The unprecedented surge in tax receipts beginning in fiscal

Forecasting Federal Individual Income Tax Receipts Challenges and Uncertainties in Forecasting Federal Individual Income Tax Receipts Abstract - Forecasting individual income receipts has been greatly complicated by the receipts surges. First, the persistence of the surges has made forecasting the short- and long-term effects of the recent receipts growth especially difficult. Second, collections data are crucial to updating the receipts forecast. However, the receipts surges, in combination with tax law changes that altered the timing of tax payments, have changed the relationship between current collections and underlying tax liability. Data from aggregate collections and individual income tax returns confirm this changing relationship. Ann D. Parcell U.S. Department of the Treasury, Office of Tax Analysis, Washington, D.C. 20220 National Tax Journal Vol. LII, No. 3 INTRODUCTION The unprecedented surge in tax receipts beginning in fiscal year (FY) 1996 has been great news for federal finances. The U.S. government posted a deficit of $21.9 billion in FY 1997, its lowest in 23 years. This was followed by a record surplus of $69.2 billion in FY 1998. The Administration and the Congressional Budget Office (CBO) forecast budget surpluses in FY 1999 of $79.3 and $107 billion, respectively. Growth in individual income taxes has been the driving force behind the growth in total federal tax receipts. As Figure 1 shows, individual income tax receipts have increased as a share of gross domestic product (GDP) from 7.8 percent in to 9.1 percent in 1997. This 1.3 percentage point rise in the ratio of individual receipts to GDP added approximately $100 billion in collections in FY 1997. Individual income tax receipts, both with and without the effect of capital gains taxes, show the same pattern when measured as a percent of National Income and Product Accounts (NIPA) personal income, as shown by Figure 2. As the CBO (1999) has noted, four factors explain the majority of excess, or surge, growth in individual income tax liability. The CBO defines excess growth as liability growth in excess of GDP growth. They find that from 1993 to 1997, about 10 percent of excess growth was explained by excess growth in taxable personal income, approximately 15 percent by such sources as retirement income and partnership and S Corporation income, almost 33 percent by the growth in capital gains realizations, and 40 percent by rising effective tax rates. 325

NATIONAL TAX JOURNAL Figure 1. Individual Income Taxes (FY) Figure 2. Individual Income Tax Liability Table 1 helps illustrate the effective tax rate and capital gains realizations growth rates using tax return data. The growth in the number of returns with real modified adjusted gross income (AGI) of $200,000 or more was more than 13 percent from to, more than 4 percent from to 1996, and more than 8 percent from 1996 to 1997. 1 As the number of returns reporting real modified AGI of $200,000 or more grew each year, the year-to-year wage growth reported for this class of returns was in excess of 17, 15, and 22 percent in, 1996, and 1997, respectively. 1 Modified AGI is AGI minus capital gains in AGI minus social security in AGI plus tax-exempt interest income. Real modified AGI is in levels. 326

327 Year 1996 1997 Year 1996 1997 Number Returns 13.16% 4.13% 8.11% Number Returns 1.86% 1.78% 1.72% Modified AGI 14.58% 14.45% 15.04% Modified AGI 5.41% 5.17% 6.80% TABLE 1 GROWTH IN TOTAL RETURNS AND TYPES OF INCOME FOR TAXPAYERS AGI 16.36% 21.08% 19.01% AGI 5.87% 3.71% 10.52% With Modified AGI of $200K or More ( Levels of Modified AGI) a,b Wages and Salaries 17.30% 15.18% 22.06% Taxable Interest 26.32% 8.22% 7.20% Dividends 18.84% 15.19% 9.53% Business Income 7.75% 3.76% 2.90% With Modified AGI of Less Than $200K ( Levels of Modified AGI) a,b Wages and Salaries 4.74% 4.51% 6.11% Taxable Interest 21.74% 6.72% 0.06% Dividends 13.10% 8.04% 15.28% Business Income 1.43% 4.39% 1.90% a Modified AGI is AGI minus capital gains in AGI minus social security income in AGI plus tax exempt interest income. b Figures for 1997 are preliminary data. Partnership Income 4.26% 22.18% 0.39% Partnership Income 6.42% 8.69% 6.26% S-Corp Income 10.54% 12.93% 7.78% S-Corp Income 7.40% 14.17% 4.13% Capital Gains in AGI 28.26% 68.45% 36.71% Capital Gains in AGI 12.48% 30.41% 30.93% Forecasting Federal Individual Income Tax Receipts

