December 9, 1993 ESTIMATING THE IMPACT OF HEALTH REFORM ON FEDERAL RECEIPTS Executive Summary The Treasury Department's Office of Tax Analysis (OTA) is responsible for preparing revenue estimates of proposals which affect Federal receipts. In general, OTA analyzes legislative proposals that change the Internal Revenue Code. OTA also analyzes the effects of certain legislative changes which do not amend the Internal Revenue Code but nonetheless affect Federal receipts. For example, changes in the laws concerning employer provision of certain fringe benefits can affect receipts because of the favorable tax status of such benefits. The tax code also provides preferential treatment for certain types of health insurance expenditures. Health insurance contributions receive preferential tax treatment under several different provisions. Employer contributions for health insurance are deductible as a business expense by the employer and are excluded from the income of employees. Through their employers, some employees may have the option of contributing to tax-preferred cafeteria plans, enabling them to pay for their portion of health costs with pre-tax dollars. Self-employed individuals can deduct 25 percent of health insurance costs from adjusted gross income. Taxpayers can also deduct qualifying medical expenses which exceed 7.5 percent of their adjusted gross income. As a consequence, changes in the financing of health insurance will have implications for Federal receipts. Estimating the effects of health reform on Federal receipts has required a cooperative effort among many agencies. The undertaking has demanded a broad understanding of the provisions contained in the proposal. Estimating the revenue impact of the proposal has required many data inputs from other Federal agencies involved in this process. To maintain consistency while estimating the costs of the health reform plan, the estimates of the revenue impact of the plan rely on certain inputs from other Federal agencies involved in this process. Because of the interaction among the provisions, a change in one or two of the basic underlying policy parameters could trigger significant changes in the revenue estimates. The health reform plan contains many non-tax provisions which may affect Federal receipts by changing the financing of health care. Each of these provisions may have very different effects on revenues, but in combination the Administration's plan results in a net increase in Federal receipts over the budget period. An alternative plan with similar features could yield very different revenue results, even if it differed from the Administration's plan in only a few key non-tax aspects. This technical note provides background as to the methods and assumptions underlying Treasury's estimates of the impact of health reform on Federal receipts. In preparing these estimates, Treasury followed long-standing estimating conventions accepted by both Administration and Congressional agencies responsible for producing estimates of the budgetary impact of legislative proposals. In analyzing the revenue impact of non-tax changes in the financing of health insurance, Treasury has used the same methodology and models which are used to estimate the effects of changes in Internal Revenue Code provisions on receipts. Individual Tax Model The Individual Tax Model (ITM) is one of the most powerful tools developed by OTA to aid in estimating changes in Federal receipts. The ITM is a large microdata simulation model. The microdata aspect of the model refers to the fact that it contains data on the income, deductions, health expenditures, and other characteristics of individual tax filing units and families. The model can simulate the taxes paid under both current law and proposed changes in law. Professional economists in OTA construct, maintain, and utilize the ITM. OTA economists share a broad background in applied microeconomics, particularly in public finance. In addition, OTA economists are also specialists in other fields, such as econometrics, health economics, labor economics, statistics, and computer programming. These skills are used to develop the inputs to the model and for examining its outputs. Interactions with Other Agencies with Interests in Tax and Health Policy Questions OTA economists communicate regularly with their counterparts at the Joint Committee on Taxation (JCT) and the Congressional Budget Office (CBO), who also use large microsimulation models. Such contacts are useful both for identifying differences among the models as well as for developing consensus among the agencies responsible for analyzing the receipts effects of proposals. In the health area, OTA staff maintains regular contact with numerous specialists. Members of OTA staff routinely discuss health modeling issues with staff from several agencies in the Department of Health and Human Services, including the Agency for Health Care Planning and Research (AHCPR), the Office of the Assistant Secretary of Planning and Evaluation (ASPE), and the Health Care Financing Administration (HCFA). In addition, OTA consults with health policy staff at the Council of Economic Advisers (CEA), the Office of Management and Budget (OMB), and the CBO. OTA's extensive contacts with the staffs of other government agencies and outside experts provide additional access to data, research, and other techniques which are generally useful for model development. In many