THE TAX EXCLUSION FOR EMPLOYER-SPONSORED HEALTH INSURANCE

Transcription

1 National Tax Journal, June 2,64 (2, Part 2), 5-53 THE TAX EXCLUSION FOR EMPLOYER-SPONSORED HEALTH INSURANCE Jonathan Gruber This paper reviews the issues raised by and the impats of the tax exlusion for employer-sponsored health insurane. After reviewing the arguments for and against this poliy, I present evidene from a miro-simulation model on the impats on federal revenue, insurane overage, and inome distribution of various reforms to the exlusion. Keywords: health insurane, tax expenditures JEL Codes: US. H5, H24 I. INTRODUCTION ^:: pending on health are is the largest and fastest growing element of govemment ^budgets in the United States. Despite our ostensibly private health are system, almost half of medial spending is done by the govemment, primarily through the $4 billion Mediare program and the $3 billion Mediaid program. Yet the third largest govemment expenditure on health are is not inluded in this alulation: the exlusion of employer-sponsored insurane (ESI) expenditures from taxable inome. In 29 U.S. state and federal governments will lose roughly $26 billion from the fat that expenditures by employers (and more than 8 perent of expenditures by employees) on ESI are not taxed as ompensation. This is by far the largest of the tax expenditures by the federal government. There are a number of oherent rationales for the ESI exlusion. In partiular, in the absene of viable pooling mehanisms outside the employment setting in the United States, the exlusion an be rationalized as the "glue" that holds employer pools together. At the same time, there are a number of problems assoiated with the exlusion. In partiular, a number of studies doument that the ESI exlusion leads to (likely ineffiient) inreases in insurane plan generosity. The exlusion is also highly regressive Jonathan Gruben MIT Department of Eonomis and National Bureau of Eonomi Researh, Cambridge MA, USA (gruberj@mit.edu)

2 52 National Tax Journal as both tax rates and ESI expenditures rise with inome. As a result, eonomists have advoated reform of this tax expenditure for years. Reently, poliy makers have taken up this harge as well. Most notably. President Bush's 28 budget proposed to replae the ESI exlusion with an individual dedution of $7,5 for individuals holding health insurane. The Senate Finane Committee initially proposed to finane part of its expansion of health insurane in 29 through a ap on the exlusion of high ost insurane plans. The Senate hose instead a "Cadilla tax," an exise tax on high-ost insurane plans (with premiums of more than $8,5 for singles and $23, for families) levied on insurers. Given that this exise tax would be likely to be passed on as higher insurane pries, it has roughly the same impat as apping the tax exlusion at the same levels as the tax is imposed. In this paper, I disuss the ESI exlusion and options for reform. I begin by providing bakground on the ESI system and its plae in the larger insurane ontext in the United States. In Setion III, I disuss the pros and ons of the ESI exlusion, and review evidene on the impats of the exlusion on individual and firm behavior. Setion IV disusses the issues in modeling reform to the ESI exlusion. Setion V presents estimates of the extent of the ESI exlusion, and disusses a variety of reform options for the exlusion. Setion VI onludes. II. BACKGROUND ON ESI AND INSURANCE COVERAGE IN THE UNITED STATES The goal of insurane providers is to reate large pools of individuals with preditable distributions of risk. These pools an be reated in many different ways; in the limit, a national health insurane plan, suh as that provided in Canada, provides pooling aross the nation. The United States has long relied on the employer as the main pooling devie for insurane. The growth in ESI dates bak to the wage and prie ontrols of World War II, whih ould be evaded through more generous provision of (untaxed) employee benefits, although the exlusion itself was not odified until several years later; these issues are reviewed in Thomasson (2) and Helms (28). Table shows the distribution of insurane overage in the United States in 28 (Fronstin, 29). Over 6 perent of the non-elderly population reeives ESI, whih is 7 perent of all individuals who are privately or publily insured and 9 perent of those who are privately insured. The Mediare program universally overs those over age 65, although a major expenditure of the ESI system is retiree overage. The major non-esi soure of overage for those under age 65 is the Mediaid program, whih provides insurane to low inome families, the disabled, and the elderly. Only perent of private insurane is provided outside of the employment setting in the non-group market. While there are variations in the extent and regulation of this market aross states, by and large it is a market where pries are high and variable, and where (in most states) individuals an be exluded entirely from overage based on their health status. At a basi level, most argue that there is a fundamental failure of