This class of returns also reported tremendous growth in net positive capital gains realizations of more than 28, 68, and 36 percent for those same years. Returns reporting real modified AGI of less than $200,000 also reported very strong growth in capital gains realizationsin excess of 30 percent in both 1996 and 1997. For forecasters of tax receipts, the recent surges have presented a number of new challenges. First, the persistence of the exceptional receipts growth is unprecedented. This has made it particularly difficult to distinguish between the temporary and permanent effects of the surges on future receipts. Several decades of tax receipts data suggest that big surges in receipts are short-lived because the causes of the surges are usually temporary. For example, a one-year revenue surge occurred in FY 1987 that was largely caused by a one-year surge in capital gains realizations in tax year 1986. This was the result of an impending increase in the tax rate on capital gains scheduled for tax year 1987. Likewise, a receipts surge took place in 1969 when a temporary surtax was imposed. As Figure 1 illustrates, however, recent receipts have shown three consecutive years of well-above-average growth, with a fourth strong year forecast for FY 1999. By definition, the underlying features of the surges have been equally persistent. We have had double digit growth in capital gains realizations in each year from to 1997, with only 1997 affected by a reduction in capital gains tax rates. Similarly, businesses appear to be continuing a trend toward performance-based compensation, which is suspected to have contributed substantially to the recent surges by rapidly increasing the effective tax rate. Yet both of these features of the receipts surges are historically short-term phenomena. As such, it has become increasingly difficult to disentangle the shortand long-term effects of the receipts surges on the receipts forecast. 328 NATIONAL TAX JOURNAL A second challenge for forecasters is that the surge in receipts has helped to cause large deviations in the composition of tax payments from historical norms. This has made it very difficult to interpret current tax collections. Net tax collectionswithheld taxes, estimated taxes, final payments, and refundsare crucial in informing and updating the forecast of tax receipts, especially before tax return data are available. Over the past half-decade, the composition of tax payments has changed greatly from the experience of the early 1990s. Tax collections are highly aggregated data that reflect many factors. Recent collections reflect the level of current income sources, such as those in NIPA personal income, as well as income not measured by NIPA variables, such as capital gains realizations. Current tax collections also reflect changes in the income distribution through the effective tax rate. At the same time, collections reflect changes in the rules that govern the timing of tax payments to the U.S. Treasury. All of the factors have contributed to changing payment patterns in the 1990s. As current collections are monitored, it is easy to confound changes in the timing of payments with changes in the factors that affect tax liability. This distinction, however, is never more important than when a revenue surge occurs and a receipts forecast must be updated. Forecasts of tax receipts have a number of important uses. Receipts forecasts produced by the U.S. Treasury Department and the CBO are used in official estimates of current law federal budget deficits/ surpluses. These estimates reflect the current tax law, spending rules, and overall budget guidelines. Proposed changes in federal receipts or spending can in turn be influenced by current law budget deficit/surplus estimates. Receipt forecasts also guide those who manage the federal cash flow and the national debt. Finally, there are those in the private sector and

Forecasting Federal Individual Income Tax Receipts in state governments who use the federal forecast of tax receipts for their own models of budget deficits/surpluses, cash flows, debt and equity markets, and the macroeconomy. This paper describes some of the difficulties of forecasting individual income tax receipts in a period of persistent high growth in receipts. The second section describes the basic methodology for forecasting individual income tax receipts. The third section details the effect that changes in tax law have had on recent collections data. The fourth section presents some data from a panel of taxpayers from 1991 to 1996. The fifth section is the conclusion. METHODOLOGY FOR FORECASTING INCOME TAX RECEIPTS Forecasting individual income tax receipts always begins by forecasting income tax liability. To produce a forecast of tax liability, we use a microsimulation model that is based on a sample of recent tax returns (Cilke, ). Currently, our receipts forecasting model is based on the Statistics of Income (SOI) sample of approximately 118,000 tax returns, which is weighted to reflect the income of the U.S. individual tax filing population. We have reweighted the file to match the return profile of individuals in 1996 and 1997. 