TAX-PREFERRED ASSETS AND DEBT, AND THE TAX REFORM ACT OF 1986: SOME IMPLICATIONS FOR FUNDAMENTAL TAX REFORM ERIC M. ENGEN * & WILLIAM G.

Transcription

1 TAX-PREFERRED ASSETS AND DEBT, AND THE TAX REFORM ACT OF 1986: SOME IMPLICATIONS FOR FUNDAMENTAL TAX REFORM ERIC M. ENGEN * & WILLIAM G. GALE ** Abstract - This paper focuses on two aspects of the tax changes enacted in the Tax Reform Act of 1986 (TRA86). First, the TRA86 phased out tax deductions for interest on consumer debt, which contributed to a marked shift in the composition of household debt. Second, the TRA86 restricted the tax deductibility of contributions to individual retirement accounts (IRAs) for some higher-income households. This appears to have contributed to, but is not solely responsible for, the shift in the composition of some households portfolios of tax-preferred saving incentive plans. This paper also discusses the interaction of household debt and 401(k) plans following the TRA86. The aspects of the TRA86 focused upon in this paper appear to represent examples of more typical responses to tax changes: changes in the composition of economic activity but with little change in the real level of * Board of Governors of the Federal Reserve System, Washington, D.C ** The Brookings Institution, Washington, D.C economic activity. This conclusion is consistent with the hierarchy of taxpayer responses suggested by Slemrod (1990, 1995) and yields some potentially relevant implications for fundamental tax reform. INTRODUCTION The current debate on fundamental tax reform is often described as merely a choice between an income tax and a consumption tax. 1 This characterization can easily lead to the general conclusion that the effects of the Tax Reform Act of 1986 (TRA86) have little to offer in the way of guidance concerning the effects of a fundamental tax reform. The TRA86 enacted broad changes in the personal and corporate income taxes by reducing marginal tax rates and by broadening the tax base via a reduction in tax preferences, but did not embody a fundamental shift from an incomebased tax toward a consumption tax. 331

2 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 While the shift to a consumption base is an important part of fundamental tax reform proposals, it may not be the only relevant issue. For example, the flat tax, which expenses investment at the business level, is appropriately considered a consumption-based tax. If the flat tax instead provided depreciation allowances for investment, rather than expensing, it would be an income-based tax (Slemrod, 1995b). Our conjecture is that, for a number of issues involved with tax reform, a switch to a flat-taxwith-expensing might not lead to dramatically different conclusions from switching to a flat-tax-with-depreciation. 2 If so, then at least for some issues, the crucial point may not be the distinction between an income tax base and a consumption tax base, but rather the desirability of broadening the current hybrid income-consumption tax base whether the ultimate base is consumption or income and flattening the current increasing tax rate structure. The latter two issues, however, are exactly the focus of the changes in the TRA86, and potentially make the effects of TRA86 relevant for understanding at least some aspects of the debate on fundamental tax reform. In this paper, we focus on two aspects of the tax changes enacted in the TRA86 and discuss some of their effects on households saving and debt decisions. First, TRA86 phased out tax deductions for interest on consumer debt. This contributed to a marked shift in the composition of household debt away from consumer borrowing and toward mortgages, the interest on which remained tax deductible. Second, TRA86 restricted the tax deductibility of contributions to Individual Retirement Accounts (IRAs) for higher-income households with a retirement plan; deductibility is phased out (eliminated) when adjusted gross income exceeds $40,000 ($50,000) for joint taxfilers and $25,000 ($35,000) for single taxfilers. This appears to have contributed to, but is not solely responsible for, the shift in the composition of the portion of households asset portfolios allocated to tax-preferred saving incentive plans, such as IRAs and 401(k) plans. Before the TRA86, IRAs constituted the majority of total saving incentive contributions. However, IRA contributions plummeted following the restrictions imposed by the TRA86. By 1992, IRA contributions comprised less than one-fifth of total saving incentive contributions, while contributions to 401(k) plans grew to constitute about three-quarters of total saving incentive contributions. Household debt and contributions to tax-based saving incentives can also interact in important ways. It is important to recall that household saving and wealth are ultimately net concepts assets minus debt. Hence, studies should also focus on how tax changes affect household wealth, rather than just assets or debt. However, these two components of household wealth, and the effects of tax changes on household asset accumulation or household debt, typically have been analyzed separately. This paper discusses the interaction of household debt and tax-preferred saving following the TRA86. The aspects of the TRA86 focused upon in this paper appear to represent examples of more typical responses to tax changes. They represent primarily changes in the composition of economic activity but with little change in the real level of economic activity. This conclusion is consistent with the hierarchy of taxpayer responses suggested by Slemrod (1990, 1995a) and yields some potentially relevant implications for fundamental tax reform. 332

