of this magnitude would raise the price of gasoline by ten cents if it were fully passed forward to consumers. Summing up, relative to a world with no taxes the current tax code provides net subsidies to nuclear and solar power. Relative to a tax system with a level playing field, conventional technologies receive very modest subsidies while subsidies for nuclear and clean coal are substantial and the subsidies for solar very substantial. The subsidies are most effective (in the sense of making electricity competitive from this
Metcalf
174
source) for IGCC plants which become competitive with conventional coal and for biomass and wind which become competitive with natural gas. 5.4 Energy Efficiency The energy efficiency incentives contained in EPACT are similar in many ways to energy tax credits contained in the Energy Tax Act of 1978, including a 15 percent tax credit (up to $300) for residential energy conservation improvements. Analyzing a panel of federal tax returns between 1979 and 1985 when the residential conservation credit expired, Hassett and Metcalf (1995) found that the credit significantly raised the probability of a household installing energy conservation capital in their home. Somewhat surprisingly, the authors found that the credit was much more successful at raising investment levels than a comparable energy price increase. They speculated that the credit program may have publicity effects that spur investment that the energy price increase does not have. The study was not able to determine to what extent credit takers were inframarginal investors—that is homeowners who would have made conservation investments in the absence of the tax credit.51 6.
Conclusion
Tax incentives are a major part of the federal government’s energy tax policy and increasingly so with the passage of the Energy Policy Act of 2005. A number of points emerge from this analysis. First, the focus of energy incentives contained in the tax code has shifted over the years from focusing almost entirely on traditional fossil fuel production to an increasing emphasis on alternative and renewable technologies. Second, those incentives are difficult to rationalize on the basis of economic efficiency or distributional goals. Production and investment tax credits, in particular, may be very costly ways to encourage the development of renewable energy technology. Third, incentives for the oil and natural gas industry are unlikely to have an appreciable impact on world energy prices despite the United States being the third largest oil producer in the world (and second largest natural gas producer). Fourth, current tax incentives are making wind and biomass cost competitive with natural gas electricity production. The 20 percent investment tax credit for IGCC in EPACT is likely to make this technology cost competitive with conventional coal
Federal Tax Policy towards Energy
175
generated electricity. Solar generated electricity continues to be very expensive. Fifth, the limited evidence suggests that the energy efficiency incentives enacted by EPACT should increase conservation investment activity. It is difficult to say, however, how much of this investment will be new investment as opposed to investment that would have taken place in the absence of the incentive programs. To the extent that inframarginal investment is a significant fraction of total investment, the costeffectiveness of this incentive is driven down. But of course this is true for all of the energy incentives described in this paper and suggests the importance of further research on the behavioral impacts of energy tax incentives. Notes This paper was prepared for the Tax Policy and the Economy conference held on September 14, 2006 in Washington, DC. I thank Tom Barthold, Alex Brill, John Navratil, Nicolas Osouf, John Parsons, Jim Poterba, and participants in the MIT Joint Program on the Science and Policy of Global Change EPPA Seminar for helpful discussions. I am grateful for support from the MIT Joint Program on the Science and Policy of Global Change which I was visiting while writing this paper. Contact information: gmetcalf@ tufts.edu. 1. The list is naturally incomplete given the complexity of the tax code. In particular I do not focus on how the tax treatment of foreign income earned by multinational corporations bears on energy production. This is potentially a major issue. For example, prior to the nationalization of oil production in the major oil producing countries, the major U.S. oil producers paid taxes to host countries that were termed income taxes but were in reality excise taxes. Standard tax treatment would provide for a deduction on the U.S. corporate income tax for those foreign tax payments. Instead, the U.S. companies were allowed a foreign tax credit for the “income” taxes paid to host countries, a preference lobbied for by—among others—the State Department. See Adelman (1995), pp. 50–55 for more on this point. 2. See Feng et al. (2005) for a discussion and evaluation of feasible alternatives to direct emission taxes for motor vehicles. 