3 The Tax Exlusion for Employer-Sponsored Health Insurane 53 Table Non-Elderly Amerians' Soure of Health Insurane Coverage Total Population Private Employment-based Individually purhased Publi Mediare Mediaid Military health Uninsured People (Millions) Perentage of Population Notes: These data are from Fronstin (29). Figures may not sum to perent due to rounding error. insurane pooling in this market.' Attempts to remedy this in states suh as New York through ommunity rating laws have led to exessively high premiums on average and an exit of healthy individuals from the market (Ameria's Health Insurane Plans, 27). The ESI exlusion initially applied only to expenditures by employers. But over time there has been a sizeable growth in Setion 25 afeteria plans, whih allow employees (sine 978) to shelter their ontributions to ESI from personal taxation as well. Currently, roughly 8 perent of those with ESI have aess to a Setion 25 aount, although overage is still very inomplete among small firms. III. BENEFITS AND COSTS OFTHE ESI EXCLUSION A. Why Have an ESI Exlusion? In this setion I review the arguments for and against an ESI exlusion. As disussed earlier, the ESI exlusion grew out of a ompensation anomaly around World War II, not any oherent rationale. Nevertheless, as we onsider reforming the ESI exlusion, it is important to ontemplate its benefits and osts. There are some dissenting voies on this issue. Pauly and Herring ( 999), for example, argue that pries in the non-group market do not vary muh more by health status than do group premiums. An important problem with their analysis, however, is that it only examines the premiums paid by those who are allowed into the non-group market; if individuals are exluded based on health onditions, as seems to be the ase, then the analysis would understate the true underlying variation in the prie of health insurane.

4 54 National Tax Journal The main argument for ontinuing the ESI exlusion is that it is the glue that holds together our existing system of employer-provided insurane. Repealing the exlusion by taxing health insurane benefits, some argue, will lead employers to stop offering health insurane to their employees. This will leave employees without aess to atuarially fair pooling mehanisms and at the whim ofthe non-group market. But the non-group market may exlude those who are sik, leading to a large welfare ost from the redution in insurane overage for those who value it most. The extent to whih this onern is valid depends on two fators. Thefirstis how large an effet removing the ESI exlusion would have on employer offering of insurane. Employees value employers as an insurane purhasing mehanism for several reasons, of whih the exlusion is only one; there are also the benefits of group purhase, the negotiating power obtained with group size, and ease of plan hoie and administration. These fators will still be present even if the ESI exlusion is ended. Indeed, virtually all medium size and large firms in the United States have offered health insurane ontinuously over the past thirty years, despite enormous swings in the effetive tax prie of health insurane. Gruber and Lettau (24) examine the impat of tax prie variation on employerprovided insurane; the larger literature on this topi is reviewed in Gruber and Madrian (24). They find that medium-sized firms are only very modestly sensitive, and large firms not at all sensitive, to the tax prie of ESI. They do find that small firms are prie sensitive, with an elastiity of small firm offering with respet to the tax prie of-.69. Therefore, while prediting the impats of removing the exlusion goes out of sample, there is no reason to think that there will be a whole sale exit of medium and large firms from ESI. I inorporate these estimates into the modeling below. The seond unknown fator is how a major influx of individuals into the non-group market will affet priing in that market. The non-group market might funtion muh better if its sale were dramatially inreased by individuals leaving employer-sponsored insurane. While this may lower overall osts, however, there is little reason to think that it would redue the enormous disparities in prie and aess by health status. Therefore, absent other reforms to make insurane available to all outside the employer setting, there is a reasonable seond-best argument for maintaining the ESI exlusion. B. Costs of an ESI Exlusion Offsetting these benefits are the major osts ofthe ESI exlusion. First is the revenue ost ofthe exlusion, estimated below. Seond, this tax expenditure is highly regressive, as doumented below, as both tax rates and ESI expenditures rise with inome. The ESI exlusion also biases individuals towards purhasing exessively generous insurane beause they are paying with after-tax dollars on the margin. There is a sizeable literature that tries to estimate the elastiity of health spending with respet to the tax exlusion, as reviewed in Gruber (25). Gruber and Lettau (24) estimate a sizeable elastiity of employer-spending among thosefirmsoffering insurane with respet to the tax prie of-.7. This, however, reflets both redutions in insurane

5 The Tax Exlusion for Employer-Sponsored Health Insuranee 55 generosity and redutions in employer ontributions that are shifted to employee ontributions. Diret evidene on plan generosity omes from Gentry and Perress (994), who use ity-level variation in tax rates to show that the more "eletive" elements of benefits pakages, suh as dental and vision overage, are very prie sensitive. Of ourse, elastiity of spending with respet to the tax prie does not neessarily imply a distortion, if there is a pre-existing bias to too little health insurane spending. But this does not appear to be the ase. Existing evidene, partiularly from the RAND Health Insurane Experiment, is lear that the optimal health insurane plan features high initial ost-sharing with protetion against extreme out of poket risk (Gruber, 26). Yet, even in today's high health ost environment, the vast majority of employer-insured individuals have very modest ost sharing, with a relative pauity of high dedutible plans in the ESI setting. While there are several ompeting explanations for this "overinsurane," a leading ontender is the tax subsidy for ESI. Finally, the promotion of the employer-sponsored insurane system is not neessarily a benefit to soiety. A large literature douments the distortions to the labor market assoiated with suh a system, as reviewed in Gruber and Madrian (24), inluding limited job-to-job mobility and distorted retirement deisions. IV. MODELING THE ESI EXCLUSION To model the ost of the existing ESI exlusion, and to onsider the impats of options for reforming the exlusion, I turn to a miro-simulation model that I have developed over the past deade to model health insurane reform. This model is desribed in great detail in Gruber (29), so I just summarize the key elements here, fousing in partiular on the newly updated mathing of employer premium osts that is entral to the revenue estimate for the ESI exlusion. The model is based on data from the February and Marh 25 Current Population Survey (CPS), realibrated to 28 populations. These data are mathed to information on health insurane premiums and health osts. Data on the premiums for employer insurane, and the distribution of premiums between employers and employees, ome from the 24 Medial Expenditure Panel Survey (MEPS). For non-group insurane, I use data from the MEPS to ompute the underlying distribution of health spending by age and health status, and then add a load fator suh that the distribution of premiums by age mathes available data. All figures in the analysis are in 29 dollars, using the tax law as of 29. I assume that the reation to reform is immediate; in reality there ould be a transition to the new equilibrium simulated in the model. These data are used to develop a miro-simulation model that omputes the effets of health insurane poliies on the distribution of health are spending and private and publi setor health are osts. This model takes as inputs both the data soures desribed above and the detailed parameterization of reform options. The model first turns these poliy rules into a set of insurane prie hanges; for example, if the poliy intervention is a tax redit for non-group insurane, then the model omputes the implied perentage hange in the prie of non-group insurane for eah individual in the model.