3 The remainder of the paper is organized as follows. The next section examines trends in debt before and after the TRA86. The subsequent section analyzes interactions between IRAs and 401(k) plans following the tax changes in the TRA86. The following section examines the interaction between households accumulation of debt and their accumulation of assets, particularly tax-preferred assets, in the period following the TRA86. The final section concludes with a discussion of some implications for the design and implementation of fundamental tax reform. TAX REFORM AND THE ALLOCATION OF HOUSEHOLD DEBT Since the early 1980s, the level of debt held by U.S. households has risen significantly. Data on aggregate trends in household debt holdings compiled by the Federal Reserve Board (1995) show that nominal household debt, defined as the sum of mortgage and consumer debt, grew at an annual rate of 12 percent in the 1970s, 10 percent between 1980 and 1986, and 8 percent since This growth in debt outpaced inflation, which (measured by the CPI-U-X1) averaged 7.1 percent in the 1970s, 4.9 percent between 1980 and , and 3.8 percent from 1986 to Moreover, household debt has undergone long-term growth relative to disposable personal income or relative to 1 household financial and housing assets. Mortgage debt defined by the presence of a residence as collateral, rather than by the use of the loan proceeds grew at annual rates of 12.7 percent in the 1970s, 9.7 percent between 1980 and 1986, and 9.1 percent from 1986 to Relative to the value of owner-occupied real estate, mortgage debt rose from 27.5 percent in 1980 to 31.6 percent in 1986 to 42.8 percent by Consumer debt also grew rapidly over much of this period, rising at an annual rate of 10.3 percent in the 1980s and 10.8 percent from 1980 to The TRA86 phased out deductions for consumer interest payments, but retained the deduction for mortgage interest for almost all taxpayers. 4 This reduced the price of borrowing through mortgages relative to consumer debt. From 1986 to 1994, consumer debt outstanding grew at only 5.1 percent per year, and fell from 21 percent of disposable personal income in 1986 to less than 19 percent in 1991 through The composition of household debt shifted toward mortgages during the 1970s housing boom, was roughly constant 2 from 1980 to 1986, and has shifted further toward mortgages since 1986, even with flat housing prices. An increase in mortgage lending can occur through increased home equity lending, refinancing of first mortgages that extract equity, and higher loan-tovalue ratios on new purchases of homes. Manchester and Poterba (1989) cite data showing that home equity loans rose from 3.6 percent of outstanding mortgages in 1980 to 10.8 percent in The U.S. General Accounting Office (1993) cites different data with similar ratios: 3.8 percent in 1981, 10.6 percent in 1986, and 12.2 percent in Most of the growth in home equity lines of credit occurred after 1986: balances rose from $1 billion in 1981 to $32 billion in 1986 and $132 billion in Eugeni (1993) estimates that outstanding home equity balances rose from $268 billion in 1988 to about $469 billion in Canner, Fergus, and Luckett (1989) estimate outstanding home equity balances in 1988 of between $210 and $265 billion. 333

4 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 These different studies are roughly consistent and suggest rapid growth of home equity lending in the 1980s. 5 Home equity lending can account for a sizable portion of the increase in mortgage debt before 1986, but a smaller portion of the increase after Nonetheless, in the post-86 period, home equity lending may have played an important role in enabling households to change the composition of their debt. Mortgage refinancing represents an alternative way to extract housing equity. In a 1989 survey reported in Canner and Luckett (1990), 20 percent of households with mortgages had refinanced their mortgage. Of those, 57 percent had extracted equity at the time of refinancing. The mean and median amounts extracted were $25,145 and $15,941. These are roughly similar to the mean and median for home equity loans, $22,534 and $15,905. Higher initial loan-to-value ratios (i.e., smaller down payments) are another channel through which mortgage debt can rise relative to the house value backing the debt. Data from the Federal Housing Finance Board (1995) shows that aggregate loan-to-value ratios on purchase money mortgages (i.e., primary mortgages that are not refinancings) rose from 72.9 percent in 1980 to 77 percent in 1984, fell to 74.4 percent by 1991, and then rose to 79.9 percent in Similar time patterns occur for the proportion of new loans with loan-to-price ratios that exceed 90 percent. Skinner and Feenberg (1990) use a panel of taxpayers from 1984 through 1987 who itemize their tax deductions. Their estimates imply that limiting the deductibility of consumer interest in TRA86 significantly reduced consumer interest payments. They also find that a dollar of reduced consumer interest is associated with a 67 to 86 cents increase in mortgage interest payments. Using interest rate data from Canner, Fergus, and Luckett (1988), these estimates 3 suggest that a dollar in reduced consumer debt raised mortgage debt by between $0.87 and $ Scholz (1994) uses cross sections from the 1983 and 1989 Surveys of Consumer Finances. He finds that household debt rose between 1983 and Among higher-income homeowners, property-backed debt rose rapidly. Among low-income renters, who presumably had less access to deductible borrowing after the TRA86, debt holdings fell. Maki (1995) uses successive cross sections of the Consumer Expenditure Survey. He shows that high income homeowners reduced their consumer debt and raised their mortgage debt after the TRA86, relative to other households. He also shows that highincome renters did not reduce their consumer debt holdings relative to other renters and that households increased their total debt. Engen and Gale (1995) used successive cross sections of the Survey of Income and Program Participation (SIPP) in 1984, 1987, and 1991 to look at household debt. Mortgage debt rose across all income and age groups. Median mortgage debt rose by 14 percent from 1984 to 1987 and by an additional 37 percent from 1987 to Virtually all of this increase was due to an increase in the ratio of mortgage debt to house value, rather than changes in house value. 7 From 1987 to 1991, homeowners with family earnings above $50,000 reduced their holdings of consumer debt, while 334

