Printer Friendly

HEALTH INSURANCE BENEFIT MANDATES AND FIRM SIZE DISTRIBUTION.

INTRODUCTION

The U.S. states have passed a large and rapidly increasing number of health insurance benefit mandates. Benefit mandates require health insurance plans to cover a specific treatment, condition, provider, or type of person (such as acupuncture, autism, chiropractor visits, or grandchildren of a planholder). These mandates are likely to increase health insurance premiums in plans that are forced to add new benefits or provide additional coverage (Kowalski, Congdon, and Showalter, 2008; Bailey, 2013). Rising health insurance premiums increases labor costs for employers who provide health insurance to their employees. Those increased costs could lead employers to reduce health insurance coverage, wages, or employment. Alternatively, firms can legally escape mandates and lower their costs by self-insuring if they are willing to bear the risk of paying employees' health insurance claims. Due to the nature of the costs associated with pooling risk, large firms can more easily take advantage of this loophole. Hence, the passage of health insurance benefit mandates may give large firms a competitive advantage relative to small firms. In this article, we use a novel empirical strategy to determine the extent to which health insurance benefit mandates affect employment and firm size in the United States.

We make four new contributions to the literature in this article. First, we examine the effect of health insurance mandates, in general, on employment using recent data. Previous work examined the effects of a single mandate (Gruber, 1994a; Wolaver, McBride, and Wolfe, 2003; Lahey, 2012; Bailey, 2014) or a handful of specific mandates (Meer and West, 2011) or examined the effect of mandates only for the 1990s (Kaestner and Simon, 2002; Mathur, 2010). This is important because the average number of state mandates has increased by more than 40 percent from the late 90s until 2010 (our sample period).

Second, we are the first to show empirically that health insurance mandates can lead to distortions in firm size distribution. As large firms are more easily able to self-insure and avoid state-level health insurance mandates, leaving small firms with relatively higher costs, health insurance mandates give a strategic advantage to large firms.

Third, our identification strategy takes into account the political forces behind the passage of mandates, rather than assuming mandates are passed at random as is often done in the prior literature. We overcome the endogeneity problem by using novel instruments, that is, the political variables that have been determined to affect the passage of mandates. (1)

Fourth, we use a data set novel to the literature on health insurance benefit mandates, the Quarterly Census of Employment and Wages (QCEW). These administrative data from the U.S. Bureau of Labor Statistics (2) have provided employment by firm size and state for virtually all firms in the United States since 1990. Meer and West (2011) argue convincingly that most previous work on the labor market effects of mandates suffers from severe attenuation because of reliance on data sets with high sampling error. The QCEW avoids this sampling error by using administrative data from the full population of firms rather than self-reported data from a sample of firms.

We find that mandates do not lead to statistically significant decreases in employment, but do distort the distribution of firm size by increasing the number of very large firms (greater than 1,000 workers) and reducing the number of small firms. Such a distortion may have a negative effect on aggregate productivity and job growth (Haltiwanger, Jarmin, and Miranda, 2013).

PREVIOUS WORK: THEORY AND EVIDENCE ON THE EFFECTS OF MANDATES

The Summers Model and the Effects of Health Insurance Benefit Mandates on Employment

Summers (1989) proposes a simple model of how mandated benefits should affect the labor market. In the Summers model, a mandated benefit decreases labor demand as if it were a tax equal to the cost of the benefit. Unlike a tax, the mandate also increases labor supply by an amount equal to the value workers receive from the benefit. For example, suppose that a mandated benefit costs employers $1,000 to provide, and that employers therefore reduce cash compensation by $1,000. If workers value the benefit fully at cost ($1,000), then the number of hours worked in equilibrium is unchanged; employers reduce money wages by $1,000 dollars, but employees perceive that the total value of their compensation is unchanged. If workers receive no value from the benefit, then they perceive that the total value of their compensation has decreased by $1,000, leading to a reduction in equilibrium hours worked (or employment); the effect is equivalent to a $1,000 per worker tax. If workers value the benefit at some fraction of its cost, say $600, then they perceive a $400 reduction in the value of their compensation and would reduce hours accordingly. The inefficient mix of cash and benefits lowers the value of employment to workers. In order to maintain the value of the compensation package, employers would have to increase their total spending per worker, which would lower the quantity of labor they demand. Likewise, if wages are downwardly sticky, the mandate will lead to increased health benefits without lower wages, raising the total cost of workers and lowering the quantity of labor demanded. (3)

Practical Considerations of the Effects of Mandates on Employment

Why should we expect health insurance benefit mandates to affect employment in the first place? After all, health insurance benefit mandates are aimed at health insurers, not employers. Several conditions must be met for health insurance benefit mandates to be binding on employers. First, the law must actually apply to the health insurance plan. Self-insured health insurance plans, commonly used by large firms, are exempt from state-level health insurance regulations by the federal Employee Retirement Income Security Act (ERISA) of 1974.