3. Two-thirds of petroleum is used in the transportation sector (U.S. Energy Information Administration 2005). 4. Partial policies can raise the cost of carbon emission reductions considerably. Pizer et al. (2006) present model results showing that focusing climate change policies only on the transportation and electricity sectors doubles the cost of a given carbon emissions reduction. Note too that the motor vehicle fuels tax is sometimes justified as a use charge for highways. To the extent this is true, the motor vehicle fuels tax is even less effective as a proxy tax for externalities. 5. That the source of the oil the United States consumes is irrelevant for oil price stability should be made clear by the fact that the United States is the third largest oil producer in
176
Metcalf
the world, with production only exceeded by the Russian Federation and Saudi Arabia. The United States produced 8.5 percent of the world’s oil in 2004. It is also the second largest producer of natural gas after the Russian Federation with a world production share of 19 percent. See BP (2006) for data. 6. Karp and Newbery (1991) provide a more sophisticated analysis to resolve a dynamic inconsistency problem with simple oil tariff expropriation stories. But the basic result holds. 7. The tax was most recently raised to 18.3¢ per gallon for gasoline on Oct. 1, 1993. See Jackson (2006) for a history of changes to this tax. 8. In fiscal year 2004, the UK collected £832 million in its Climate Change Levy, approximately £1,614 million in VAT on energy related sales, and £22,786 in its hydrocarbons tax. Data are from excise tax sheets published by HM Revenue & Customs and available at http://www.uktradeinfo.com. 9. The mileage rating is calculated approximately as 55 percent of the EPA city mileage rating and 45 percent of the highway rating. 10. I have not included tax expenditures associated with transportation (e.g., exclusion from income for employer reimbursed parking). Nor do I consider state or local energy tax incentives in this paper. 11. As GAO points out, one cannot simply add tax expenditures given the interactions among different provisions of the tax code. But the summation indicates the relative importance of the provisions when making comparisons across time. 12. It is unclear whether this tax expenditure has any incentive effect now that ethanol use is mandated in motor fuels by the Energy Policy Act of 2005. I thank John McLelland for pointing this out. 13. This brief description draws on an excellent overview by Lazzari (2006). 14. The American Jobs Creation Act of 2004 (PL 108–357) provided a major expansion of the production tax credits. 15. The production tax credit for wind and biomass briefly expired in 2003. According to the American Wind Energy Association, wind power capacity additions fell from 1,687 MW in 2003 to 389 MW following the temporary lapse of this tax provision. 16. In general I do not discuss energy tax incentives that have expired. See Edwards et al. (1998) for some discussion of energy tax incentives related to global warming that existed prior to 1998. This includes the major incentives that have expired. I also generally do not provide information about sunset provisions for the various incentives since historically sunset dates have been extended for most energy-related tax incentives. 17. The recovery period is the number of years over which an asset may be depreciated for tax purposes. Certain assets must be depreciated under the Alternative Depreciation System (ADS). See U.S. Internal Revenue Service (2006) for more information. 18. Exploratory wells continue to have high failure rates. In 2003, 55 percent of exploratory wells were dry holes and 9 percent of development wells were dry holes. But less than 2,700 exploratory wells were drilled that year compared to over 32,200 development wells. Roughly the same number of development wells were drilled in 1960 but with a dry hole rate of 25 percent. However, 11,700 exploratory wells were drilled with over 80
Federal Tax Policy towards Energy
177
percent of them being dry holes. See tables 4.5–4.7 in U.S. Energy Information Administration (2005). 19. The caps for 2005 were $2,960 in the first year, $4,700 in the second year, $2,850 in the third year, and $1,675 in subsequent years for passenger cars. For light trucks weighing less than 6,000 pounds (including minivans, SUVs, and pickup trucks) the limits are $3,260 in the first year, $5,200 in the second year, $3,150 in the third year, and $1,875 in subsequent years. 20. Independent producers are defined as producers who do not engage in refining or retail operations. EPACT increased the amount of oil a company could refine before it was deemed to engage in refining for this purpose from 50,000 to 75,000 barrels per day. 21. Amounts in excess of the 65 percent rule can be carried forward to subsequent tax years. The net income limitation has been suspended in years past but the suspension lapsed as of this year. 22. This example presumes that the net income from the first year’s operation exceeds $94,500. If not, the deduction would be reduced accordingly. For purposes of computing the net income limitation, costs are computed without any depletion deduction considered. 