2 To produce a forecast of liability for each year of the budget period, we use the macroeconomic forecast prepared for the President s latest budget. The macroeconomic forecast guides our reweighting of the base SOI file in the microsimulation model to achieve population, income, and itemized deduction growth in each of the forecast years. The macroeconomic forecast also dictates the path for inflation, which indexes many individual income tax parameters. For the FY 2000 President s budget, for example, we simulated the tax return profile for each year from 1998 to 2009 with the help of the macroeconomic forecast; we then applied a tax calculator to each simulated year of tax returns to calculate tax liability. Tax return data have very precise detail on income and deduction sources, as well as the income distribution. However, the latest return data available are usually several years old due to the lag in processing tax returns for a given year. This has been more of a problem in recent years because the composition of income and deductions and the distribution of income has changed dramatically from year to year. As a result, we have placed greater reliance in recent years on estimates of liability from the most recent collections data. Because collections liability is much more timely than tax return liability, we use this as our jump-off point for forecasting liability. Thereafter, collections liability is grown by the growth in liability from our tax return based microsimulation model. A receipts forecast is produced by estimating the cash flow of the forecast liability across fiscal years. In times of revenue surges, collections liability will exceed our microsimulation model estimate of tax liability. For the FY 2000 budget, for example, we had substantial information on 1998 collections but no tax return information. Our tax return based simulation of 1998, using prior year relationships of tax liability to macroeconomic aggregates, yielded a lower liability than what collections suggested. When collections liability exceeds the simulated tax return liability, the forecaster must make several judgments. They must decide if the additional collections represent a pure timing change (no in- 2 Although the final 1997 SOI file is not yet available, we target our model to simulate as much as we know about the 1997 tax returns. 329

NATIONAL TAX JOURNAL crease in liability), a one-year surge in liability, or a surge that will last more than one year. Because it is generally not known what is causing the surge at the time it occurs (e.g., type of income earned, change in the effective tax rate, or timing change), the interpretation of the collections surge becomes the most problematic part of the forecast. A comparison of our forecasts of recent years illustrates how much we interpreted the revenue surges as short-run phenomena. For example, Table 2 shows that in January, we had forecast individual income tax liability to be $572.6 billion. By the summer of 1996, after the April 1996 revenue surge, we had revised our liability to $588 billion. The actual collections-based liability has subsequently been calculated at $582 billion. 3 The forecasts of 1996 liability from January to January 1998 show far more striking changes. We initially forecast 1996 liability based on collections of $602.1 billion, whereas the final calculation was $661.3 billion. The actual growth rate of collections liability from to 1996 was almost triple that which we expected in January. A similar story holds for 1997 liability, which was subject to law changes (namely, the decrease in the long-term capital gains tax rate) from our January to our January 1999 forecast. Because we have a growth rate forecast of liability, revenue surges pose many forecasting difficulties in the out-years of the budget window. If the entire amount of a surge is allowed to affect the outyears, then the surge value acts as an intercept adjustment to our growth rate forecast. By contrast, the more the surge is considered temporary, the less it will