Second, the mandate must increase the costs of insurers enough to be reflected in increased premiums. The empirical literature has found mixed evidence of this for individual health insurance: LaPierre et al. (2009) find that mandates do not increase premiums; Kowalski, Congdon, and Showalter (2008) find that they do; Gohmann and McCrickard (2009) find that some mandates raise premiums, whereas others do not. Most relevant for our analysis, the only researcher to analyze the effect of mandates on employer-based plans (rather than individual plans) finds that the average mandate does significantly increase employer premiums by 0.44 percent to 1.11 percent (Bailey 2013).

Finally, for mandates to affect employers in the sense described in the Summers model, they must offer health insurance in the first place. This is generally a reasonable assumption. According to Medical Expenditure Panel Survey data summarized by the Kaiser Family Foundation, (4) in 2012, 50 percent of private sector establishments offered health insurance to their employees. However, firm size makes an important difference in the likelihood of employees being offered health insurance: according the same data, 35 percent of establishments with fewer than 50 employees offer health insurance and 96 percent of those with 50 or more employees do.

Though not all employers offer health insurance, and loopholes and cost sharing dissipate some of the cost of mandates on employers, it still seems reasonable to expect that mandates could cause disemployment at some firms. Cutler and Madrian (1998, p 509) go so far as to say that "increases in the cost of providing health insurance must have some effect on labor markets, either in lower wages, changes in the composition of employment, or both." We expect this effect to be the largest on medium-sized firms, which are most likely to be purchasing health insurance that is subject to mandates. Most small firms do not offer health insurance in the first place and so experience no direct effect on their labor costs when health insurance premiums increase. (5) The largest firms are usually able to self-insure, (6) taking direct financial responsibility for their employees' health insurance claims instead of paying a fixed premium. This exempts them from state-level regulations, including mandates.

Empirical Work on Various Effects of Health Insurance Benefit Mandates

Work on the effects of mandates has largely focused on whether mandates decrease the chance that employers offer health insurance and on whether they decrease the number of individuals with employer-provided health insurance. A few researchers have examined labor market effects of mandates in general on outcomes like wages. (7)

Other researchers have looked for the labor market effects of particular mandates aimed at a specific identifiable group. These studies have tended to find strong effects, presumably because it is easier for employers to substitute from one group of workers to another than to substitute away from workers in general. (8) Gruber (1994a) finds that maternity-care mandates resulted in reduced wages, although not reduced employment, for women of child-bearing age. Lahey (2012) and Bailey (2014) find that group-specific mandates (for infertility treatments and prostate cancer screening, respectively) decreased employment for the groups the mandates were meant to benefit.

Empirical Work on the Effect of Mandates on Employment

To date, three groups of researchers have examined the effects of mandates, in general, on employment. All focused on small firms, (9) none had an explicit strategy to overcome the endogeneity of mandates, and only one found statistically significant disemployment effects.

Kaestner and Simon (2002) use Blue Cross Blue Shield Association data (10) on mandates to cover treatments and conditions (though not providers) from 1993, 1997, and 1998, along with data on hours worked, weeks worked, wages, health insurance, and employment from the Current Population Survey (CPS) for March 1989-1998. After running individual-level ordinary least squares (OLS) regressions with state and year fixed effects, they find that mandates had no statistically significant effect on employment.

Mathur (2010) uses 1993-1995 Blue Cross Blue Shield Association data on mandates and 1993-1995 Survey of Income and Program Participation data on employment in small businesses (1-50 employees). Mathur runs an ordered probit regression with no state fixed effects to determine number of employees in small firms of various sizes and finds that mandates do significantly reduce employment at small firms.

Meer and West (2011) investigate the effect of nine high-cost mandates on employment in small firms (0-20 and 20-99 employees). They use 1989-2008 Statistics of the U.S. Business (SUSB) data on employment, which is the U.S. Census Bureau product closely related to the QCEW that we use in our study. Using fixed effects estimation of state-year-level data, they find that most of their nine mandates do not reduce employment; however, they find some evidence that mandates for drug abuse and mental health treatment do.