cases, these contacts are long-standing. Description of the Individual Tax Model [ For a more detailed discussion of the Individual Tax Model, see James Cilke and Roy A. Wyscarver, The Treasury Individual Income Tax Simulation Model. Department of the Treasury, Office of Tax Analysis, March 1990.] Data: The current version of the ITM was constructed from a sample of 110,000 individual income tax returns filed in 1989. The data base is a stratified probability sample of tax returns prepared by the Internal Revenue Service's Statistics of Income (SOI) Division. [Under Section 6103 of the Internal Revenue Code, OTA and JCT have access to tax return data, including the complete SOI file. A public-use computer tape file is available to other analysts, but it has fewer data items and taxpayer records and does not contain information which might violate the confidentiality of taxpayers. In particular, the public use file blurs information on high-income taxpayers.] This is the same sample employed by the SOI to produce the tabulations published in the Statistics of Income - 1989 Individual Income Tax Returns. When weighted, these data represent the total population of taxpayers in the United States. Tax returns contain extensive information on the components of taxable income. In addition, tax returns provide information about taxpayers' marital status and family size. However, tax returns do not contain information on other demographic characteristics, on non-taxable forms of income such as welfare benefits or earnings on pension and other retirement savings, or on expenditures made by the taxpayer. Nor do tax returns provide any information on families outside the tax system. [Nonfilers are predominantly low-income persons who do not have an income tax liability and do not file a return to claim a refund.] More comprehensive information than is provided on tax returns is needed to analyze the impact of proposals which extend the current income tax base and to analyze payroll, excise and other taxes. To add more information, the SOI tax return data are first matched to age data from Social Security records and then statistically merged with records from the Current Population Survey (CPS) conducted annually by the Bureau of the Census. The records in the ITM are grouped into family units as well as income tax return units, and are weighted to represent the entire filing population and noninstitutionized nonfiling population. The SOI file is also statistically merged with records from the Bureau of Labor Statistics' Consumer Expenditure Survey. In addition, imputations of other critical income, asset, expenditure, employment, and demographic measures are made using a variety of sources (e.g., the Federal Reserve Board's Survey of Consumer Finances). The data sources described above contain only limited information on expenditures on health care. From tax returns, information is available on the amount of health insurance purchased by self-employed persons who claim a 25 percent deduction. Tax returns also contain information on certain health expenditures, but only for those filers who itemize deductions and whose expenditures exceed 7.5 percent of their adjusted gross income. The Current Population Survey provides information on the insured status of individuals, including whether the insurance is provided through private or public sources. Lacking from these surveys is information on a family's total expenditures (including any employer contributions) on health insurance, characteristics of their insurance policy, and the health status of family members. To supplement these data sources, OTA statistically matched the data from the 1987 National Medical Expenditure Survey (NMES) to the ITM. The NMES is an extensive survey of approximately 14,000 households representing the civilian, noninstitutionalized population of the United States. It was conducted under the auspices of the Department of Health and Human Services' Agency for Health Care Planning and Research (AHCPR). The 1987 survey updates and expands previous surveys conducted in 1977 and 1980. The surveys collected information about participants' utilization and expenditures for health services, health insurance coverage, health status, and employment and income. Data from the household survey are supplemented by information from medical providers, employers, and insurers. Extrapolation: The complete data file is then extrapolated to future years based on the economic forecasts used in the Budget. [ At the time of the release of the Administration's health reform plan, the estimates were based on the Administration's economic assumptions contained in the 1993 mid-session review. The economic assumptions were extended through the year 2000 by OMB for purposes of determining the longer-term budgetary impact of health reform.] The extrapolation is done in two stages. The first stage adjusts for anticipated economic growth and inflation. This is accomplished by multiplying the various income, deduction, and credit items on each return by forecasts based on per-capita growth rates estimated from the economic forecasts. In the second stage, the weights assigned to the records in the file are changed to hit separately determined targets for key variables, including the size distribution of adjusted gross income. The growth rates for health data are generally based on projections contained in the Federal Budget or the National