affect the level of receipts in the out-years. A result of multiple year surges is that the relationship of liability to GDP moves far above its historical average. To return to a historical relationship, the growth path of liability requires some rather extreme assumptions. Figure 3 shows the growth path individual income tax liability might follow to 2001 to achieve the ratio of receipts to GDP that occurred in. Note that it requires three consecutive years of decline in the level of liability of at least two percent a year to achieve the ratio. COLLECTIONS LIABILITY AND TAX LAW CHANGES: CALENDAR YEAR (CY) 1992 TO THE PRESENT There were numerous tax law changes, including those that alter the timing of tax payments, that have affected both the level and the timing of collections through this surge period. The main recent tax law changes that affected individual income tax liability came about in the Omnibus Budget and Reconciliation Act of 1993 (OBRA 93) and the Taxpayer Relief Act of 1997 (TRA 97). OBRA 93 added two marginal tax rate brackets, of 36 and 39.6 percent, to the prior 31 percent top marginal rate, giving an incentive to higher income taxpayers who anticipated the change to shift income from 1993 into 1992. The rate on alternative minimum taxable income (AMTI) was also increased from 24 percent to a progressive rate structure of 26 and 28 percent, which was mitigated somewhat by an increase in the AMTI exclusion. 4 OBRA 93 also permanently extended the phaseout of personal exemptions and the limitation on itemized deductions (PEP and PEASE). Also, the base for Hospital Insurance (HI) taxes was increased beginning in tax year by making all HI covered earnings subject to the combined employer/employee tax rate of 2.9 3 Collections liability may be revised for several years as withheld tax collections are reclassified as individual or employment taxes. 4 On the tax relief side, there was an expansion of the earned income tax credit, which phased in from to 1996. This expansion, however, largely affected the outlay side of the budget. 330

331 Budget FY 1996 January budget FY 1996 Midsession review FY 1997 January budget FY 1997 Midsession review FY 1998 January budget TABLE 2 CY BUDGET FORECASTS VERSUS ACTUAL INCOME TAX LIABILITY (IN $ BILLIONS) Time of budget January Summer January 1996 Summer 1996 January 1997 CY Individual Income Tax Liability in $ Billions 1996 1997 1998 1999 2000 575.6 575.4 570.6 588.1 585.8 602.1 4.6% 600.4 4.3% 611.6 7.2% 618.6 5.2% 632.0 7.9% 637.1 5.8% 634.4 5.7% 642.1 5.0% 652.1 5.4% 657.3 4.0% 675.0 6.0% 670.3 5.7% 676.9 5.4% 687.8 5.5% 687.5 4.6% 716.7 6.2% 712.2 6.3% 711.7 5.1% 726.9 5.7% 719.4 4.6% 760.5 6.1% 756.3 6.2% 751.3 5.6% 769.2 5.8% 759.7 5.6% Forecasting Federal Individual Income Tax Receipts FY 1998 Midsession review Summer 1997 582.4 660.3 13.4% 711.8 7.8% 729.8 2.5% 753.4 3.2% 786.5 4.4% FY 1999 January budget January 1998 661.3 724.4 9.6% 740.9 2.3% 754.0 1.8% 777.8 3.2% FY 1999 Midsession review May 1998 735.1 11.2% 780.7 6.2% 797.0 2.1% 820.1 2.9% FY 2000 January budget January 1999 739.1 11.8% 806.8 9.2% 841.5 4.3% 863.0 2.6%

NATIONAL TAX JOURNAL Figure 3. Possible Path for Individual Income Tax Liability with a Return to 7.9 Percent of GDP by 2001 percent. This gave an incentive to taxpayers to shift earnings from into 1993. Finally, the maximum amount of social security benefits that could be included in AGI was increased from 50 to 85 percent. TRA 97 cut taxes for most individuals. TRA 97 lowered the tax rates on long-term capital gains, including a retroactive change in the rates in 1997. The law also provided tax credits for taxpayers with children age 17 and under. Tax credits became available for higher education. There were a number of new individual retirement account provisions contained in the law, although the revenue consequences of these provisions tend to be largest in future decades. There were explicit changes in the rules governing the timing of tax payments throughout the 1990s. Shortly into tax year 1992, the withholding tables were changed by an administrative action. The result was that unless taxpayers refiled a form W-4, less individual income tax was withheld per paycheck. Because the action did not relieve the tax liability, however, more liability was expected to be paid in April, 1993. In addition, in March, 1992, the prior year safe harbor rule was changed for the 332 remainder of tax year 1992 and tax year 1993. The prior year safe harbor rule generally allows taxpayers to use their prior year s tax liability, or