Meer and West (2011) argue that previous researchers had been unable to identify an effect of mandates on small business employment because they used smaller data sets, such as the CPS data, that suffer from sampling bias. While the CPS seems large, with about 170,000 observations annually, it does not have many employees of small businesses in any given state-year bin. Meer and West show that state-year estimates of small business employment are incredibly noisy in the CPS compared to data sets, such as the SUSB, that capture the full population of employer firms (about 7 million annually) rather than a small sample of employees. SUSB data show that small business employment in a state essentially never changes by more than 10 percent from one year to the next, whereas CPS data show this happening in 38 percent of all state-year observations.

Bhattacharya and Vogt (2000), like Meer and West (2011), show how previous work using the CPS had little chance of detecting the true effect of mandates on employment. They calculate that Gruber (1994a) and Kaestner and Simon (2002) use CPS data that did not have the statistical power available to detect the effects the researchers were testing for. Although these articles found no statistically significant disemployment effects of mandates, their 95 percent confidence intervals include employment decreases of at least 2 percent.

The Effect of Mandates on Firm Size and Self-Insurance

To our knowledge, ours is the first article to investigate the effect of health insurance benefit mandates on firm size. A range of previous work has studied various other regulations that target businesses above or below a certain threshold number of employees. Kapur et al. (2012) find that nonmandate U.S. health reforms that explicitly targeted small-group insurance, such as guaranteed-issue laws, led to a decrease in the number of small firms. Garicano, Lelarge, and Van Reenen (2012) examine a French labor law that increases firing costs for firms with 50 or more workers. They find that the law led to a distortion in firm size, which in turn decreased overall productivity.

We expect mandates to lead to a distortion in firm size primarily because of ERISA. This federal law allows firms that self-insure to be exempt

from many state-level health insurance regulations, including health insurance benefit mandates. Self-insuring means directly bearing financial responsibility for employees' health insurance claims rather than paying predictable premiums to a separate health insurer who pays the claims as they arrive. Self-insured firms may still hire third-party administrators to process claims. According to Fronstin (2012), 57.5 percent of those with employer-sponsored health insurance were covered by self-insured plans in 2010, up from 46 percent in 1996 (when the data used in our article begin).

Health insurance claims are variable, which creates risk for those paying them. That is why so many individuals and firms buy insurance in the first place; they want someone else to cover the risk. Insurers, by pooling many individuals with near-uncorrelated risks, can face a relatively smooth stream of claim payments. This logic of insurance means that larger firms will find it easier to self-insure. The more workers they can pool, the lower the chance that a large proportion of them will become catastrophically ill in the same year. This logic is borne out clearly in the data. According to 2010 Medical Expenditure Panel Survey data summarized by Fronstin (2012), firms with fewer than 50 employees have only 12.5 percent of covered workers in self-insured plans compared to 37.6 percent for firms with 100-999 employees and 83.6 percent for firms with 1,000 or more employees.

Because self-insured firms are exempt from mandates and large firms are much more likely to self-insure, we expect mandates to prove costly for smaller firms while providing a relative advantage for large firms (especially those with 1,000 or more employees).

Our Contribution to Understanding the Effects of Mandates on Employment

Our main contribution to research on the effect of health insurance mandates is to use novel instrumental variables for mandates, taking into account that mandates are the outcome of a political process and are not passed randomly as natural experiments. The only step toward addressing the endogeneity problem in previous work was the use of state fixed effects by Kaestner and Simon (2002) and Meer and West (2011).

In addition, like Meer and West but unlike other previous researchers, we use an employment data set that draws from Internal Revenue Service payroll data to capture nearly the full population of firms and workers rather than relying on survey data that contact only a small sample of workers. Our new identification strategy allows us to more cleanly identify the effect of mandates on employment and firm size. Finally, we are the first to estimate the extent to which mandates distort firm size.

DATA AND EMPIRICAL STRATEGY

Health Insurance Benefit Mandates

Health insurance benefit mandates are laws commonly passed by states that specify what insurers must cover. They may require insurers to cover specific treatments or procedures (e.g., rehabilitation for alcoholism) or to cover certain types of providers (such as midwives or acupuncturists). Less commonly, they may require insurers to offer coverage to certain sorts of people (e.g., adult dependent children in a family plan).