Health Accounts. Where relevant, the targets reflect significant changes in participation or expenditures since 1987 (the base year for the NMES) or 1989 (the base year for the CPS). For example, a major expansion in Medicaid will affect participation rates during the mid-nineties. To ensure consistency with the Administration's Budget estimates, the Health Care Financing Administration's projections for persons insured by Medicaid are used to estimate targets in the extrapolation of these items in the ITM. Tax Calculator: Using the extrapolated files, the tax laws for each year in the Budget period are simulated. In combination, these simulation programs are referred to as the "tax calculator" or simply, the "calculator." The calculator takes information from each potential tax filing unit in the data file, and using a set of specified tax parameters, computes that unit's Federal individual income tax liability under the proposed change in law. Two basic revenue estimating assumptions are embedded in the calculator for computing tax liabilities. First, all filers are assumed to choose tax options which minimize their tax liabilities. Second, variables such as the level and distribution of total pre-tax income or total expenditures are held constant when simulating a tax policy change. The calculator computes the values of a number of variables that are endogenous to the model -- that is, these are tax variables, in addition to liabilities, which may be affected by a proposal and which, in turn, can affect the calculation of tax liabilities. In general, the ITM can trace through most of the interactions between any income source and the various provisions of the Internal Revenue Code. Appropriate behavioral responses have been incorporated into the tax model. In addition, as will be discussed further below, off-model adjustments are often made by the analysts to incorporate other anticipated behavioral changes in response to a proposed tax change. Uses and Limitations of Model Output As noted above, the ITM is a powerful tool which enables OTA economists to better analyze the effects of various proposals. There are several important distinctions, however, between the output of the tax model and the final analyses prepared by OTA. First, even in the simplest case, output from the ITM does not go unexamined. Output is subject to a reality check. Users of the ITM check carefully the results to determine if they appear reasonable. For example, users may compare the extrapolation of a particular variable with data which has become available since the initial construction of the tax model. Such information may include data from more recent samples of tax returns (e.g., the 1991 SOI sample of tax returns), other government organizations (e.g., the Bureau of Labor Statistics' Surveys of Employee Benefits; the Census Department's Survey of Income and Program Participation), trade associations (e.g., surveys conducted by the Health Insurance Association of America), and independent consulting organizations. Second, the ITM is best utilized to analyze the effects of changes in the tax code that affect broad groups of taxpayers and involve current law tax rules. The tax model cannot be relied upon exclusively to estimate changes in the tax code which affect narrow populations or introduce new income tax rules. In these instances, OTA economists may rely on "spreadsheet" models to produce estimates of tax changes. Often, these spreadsheet models are, themselves, quite extensive and sophisticated. In many cases, information from the ITM (e.g., the marginal tax rate faced by a comparable group of taxpayers) may be used as input into these spreadsheet models. The ITM is also not used to analyze proposals affecting tax units other than individuals. For these purposes, OTA maintains several other tax models, including a corporate model, a depreciation model, and an estate model. Third, subject to certain budget estimating conventions, estimates of the revenue effects of tax changes include assumptions about changes in taxpayers' behavior induced by changes in tax policy. Given the set of macroeconomic assumptions used to prepare the Budget, major GDP components -- such as real and nominal GDP -- are assumed to be fixed for purposes of estimating the deficit impact of a proposed change in legislation. Thus, for revenue estimates, behavioral effects are constrained by this "fixed GDP" assumption. Behavioral assumptions which affect the composition of GDP, but not its level, are integral to the revenue estimates. Off-model adjustments are generally necessary to account for the full range of potential behavioral effects. Fixed GDP Assumption The "fixed GDP" budget estimating convention is a long standing rule and is followed by the Office of Management and Budget, Treasury, and all other Executive Branch agencies. The Congressional Budget Office (CBO) and the Joint Committee on Taxation (JCT) follow a similar convention, using the economic assumptions contained in CBO's budget analyses. The "fixed GDP" assumption allows policymakers to view the effects of proposed change in law on the deficit, as forecasted in the most recent Budget. [ Under the Budget Enforcement Act of 1990, the estimates of legislative proposals are based on the economic assumptions contained in the President's Budget. At the time of the introduction of the Administration's health reform bill in the fall of 1993, it did not