5 renters in the same earnings classes raised their consumer debt holdings. These studies have important implications for studying household saving behavior and its response to tax changes. First, these studies are consistent with the conclusion that the composition of household debt portfolios is sensitive to the tax treatment of different types of debt. Second, the trends suggest the importance of analyzing the effects of tax reforms on broad measures of household wealth which include liabilities as well as assets. TAX REFORM AND THE ALLOCATION OF TAX-PREFERRED SAVING In this section, we examine the relationship between IRAs and 401(k) plans. Since the early 1980s, tax-based saving incentives plans, such as IRAs and 401(k) plans, gained tremendous popularity. 8 Total contributions to IRAs, 401(k)s, SEPs, and Keoghs increased from less than 4 percent of personal saving in 1981 to about 35 percent of personal saving by 1985, and then continued to comprise about one-third of personal saving through However, the restrictions imposed by the TRA86 on deductible IRA contributions significantly changed the composition of saving incentive contributions. Before the TRA86, IRAs constituted the majority of total saving incentive contributions. However, contributions plummeted following the change in the treatment of IRAs enacted in TRA86. By 1992, IRA contributions comprised less than 20 percent of total saving incentive contributions. In contrast, contributions to 401(k) plans grew steadily since their tax treatment was clarified in 1981 and constituted about three-quarters of total saving incentive contributions by IRAs and 401(k) plans are unlikely to be perfect substitutes. 401(k) plans are tied to the work place, while IRAs are not. 401(k)s usually have employer matching of employee contributions, while IRAs do not. 401(k)s and IRAs generally have different contribution limits, withdrawal and loan provisions, and investment opportunities. 401(k) contributions tend to be regular salary reductions; IRAs can be funded anytime. For example, people may want to put some money into a 401(k) but then keep the flexibility of contributing more to an IRA at a later date in the tax year, depending on their income and expenses. Nevertheless, IRAs and 401(k)s could very well be good substitutes for some people (particularly those who hold both 401(k)s and IRAs). The plans represent alternative ways to save for retirement, and 401(k)s should be particularly attractive for high-income households after the removal of tax deductibility for IRA contributions in One way to test for substitutability is to exploit the removal of tax deductibility of IRA contributions for some higher income families in the TRA86. If households find IRAs and 401(k)s to be substitutes, controlling for other factors, the decline in IRA participation after TRA86 should be larger for 401(k) eligible households than for noneligible households among income groups who lost deductibility of IRA contributions. Poterba, Venti, and Wise (1992) present evidence related to this issue using data from the Employee Benefits Supplement of the 1988 Current Population Survey. They find that for families with incomes above $75,000, the IRA participation rate dropped substantially more for 401(k) eligibles than for noneligibles. In this group, the IRA participation rate fell by 57 percentage points for eligibles compared to 27 percentage points for ineligibles. 335

6 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 Engen, Gale, and Scholz (1994) examined the relationship between IRAs and 401(k)s after TRA86, using data from the SIPP. If 401(k)s and IRAs are substitutes, 401(k)-eligible families whose IRA deductibility was restricted in 1986 should funnel more money into 401(k) plans from 1987 to 1991 than eligible families whose deductibility was not removed, controlling for other factors such as income and net worth. They find that among 401(k)-eligible families, not being allowed to make deductible IRA contributions raised 401(k) balances by $2,538 from 1987 to This result is consistent with the hypothesis that there is substitution between 401(k) plans and IRAs. Moreover, they note that the average real 401(k) balance among eligibles in the effected group rose by about $4,300 between 1987 and Hence, the removal of deductibility potentially can account for half of the increase in 401(k) balances in the higher-income group affected by the removal of IRA tax deducibility in the TRA86. Thus, there exists some evidence implying that some households substituted away from IRAs into 401(k)s following the TRA86. The TRA86 probably contributed to, but is almost certainly not solely responsible for, the shift in the composition of households holdings of tax-preferred financial assets. This is consistent with the conclusion that the composition of household asset portfolios is sensitive to the tax treatment of different types of assets. INTERACTIONS BETWEEN ASSETS AND DEBT Engen and Gale (1995) examined interactions between household assets and liabilities by examining the effects of eligibility for, and participation in, 401(k) plans on households accumulation of wealth, where wealth is defined as the sum of net financial assets (financial assets minus consumer debt) and housing equity (house value minus morgage debt). Their analysis used data from successive cross sections in 1987 and 1991 of the SIPP. They obtained several results that point to important interactions between household assets and liabilities following the TRA86. 9 First, between 1987 and 1991, controlling for IRA status and other factors, households with at least one worker eligible for a 401(k) plan did accumulate more financial assets and net financial assets than observationally equivalent households that did not contain a 401(k)-eligible worker. 10 But they also show that 401(k)-eligible households did not accumulate more wealth when housing equity is included. House values rose for eligibles relative to noneligibles, but mortgage debt grew even faster, so that housing equity fell for 401(k)- eligible households relative to ineligible households. Second, they split the sample into homeowners and renters. For renters, being eligible for a 401(k) has no effect on holdings of financial assets or net financial assets. For homeowners, 401(k) eligibility raises gross financial assets and net financial assets, but has no effect on a broader measure of wealth defined to include housing equity. Third, similar results apply to the effects of 401(k) participation in a set of analogous tests that control for IRA status and other factors. The results imply that groups of families that were eligible for 401(k)s or participated in 401(k)s and that had access to mortgage debt, which was still tax deductible after TRA86, raised their financial assets and net financial assets relative to other groups; but this 336