We use data on the number of mandates in force in each state from 1996 to 2010, compiled by the Blue Cross Blue Shield Association and presented by Laudicina, Gardner, and Holland (2011). These data on mandates have often been used in academic research, for example, by Kaestner and Simon (2002) and Bailey and Webber (2014). The data, summarized in Table 1, reveal a great deal of variation among states, as well as the general trend of an increase in the number of mandates over time. The average state had 20 separate benefit mandates in 1996, increasing to 36 by the end of our sample in 2010.

Employment and Establishment Size

Our data on employment come from the QCEW. This administrative data set, compiled by the U.S. Bureau of Labor Statistics using reports from all establishments that report to unemployment insurance programs, reaches essentially all businesses in the United States (approximately 97 percent). Bhattacharya and Vogt (2000) argue that survey data sets, such as the CPS, do not have the statistical power to determine the effect of mandates on employment; Meer and West (2011) concur and argue that administrative data are superior. These arguments, which we discuss in greater detail in the section "Empirical Work on the Effect of Mandates on Employment," demonstrate why our use of the QCEW is appropriate.

The QCEW provides data broken out by establishment size rather than firm size. An establishment is "typically at one physical location and engaged in one, or predominantly one, type of economic activity for which a single industrial classification may be applied." (11) A firm may consist of multiple establishments. Whether firm-level data would be more appropriate than establishment level depends on how often multi-establishment firms use the same health insurance across establishments. If firms commonly use the same plan across establishments, especially if that plan is self-insured, then firm-level data would be more appropriate than the establishment-level data we use. However, it appears this is not the case, because small establishments within larger firms are actually less likely to offer health insurance than single-establishment firms of the same size (Glied, Lambrew, and Little, 2003, p. 22).

We use annual QCEW data on employment by establishment size in each state from 1996 to 2010. The administrative data used to build the QCEW are incredibly rich, containing data on the exact size of almost every firm in the United States. Unfortunately, to protect the privacy of firms, the QCEW provides these data aggregated only up to the state level. The QCEW breaks establishments down into nine size categories, ranging from establishments with fewer than 5 employees to those with more than 1,000. Thus, we observe the total number of establishments and the number of employees in each size category in each state and year. Tables 2 and 3 show summary statistics for the number of establishments and total employment by establishment size.

Instrumental Variables: The Political Roots of Health Insurance Benefit Mandates Most researchers studying the effects of mandates have used empirical strategies that assume the passage of mandates is exogenous. Usually, researchers treat mandates as a right-hand-side variable in an OLS or fixed effects regression and either ignore the endogeneity problem or argue that it is not important in their particular case. But mandates are passed by state legislatures, the outcome of a political process that may well be correlated with a wide variety of health-care and labor market outcomes.

Bailey and Webber (2014) model the political process that gives rise to health insurance benefit mandates. They consider whether the passage of mandates is driven more by political parties and ideology, differing state institutions, producer interest groups, or greater need from consumers and patients. They run fixed effects regressions with state-level data from 1996 to 2010 and find that producer interest groups have a substantial impact on the passage of health-care mandates, whereas political parties, the insurance status of patients, and state institutions, such as mandate review boards, do not significantly affect the number of mandates. In particular, states with a higher number of doctors per capita pass more mandates, and states with greater political contributions by insurers pass fewer mandates. Doctors benefit from mandates because health insurance increases the demand for their services. (12) Insurers, by contrast, would prefer to have the freedom to design their own plans as they see fit. Bailey and Webber find the relationship between doctors per capita and mandates to be particularly strong--one additional doctor per 1,000 individuals (the difference between a state with few doctors like Idaho or Mississippi and an average state like Louisiana or Virginia) leads a state to have 10 additional mandates.

We use these variables, doctors per capita and political contributions by insurers, as instruments for mandates. We use the same data sources for these variables as used by Bailey and Webber (2014). Data on physicians per capita come from the U.S. Department of Health and Human Services' Area Resource File. These data are available on the state level from 1995 to 2011, with the exception of 2009. Data on political contributions come from the National Institute on Money in State Politics. These data are available back to 1996. (13)

Empirical Strategy: Panel Instrumental Variables

We utilize a panel of annual, state-level data on the number of health insurance benefit mandates in effect and the number of employees at firms of various sizes from 1996 to 2010. We begin by running nave fixed effects regressions of mandates on employment of the following form:

[Employment.sub.st] = [Mandates.sub.st] + [[tau].sub.t] + [[theta].sub.s] + [[epsilon].sub.st],

where [Employment.sub.st] is total employment at firms in a specific size range in each state and year (we run separate regressions to determine how mandates affect employment at firms of various sizes), [Mandates.sub.st] is the number of health insurance benefit mandates in effect in that state and year, [[tau].sub.t] represents year dummies, [[theta].sub.s] represents state dummies, and [[epsilon].sub.st] is the error term.