6 56 National Tax Journal The effets of these prie hanges are then simulated, using a detailed set of behavioral assumptions about how hanges in the absolute and relative pries of various types of insurane affet individuals, families, and businesses. The key onept behind this modeling is that the impat of tax reforms on the prie of insurane ontinuously determines behaviors suh as insurane take-up by the uninsured and insurane offering by employers. The model assiduously avoids "knife-edge" type behavior, where some ritial level is neessary before individuals respond, and beyond whih responses are very large. Instead, behavior is modeled as a ontinuous funtion of how poliy hanges (net of tax) insurane pries. An important element of the modeling is therefore properly measuring the impat of the tax exlusion on the prie of employer-sponsored insurane. The impat of the exlusion is to alter the tax prie of employer sponsored insurane to TP=- + T + T^^ where î^is the federal inome tax marginal rate; r is the state inome tax marginal rate; r^ is the marginal payroll tax rate for the OASDI program (the 6.2 perent tax rate that is levied equally on employees and employers); and T^^, is the marginal payroll tax rate for the Mediare HI program.- differentiate the latter two programs beause, beginning in the early 99s, the taxable maximum for the HI program was inreased above that for the OASDI program (and was eventually removed altogether); by omparison, the marginal tax rate is zero above the taxable maximum for payroll taxation under the OASDI program. For a typial worker in the 5 perent tax braket, faing a 5 perent state tax rate and a 5.3 perent ombined payroll tax rate, this tax prie is roughly.65; that is, a dollar of health insurane osts 35 ents less than a dollar of other goods purhased with after-tax wages. In doing this type of analysis, a number of assumptions must be made about how individuals will respond to tax subsidies, through their effet on the prie of insurane. These assumptions have been developed based on the available empirial evidene reviewed above, although there are many gaps in this literature that must be filled in order to be able to aurately simulate the effets of poliy hanges. These assumptions are reviewed in detail in Gmber (29). A key aspet of modeling health insurane poliy is aurately modeling the deisions of firms. Eonomists tend to model firm deision-making as refieting the aggregation of worker preferenes within the firm. The exat aggregation funtion is unlear, as reviewed in Gruber (22); in my model I assume that the mean inentive for the firm (e.g., the average subsidy rate for non-group insurane) is what matters for firm deision-making. The reason that the payroll tax rate is additive in the denominator is that the employer is indifferent between purhasing one dollar of benefits or paying wages of /(+ r^+ r,,^.), sine eah dollar of wages requires a payroll tax payment as well.

7 The Tax Exlusion for Employer-Sponsored Health Insurane 57 The fundamental problem faed by individual-based miro-simulation models is that data on individuals do not reflet the nature of their oworkers, so that it is impossible to ompute exatly onepts suh as the average non-group subsidy in a worker's firm. I address this problem by building "syntheti firms" in the CPS, assigning eah CPS worker a set of oworkers seleted to represent the likely tme set of oworkers in that firm. The ore of this omputation are data from the Bureau of Labor Statistis that show, for workers of any given eamings level, the earnings distribution of their oworkers, separately by firm size, region of the ountry, and health insurane offering status. Using these data, I randomly selet 99 individuals in the same firm size/region/ health insurane offering ell as a given CPS worker in order to statistially repliate the eamings distribution for that worker's eamings level. These 99 workers then beome the oworkers in a worker's syntheti firm. These syntheti firms then fae three deisions about insurane: ( ) offering (whether to offer, if now not offering, or whether to drop, if now offering); (2) the division of osts between employer and employees; and (3) the level of insurane spending. Eah of these deisions is influened by the tax treatment of ESI expenditures. For example, if both employer and employee ESI expenditures are subjet to taxation, this will lower offering, lead to some shifting of premiums to individuals (sine there was muh less than full Setion 25 overage so the existing exlusion led to a bias in the aggregate to greater employer spending), and lead to a redution in the generosity of ESI. Taxing just employer spending on ESI (while leaving employee ontributions untaxed if through a Setion 25 plan), however, will lead to a small redution in offering and plan generosity, but a muh larger shift from employer to employee finaning of the premiums. Likewise, removing the Setion 25 tax shield but retaining the exlusion for employer spending will lead to a smaller redution in offering and plan generosity, but a shift from employee to employer finaning of premiums. While there are a large number of assumptions underlying the model, I fous here on the key assumptions about firm behavior. Firm deisions to offer insurane if the tax prie falls (or to drop insurane, if the tax prie rises) are based on the firm-sizespeifi elastiities of insurane offering estimated in Gruber and Lettau (24). In partiular, that paper finds that small firms (with fewer than employees) are fairly prie elasti in their insurane deisions, with an elastiity of insurane offering with respet to its prie of-.69. On the other hand, they estimate only a small and insignifiant impat on medium-size firms, and an even smaller effet for large firms (more than, employees). A key assumption for this type of modeling onems the inidene on wages of hanges in employer-insurane spending. Gruber (22) reviews the literature on inidene, and onludes that there is strong evidene for full shifting to wages of firm-wide hanges in insurane osts, with some evidene of shifting to sub-groups within the workplae as well. I make a mixed inidene assumption for this model. Any firm-wide reation, suh as dropping insurane or lowering employee ontributions, is diretly refleted in wages. Yet any individual's deision, suh as swithing from group to non-group insurane, is not refleted in that individual's wages; rather, the savings