7 increase in one component of their portfolio was fully offset by reductions in housing equity, relative to the other groups. 401(k)-eligible families, or participants, that did not have access to tax-preferred debt after the TRA86 did not raise their financial or net financial assets relative to other groups. In summary, we find that no group of families that was eligible for, or participated in, 401(k)s significantly raised its wealth (including housing equity) from 1987 to 1991 relative to the observationally equivalent group of families that was not eligible or did not participate. In showing that borrowing and saving at the household level can interact in important ways, these results have direct implications for the effects of tax changes on household portfolios and saving. DISCUSSION Slemrod (1990, 1995a) has categorized a hierarchy of taxpayer responses to tax reform. The behavioral response most sensitive to tax reform involves the timing of economic transactions; altering the timing of economic activity in reponse to permanent variations in the time pattern of tax rates or in anticipation of future changes in the tax law. The next most sensitive type of taxpayer response involves avoidance activities, including the reallocation of asset and debt portfolios outlined above. The smallest responses should be expected in real economic behavior: saving, labor supply, etc. 11 The evidence discussed in this paper is consistent with this hierarchy. We note as well that, typically, the timing and avoidance effects are unintended consequences of tax reform, while the real effects are intended. This suggests some implications for the design and implementation of fundamental tax reform. First, the anticipation of tax reform may have important effects on economic activity that serve to offset some of the intended effects. Virtually all analyses of fundamental tax reform assume that the new system is unanticipated and is imposed immediately. This eliminates the possibility of timing responses in the economic models being used, but does not eliminate those responses in the real world. Second, if a fundamental reform from the current tax system to a consumption-based tax includes transitional relief for existing capital, then this will create differential tax treatment across assets. The studies outlined above suggest that these tax differences could create large reshuffling of portfolios. A case in point is the mortgage interest deduction. If the deduction is removed, then we expect households to substitute out of mortgage debt, either by raising their invest- ment debt (Maki, 1996) or by paying down their loan with accumulated assets. If the mortgage deduction is retained in a switch to a consumptionbased tax, then tax reform will have created a conduit for tax-deductible borrowing in a system that does not tax interest income. Creating such arbitrage possibilities could lead to large shifting into mortgages. A final note of caution is warranted. The top few percent of households, by income or wealth, account for a very large share of the net wealth accumulation in the United States. There is little conclusive evidence on how such house holds responded to the TRA86 (Slemrod, 1995a), but their impact on assets, debt, and saving in any fundamental tax reform proposal could prove decisive. ENDNOTES We thank Jane Gravelle for her comments on this paper. All opinions are our own and should not be 337

8 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 ascribed to the staff, officers, or trustees of the Federal Reserve Board or The Brookings Institution. 1 A third possibility is a wage tax. Because future consumption may be financed only from existing assets and future wage income, a consumption tax that allows for transitional tax relief on consumption financed out of existing assets is essentially a wage tax. 2 Note also that a consumption-based flat tax would only exempt the portion of the overall return to capital that reflects the opportunity cost of capital. The portion of the return to capital that reflects rents would be taxed in the same way under both a consumption-based and an income-based flat tax (Bradford, 1996). 3 See Engen and Gale (1995) for a more detailed discussion of these trends in debt. 4 The phaseout of deductibility of consumer interest occurred over five years. Deductibility was retained on debt backed by a taxpayer s first or second residence, up to the taxpayer s basis in the property. The Omnibus Budget Reconciliation Act of 1987 restricted deductibility of mortgage interest to the first $1 million of acquisition debt and the first $100,000 of home equity loans. Maki (1995) provides further information. 5 Besides tax policy, the growth in home equity lending has been attributed also to changes in home values, changes in banking laws, and other factors. See GAO (1993) and Canner, Fergus, and Luckett (1988, 1989). 6 Canner, Fergus, and Luckett (1988) show average rates in 1984 through 1987 of 11.6 percent for home equity lines of credit, 18.4 percent for credit card debt, 12.1 percent for new 48-month car loans, and 15.3 percent for 24-month personal loans. If the average rate on consumer debt (the last three categories) is 15 percent, the estimates in Skinner and Feenberg (1990) suggest that a dollar reduction in consumer interest corresponds to a $6.67 decline in consumer debt. At an 11.6 percent rate on home equity loans, a 67 (86) cent increase in mortgage interest corresponds to an increase of $5.78 ($7.41) in mortgage debt. Hence, reducing consumer debt by a dollar would raise mortgage debt by between 87 cents (5.78/ 6.67) and $1.11 (7.41/6.67). 7 The median house value (1991 dollars) in the SIPP was about $77,000 in 1984, $79,000 in 1987, and $75,000 in These patterns are consistent with data reported by Poterba (1991) on U.S. housing prices. The mean ratio of mortgage to house value in the SIPP rose 9 percent from 1984 to 1987 and 24 percent from 1987 to See Engen, Gale, and Scholz (1994, 1996) for a more detailed discussion of these trends in taxbased saving incentive plans. 9 See Engen, Gale, and Scholz (1996) also for more discussion of the interaction between households contributions to tax-preferred 401(k) plans and their accumulation of tax-preferred mortgage debt. 10 Poterba, Venti, and Wise (1995) obtained similar findings for gross financial assets using similar data from the SIPP. 11 Engen (1994) and Engen and Gale (1996) explicitly analyze the potential real effects of fundamental tax reform on household saving. These studies conclude that switching to a consumption-based tax would modestly increase household saving. In assessing the effects of fundamental tax reform, Engen and Gale (1996) note the importance of realizing that in the current hybrid incomeconsumption tax, much of household saving is already done through tax-preferred forms such as pensions, saving incentive acounts, and owneroccupied housing. Sabelhaus (1996) documents the importance of tax-preferred saving in aggregate household saving. REFERENCES Board of Governors of the Federal Reserve System. Balance Sheets for the U.S. Economy Washington, D.C.: Board of Governors of the Federal Reserve System, June, Bradford, David F. Consumption Taxes: Some Fundamental Transition Issues. In Frontiers of Tax Reform edited by Michael J. Boskin. Stanford: Hoover Institution Press, Canner, Glenn B., James T. Fergus, and Charles A. Luckett. Home Equity Lines of Credit. Federal Reserve Bulletin 74 (June, 1988): Canner, Glenn B., James T. Fergus, and Charles A. Luckett. Home Equity Lending. Federal Reserve Bulletin 75 (May, 1989): Canner, Glenn B., and Charles A. Luckett. Mortgage Refinancing. Federal Reserve Bulletin 76 (August, 1990): Engen, Eric M. Precautionary Saving and the Structure of Taxation. Federal Reserve Board. Mimeo, Engen, Eric M., and William G. Gale. Debt, Taxes, and the Effects of 401(k) Plans on Household Wealth Accumulation. Federal Reserve Board and The Brookings Institution. Mimeo, Engen, Eric M., and William G. Gale. The Effects of Fundamental Tax Reform on Saving. In The Economic Effects of Fundamental Tax Reform, edited by Henry Aaron and William Gale. Washington, D.C.: The Brookings Institution, 1996 (forthcoming). Engen, Eric M., William G. Gale, and John Karl Scholz. Do Saving Incentives Work? Brookings Papers on Economic Activity 1 (1994):