some specified amount in excess of last year s liability, as a guide for making large enough withheld and estimated tax payments on the current year s liability to avoid an estimated tax penalty. The prior year safe harbor rules thus relieve taxpayers of having to make precise estimates of their current tax liability while giving them a way to make timely current tax payments. These rules are especially useful for taxpayers who have limited use of the withholding tax system, such as self-employed individuals, or taxpayers who have substantial amounts of unearned income on which there is little or no tax withholding. For tax year 1991, taxpayers could have paid at least 90 percent of their current year s taxes through withholding and estimated tax payments to avoid an estimated tax penalty. Alternatively, they could have paid 100 percent of their 1990 tax liability (including self-employment tax liability and subtracting any credits against tax) in withholding and timely estimated tax payments to avoid an esti-

Forecasting Federal Individual Income Tax Receipts mated tax penalty. The March, 1992 prior year safe harbor rule change, effective for tax year 1992, created a rather complicated rule that in some cases limited the use of a prior year safe harbor for taxpayers with current year AGI in excess of $75,000, and whose current year modified AGI exceeded the prior year s AGI by more than $40,000, and who had made an estimated tax payment in the three preceding years or were assessed a penalty for failure to pay estimated tax in such years. 5 The March, 1992 rule was in place for tax years 1992 and 1993 only. OBRA 93, TRA 97, and the Tax and Trade Relief Extension Act of 1998 further changed the safe harbor rule for tax years through 2003. For tax year 1998, taxpayers could have used one of three conditions to avoid an estimated tax penalty. Taxpayers could have paid at least 90 percent of their 1998 liability after credits (including self-employment taxes and any other taxes paid on a form 1040) through withholding and/or estimated tax payments. They also could have avoided an estimated tax penalty if they owed less than $1,000 in tax liability by April 15, 1999. 6 Finally, they could have avoided penalty if they had paid at least 100 percent of 1997 liability through withholding and/or estimated tax payments (the so-called safe harbor rule). For tax years 7, taxpayers had to pay at least 110 percent of the prior year s liability to meet the safe harbor if their prior year s AGI was more than $150,000. For tax years 1999, 2000, 2001, and 2002, the prior year safe harbors for taxpayers with prior year AGI of more than $150,000 are 105, 106, 106, and 112 percent, respectively. For tax years 2003 and after, the prior year safe harbor is 110 percent for taxpayers with prior year AGI in excess of $150,000. To interpret current tax collections, it is important to understand how prior year safe harbor rules may be affecting the forms of payments (withholding, estimated tax payments, final payments, and refunds). In particular, it would be very useful to know if current withholding and estimated tax payments reflect current tax liability or if they reflect taxpayer responses to a prior year safe harbor rule. The more taxpayers use a prior year safe harbor rule, the less current collections may indicate how current tax liability is really growing. Such a distinction is crucial for the estimate of current tax liability. Consider the taxpayer s optimal decision concerning the timing of tax payments to the government. A simplified version of the 1997 prior year safe harbor can be summarized as S = min (0.9 T t Y t, 1.1 T t 1 Y t 1 ), where S is the optimal, penalty-free tax payment on year t tax liability; Y t is taxable income and T t is the tax rate in year t; and Y t 1 is taxable income and T t 1 is the tax rate in year t 1. If a taxpayer s tax liability T Y grows year over year by more than 22 percent, then the taxpayer will be better off by using the 110 percent safe harbor rule for calculating this year s estimated tax payments to avoid an underpayment penalty. The break-even tax liability growth rate decreases to 11 percent when the safe harbor falls to 100 percent of last year s liability, as it did for tax year 1998. This analysis makes several simplifying assumptions. First, it does not factor in tax credits, so that for the rule to hold, credits 5 Modified AGI in this context referred to AGI for the current year less any gain from the sale or exchange of a principal residence or from an involuntary conversion. Items from pass-through entities for the prior year were to be taken into account in determining modified AGI, and any pass-through items for the current year were to be disregarded, unless the taxpayer was a general partner or ten percent owner in a pass-through entity. 6 Prior to 1998, the taxpayer could owe no more than $500 by April 15 to avoid penalty. This amount was raised to $1,000 in TRA 97. 333