8 58 National Tax Journal to the firm (or the ost to the firm) are passed along on average to all workers in the firm. This analysis is subjet to three major limitations. First, I approximate the tax expenditure assoiated with retirees that reeive tax-favored employer spending. It is diffiult to estimate the ost of this element of the tax expenditure, but based on estimates from the Joint Tax Committee I estimate that firms spend perent of their ex-ante spending on retirees. Therefore, I alulate the effets of tax poliy in the ex-ante world, and adjust overall tax hanges upward by perent ofthat ex-ante amount. Seond, I do not onsider any feedbak effets due to the hanging omposition of insurane pools. In urrent employer-sponsored insurane pools, healthy and young workers ross-subsidize their sik and older oworkers by paying ommunity-rated insurane premiums that do not vary by age or health. Part of the reason that they are willing to undertake this ross-subsidization is that it is the only tax-subsidized route to health insurane. If that tax subsidy is removed (or mitigated), then healthy workers may find better pries in a more losely experiene-rated non-group market, and to some extent abandon the ross-subsidized employer pools. This will raise the prie of ESI, whih ould exert further pressure on healthy workers to exit. This potentially important spiral of rising premiums is not inluded in the analysis. This effet ould be reinfored through the redued influene of non-disrimination mies that are enfored indiretly through the tax exlusion. Third, and most importantly, there is an enormous amount of unertainty assoiated with suh a modeling exerise. A number of the key parameters on whih these results depend, suh as the tax-prie elastiity estimates in Gmber and Lettau (24), are estimated based on within sample variation in tax rates, whih is very modest relative to the hange that would our with omplete elimination of the tax exlusion. As a result, the farther out of sample is the projeted poliy, the larger is the unertainty attahed to the estimates below. V. THE ESI EXCLUSION: COSTS AND REFORM OPTIONS A. Base Case The base results for the ost and distributional impats of the ESI exlusion are presented in Table 2. The first row provides the total federal revenue ost of the ESI exlusion, whih is $263 billion. The next two rows divide that into federal inome tax and federal payroll tax omponents; roughly 6 perent of the revenue ost of the exlusion is through redutions in federal inome taxes. The next row shows as a note that there is a major state inome tax revenue loss from the exlusion as well of over $3 billion in 29 dollars. The analysis in the remainder of the paper will fous on federal revenue effets only, but this is an important onsideration for states if the reforming the exlusion is used to finane overage expansions for whih the states bear some osts.

9 The Tax Exlusion for Employer-Sponsored Health insurane 59 The Cost and Distributional Innpliations ofthe ESI Exlusion (SMillion) Federal revenue raised Federal inome taxes Federal payroll taxes Note: State inome taxes 263, 57, 6, 3, Revenue Burden Dollars Perentage Deile Deile 2 Deile 3 Deile 4 Deile 5 Deile 6 Deile 7 Deile 8 Deile 9 Deile Deile Deile 2 Deile 3 Deile 4 Deile 5 Deile 6 Deile 7 Deile 8 Deile 9 Deile 2, 2, 6, 3, 2, 26, 34, 42, 55, 62, Ex-Ante ESI Spending 3, 2, 25, 43, 58, 78,, 5, 34, 5, Average Marginal Tax Rate Notes: These results are from the miro-simulation model desribed in the text. All figures are in 29 dollars. The top panel shows revenue raised and its subdivision into soures. The seond panel shows the distribution ofthe inreased revenue aross deiles ofthe inome distribution. The third panel shows the distribution of aggregate ex-ante ESI spending and average marginal tax rate aross deiles ofthe inome distribution. Deiles are defined separately for single and married taxpayers