9 Engen, Eric M., William G. Gale, and John Karl Scholz. Effects of Tax-Based Saving Incentives on Saving and Wealth: A Critical Review of the Literature. Federal Reserve Board, The Brookings Institution, and the University of Wisconsin Madison. Mimeo, Eugeni, Francesca. Consumer Debt and Home Equity Borrowing. Economic Perspective 17 No. 2 (March April, 1993): Federal Housing Finance Board. News. FHFB 95-55, Washington, D.C., September, Maki, Dean M. Household Debt and the Tax Reform Act of Federal Reserve Board. Mimeo, Maki, Dean M. Portfolio Shuffling and Tax Reform. National Tax Journal 49 No. 3 (September, 1996): Manchester, Joyce M., and James M. Poterba. Second Mortgages and Household Saving. Regional Science and Urban Economics 19 No. 2 (May, 1989): Poterba, James M. House Price Dynamics: The Role of Tax Policy and Demography. Brookings Papers on Economics Activity 2 (1991): Poterba, James M., Steven F. Venti, and David A. Wise. 401(k) Plans and Tax-Deferred Saving. NBER Working Paper No Cambridge, MA: National Bureau of Economic Research, Poterba, James M., Steven F. Venti, and David A. Wise. Do 401(k) Contributions Crowd Out Other Personal Saving? Journal of Public Economics 58 (September, 1995): Sabelhaus, John. Public Policy and Saving Behavior in the U.S. and Canada. Congressional Budget Office. Mimeo, Scholz, John Karl. Tax Progressivity and Household Portfolios: Descriptive Evidence from the Surveys of Consumer Finances. In Tax Progressivity and Income Inequality, edited by Joel Slemrod, Cambridge: Cambridge University Press, Skinner, Jonathan, and Daniel Feenberg. The Impact of the 1986 Tax Reform on Personal Saving. In Do Taxes Matter?: The Impact of the Tax Reform Act of 1986, edited by Joel Slemrod, Cambridge, MA: MIT Press, Slemrod, Joel. The Economic Impact of the Tax Reform Act of In Do Taxes Matter?: The Impact of the Tax Reform Act of 1986, edited by Joel Slemrod, Cambridge, MA: MIT Press, Slemrod, Joel. Income Creation or Income Shifting? Behavioral Responses to the Tax Reform Act of American Economic Review 85 No. 2 (May, 1995a): Slemrod, Joel. What Makes Some Consumption Taxes So Simple, and Others So Complicated? Paper presented at the Conference on Fundamental Tax Reform, sponsored by the Center for Economic Policy Research at Stanford University, Stanford, CA, 1995b. U.S. General Accounting Office. Tax Policy: Many Factors Contributed to the Growth in Home Equity Financing in the 1980 s. GAO/ GGD-93-63, Washington, D.C., March,

10 TAX-PREFERRED ASSETS AND DEBT, AND THE TAX REFORM ACT OF 1986: SOME IMPLICATIONS FOR FUNDAMENTAL TAX REFORM ERIC M. ENGEN * & WILLIAM G. GALE ** Abstract - This paper focuses on two aspects of the tax changes enacted in the Tax Reform Act of 1986 (TRA86). First, the TRA86 phased out tax deductions for interest on consumer debt, which contributed to a marked shift in the composition of household debt. Second, the TRA86 restricted the tax deductibility of contributions to individual retirement accounts (IRAs) for some higher-income households. This appears to have contributed to, but is not solely responsible for, the shift in the composition of some households portfolios of tax-preferred saving incentive plans. This paper also discusses the interaction of household debt and 401(k) plans following the TRA86. The aspects of the TRA86 focused upon in this paper appear to represent examples of more typical responses to tax changes: changes in the composition of economic activity but with little change in the real level of * Board of Governors of the Federal Reserve System, Washington, D.C ** The Brookings Institution, Washington, D.C economic activity. This conclusion is consistent with the hierarchy of taxpayer responses suggested by Slemrod (1990, 1995) and yields some potentially relevant implications for fundamental tax reform. INTRODUCTION The current debate on fundamental tax reform is often described as merely a choice between an income tax and a consumption tax. 1 This characterization can easily lead to the general conclusion that the effects of the Tax Reform Act of 1986 (TRA86) have little to offer in the way of guidance concerning the effects of a fundamental tax reform. The TRA86 enacted broad changes in the personal and corporate income taxes by reducing marginal tax rates and by broadening the tax base via a reduction in tax preferences, but did not embody a fundamental shift from an incomebased tax toward a consumption tax. 331