23. EPACT set the recovery period at two years but the Tax Increase Prevention and Reconciliation Act of 2005 (PL 109–222) extended the period to five years for the major integrated oil companies. 24. Section 29 is relabeled as section 45K by EPACT. 25. For an overview and analysis of the Windfall Profit Tax, see Lazzari (1990). 26. Most energy tax credits were part of the general business credit. Prior to EPACT, the section 29 credits were an exception and so any unused credits were lost. As part of the general business credit, excess credits can be carried backward one year and forward 20 years. 27. The reference price for oil in 2005 was $50.26 and so the full credit could be taken. The credit amount and reference price are published annually in the Federal Register. With the reference oil price currently at $62.51 (April 2006 crude oil domestic first purchase price), it is unlikely that firms will be able to take the full section 29 credit in 2006. 28. A closed-loop biomass is plant material grown specifically for use in a biomass generator. 29. Open-loop biomass is eligible for a 0.75¢ in 1992 dollars per kWh. 30. Refined coal is a synthetic fuel produced from coal with lower emissions of certain pollutants. 31. State and local tax exempt financing is also available for qualified energy facilities. These bonds are subject to a state’s private-activity volume cap. 32. The section 29 reference price is used to determine eligibility for this credit. 33. Originally, the law provided a full exemption from the then $.04 per gallon tax. As the motor fuels excise tax was raised over time, the exemption did not keep pace with the excise tax rate. See General Accounting Office (1997) for an early chronology of events related to this tax exemption.
178
Metcalf
34. The American Jobs Creation Act of 2004 subsequently eliminated the tax exemption in favor of the tax credit. 35. This would occur if the taxpayer’s regular tax liability after foreign tax credits but before other tax credits exceeded the tentative minimum tax but its regular tax liability after tax credits was less than the tentative minimum tax. Foreign tax credits are included in the tentative minimum tax and thus not subject to AMT limitations. 36. In addition to the AMT’s impact on tax credits, the AMT treats mining exploration and development costs, depletion including percentage depletion (unless an independent producer) and intangible drilling costs as AMT preferences. 37. The equipment need not be placed in service in that year. 38. Doyle and Samphanthark report other studies showing complete forward shifting of federal motor fuel taxes. 39. Oil differs in its transportation costs as well as product characteristics (sulfur content, viscosity, etc.) In the long run, however, these cost and characteristic differences have little impact on the final product costs. 40. Cooper (2003) reviews estimates of the long run demand elasticity and Greene et al. (1998) reviews long run demand and supply elasticities. 41. This is a high estimate. The GAO estimates for FY2003 tax preferences for the section 29 and enhanced oil recovery credits, the excess of percentage over cost depletion, expensing of IDCs, and the rules on passive loss limitations equal just over 2 percent of the value of domestic crude oil and natural gas production in that year. 42. Natural gas imports as a percentage of total U.S. consumption have risen from 4.7 percent in 1980 to 15.3 percent in 2004 (U.S. Energy Information Administration 2005). 43. The price is a constant real price received over the life of the plant to cover lifetime fixed and variable costs. 44. See Appendix A for a complete listing of parameter values and additional detail about the calculations. These parameter values should be viewed with some caution. A degree of uncertainty underlies many of the values. The overnight cost for nuclear power, for example, is highly uncertain given the limited recent experience with construction in the United States and the uncertainties of the regulatory process. 45. I have not adjusted the wind and solar capacity factors to account for the intermittency of these power sources. 46. I have not assumed any limitations on credits from the Alternative Minimum Tax in table 5.8. 47. Given the cost differential between coal and nuclear, the current interest in nuclear power reflects in part a bet that a U.S. carbon policy will eventually raise the cost of coal power plants. 48. I’ve also modeled economic depreciation for the assets according to the depreciation rates estimated by Fraumeni (1997) in a modified one-hoss-shay model. I assume geometric depreciation using Fraumeni’s rates over the life of the asset with all remaining basis depreciated in the final year. The levelized costs under this approach are very similar to those calculated when economic depreciation is modeled as straight-line.