10 52 National Tax Journal The seond panel of the table displays the distribution of the benefits of the ESI exlusion. About one-sixth of the benefits of the exlusion go to those in the lower half of the inome distribution, and about five-sixths to the upper half. The share of the benefits that go to the top inome deile is more than thirty times as large as that to the bottom inome deile. As noted earlier, this skewed distribution reflets both rising tax rates and rising ESI expenditures with inome. This is illustrated in the third panel, whih shows that both aggregate ESI expenditures and tax rates rise with inome. B. Options for Repeal Table 3 extends the analysis to onsider various reforms that would repeal or limit the ESI exlusion. The seond olumn of Table 3 shows the results of repeal of the ESI exlusion. The finanial results are idential to Table 2, but I also show impats on insurane spending and overage; the first olumn provides the ex-ante means of these variables. Reall that these findings must be interpreted with onsiderable aution, as they are using the prie elastiity estimated from existing variations in the tax prie to estimate the impat of a muh more radial hange in tax prie due to repeal of the exlusion. I find that this poliy leads to a redution in employer insurane spending of almost $ 83 billion, or about one-third of ex-ante employer spending. I alsofindthat employee insurane spending falls by $7 billion, or about perent of ex-ante employee spending. The last row of the seond set of figures shows the average employer annual premiums and employee annual ontributions for a single plan. At baseline, on average in our sample, employers ontribute $6,85 and workers ontribute $2,3. When the exlusion is removed, the employer ontribution falls dramatially, while the worker ontribution rises slightly; overall worker spending falls, however, due to the redution in the ESI overed population. I estimate that the removal of the ESI exlusion leads to a redution in the number of individuals with ESI of 5 million, whih is roughly perent of the number of ex-ante individuals with employer-provided insurane. This is a large number relative to the ex-ante stok of uninsured individuals, 49 million, but is modest relative to the ex-ante stok of individuals with employer-provided insurane. I also estimate that a number of those losing employer insurane will gain insurane through other hannels; roughly 3 perent of those losing ESI will hoose to purhase non-group insurane or move to publi overage. So only about 7 perent of those losing ESI beome uninsured aording to these estimates. Given the results in the Gruber and Lettau (24) paper that drive the estimated employer responses, most of the reation by employers happens in the small employer setor; large employers have offered insurane fairly onsistently over time as the tax prie has varied. There is relatively little hange in employee take-up most of the reation is in employer offering. One again, as with other estimates, it is important

11 The Tax Exlusion for Employer-Sponsored IHealth Insurane 52 Table 3 Revenue and Population Impats of Reforming the ESI Exlusion Inome & Payroll Taxes Complete Employer Employer Changes in: Ex-Ante Levels Repeal Share Only Share Only Federal revenue (Smillion),77, 263, 84, 42, Federal inome taxes 7, 57, 9, 27, Federal payroll taxes 833, 6. 74, 5, ESI (millions of people) Uninsured (millions of people) 49 Employer spending (Smillion) 563, -83, -72, 27, limployee spending (Smillion) 67, -7, 66, ^2, Firm/worker premium (S) 6,85/2,3 5,8/2,6 5,33/3,5 7,28/,56 Revenue burden Deile utoff Deile 2 utoff Deile 3 utoff Deile 4 utoff Deile 5 utoff 5 g g 7 Deile 6 utoff Deile 7 utoff Deile 8 utotï Deile 9 utott Deile utoff Ineome Taxes Only Changes in: Ex-Ante Levels Total Repeal Boss Repeal Worker Repeal Federal revenue (Smillion),77, Federal inome taxes 7, Federal payroll taxes 833, 72, 55, 8, 2,,, 32, 3, 2, ESI (millions of people) 52 Uninsured (millions of people) 49 Employer spending (Smillion) 563, Employee spending (Smillion) 67, Finn/worker premium ($) 6,85/2,3 - g -8, -4, -g 6-2, 4, -2 7, -3, 5,8/,99 5,9/2,72 7,3/,7

12 522 National Tax Journal Changes in: Revenue burden Deile utoff Deile 2 utoff Deile 3 utoff Deile 4 utoff Deile 5 utoff Deile 6 utoff Deile 7 utoff Deile 8 utoflf Deile 9 utoff Deile utoff Table 3 (Continued) Revenue and Population Impats of Reforming fhe ESI Exlusion Ex-Ante Levels Total Repeal Inome Taxes Only Boss Repeal Worker Repeal Notes: These results are from the miro-simulation model desribed in the text. All dollar figures are in 29 dollars. The top panel shows hanges to both the inome and payroll tax treatment of ESI spending; the bottom panel shows reforms to the inome tax treatment only. The first olumn shows ex-ante taxation/spending and population distribution; the seond olumn shows the hanges from total repeal of the exlusion; the third olumn shows the hange from repealing the exlusion only for the employer's share of ESI expenditures; the last olumn shows the hange from repealing only the employee's share of ESI expenditures. The top rows in eah panel show the ex-ante levels and hanges in tax revenues, in total and by soure, in millions of dollars. The next rows show ex-ante levels and hanges in insurane overage, by soure, in millions of persons. The next rows show ex-ante levels and hanges in insurane spending by employers and employees, overall, and in terms of average premiums. The remaining rows show the distribution of the inreased revenue aross deiles of the inome distribution; deiles are defined separately for single and married taxpayers to remember that this is a largely out of sample predition. In partiular, if there is signifiant adverse seletion in the ESI pool of the type desribed above, the redution in ESI ould be larger. Nevertheless, the poliy leads to an inrease in the uninsured population of roughly 22 perent of the ex-ante number of uninsured. Table 4 shows the age and health harateristis of those ex-ante uninsured and group insured, and then Table 3 shows for eah mn the omparable harateristis of those who beome uninsured and those who move to non-employer insurane (non-group or publi). Ex-ante, the uninsured are somewhat younger than the employer-insured, but in signifiantly worse health: 75 perent of those on ESI are in exellent or very good health, ompared to only 6 perent of the uninsured, and only 5 perent of those with ESI are in fair or poor health, ompared to over perent of the uninsured.