NATIONAL TAX JOURNAL would have to grow at the same rate as taxable income. Second, the break-even rule assumes that taxpayers know year t income with certainty. This is not usually the case. Third, and perhaps most importantly, the tax rate T must also be known. This is also not usually the case, especially in years of revenue surges. If taxpayers are uncertain about current income and effective tax rates, and have an aversion to penalties, they will rely more heavily on the prior year safe harbor. This lowers the tax liability growth rate that makes the prior year safe harbor the optimal method for calculating the current year s estimated taxes. One result of these tax law changes is that the forms of payments have changed quite dramatically as a percent of net liability during the surge period. Table 3 shows forms of payment as a percent of net liability for 1990 7. The changes in the estimated tax safe harbor rules helped change estimated tax payments as a percent of net liability from a high of 19.2 percent in 1992 to a low of 16.1 percent in, and back to 17.1 percent in 1997. Over this period, a percentage point of net liability grew from $4.6 billion in 1992 to almost $7.4 billion in 1997. 7 In addition, the interaction of the safe harbor rules and fast-growing income, especially among higher income taxpayers, helped mask the size of total liability growth as measured by withholding and estimated tax payments. For example, suppose that performance bonuses, the exercising of stock options, and capital gains realizations caused growth in income from year to year in excess of ten percent. It is possible that the prior year safe harbor would be met by levels of withholding and estimated tax payments that suggested substantially smaller total tax liability than was actually the case. These taxpayers could then make quite sizable final payments without penalty. In addition, taxpayers with fast-growing incomes whose prior year AGI was $150,000 or less could have an even more dramatic switch from withheld and estimated tax payments to final payments because their prior year safe harbor was 100 percent. As Table 3 shows, withholding as a percentage of net liability did indeed drop significantly from to 1997, making it very difficult to gauge the year-to-year growth in net liability from withholding data alone. Typically, however, withholding is the form of payment for which we have the most complete data at the time of preparation of a new budget forecast. When the percent of net liability attributable to withholding dropped from 86.3 percent in to 83.1 percent in 1996, more than $21 billion in receipts shifted from withholding to higher final payments and lower refunds. The same declining pattern holds for the combined withholding and estimated tax payment percentage of net liability over the same period. Interestingly, even if withholding makes up a smaller portion of net liability than the prior year, the year-to-year growth in withholding can be quite large if the growth in total liability is sufficiently large. Withholding grew in excess of nine 7 The and 1996 SOI samples highlight the importance of estimated tax payments, especially the portion paid by higher income taxpayers. The data show that estimated tax payments reported on tax returns totaled $122.5 and $141.6 billion in and 1996, respectively. These payments include amounts for Self Employment Contributions Act taxes, although the majority represent payments on individual income tax liability. The SOI data also show that among taxpayers who made estimated tax payments, those with AGI of $200,000 or more made 65.2 percent of the payments, although they made up only 5.5 percent of the returns reporting estimated tax payments. Estimated taxpayers with AGI of $1,000,000 or more made almost 31 percent of the estimated tax payments, although they made up only 0.5 percent of total estimated taxpayers. In 1996, those estimated taxpayers with AGI of $200,000 or more made 56.4 percent of the payments, yet made up just 6.4 percent of estimated taxpayers. The same statistics for estimated taxpayers with AGI of $1,000,000 or more are 26.6 and 0.6 percent, respectively. 334