seem likely that the legislative action would be completed by the end of the year. As a consequence, there was an Administration-wide decision to use the recent economic assumptions contained in the Mid-session review. ] Without such a convention, dozens of analysts in each agency could derive their own independent forecast of GDP each time they estimated the deficit impact of a proposed change in legislation. [ The fixed GDP assumption is discussed in detail in Howard W. Nester, "A Guide to Interpreting the Dynamic Elements of Revenue Estimates." Compendium of Tax Research 1987, C. Eugene Steuerle and Thomas Neubig, eds. Washington, D.C.: Government Financing Office, 1987, pp. 13 - 41.] As a consequence of the fixed GDP assumption, Treasury, CBO, and JCT assume that total employee compensation remains unchanged in response to a requirement that employers provide a new fringe benefit to their workers. This identity is derived from the Census Bureau's National Income Accounts (NIA). Under these assumptions, if employer contributions for health insurance increase, then other forms of labor income -- wages or other fringe benefits - must decline in order for GDP to remain constant. [ Some benefits -- such as employer contributions for social insurance -- are linked by law to wages and thus change as wages change.] However, if wages and salaries decline, income taxes and employment taxes must decline as well. This effect is sometimes referred to as the "income offset." [ For a discussion of the "income offset," see George Tolley and C. Eugene Steuerle, "The Effects of Excises on the Taxation and Measurement of Income," 1978 Compendium of Tax Research. Washington, D.C.: Government Printing Office, 1978, pp. 67 - 78; and Sonia Conly and Linda Radey, "Changes in Excise and Payroll Taxes and Their Effect on Total Budget Receipts," paper presented at the 1988 Eastern Economic Association Meetings, Boston, Massachusetts.] Federal income and employment tax liabilities are affected by these compositional changes, as the allocation between taxable cash wages and non-taxable compensation (including the employer portion of payroll taxes) shifts. Based on observable relationships within the NIA, wages and salaries would appear to fall by almost the full amount of an increase in employer contributions for health insurance. Wages and salaries do not fall by the full amount for several reasons. First, the reduction in wages automatically causes employer contributions for social insurance (another form of labor compensation) to decline. Further, to some extent, employer contributions for other fringe benefits, such as pensions and life insurance, will also fall. Estimating the Effect of Required Employer Contributions for Health Insurance, Premium Discounts, and Cost Containment on Federal Receipts Under the health reform plan, employers would be required to contribute towards the purchase of a comprehensive health insurance benefit plan for their workers. This package may cost either more or less than the health insurance plan currently provided by the firm, and its scope may also differ markedly from the firm's current plan. Employers' response to a requirement that they contribute toward their employees' health insurance will depend, in large part, on how the guaranteed comprehensive benefit package differs both in costs and generosity from their current plans and the extent to which they may be entitled to premium discounts under the proposal. [ The health reform plan affects revenues largely through its impact on the allocation between taxable wages and non-taxable health benefits. The plan can affect revenues in other ways as well. As insurance coverage expands as a consequence of the plan, some taxpayers will not incur large out-of-pocket expenditures for uncovered medical expenses, and deductible medical expenditures will also fall. Expansion of the Medicare benefit package to include prescription drugs may also reduce deductions for medical expenses. ] Five key pieces of information are necessary to evaluate the impact of the health reform proposal on Federal receipts. These include: * Initial cost of the comprehensive benefit package; * Rate of growth in the cost of the comprehensive benefit package; * Degree to which employers' costs are offset by premium discounts; * Employees' demand for health insurance in excess of the comprehensive benefit package; and * Employees' ability to negotiate with employers to obtain tax-preferred methods of paying for supplemental coverage and the employee share of the cost of the comprehensive benefit plans. The data sources and the key underlying assumptions for each of these items are described briefly below. Costs of the Benefit Package: The Health Care Financing Administration (HCFA) provided estimates of the costs of the benefit package at 1994 levels, assuming that the plan was fully effective in that year. Their estimates included the effects of moving to a system of universal coverage. Rate of Growth in the Costs of the Benefit Package: All agencies involved in estimating the budgetary impact of the health reform plan used the same assumptions regarding the rate of growth in the cost of the benefit package. Under these assumptions, the basic benefit package was assumed to grow at a rate consistent with private health insurance between 1994 and 1996. Beginning in 1996, the costs of the plan were assumed to grow at the targeted rates of growth