تقاضای شرکت بیمه درمانی چگونه کشسانی است؟

We investigate the impact of tax subsidies on the firm’s decision to offer insurance, and on conditional firm spending on insurance. We do so using the micro-data underlying the Employment Cost Index, which has a major advantage for this exercise: the matching of very high quality compensation data with information on a sample of workers in the firm. We find that, overall, there is a moderately sized elasticity of insurance offering with respect to after-tax prices (−0.25), and a larger elasticity of insurance spending (−0.7). We also find that the elasticities are driven primarily by small firms, for whom the elasticity is larger. And we find that there is significant value added to this employer-based data: replicating methods from standard micro-data sources in our data lead to misleading estimates of these key parameters. Our simulation results suggest that major tax reform could lead to an enormous reduction in employer-provided health insurance spending.

The dominant feature of the health insurance market in the U.S. is the provision of private health insurance through the workplace. In 1998, 91% of the privately insured non-elderly population, representing 65% of the total non-elderly population, received their insurance through the workplace (EBRI, 2000). There are a number of reasons why health insurance may naturally be provided through the workplace in the U.S. There may be substantial economies of scale in administering insurance which increases the value of pooling mechanisms. Workplaces provide a natural pooling mechanism along dimensions largely exogenous to health status. And, finally, the U.S. tax code subsidizes health insurance purchase through the firm relative to the non-group market by excluding the value of that insurance from an individual’s income, a tax exclusion estimated to cost more than $100 billion in foregone federal, state and local tax revenues in 1999 (Sheils and Hogan, 1999). This tax subsidization of employer-provided insurance has been criticized along a number of dimensions. First, relative to a flat credit, a deduction for health insurance expenditures is regressive, providing the largest tax break for the most well off employed persons. Second, the tax system subsidizes insurance purchases over purchases of other goods, distorting individuals towards increased insurance purchase; some (e.g. Feldstein, 1973) have claimed that this leads to ‘overinsurance’ and, perhaps, ultimately to excess cost inflation in the health care sector. While the first of these criticisms is unassailable, assessing the validity of the second requires understanding the impact of the tax subsidy to employer-provided insurance on the amount and nature of employer-provided insurance. As a result, over the past 25 years a large number of articles have assessed the responsiveness of employer-provided insurance to its after-tax price. Unfortunately, this literature has led to a wide range of estimates, from roughly zero to almost −6. This wide range likely reflects two important limitations of the previous literature. One is that many articles in this literature have been unable to control for confounding factors correlated with both the after-tax price of insurance and the demand for insurance; this limitation has been addressed to some extent by more recent work. The second is that all of the work in this literature has either used data at the level of the firm, or at the level of randomly selected workers in firms (through individually based micro-data sets). The disadvantage of the first of these is that we do not observe the characteristics of the workers to whom the firm is responding. The disadvantage of the second is that, with any workforce heterogeneity, the firm may not be responding solely to the demand of the randomly observed worker. While this weakness has been recognized, it has not been effectively addressed, due to the well-known (and well-lamented) absence of data with information on both firm benefits provision and details on a sample of workers in that firm. In this paper we address these difficulties by drawing on an excellent data source that provides high quality information on insurance offering, insurance spending, and, most importantly, information on the distribution of characteristics of workers in each firm. These are the micro-data of the National Compensation Survey, which underlie the well-known Employment Cost Index (ECI) measure of inflation. We match these data, from the 1983 through 1995 period, to information on the tax subsidy to insurance faced by the workers. This tax subsidy varies across workers of different income levels, over time, and across states, providing substantial exogenous variation in the after-tax price of insurance. And we use this unique information on the workers within a firm to create firm-specific tax subsidy measures that more closely correspond to the appropriate firm subsidy. We have four findings of interest. First, we estimate that there is a moderately sized response of firm offering, and an even more elastic response of firm insurance spending, to after-tax prices. Our central estimates indicate that the elasticity of firm offering with respect to the tax price of the median worker is −0.25 and that the elasticity of firm spending conditional on offering is −0.7. Second, we also find that the responsiveness of firm offering and conditional spending is larger for small firms. Third, we show that estimates from alternative approaches, which are feasible in traditional micro-data sources, lead to misleading conclusions about parameter values. This suggests that it is important that data such as these be used for estimating the impact of taxes on benefits decisions. Finally, we present simulations to show that our estimates imply sizable impacts of tax reform on health insurance offering and spending in the U.S.; for example, our results imply that complete removal of the tax subsidy to health insurance spending would lead to about 15 million fewer workers being offered health insurance, and a total reduction in insurance spending on the order of 45 percent. Our paper proceeds as follows. Section 2 reviews the literature on the elasticity of employer-provided health insurance, highlighting the value added from our approach. Section 3 provides a brief theoretical discussion of the appropriate specification of the tax incentives in our firm data. Section 4 introduces our data source, and discusses the empirical strategy. Section 5 presents our results. Section 7 presents simulations of the impact of tax reform on insurance offering and spending, given our results. Section 8 concludes.