11 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 While the shift to a consumption base is an important part of fundamental tax reform proposals, it may not be the only relevant issue. For example, the flat tax, which expenses investment at the business level, is appropriately considered a consumption-based tax. If the flat tax instead provided depreciation allowances for investment, rather than expensing, it would be an income-based tax (Slemrod, 1995b). Our conjecture is that, for a number of issues involved with tax reform, a switch to a flat-taxwith-expensing might not lead to dramatically different conclusions from switching to a flat-tax-with-depreciation. 2 If so, then at least for some issues, the crucial point may not be the distinction between an income tax base and a consumption tax base, but rather the desirability of broadening the current hybrid income-consumption tax base whether the ultimate base is consumption or income and flattening the current increasing tax rate structure. The latter two issues, however, are exactly the focus of the changes in the TRA86, and potentially make the effects of TRA86 relevant for understanding at least some aspects of the debate on fundamental tax reform. In this paper, we focus on two aspects of the tax changes enacted in the TRA86 and discuss some of their effects on households saving and debt decisions. First, TRA86 phased out tax deductions for interest on consumer debt. This contributed to a marked shift in the composition of household debt away from consumer borrowing and toward mortgages, the interest on which remained tax deductible. Second, TRA86 restricted the tax deductibility of contributions to Individual Retirement Accounts (IRAs) for higher-income households with a retirement plan; deductibility is phased out (eliminated) when adjusted gross income exceeds $40,000 ($50,000) for joint taxfilers and $25,000 ($35,000) for single taxfilers. This appears to have contributed to, but is not solely responsible for, the shift in the composition of the portion of households asset portfolios allocated to tax-preferred saving incentive plans, such as IRAs and 401(k) plans. Before the TRA86, IRAs constituted the majority of total saving incentive contributions. However, IRA contributions plummeted following the restrictions imposed by the TRA86. By 1992, IRA contributions comprised less than one-fifth of total saving incentive contributions, while contributions to 401(k) plans grew to constitute about three-quarters of total saving incentive contributions. Household debt and contributions to tax-based saving incentives can also interact in important ways. It is important to recall that household saving and wealth are ultimately net concepts assets minus debt. Hence, studies should also focus on how tax changes affect household wealth, rather than just assets or debt. However, these two components of household wealth, and the effects of tax changes on household asset accumulation or household debt, typically have been analyzed separately. This paper discusses the interaction of household debt and tax-preferred saving following the TRA86. The aspects of the TRA86 focused upon in this paper appear to represent examples of more typical responses to tax changes. They represent primarily changes in the composition of economic activity but with little change in the real level of economic activity. This conclusion is consistent with the hierarchy of taxpayer responses suggested by Slemrod (1990, 1995a) and yields some potentially relevant implications for fundamental tax reform. 332

12 The remainder of the paper is organized as follows. The next section examines trends in debt before and after the TRA86. The subsequent section analyzes interactions between IRAs and 401(k) plans following the tax changes in the TRA86. The following section examines the interaction between households accumulation of debt and their accumulation of assets, particularly tax-preferred assets, in the period following the TRA86. The final section concludes with a discussion of some implications for the design and implementation of fundamental tax reform. TAX REFORM AND THE ALLOCATION OF HOUSEHOLD DEBT Since the early 1980s, the level of debt held by U.S. households has risen significantly. Data on aggregate trends in household debt holdings compiled by the Federal Reserve Board (1995) show that nominal household debt, defined as the sum of mortgage and consumer debt, grew at an annual rate of 12 percent in the 1970s, 10 percent between 1980 and 1986, and 8 percent since This growth in debt outpaced inflation, which (measured by the CPI-U-X1) averaged 7.1 percent in the 1970s, 4.9 percent between 1980 and , and 3.8 percent from 1986 to Moreover, household debt has undergone long-term growth relative to disposable personal income or relative to 1 household financial and housing assets. Mortgage debt defined by the presence of a residence as collateral, rather than by the use of the loan proceeds grew at annual rates of 12.7 percent in the 1970s, 9.7 percent between 1980 and 1986, and 9.1 percent from 1986 to Relative to the value of owner-occupied real estate, mortgage debt rose from 27.5 percent in 1980 to 31.6 percent in 1986 to 42.8 percent by Consumer debt also grew rapidly over much of this period, rising at an annual rate of 10.3 percent in the 1980s and 10.8 percent from 1980 to The TRA86 phased out deductions for consumer interest payments, but retained the deduction for mortgage interest for almost all taxpayers. 4 This reduced the price of borrowing through mortgages relative to consumer debt. From 1986 to 1994, consumer debt outstanding grew at only 5.1 percent per year, and fell from 21 percent of disposable personal income in 1986 to less than 19 percent in 1991 through The composition of household debt shifted toward mortgages during the 1970s housing boom, was roughly constant 2 from 1980 to 1986, and has shifted further toward mortgages since 1986, even with flat housing prices. An increase in mortgage lending can occur through increased home equity lending, refinancing of first mortgages that extract equity, and higher loan-tovalue ratios on new purchases of homes. Manchester and Poterba (1989) cite data showing that home equity loans rose from 3.6 percent of outstanding mortgages in 1980 to 10.8 percent in The U.S. General Accounting Office (1993) cites different data with similar ratios: 3.8 percent in 1981, 10.6 percent in 1986, and 12.2 percent in Most of the growth in home equity lines of credit occurred after 1986: balances rose from $1 billion in 1981 to $32 billion in 1986 and $132 billion in Eugeni (1993) estimates that outstanding home equity balances rose from $268 billion in 1988 to about $469 billion in Canner, Fergus, and Luckett (1989) estimate outstanding home equity balances in 1988 of between $210 and $265 billion. 333