Federal Tax Policy towards Energy
179
49. One could also impose renewable portfolio standards as many states have done. Palmer and Burtraw (2005) argue that portfolio standards are more cost effective at achieving given renewable shares in electricity generation than production tax credits given the social cost of raising revenue to finance the subsidies. Note too the different incidence of production tax credits and renewable portfolio standards. The former are borne by taxpayers while the latter are borne by electricity consumers in the form of higher electricity prices. 50. Metcalf (2005) discusses how a carbon tax could be used to finance corporate tax integration. The advantage of taxes over subsidies for clean power extends beyond the distortionary cost of financing the subsidies. The subsidies lower the cost of electricity and so encourage increased consumption. 51. In the energy conservation literature, this is referred to as free-riding. See Metcalf (2006) for a discussion of behavioral responses to energy conservation initiatives and the difficulty in assessing the cost-effectiveness of these programs. 52. Many renewable plants are built at considerably smaller capacity. The cost assumptions used here are based on a plant of optimal size. My approach follows that of Sekar, et al. (2005).
References Adelman, M.A. (1995). The Genie out of the Bottle. Cambridge, MA: The MIT Press. American Petroleum Institute (2006). “Gasoline and Diesel Taxes,” http://api-ec.api. org/filelibrary/2006-gasoline-diesel-taxes-summary.pdf, Accessed on June 15, 2006. Bergstrom, Theodore C. (1982). “On Capturing Oil Rents with a National Excise Tax,” American Economic Review, 72(2):194–201. BP. (2006). “Statistical Review of World Energy 2006,” London: BP. Carlson, Curtis P. (2005). “The Corporate Alternative Minimum Tax Aggregate Historical Trends,” Department of the Treasury Office of Tax Analysis, OTA Paper 93. Cooper, John C.B. (2003). “Price Elasticity of Demand for Crude Oil: Estimates for 23 Countrie,” OPEC Review, 27(1):1–8. Deutch, John, and Ernest J. Moniz, (eds.) (2003). The Future of Nuclear Power. Cambridge, MA: Massachusetts Institute of Technology. Doyle, Joseph J., Jr., and Krislert Samphantharak (2006). “$2.00 Gas! Studying the Effects of a Gas Tax Moratorium,” NBER working paper no. 12266. Edwards, Chris, Ada Rousso, Peter Merrill, and Elizabeth Wagner (1998). “Cool Code: Federal Tax Incentives to Mitigate Global Warming,” National Tax Journal, 51(3):465–484. Energy Information Administration (2005). “Annual Energy Review 2004,” Energy Information Administration. Feng, Ye, Don Fullerton, and Li Gan (2005). “Vehicle Choices, Miles Driven, and Pollution Policies,” NBER working paper no. 11553. Fraumeni, Barbara M. (1997). “The Measurement of Depreciation in the U.S. National Income and Product Accounts,” Survey of Current Business, 77(7):7–23.