The central behavioral parameter for estimating the impact of fundamental changes to the tax treatment of health insurance is the firm’s elasticity of demand for health insurance. Unfortunately, the large body of past work in this area has been unable to deliver a consistent, and perhaps even a convincing, estimate of this critical parameter. This is largely because of the absence of the ‘right’ data for this exercise, and related difficulties in deriving an empirical framework that can deliver cleanly identified estimates of the relevant elasticities. We have attempted to address this deficiency by using data from the ECI, which include information not only on firm insurance offering and spending, but also data on the distribution of workers within the firm. We estimate that firms are fairly responsive to tax prices in their benefits decisions. We obtain central estimates of the elasticity of firm offering of about −0.25 and of conditional insurance spending of −0.7. We also estimate that small firms are particularly responsive to taxes in their offering decisions, while larger size firms are more sensitive in their spending decisions. And we find that alternative approaches yielding more commonly available micro-data yield misleading estimates of the elasticities of insurance offering or spending. Our estimates have important implications for tax policy towards firm benefits provision. Most importantly, they suggest that the tax subsidy towards employer-provided health insurance is a critical force holding together the employer-provided insurance system. Without this subsidy, we estimate, both employer offering and spending would be dramatically lower than they are today. Our findings do suggest that the majority of the impact of changes in tax policy towards health insurance would be on the level of spending on insurance, not on firm provision of insurance. This is an important conclusion because one of the major benefits claimed for removing the tax exclusion is that it would lead to a reduction in inefficiently high spending, while the major concern about this policy is that it could lead to sizeable reductions in insurance coverage. Our findings suggest that the impact on insurance coverage is likely to be small relative to the reduction in spending. Nevertheless, the reductions in employer-provided insurance offering that we estimate are non-trivial. For a reform that removes the income tax subsidy to insurance (while maintaining the payroll tax subsidy), we estimate that offering would fall by almost 10 percent. There are currently 155 million workers and dependents covered by employer-provided health insurance (EBRI, 2000). A 10 percent reduction in coverage of this population is therefore 15 million persons, which amounts to almost 40% of the size of the current pool of uninsured. A critical question for future research is whether the persons who lose their employer-provided coverage will find coverage from other sources, such as the non-group market, or through spouses. In addition, one limitation of our work for analyzing policies such as removing the entire exclusion of employer-provided insurance from taxes is that there may be more substantial responses to major policy changes such as this than there are to the minor variation in tax rates used in our analysis. It is also important to get inside the black box of ‘total spending’ on insurance and understand which elements of spending are responding so elastically to tax prices. Gentry and Peress (1994), in their cross-sectional study, do find that elements of insurance which are likely to be more elastically demanded, such as dental and vision care, were more elastic with respect to tax price differences. Further work of this nature, particularly focusing on the choice of form of insurance (traditional fee for service vs. managed care), is a high priority. Finally, it would be useful to understand the corresponding impact of tax policy on the provision of other fringe benefits by the firm. Is the provision of health insurance more or less tax sensitive, for example, than the provision of pension benefits? Ultimately, data such as the ECI data used here can allow us to assess the sensitivity of the entire compensation package to taxation.