335 Level of collections (CY) Withheld taxes Refunds Total nonwithheld Estimated payments Final payments Total net liability Percent of net liability Withheld taxes Refunds Total nonwithheld Estimated payments Final payments Total TABLE 3 INDIVIDUAL INCOME TAX LIABILITY (BASED ON TAX COLLECTIONS) FOR THE FY 2000 JANUARY BUDGET BY FORM OF PAYMENT 1991 1992 1993 1996 1997 404.2 81.3 118.6 71.2 47.4 441.5 91.5% 18.4% 26.9% 16.1% 10.7% 100.0% 410.0 75.5 128.2 89.0 39.2 462.7 88.6% 16.3% 27.7% 19.2% 8.5% 100.0% 438.3 77.1 134.5 89.6 44.8 495.6 88.4% 15.6% 27.1% 18.1% 9.0% 100.0% CY, in $ Billions 467.2 85.6 145.9 87.5 58.4 527.5 88.6% 16.2% 27.7% 16.6% 11.1% 100.0% 502.6 88.9 168.6 93.7 75.0 582.4 86.3% 15.3% 29.0% 16.1% 12.9% 100.0% 549.3 93.6 205.5 108.6 96.9 661.3 83.1% 14.1% 31.1% 16.4% 14.7% 100.0% 600.6 99.7 238.2 126.6 111.6 739.1 81.3% 13.5% 32.2% 17.1% 15.1% 100.0% Forecasting Federal Individual Income Tax Receipts Year-to-year growth rates in form of payments Withheld taxes Refunds Total nonwithheld Estimated payments Final payments Total net liability 1.4% 7.0% 8.1% 25.0% 17.3% 4.8% 6.9% 2.0% 4.9% 0.7% 14.3% 7.1% 6.6% 11.1% 8.5% 2.4% 30.3% 6.4% 7.6% 3.8% 15.6% 7.0% 28.4% 10.4% 9.3% 5.2% 21.9% 16.0% 29.2% 13.5% 9.3% 6.5% 15.9% 16.6% 15.2% 11.8%

NATIONAL TAX JOURNAL percent from to 1996 despite a drop of 3.2 percentage points in the portion of net liability attributable to withholding over the same period. This occurred because overall net liability grew by 13.5 percent between and 1996. RESULTS FROM A PANEL OF TAX RETURNS The interaction of the safe harbor rule changes and fast-growing incomes can be seen from tax returns that appeared in each SOI sample from 1991 to 1996. Table 4 shows the and average forms of payments as a percent of net liability for taxpayers in this sample with AGI of less than $150,000 in. Note that these taxpayers were subject to a 100 percent prior year safe harbor rule in. The data show an average decline in the amount of net liability paid through withholding and estimated tax payments of almost five percentage points. On average, these taxpayers were due a refund (negative balance due), as reported on their tax return. Hence, on average, the lower percent of net liability paid through withholding and estimated tax payments raised extension payments and lowered refunds instead of increasing final payments. 8 Table 4 illustrates how the pattern of shifting average form of payments be- Tax Year Difference Combined switch b TABLE 4 FORM OF PAYMENTS AS A PERCENT OF NET PAYMENTS FOR THE AVERAGE SAMPLE TAXPAYER WITH AGI OF $150,000 OR LESS IN a Withheld Payments 99.99% 95.78% 4.21% Estimated Payments 13.48% 12.79% 0.68% 4.89% Extension Payments 0.86% 1.24% 0.38% Balance Due (Refund) 14.33% 9.82% 4.51% 4.89% Total Net Payments Difference Combined switch b With AGI Growth of at Least Five Percent from to c 103.33% 90.65% 12.68% 15.47% 11.01% 4.45% 17.13% 0.84% 1.38% 0.54% 19.64% 3.05% 16.59% 17.13% Difference Combined switch b With AGI Growth of at Least Ten Percent from to c 102.17% 83.18% 18.99% 18.77% 13.39% 5.38% 24.36% 1.09% 2.01% 0.92% 22.03% 1.42% 23.45% 24.36% With AGI Growth of at Least 15 Percent from to c Difference Combined switch b 101.63% 79.25% 22.38% 22.07% 14.33% 7.74% 30.12% 1.21% 2.29% 1.07% 24.92% 4.13% 29.05% 30.12% a Taxpayers had to appear in each SOI file from 1991 to 1996. Weighted sample sizes of those with AGI of $150,000 or less in and are 78.7 and 78.4 million, respectively. Returns were weighted by their maximum weight over 1991 6. b The combined switch is the sum of withheld and estimated tax payments and the sum of extension payments and balance due (refunds). c The subsample sizes of those with 5 AGI growth of 5, 10, and 15 million are 51.6, 40.2, and 32.4 percent, respectively. 8 Although not shown in Table 4, the same patterns hold for the average forms of payments for taxpayers in this sample with AGI of less than $150,000 in, a group that was subject to a 100 percent prior year safe harbor rule in 1996. 336