13 The Tax Exlusion for Employer-Sponsored Health Insurane 523 When the exlusion is repealed, those beoming uninsured look very muh like those who are ex-ante employer-insured. They are slightly older, but of omparable health. Relative to the ex-ante uninsured, this is a muh healthier population, so the overall health of the uninsured improves. Those beoming non-group or publily insured are muh healthier than the average person in ESI; as noted earlier, it is the young and healthy who will find it most attrative to "peel off' from ESI to other soures of overage when the tax subsidy is eroded. But this is a small share of those who were on ESI. Overall, there does not appear to be muh of a hange in the health mix of those remaining on ESI. This mitigates against any onem that we are missing dynami impats on ESI premiums from an eroding ESI health pool. The third olumn of Table 3 (and the third set of olumns of Table 4) onsiders the impat of removing the subsidy to employer spending only, maintaining the dedutibility of setion 25 aounts. Suh a poliy raises only $84 billion in new revenues, or about 7 perent of the total from removing the exlusion on both employer and employee spending. This is lower than the ex-ante proportion of insurane spending that is done by employers, 75 perent, beause of a shift of spending from newly taxed employer spending to tax-sheltered employee spending. Indeed, employer insurane spending falls by almost as muh as in the previous olumn, while employee insurane spending rises by $66 billion. This highlights the leakages in revenue raising that an arise from partial reform. Moreover, this is likely an underestimate of suh leakage, sine the model does not endogenize adoption of Setion 25 aounts. Suh a poliy would likely lead to an expanded use of Setion 25 aounts, and thereby even further shifting to employee ontributions (and out of taxable employer ontributions). Column 4 shows the impat of the omplementary poliy: retaining the exlusion for employer spending but removing the tax dedutibility of employee ontributions through Setion 25 aounts. This poliy hange raises $42 billion in revenues, or only 6 perent as muh as the full removal of the subsidy, despite employee ontributions being 25 perent of employer spending ex-ante. One again, the reason is an endogenous shift from employee to employer spending under this poliy; employer spending atually rises while employee spending falls by $42 billion, more than twie the amount than in the ase where the exlusion is ftjlly repealed. The distribution of impats is fairly similar in these two mns to the base ase mn, shown in olumn 2. The repeal of setion 25 is slightly more progressive than the other reforms, refleting the fat that tax-favored employee payments are even more onentrated with inome than is employer spending. The bottom of Tables 3 and 4 show the results of repeating the analysis, onsidering only the removal of the tax exlusion for inome tax purposes, but retaining the exlusion for payroll tax purposes. If employer and/or employee spending on insurane was inluded in the wage base and taxed for payroll tax purposes, then there would be some offsetting inrease in the soial insurane benefits finaned by those taxes. We may therefore overstate the net revenue gain by onsidering the revenues raised by inlusion of ESI spending in both the inome and payroll tax bases.

14 ^ 524 National Tax Journal o 2 d o D- E UJ : ovi s > o /Pu bli Z / o öl Only Shai yer o D, tu _^Í. Sing s o o ovin ubli OH O ' u /Pub>li O Z e s"hp 5 u oo > n O o Q. C o JS Q. O Q. Qj u a> Où dû Où? CÜ ^ C C "

15 The Tax Exlusion for Employer-Sponsored Health Insurane 525 ^ b ts I J o Ü C.J2 r-~ r^ o 3 <u ESIsp' before ) movi mt roll o ;heii to botb é refo o "S 3 o E 5 3 g. u S. «^ <l> u *-' JZ o er-ins lat re linsured 3 U O teristis number fhe next re O C } s ateristi il of G 5 JZ SULU :3 O 3 (^ ÎJ O and sets o Tf I 8 I "^ in "^ S Ü 3 ^ V- Í 8 ëf tî^ i ^ en (3 E (U tn 'S î^ ë :e >."3 s = 5) = ë i o re S Sí i o é l Ë '5 Uli Xi [7 T3 I áü " 2 g S) ^ Ö g" " ä I Ja' i P oo -S.f3.S B.S ^ ^ ^ U Ci. a. o. Q- Cu Ê z -s.s s