180
Metcalf
General Accounting Office (1997). “Effects of the Alcohol Fuels Tax Incentives,” U.S. General Accounting Office, GAO/GGD-97–41. General Accounting Office (2005). “National Energy Policy: Inventory of Major Federal Energy Programs and Status of Policy Recommendations,” U.S. General Accounting Office, GAO–05–379. Greene, David L., Donald W. Jones, and Paul N. Leiby (1998). “The Outlook for U.S. Oil Dependence,” Energy Policy, 26(1): 55–69. Greene, David L., Philip D. Patterson, Margaret Singh, and Jia Li (2005). “Feebates, Rebates and Gas-Guzzler Taxes: A Study of Incentives for Increased Fuel Economy,” Energy Policy, 33:757–775. Hassett, Kevin, and Gilbert E. Metcalf (1995). “Energy Tax Credits and Residential Conservation Investment: Evidence from Panel Data,” Journal of Public Economics, 57:201–217. Jackson, Pamela J. (2006). “The Federal Excise Tax on Gasoline and the Highway Trust Fund: A Short History,” Congressional Research Service, RL30304. Karp, Larry, and David M. Newbery (1991). “Optimal Tariffs on Exhaustible Resources,” Journal of International Economics, 30(2): 285–299. Lazzari, Salvatore (1990). “The Windfall Profit Tax on Crude Oil: Overview of the Issues,” Congressional Research Service, 90–442 E. Lazzari, Salvatore (2005). “Energy Tax Policy: An Economic Analysis,” Congressional Research Service, RL30406. Lazzari, Salvatore (2006). “Energy Tax Policy,” Congressional Research Service, CRS Issue Brief IB10054. Metcalf, Gilbert E. (2005). “Tax Reform and Environmental Taxation,” NBER working paper no. 11665. Metcalf, Gilbert E. (2006). “Energy Conservation in the United States: Understanding Its Role in Climate Policy,” NBER working paper no. 12272. Newbery, David M. (1976). “A Paradox in Tax Theory: Optimal Tariffs on Exhaustible Resources,” Department of Economics, Stanford University, SEER Technical Paper. Newbery, David M. (2005). “Why Tax Energy? Towards a More Rational Policy,” The Energy Journal, 26(3):1–39. Palmer, Karen, and Dallas Burtraw (2005). “Cost-Effectiveness of Renewable Electricity Policies,” Resources For The Future, RFF DP05–01. Parry, Ian, and Kenneth A. Small (2005). “Does Britain or the United States Have the Right Gasoline Tax?” American Economic Review, 95:1276–1289. Pizer, William, Dallas Burtraw, Winston Harrington, Richard Newell, and James Sanchirico (2006). “Modeling Economy-Wide Vs Sectoral Climate Policies Using Combined Aggregate-Sectoral Models,” The Energy Journal, 27(3):135–168. Poterba, James M. (1991). “Is the Gasoline Tax Regressive?” Tax Policy and the Economy, 5: 145–164. The Royal Academy of Engineering (2004). “The Costs of Generating Electricity,” The Royal Academy of Engineering.
Federal Tax Policy towards Energy
181
Sekar, Ram C., John E. Parsons, Howard J. Herzog, and Henry D. Jacoby (2005). “Future Carbon Regulations and Current Investments in Alternative Coal-Fired Power Plant Designs,” MIT Joint Program on the Science and Policy of Global Change, No. 129. Tolley, George, and Donald Jones (2004). “The Economic Future of Nuclear Power,” University of Chicago. U.S. Census Bureau (2006). Statistical Abstract of the United States. Washington, DC: Government Printing Office. U.S. Energy Information Administration (2004). “Analysis of Five Selected Tax Provisions of the Conference Energy Bill of 2003,” Office of Integrated Analysis and Forecasting. U.S. Energy Information Administration (2005). “Annual Energy Review 2004,” Energy Information Administration. U.S. Energy Information Administration (2006). “Assumptions to the Annual Energy Outlook 2006,” Energy Information Administration, DOE/EIA-0445(2006). U.S. Internal Revenue Service (2006). “How to Depreciate Property,” Internal Revenue Service, Publication 946.
Appendix. Calculating the Levelized Cost of Electric Power Generating Plants The levelized cost of an electricity generating power plant is the price per kWh that the plant must receive for electricity sold that will cover all costs of production including a return to equity holders. I construct levelized costs for various technologies for a hypothetical power plant with a 1,000 MW capacity.52 Table 5.A1 provides a full list of the plant-specific parameters used in the analysis. In addition, I assumed an inflation rate of 3 percent, a combined federal and state tax rate of 40 percent, a nominal bond return of 8 percent and a nominal return to equity of 15 percent. Most plant-specific parameter values are taken from the U.S. Energy Information Administration (2006) Annual Energy Outlook (AEO). Fuel costs are based on projected fuel costs from the AEO which assumes real growth in fuel prices over the 40 year expected life of all power plants. The debt-equity ratio for nuclear and coal is based on assumptions in Deutch and Moniz (2003) and I assume the same ratio for renewables as for coal. Periodic capital spending is required for all technologies for capital additions and upgrades. AEO assumes increased capital upgrade spending in the last ten years of the plant’s life. For simplicity I treat these expenditures as operating expenditures rather than capitalize them over the remaining life of the plant. The economic life of the various plants is taken from The Royal Academy of Engineering (2004). I also use this source for the capacity factor for wind based on its analysis of wind production in Europe.