13 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 These different studies are roughly consistent and suggest rapid growth of home equity lending in the 1980s. 5 Home equity lending can account for a sizable portion of the increase in mortgage debt before 1986, but a smaller portion of the increase after Nonetheless, in the post-86 period, home equity lending may have played an important role in enabling households to change the composition of their debt. Mortgage refinancing represents an alternative way to extract housing equity. In a 1989 survey reported in Canner and Luckett (1990), 20 percent of households with mortgages had refinanced their mortgage. Of those, 57 percent had extracted equity at the time of refinancing. The mean and median amounts extracted were $25,145 and $15,941. These are roughly similar to the mean and median for home equity loans, $22,534 and $15,905. Higher initial loan-to-value ratios (i.e., smaller down payments) are another channel through which mortgage debt can rise relative to the house value backing the debt. Data from the Federal Housing Finance Board (1995) shows that aggregate loan-to-value ratios on purchase money mortgages (i.e., primary mortgages that are not refinancings) rose from 72.9 percent in 1980 to 77 percent in 1984, fell to 74.4 percent by 1991, and then rose to 79.9 percent in Similar time patterns occur for the proportion of new loans with loan-to-price ratios that exceed 90 percent. Skinner and Feenberg (1990) use a panel of taxpayers from 1984 through 1987 who itemize their tax deductions. Their estimates imply that limiting the deductibility of consumer interest in TRA86 significantly reduced consumer interest payments. They also find that a dollar of reduced consumer interest is associated with a 67 to 86 cents increase in mortgage interest payments. Using interest rate data from Canner, Fergus, and Luckett (1988), these estimates 3 suggest that a dollar in reduced consumer debt raised mortgage debt by between $0.87 and $ Scholz (1994) uses cross sections from the 1983 and 1989 Surveys of Consumer Finances. He finds that household debt rose between 1983 and Among higher-income homeowners, property-backed debt rose rapidly. Among low-income renters, who presumably had less access to deductible borrowing after the TRA86, debt holdings fell. Maki (1995) uses successive cross sections of the Consumer Expenditure Survey. He shows that high income homeowners reduced their consumer debt and raised their mortgage debt after the TRA86, relative to other households. He also shows that highincome renters did not reduce their consumer debt holdings relative to other renters and that households increased their total debt. Engen and Gale (1995) used successive cross sections of the Survey of Income and Program Participation (SIPP) in 1984, 1987, and 1991 to look at household debt. Mortgage debt rose across all income and age groups. Median mortgage debt rose by 14 percent from 1984 to 1987 and by an additional 37 percent from 1987 to Virtually all of this increase was due to an increase in the ratio of mortgage debt to house value, rather than changes in house value. 7 From 1987 to 1991, homeowners with family earnings above $50,000 reduced their holdings of consumer debt, while 334

14 renters in the same earnings classes raised their consumer debt holdings. These studies have important implications for studying household saving behavior and its response to tax changes. First, these studies are consistent with the conclusion that the composition of household debt portfolios is sensitive to the tax treatment of different types of debt. Second, the trends suggest the importance of analyzing the effects of tax reforms on broad measures of household wealth which include liabilities as well as assets. TAX REFORM AND THE ALLOCATION OF TAX-PREFERRED SAVING In this section, we examine the relationship between IRAs and 401(k) plans. Since the early 1980s, tax-based saving incentives plans, such as IRAs and 401(k) plans, gained tremendous popularity. 8 Total contributions to IRAs, 401(k)s, SEPs, and Keoghs increased from less than 4 percent of personal saving in 1981 to about 35 percent of personal saving by 1985, and then continued to comprise about one-third of personal saving through However, the restrictions imposed by the TRA86 on deductible IRA contributions significantly changed the composition of saving incentive contributions. Before the TRA86, IRAs constituted the majority of total saving incentive contributions. However, contributions plummeted following the change in the treatment of IRAs enacted in TRA86. By 1992, IRA contributions comprised less than 20 percent of total saving incentive contributions. In contrast, contributions to 401(k) plans grew steadily since their tax treatment was clarified in 1981 and constituted about three-quarters of total saving incentive contributions by IRAs and 401(k) plans are unlikely to be perfect substitutes. 401(k) plans are tied to the work place, while IRAs are not. 401(k)s usually have employer matching of employee contributions, while IRAs do not. 401(k)s and IRAs generally have different contribution limits, withdrawal and loan provisions, and investment opportunities. 401(k) contributions tend to be regular salary reductions; IRAs can be funded anytime. For example, people may want to put some money into a 401(k) but then keep the flexibility of contributing more to an IRA at a later date in the tax year, depending on their income and expenses. Nevertheless, IRAs and 401(k)s could very well be good substitutes for some people (particularly those who hold both 401(k)s and IRAs). The plans represent alternative ways to save for retirement, and 401(k)s should be particularly attractive for high-income households after the removal of tax deductibility for IRA contributions in One way to test for substitutability is to exploit the removal of tax deductibility of IRA contributions for some higher income families in the TRA86. If households find IRAs and 401(k)s to be substitutes, controlling for other factors, the decline in IRA participation after TRA86 should be larger for 401(k) eligible households than for noneligible households among income groups who lost deductibility of IRA contributions. Poterba, Venti, and Wise (1992) present evidence related to this issue using data from the Employee Benefits Supplement of the 1988 Current Population Survey. They find that for families with incomes above $75,000, the IRA participation rate dropped substantially more for 401(k) eligibles than for noneligibles. In this group, the IRA participation rate fell by 57 percentage points for eligibles compared to 27 percentage points for ineligibles. 335