16 526 National Tax Journal Inluding ESI expenditures in the inome tax base leads to an inrease in Federal revenues of almost $ 72 billion, whih is roughly 65 perent of the revenues from inluding payroll taxes as well. The impats on insurane overage, employer spending, and employee spending are likewise proportionately smaller. Inluding ESI spending in the inome tax base leads to a redution in individuals overed by ESI of almost million persons, and a rise in the uninsured of almost 8 million. Employer insurane spending falls by $8 billion, and employee insurane spending falls by $4 billion. The distributional impats of this poliy differ signifiantly, however, from inluding ESI in both the payroll and inome tax bases. This poliy is muh more progressive beause inome tax rates are muh more progressive than those of the flat (and apped) payroll tax. If only the inome tax exlusion is removed, then the top half of the inome distribution bears 85 perent of the revenue burden. The remaining olumns of the seond panel show the effets of inluding only employer spending in the tax base, and of inluding tax-sheltered employee spending (i.e., repealing setion 25 for these purposes). One again, eah of these partial reforms is somewhat blunted by the shift aross types of insurane spending. The sum of the revenues raised by these partial reforms, $52 billion, is about perent less than the revenue raised by full repeal, or $ 72 billion. C. Options for Capping the Exlusion Full repeal of the ESI exlusion is pereived as a radial poliy presription and faes enormous politial opposition. A natural altemative is to ap the ESI exlusion at some level, perhaps high at first, and then gradually reduing the ap. This was the approah endorsed by the Report of the President's Advisory Panel on Federal Tax Reform (25), whih proposed that ESI spending above the typial ost of a Federal Employee Plan (at the time, $,5 for a family) be inluded in the base of taxable inome. And 28 Presidential andidate Hilary Clinton inluded a ap on the exlusion of ESI premiums from taxation at the average premium level for families making more than $2, per year. In Table 5, examine the impat of proposals to ap the exlusion rather than remove it all together. assume initially that there is a total ap on employer plus employee spending, and then vary this assumption below. There are many possible levels at whih the exlusion ould be apped, and for this exerise I hoose the median national level of premiums; using the mean instead of the median yields similar results but about 7 perent less revenues. In this table, I show paired results for () apping for both inome and payroll tax purposes, and (2) apping just for inome tax purposes. The ap is at $5,76 for single plans and $3,675 for family plans. I estimate that apping the ESI exlusion at this level would raise about $47 billion if the ap were applied to both inome and payroll taxes, and about $32 billion if the ap were applied just to inome taxes. This is about 8 perent of the amount raised

17 The Tax Exlusion for Employer-Sponsored Health Insurane 527 Table 5 Capping the Exlusion Inome and Payroll Tax Inome Tax Only Changes in: Federal revenue (Smillion) Federal inome taxes Federal payroll taxes Employer Cap Only No Employee Both Capped Repeal Both Capped 47, 36, 32, 3, 24, 29, 7, 2, 3, Employer Cap Only No Employee Repeal 25, 22, 3, ESI (millions of people) Uninsured (millions of people) _2 Employer spending (Smillion) -26, -34, -7, -22, Employee spending (Smillion) -7, 8, -7, 3, Firm/worker premium ($) 6,6/,98 6,57/2,9 6,88/,97 6,67/2, Revenue burden Deile I ulolt Deile 2 utoff Deile 3 utoff Deile 4 utoff Deile 5 utoff Deile 6 utoff Deile 7 utoff Deile 8 utoff Deile 9 utoff Deile utoff Notes: These results are from the miro-simulation model desribed in the text. All dollar figures are in 29 dollars. The first set of olumns shows hanges to both the inome and payroll tax treatment of ESI spending; the seond set shows reforms of the inome tax treatment only. The first olumn within eah set shows the impat of apping the sum of the employer and employee exlusion at the median employer plus employee ontribution to ESI; the seond shows the impat of apping the employer exlusion at the median employer ontribution to ESI. The top rows in eah panel show the ex-ante levels and hanges in tax revenues, in total and by ourse, in millions of dollars. The next rows show ex-ante levels and hanges in insurane overage, by soure, in millions of persons. The next rows show ex-ante levels and hanges in insurane spending by employers and employees overall, and in terms of average premiums. The remaining rows show the distribution of the inreased revenue aross deiles of the inome distribution; deiles are defined separately for single and married taxpayers

18 528 National Tax Journal from full repeal. On the other hand, apping the exlusion is muh more progressive than removing it. For example, while about 7 perent of the revenues from repeal for both inome and payroll tax purposes are raised from the lower half of the inome distribution, only 6 perent of revenues from apping are raised from the lower half of the inome distribution. Capping does lead to a large redution in employer and employee spending, but it is only a fration of the earlier amounts. Capping also has the virtue of being muh less dismptive to existing insurane arrangements. Modeling the impat of the ap on insurane overage is hallenging, of ourse, beause, in priniple, apping at the median should not ause any firm to offer insurane, but in pratie some firms will be uninterested in offering insurane unless it is very generous and some firms will not be able to find an insurane poliy that will be taken up by their employees unless it is very generous. In those ases, apping the exlusion ould lead firms to drop insurane overage. I model the impat of a ap as proportional to the impat of full repeal; that is, I ompute the impat of the ap on the effetive tax subsidy faing firms and have firms reat to the effetive redution in their tax subsidy. Doing so, I find a fairly modest impat of apping on insurane arrangements, as only about 2 million individuals lose ESI, and the rolls of the uninsured inrease by only about million. In the next set of olumns in the table, I show the impat of apping the exlusion solely for employer spending at its median level, $4,378 for singles and $,77 for families. Capping only employer spending raises only three-quarters as muh as apping total spending, due to a large evasion of the ap through shifting to employee ontributions. As noted above, this estimate likely understates the shifting sine do not endogenize the Setion 25 deision. A major ontroversy with suh tax aps, however, would arise around the issue of who bears the burden of the aps. Taxing the highest levels of insurane spending will redue the inentives for exessively generous insurane but it will also lead to the largest burdens in high ost states and for firms with high ost workers. One response to this point is to note that these groups are exatly the ones who have benefited for years from this tax exlusion, and all this poliy does is remove a tax bias in their favor. This argument faes politial diffiulties, however, as individuals will inevitably ompare the effets of the poliy hange to a baseline with the full exlusion in plae. Some of this disparity ould be readily addressed through adjustments to the level of the ap. For example, the aps ould be set on a state-speifi basis, with the levels refleting the differene in premiums in that state relative to the national average, while holding total revenues onstant. The ap ould also readily be adjusted for the average age mix of employees in a health plan. These adjustments are not perfet, as they would not apture variations in worker health other than age. But if the tax exlusion were removed in the ontext of a larger reform that moved the insurane system toward ommunity rating, these onems would be minimized.