182
Table 5.A1 Plant Specific Parameters for Levelized Cost Analysis
Nuclear
Coal PC
Coal IGCC
Gas-CC
Biomass
Wind
Solar Thermal
PV
Capacity factor
85%
85%
85%
85%
83%
35%
31%
21%
Construction time
6
4
4
3
4
3
3
2
Fuel cost ($/MMbtu)
0.47
0.994
0.994
5.94
2.15
0
0
0
Heat rate (BTU/kWh)
10,400
8,844
8,309
7,196
8,911
10,280
10,280
10,280
fixed O&M ($/kW/yr)
61.82
25.07
35.21
11.37
48.56
27.59
51.70
10.64
variable O&M ($/kWh)
0.00045
0.00418
0.00265
0.00188
0.00313
0
0
0
Decommissioning ($million)
350
na
na
na
na
na
na
na
Capital increment ($/kW)
18
15
15
6
0
0
0
0
K Increment (yrs 30+)
44
21
21
12
0
0
0
0
% Debt finance
50%
60%
60%
60%
60%
60%
60%
60%
% Equity finance
50%
40%
40%
40%
40%
40%
40%
40%
Discount rate
11.5%
10.8%
10.8%
10.8%
10.8%
10.8%
10.8%
10.8%
Overnight cost ($/kW)
2,014
1,249
1,443
584
1,809
1,167
3,047
4,598
Economic life
40
30
25
25
20
20
20
20
MACRS life
15
15
15
15
5
5
5
5
Metcalf
Federal Tax Policy towards Energy
183
The steps required for this calculation are: 1. Compute the present discounted value of costs in each year over life of a project. This includes all capital and operating costs net of tax deductions. 2. Sum all costs over life of project. This is the present discounted value of the project’s costs. 3. Compute the amount of constant real before-tax revenue required each year that will equal the total present discounted value of costs over the life of the project. 4. Divide this required revenue value by total kilowatt-hours produced by plant to obtain a cost per kWh. Construction time differs across the technologies. I assume construction costs follow a sinusoidal pattern (increasing, peaking and then declining) over the time period with plant construction beginning in 2005. To illustrate how construction costs are handled, consider a nuclear power plant built over six years. Table 5.A2 provides the data for an overnight cost of $2,014 per kW. The overnight cost for the plant is distributed over the six year construction period with construction costs peaking mid-way through the construction period. Spending is converted to nominal dollars using the last year of construction as the base year. Nominal cash flows are discounted using the firm’s discount rate (column 3). These discounted numbers enter the summation in step 2. Once the plant begins operation fixed and operating costs (including fuel costs, maintenance, nuclear decommissioning costs, capital increments, bond interest payments and taxes) are summed and then discounted to year zero values. Depreciation and bond interest costs are allowed as a tax deduction and so reduce the costs by the value of the tax shield (tax rate times deduction). The sum of the present discounted costs is converted to a real levelized cost that is a constant real annual payment by the firm to cover all costs. For the nuclear power plant, the constant real annual before-tax revenue required to Table 5.A2 Construction Costs for Nuclear Power Plant Year
Real
Nominal
Discounted
1
135
116
203
2
369
327
511
3
503
461
643
4
503
475
593
5
369
358
400
6
135
135
135
2014
1872
2484
Total
184
Metcalf
match the sum of the present discounted costs (including the value of the tax shields) over the life of the plant is $321 million per year. Finally in step 4 this is converted to a cost per kWh. Total annual production for the plant is the number of hours in the year times its capacity factor. Dividing this into the annual levelized cost yields a cost per kWh. Based on a capacity factor of 85 percent, the annual $321 million cost translates to a real levelized cost of 4.31¢ per kWh.