But for this approach to produce an unbiased estimate of the effect of mandates on employment, the decision to pass mandates must be uncorrelated with expected employment changes. In reality, it may be the case that a trend or expectation of increasing employment could lead to more mandates being passed because of the perception that firms could afford the cost; this would bias estimates of the effect of mandates on employment upward. In regressions reported in Appendix Table A2, we find that the passage of mandates over the 1996-2010 period is associated with a statistically significant "decline" in employment and the number of establishments for firms of most sizes at the very beginning of the period, suggesting that standard regressions suffer from reverse causality.

To overcome the endogeneity problem and clearly identify the causal effect of mandates on employment and firm size, we developed instruments for mandates using doctors per capita and the natural log of political contributions by insurers, variables that were previously determined to affect the passage of mandates (Webber and Bailey, 2014). Our first-stage equation is as follows: (14)

[Mandates.sub.st] = [Doctors.sub.st] + Insurer[Contributions.sub.st] + [[tau].sub.t] + [[theta].sub.s] + [[epsilon].sub.st],

where [[tau].sub.t] represents year dummies, [[theta].sub.s] represents state dummies, and [[epsilon].sub.st] is the error term.

The second stage of our instrumental variables (IV) regression is

[Employment.sub.fst] = [Mandates.sub.st] + [[tau].sub.t] + [[theta].sub.s] + [[epsilon].sub.st],

where [Employment.sub.fst] is total employment at firms in a specific size range in each state and year (we run separate regressions to determine how mandates affect employment at firms of various sizes), [Mandates.sub.st] is the first-stage equation's estimate of number of health insurance benefit mandates in effect in that state and year, [[tau].sub.t] represents year dummies, [[theta].sub.s] represents state dummies, and [[epsilon].sub.st] is the error term. In addition to studying the effect of mandates on total employment at firms of various sizes, we study the effect of mandates on the total number of firms of various sizes.

RESULTS

Effect on Employment

Table 4 shows our main results for the effect of mandates on employment at firms of various sizes. The first column shows the results of an OLS regression of mandates on employment, with no controls. It demonstrates that the simple correlation between mandates and the number of firms is positive and statistically significant for firms of all sizes. The magnitudes shown suggest that each mandate is associated with thousands of additional employees. The second column adds state fixed effects and controls to the regression. Once the controls are taken into account, the estimated effect of mandates becomes much smaller and statistically insignificant for firms of all sizes, though most point estimates remain positive. Although the nave OLS results could be used to argue that mandates actually encourage employment, the fixed effects results with year dummies reveal that the OLS results were biased by the omission of key variables.

The final column of Table 4 shows the results of the IV regression, which accounts for the reasons mandates are passed. The IV estimates largely turn negative, suggesting that each mandate may reduce employment by hundreds of workers per state in most firm-size ranges. For example, the 95 percent confidence intervals suggest that state-level employment among firms with 20-49 workers may fall by as much as 3,826 (1.1 percent) for each mandate passed. However, the IV results are imprecisely estimated, and we cannot reject the hypothesis that mandates have no effect, or even a small positive effect, on employment.

Effect on Firms

Table 5 shows our main estimates of mandates' effect of on firm size using regressions of the same form as before, but with the natural log of the number of establishments in each size range as the dependent variable instead of total employment at establishments in each size range. Our first column shows the results of an OLS regression of mandates on firm size, with no controls. It demonstrates that the simple correlation between mandates and the number of firms is positive and statistically significant for all firm sizes. This could imply that states with more firms are more likely to pass mandates or that mandates lead to more firms or that, in a spurious correlation, the number of mandates and firms have both grown over time, even though there is no causal relationship between them. Our second column shows the results of a fixed effects regression that controls for year dummies. We report no statistically significant effects in this column, and the magnitudes become much smaller and mostly negative.