19 The Tax Exlusion for Employer-Sponsored Health Insurane 529 VI. CONCLUSIONS While the publi and poliymakers often debate the struture of the two largest government expenditure programs for health are. Mediare and Mediaid, there has traditionally been little disussion of the third largest government (tax) expenditure, the exlusion of ESI premiums from inome and payroll taxation. However, reforms ofthe ESI exlusion have attrated inreased attention in reent years. In this paper I disuss the impliations ofthe existing exlusion and the impats of various reforms. I onlude that the existing ESI exlusion is both very expensive and highly regressive, with five-sixths ofthe benefits flowing to the top half of the inome distribution. Repealing or apping the exlusion ould result in signifiant inreases in government revenues and an inrease in the progressivity ofthe tax system. Yet it would also lead to a signifiant redution in insurane overage. Thus, repeal ofthe exlusion would make the most sense in the ontext of a system-wide reform that provided broader non-esi options for insurane overage. ACKNOWLEDGMENTS I am grateful to Len Burman, Jim Poterba, and onferene partiipants in the NBER Tax Expenditure Projet onferene for helpful disussions. REFERENCES Atneria's Health Itisurane Plans, 27. "Itidividual Health Insurane 26-27: A Comprehensive Survey of Premiums, Availability and Benefits." Center for Poliy and Researh, Ameria's Health Insurane Plans, Washington, DC. Fronstin, Paul, 29. "Soures of Health Insurane and Charateristis ofthe Uninsured: Analysis ofthe Marh 29 Current Population Survey." EBR] Brief No Employee Benefits Researh Institute, Washington, DC. Gentry. William, and Eri Peress, 994. "Taxes and Fringe Benefits Oflfered by Employers." NBER Working Paper No National Bureau of Eonomi Researh, Cambridge MA. Gruber, Jonathan, 22. "Taxes and Health Insurane." In Poterba, James M. (ed.). Tax Poliy and the Eonomy Volume 6, MIT Press, Cambridge, MA. Gruber, Jonathan, 25. "Tax Poliy for Health Insurane." In Poterba, James M. (ed.). Tax Poliy and the Eonomy 9, MIT Press, Cambridge, MA. Gruber, Jonathan, 26. "The Role of Consumer Copayments for Health Care: Lessons from the RAND Health Insurane Experiment and Beyond." Report for the Kaiser Family Foundation.The Kaiser Family Foundation, Menlo Park, CA,

20 53 National Tax Journal Gruber, Jonathan, 29. Doumentation for GMSIM Model. Unpublished manusript. MIT, Cambridge, MA. Gruber, Jonathan, and Mihael Lettau, 24. "How Elasti is the Finn's Demand for Health insmanet' Journal of Publi Eonomis 88 (7-8), Gruber, Jonathan, and Brigitte C. Madrian, 24. "Health Insurane, Labor Supply, and Job Mobility: A Critial Review of the Literature." In MLaughlin, Catherine (ed.). Health Poliy and the Uninsured, Urban Institute Press, Washington, DC. Helms, Robert, 28. "Tax Poliy and the History of the Health Insurane Industry." In Aaron, Henry J., and Leonard E. Burman (eds.). Using Taxes to Reform Health Insurane: Pitfalls and Promises, Brookings Institution Press, Washington, DC. Pauly, Mark V., and Bradley Herring, 999. Pooling Health Insurane Risks. Amerian Enterprise Institute Press, Washington, DC. President's Advisory Panel on Federal Tax Reform, 25. Simple, Fair, and Pro-Growth: Proposals to Fix Ameria's Tax System. U.S. Govemtnent Printing Offie, Washington DC. Thomasson, Melissa, 2. "The Importane of Group Coverage: How Tax Poliy Shaped U.S. Health Insurane." NBER Working Paper No National Bureau of Eonomi Researh, Cambridge, MA.

21 Copyright of National Tax Journal is the property of National Tax Assoiation and its ontent may not be opied or ed to multiple sites or posted to a listserv without the opyright holder's express written permission. However, users may print, download, or artiles for individual use.