specified in the health reform plan (CPI+1.5 percentage points in 1996, CPI+1.0 percentage points in 1997, CPI+0.5 percentage points in 1998, and CPI in 1999 and 2000). These rates of growth are based on the assumption that the cost containment initiatives contained in the plan are effective. Premium Discounts: Under the plan, premium discounts are provided to ease the burden for some employers. First, small firms with fewer than 75 employees and average wages below $24,000 will be entitled to significant premium discounts. Second, the Federal government will provide premium discounts for other firms within the regional alliance if the cost of providing the comprehensive benefit package exceeds 7.9 percent of their payroll. Some employers will receive premium discounts even though they provided health insurance in the past. These employers are expected to pass the discounts back to workers in the form of higher wages and other benefits. Receipt of premium discounts, then, could affect the estimates of the plan on Federal receipts. HCFA is responsible for producing the official estimates of the costs of the premium discounts. Using Treasury's Individual Tax Model, it is also possible to simulate the receipt of the premium discounts by individuals (as passed back to them by their employers). Treasury's estimates of the premium discounts were used solely as an input into the analysis of the effect of the plan on Federal receipts. As a check, OTA's estimates of the premium discounts are reconciled to those produced by HCFA. Demand for Supplemental Coverage: Workers' demand for supplemental coverage is estimated largely as a function of expenditures on medical services for items not within the scope of the comprehensive benefit package. Data on reimbursable expenditures on health insurance, as well as current health insurance expenditures, are used to determine the value of supplemental health insurance coverage. Estimates of the costs of administering health insurance (the "load factor") under the current system were provided by HCFA. The estimates also account for changes in the price and demand for supplemental coverage following health reform. Cafeteria Plans and Other Tax-Preferred Arrangements with Employers: Under the Administration's health plan, individuals may be responsible for a portion of the cost of the comprehensive benefit package. They may be liable for the difference between the cost of the plan which they select and eighty percent of the weighted average cost of a plan within their region. As under current law, workers are generally required to pay for health insurance premiums out of after-tax income. However, the current system provides workers with several opportunities to reduce their health insurance costs by paying with pre-tax dollars. To the extent that workers can take advantage of these options, tax receipts will fall. [ As will be discussed further below, the Administration's plan would restrict contributions to cafeteria plans. The estimates of the effects of the required employer contribution do not reflect these proposed restrictions. The effects of these restrictions are estimated separately, under the assumption that employee behavior has changed in the ways described in this section.] The estimates of the required employer contribution (with premium discounts and cost containment) took into account the likelihood that individuals may seek ways to shelter, on net, more of their health insurance premiums through cafeteria plans and other informal arrangements with employers. The estimates also took into account other offsetting factors (such as some reductions in contributions which, under the current system, cover out-of-pocket reimbursements). Estimating the Effect of Restricting Contributions for Health Insurance Under the plan, employer contributions for the comprehensive (i.e., standard) benefit package (up to 100 percent of the costs of the package) would be excluded from income for purposes of calculating individual income and employment taxes. Employer-paid premiums on supplemental plans would now be included in employees' taxable income. While this provision would generally become effective January 1, 2004, contributions for health benefits through cafeteria plans would be disallowed, effective January 1, 1997. As a consequence, the seven-year estimates of the revenue impact of health reform only show the impact of the restrictions on employer contributions through cafeteria plans. OTA's estimates of the effects of the restrictions on cafeteria plans are "stacked" after the combined effects of the required employer contribution, cost containment, and subsidies have been taken into account. In other words, the baseline for cafeteria plans, in these estimates, assume that individuals have already made certain adjustments to other aspects of health reform. Thus, for example, the baseline would reflect changes in the utilization of the cafeteria plans in response to the required employer contribution. When contributions to cafeteria plans are restricted, individuals may have alternative opportunities to shelter income through other tax-preferred arrangements with their employers (e.g., the employer may agree to pay the full amount of the employee contribution and, in turn, explicitly reduce wages by an offsetting amount). These alternatives for sheltering income are taken into account in the revenue estimates for restricting cafeteria plans.