The final column shows our IV results. Once we take into account the endogenous passage of mandates (as well as state and year fixed effects), we find statistically significant negative effects of mandates on the number of small firms with 5-9, 10-19, 20-49, and 50-99 workers. We also find negative, statistically insignificant effects on the number of very small (fewer than 5 workers) and medium-sized (100-249, 249-499, 500-999 workers) firms, but a positive and statistically significant effect on number of the largest firms (1,000+ workers).

The IV coefficient of -0.010 for firms with 5 to 9 employees should be interpreted as one additional mandate leading to a 1 percent reduction in the number of firms with 5 to 9 workers. This implies a loss of 272 firms with 5 to 9 workers in an average state. This loss is counteracted by a statistically significant 4.3 percent increase in the number of very large establishments (implying a gain of five establishments with 1,000 or more workers per mandate in an average state). (15)

Robustness

As with any empirical analysis using observational data, there are barriers to interpreting our results as causal.

First, mandates are heterogeneous. Certain mandates apply only to small groups (such as including partners in a civil union on family plans) or require insurers to cover inexpensive treatments; others are mandates for costly treatments used by many people (such as dentistry). We estimate the effect of the average mandate on employment and the number of firms of various sizes, but the true effect of any given mandate may differ greatly from our estimates. A mandate that is more costly than average is likely to prompt a greater response from firms.

Second, whereas the QCEW data we use for employment and the number of firms of various sizes are built from rich, establishment-level data, the public version of the data we were able to access is aggregated to nine firm-size bins at the state-year level. This means that we can observe that mandates seem to reduce the number of firms with 50-99 employees, but we cannot say whether this effect holds equally for all firms in this category or whether our result is being driven by firms with 50-59 or 90-99 workers. Furthermore, the public QCEW data do not contain information gleaned from following specific firms over time. This means we cannot say the mechanism by which firms with 50-99 workers are disappearing. They could be closing entirely or shrinking to fewer than 50 workers or growing to more than 99 workers or any combination of the above. The quality of our data limits what we are able to say with confidence.

Finally, the validity of our IV results depends on the validity of our instruments. Instruments must satisfy both the relevance and exclusion restrictions, meaning in this case that the number of doctors per capita and the level of political contributions by insurers are related to the number of mandates passed at the state level (which is confirmed from the first stage of our IV model) and have no direct effects on employment or the number of firms. In other words, they must influence our employment outcomes only through the mechanism of increased mandates. Because we have two instruments, we are able to test overidentifying restrictions. Although our model does "pass" this test (Hansen's J of 1.87), this is still only suggestive evidence regarding the validity of our instruments.

As the proposed instruments are not generated through some sort of random assignment (just as the vast majority of instruments used in empirical applications are not), it is certainly fair to question whether they satisfy the exclusion restriction. One potential mechanism that could violate excludability in the model that tests the impact of mandates on total employment is the possibility that doctors per capita could be correlated with product--demand factors, and hence employment. The argument is that a higher proportion of doctors represents a wealthier locality, which demands more goods and services and therefore has a higher level of local employment. While state fixed effects capture time-invariant market demand factors, there could still be an endogenous link between doctors per capita and aggregate demand. As suggestive evidence against this hypothesis, we rerun all analyses including a measure of average income as an additional control variable, and find no qualitative difference in any of the results. (16) Furthermore, the firm-size distribution results, which we feel represent the biggest contribution of this article, would still remain valid even if there were a link between our instruments and aggregate demand. For these results to fail the exclusion restriction, there must be some relationship that causes firms to increase or decrease in size while holding total employment constant.

DISCUSSION

Our IV estimation finds weak (i.e., not statistically significant) evidence that mandates decrease overall employment at small- and medium-sized firms. We find stronger evidence that mandates reduce the overall number of small firms (with 5-99 workers), and increase the number of large firms (with 1,000 or more workers), while having no net effect on medium-sized firms (with 100-999 workers). Eighty-six percent of firms with 1,000 or more workers are self-insured, compared to 35 percent of firms with 100-999 workers and approximately 12 percent of firms with 99 or fewer workers (Fronstin, 2012). This is consistent with the idea that mandates are costly for the firms they apply to but that these costs can be escaped by self-insuring under ERISA, which is much easier for large firms. Mandates appear to push firms toward the sizes that best enable them to escape the regulation.