15 NATIONAL TAX JOURNAL VOL. XLIX NO. 3 Engen, Gale, and Scholz (1994) examined the relationship between IRAs and 401(k)s after TRA86, using data from the SIPP. If 401(k)s and IRAs are substitutes, 401(k)-eligible families whose IRA deductibility was restricted in 1986 should funnel more money into 401(k) plans from 1987 to 1991 than eligible families whose deductibility was not removed, controlling for other factors such as income and net worth. They find that among 401(k)-eligible families, not being allowed to make deductible IRA contributions raised 401(k) balances by $2,538 from 1987 to This result is consistent with the hypothesis that there is substitution between 401(k) plans and IRAs. Moreover, they note that the average real 401(k) balance among eligibles in the effected group rose by about $4,300 between 1987 and Hence, the removal of deductibility potentially can account for half of the increase in 401(k) balances in the higher-income group affected by the removal of IRA tax deducibility in the TRA86. Thus, there exists some evidence implying that some households substituted away from IRAs into 401(k)s following the TRA86. The TRA86 probably contributed to, but is almost certainly not solely responsible for, the shift in the composition of households holdings of tax-preferred financial assets. This is consistent with the conclusion that the composition of household asset portfolios is sensitive to the tax treatment of different types of assets. INTERACTIONS BETWEEN ASSETS AND DEBT Engen and Gale (1995) examined interactions between household assets and liabilities by examining the effects of eligibility for, and participation in, 401(k) plans on households accumulation of wealth, where wealth is defined as the sum of net financial assets (financial assets minus consumer debt) and housing equity (house value minus morgage debt). Their analysis used data from successive cross sections in 1987 and 1991 of the SIPP. They obtained several results that point to important interactions between household assets and liabilities following the TRA86. 9 First, between 1987 and 1991, controlling for IRA status and other factors, households with at least one worker eligible for a 401(k) plan did accumulate more financial assets and net financial assets than observationally equivalent households that did not contain a 401(k)-eligible worker. 10 But they also show that 401(k)-eligible households did not accumulate more wealth when housing equity is included. House values rose for eligibles relative to noneligibles, but mortgage debt grew even faster, so that housing equity fell for 401(k)- eligible households relative to ineligible households. Second, they split the sample into homeowners and renters. For renters, being eligible for a 401(k) has no effect on holdings of financial assets or net financial assets. For homeowners, 401(k) eligibility raises gross financial assets and net financial assets, but has no effect on a broader measure of wealth defined to include housing equity. Third, similar results apply to the effects of 401(k) participation in a set of analogous tests that control for IRA status and other factors. The results imply that groups of families that were eligible for 401(k)s or participated in 401(k)s and that had access to mortgage debt, which was still tax deductible after TRA86, raised their financial assets and net financial assets relative to other groups; but this 336

16 increase in one component of their portfolio was fully offset by reductions in housing equity, relative to the other groups. 401(k)-eligible families, or participants, that did not have access to tax-preferred debt after the TRA86 did not raise their financial or net financial assets relative to other groups. In summary, we find that no group of families that was eligible for, or participated in, 401(k)s significantly raised its wealth (including housing equity) from 1987 to 1991 relative to the observationally equivalent group of families that was not eligible or did not participate. In showing that borrowing and saving at the household level can interact in important ways, these results have direct implications for the effects of tax changes on household portfolios and saving. DISCUSSION Slemrod (1990, 1995a) has categorized a hierarchy of taxpayer responses to tax reform. The behavioral response most sensitive to tax reform involves the timing of economic transactions; altering the timing of economic activity in reponse to permanent variations in the time pattern of tax rates or in anticipation of future changes in the tax law. The next most sensitive type of taxpayer response involves avoidance activities, including the reallocation of asset and debt portfolios outlined above. The smallest responses should be expected in real economic behavior: saving, labor supply, etc. 11 The evidence discussed in this paper is consistent with this hierarchy. We note as well that, typically, the timing and avoidance effects are unintended consequences of tax reform, while the real effects are intended. This suggests some implications for the design and implementation of fundamental tax reform. First, the anticipation of tax reform may have important effects on economic activity that serve to offset some of the intended effects. Virtually all analyses of fundamental tax reform assume that the new system is unanticipated and is imposed immediately. This eliminates the possibility of timing responses in the economic models being used, but does not eliminate those responses in the real world. Second, if a fundamental reform from the current tax system to a consumption-based tax includes transitional relief for existing capital, then this will create differential tax treatment across assets. The studies outlined above suggest that these tax differences could create large reshuffling of portfolios. A case in point is the mortgage interest deduction. If the deduction is removed, then we expect households to substitute out of mortgage debt, either by raising their invest- ment debt (Maki, 1996) or by paying down their loan with accumulated assets. If the mortgage deduction is retained in a switch to a consumptionbased tax, then tax reform will have created a conduit for tax-deductible borrowing in a system that does not tax interest income. Creating such arbitrage possibilities could lead to large shifting into mortgages. A final note of caution is warranted. The top few percent of households, by income or wealth, account for a very large share of the net wealth accumulation in the United States. There is little conclusive evidence on how such house holds responded to the TRA86 (Slemrod, 1995a), but their impact on assets, debt, and saving in any fundamental tax reform proposal could prove decisive. ENDNOTES We thank Jane Gravelle for her comments on this paper. All opinions are our own and should not be 337