Forecasting Federal Individual Income Tax Receipts comes more extreme as the growth in income from year to year increases. For taxpayers in this sample with 5 AGI growth of at least 5 percent, the average form of payments shift from withholding/estimated tax payments to extension/ lower refund is approximately 17 percent. For taxpayers with AGI growth of at least 10 percent, the average form of payments shift from withholding/estimated tax payments to extension/balance due is more than 24 percent. In addition, the average final settlement changes from refund to balance due. Finally, for those in the sample with AGI growth of at least 15 percent, the average form of payment shift is approximately 30 percent. A second group of taxpayers, those with persistently high incomes, also illustrates the shift in the forms of payments from withheld and estimated tax payments to final payments. Table 5 shows the average forms of payments for taxpayers who appeared in each SOI sample from 1991 to 1996 but who had AGI in excess of $150,000, $250,000, $500,000, $1,000,000, or $2,000,000 in each sample year. 9 By design, these data represent very high income tax- TABLE 5 FORM OF PAYMENTS AS A PERCENT OF NET PAYMENTS FOR THE AVERAGE SAMPLE TAXPAYER WITH AGI OF MORE THAN $150,000 IN EACH YEAR 1991 6 a Above $150K Tax Year Withheld Estimated Extension 1996 Difference 6 Combined switch 50.64% 48.57% 46.28% 4.36% 7.68% 10.06% 11.01% 3.33% Balance Due (Refund) 0.26% 2.05% 3.17% 3.42% 6.75% Total Net Payments Above $250K 1996 Difference 6 Combined switch 45.34% 43.56% 40.22% 5.12% 9.12% 12.23% 14.14% 5.02% 0.90% 0.90% 2.30% 3.21% 8.23% Above $500K 1996 Difference 6 Combined switch 37.38% 35.37% 32.66% 4.72% 11.97% 15.04% 18.59% 6.62% 2.83% 1.55% 0.10% 2.93% 9.55% Above $1M 1996 Difference 6 Combined switch 29.59% 28.48% 25.97% 3.62% 14.43% 18.40% 21.81% 7.38% 4.47% 4.21% 2.07% 2.40% 9.78% Above $2M 1996 Difference 6 Combined switch 24.73% 23.27% 20.85% 3.88% 9 1991 was selected as the beginning year of the panel to avoid the effect of the income shifting that appears to have taken place across tax years 1992 4 in response to the provisions in OBRA 93 described above. 337 16.72% 21.90% 25.67% 8.95% 6.76% 7.00% 4.30% 2.46% 11.41% a Taxpayers had to appear in each SOI sample from 1991 6. The weighted sample size of those with AGI of more than $150,000, $250,000, $500,000, $1,000,000, and $2,000,000 in each sample year are 728,075; 263,405; 64,523; 17,490; and 5,062, respectively. Returns were weighted by their maximum weight over 1991 6.

NATIONAL TAX JOURNAL payers who were subject to the 110 percent prior year safe harbor rule for tax years 6. As a result, the forms of payments displayed in this higher income sample differ from the forms of payments paid by the more modest income sample shown in Table 4 and from the aggregate forms of payments shown in Table 3. In particular, these taxpayers pay far more of their tax liability through estimated tax payments, which is consistent with the fact that unearned income makes up a far larger portion of their total income. The higher income sample data are consistent with some features of the aggregate data. Between and 1996, the switch in the percentage of net liability from withheld/estimated tax payments to extension/balance due is just under seven percent for sample taxpayers with AGI in excess of $150,000 in each of the sample years. It rises to 11.4 percent for taxpayers who have AGI in excess of $2 million in each of the sample years. CONCLUSIONS The individual income tax receipts surges of the latter 1990s have complicated the forecasting of these receipts in a number of ways. First, the persistence of the surges has made separating temporary and permanent components of the growth in receipts especially difficult. Second, collections data are heavily relied upon to update the receipts forecast. However, numerous tax law changes that simultaneously altered both tax liability and the timing of tax payments throughout the 1990s are reflected in the collections data. In addition, the recent surges in tax liability have interacted with the tax payment timing rules so that collections data have frequently masked the true underlying growth in tax liability. Thus, it is imperative that forecasters continue to discover ways to distinguish between the temporary and permanent components of the receipts surges, and to monitor carefully the effects of surges on the forms of payments that make up collections data. Acknowledgments The author thanks Gerald Auten, Len Burman, Lowell Dworin, and Joel Platt for many helpful comments and suggestions. The views expressed here are those of the author and should not be interpreted as those of the U.S. Department of the Treasury. REFERENCES Cilke, James. The Treasury Individual Income Tax Simulation Model. Washington, D.C.: U.S. Department of the Treasury, Office of Tax Analysis,. U.S. Congressional Budget Office. The Economic and Budget Outlook: Fiscal Years 2000 2009. Washington, D.C.: Government Printing Office, 1999. 338