How is it possible for mandates to reduce the number of firms without reducing employment? Our results suggest that employment is being pushed toward larger firms. This is clearly occurring across firm-size bins: we find that mandates lead toward a statistically significant increase in the number of large (1,000-plus workers) firms in a state.

Though our data do not allow us to observe the exact size of firms, our results also suggest that mandates result in larger firms within each size category. We observe that mandates lead to fewer firms with 50-99 workers but do not cause a (statistically significant) decrease in total employment at firms with 50-99 workers. These findings can be reconciled if we are losing many firms with 50-60 workers while gaining some firms with 90-99 workers.

In one sense, our results are similar to those in the literature on group-specific health insurance benefit mandates. Lahey (2012) and Bailey (2014) find that when a mandate meant to benefit a group causes an increase in the cost of employing that group (e.g., an infertility treatment mandate and 28- to 42-year-old women), the mandate can decrease employment among that group without decreasing employment overall. In this study, we find that mandates change the distribution of employment by firm size without decreasing employment overall.

Suppose we are correct that the growing prevalence of health insurance benefit mandates, together with the ERISA self-insurance loophole that larger firms can more easily take advantage of, has led to the growth of large firms at the expense of small ones. Is this truly a problem? We argue that it is.

First, self-insurance is a costly way to escape mandates. Individuals and firms buy insurance for a reason--they value how it protects them from risk. Self-insuring means accepting the risk that some workers will get very sick all at once, requiring you to make much higher than expected payments for their claims. For truly large firms, this risk is very small. But for medium-sized firms, with say 100 workers, self-insuring still means accepting substantial risk, risk that firms would prefer to insure against if mandates did not make premiums too expensive.

Second, regulations that favor large firms over small ones contribute to a general decrease in economic dynamism. Economic dynamism--the birth and death of firms that embody creative destruction--has been declining in the United States since the early 1980s (Decker et al., 2014). Recent work has found that this creative destruction, in particular entry by small, new firms, is crucial for job growth (Haltiwanger, Jarmin, and Miranda, 2013), productivity (Syverson, 2011, and Garicano, Lelarge, and Van Reenen, 2012, note the particular role of cost-increasing regulations in reducing productivity), and overall economic growth (Aghion et al., 2013). The incredible importance of economic dynamism suggests that policymakers should be cautious about creating cost-increasing regulations and loopholes, such as health insurance benefit mandates with the ERISA self-insurance exemption that fall disproportionally on small firms.

CONCLUSION

We find that health insurance benefit mandates, laws, which increase the cost of providing health insurance, do not lead to statistically significant decreases in employment. However, these laws do lead to distortions in firm-size distribution. Because large firms can much more easily self-insure, exempting their health insurance plans from state-level mandate regulations provides a competitive advantage for large firms. This distortion in firm size may lead to reduced dynamism and productivity growth. The use of self-insurance to escape mandates also means that firms take on risk that they would otherwise prefer to insure away.

APPENDIX
TABLE A1 Details of Instrumental Variable Estimation--First-Stage
Results

                                         Instrumenting for Total Mandate

Doctors per capita                                4.748 (***)
                                                 (0.53896)
Log political contributions by insurers          -0.2777
                                                 (0.18319)
Observations                                    620
F-statistic                                      41.60 (***)

Notes: Coefficients are omitted for controls (year fixed effects, state
fixed effects, state population). Standard errors are given in
parentheses. (***) p < 0.01.

TABLE A2 Robustness Check: Do Mandates in 2010 Affect Establishments in
1996-1998?

Firm Size      Employment   Number of Establishments

   <5           6,471 (**)           0.0413 (***)
               (3,068)              (0.0134)
    5-9         6,611 (**)           0.0360 (***)
               (3,209)              (0.0133)
   10-19        8,632 (**)           0.0364 (***)
               (4,241)              (0.0134)
   20-49       13,909 (*)            0.0387 (***)
               (6,976)              (0.0141)
   50-99       11,121 (*)            0.0416 (***)
               (5,594)              (0.0145)
  100-249      13,479 (**)           0.0420 (***)
               (6,703)              (0.0147)
  250-499       7,230 (*)            0.0414 (**)
               (3,838)              (0.0157)
  500-999       4,817 (*)            0.0338 (**)
               (2,777)              (0.0154)
1,000+          7,642 (*)            0.0368 (**)
               (4,471)              (0.0161)
Year dummies      No                   No
State dummies     No                   No
Observations      110              110

Notes: (*) p < 0.1, (**) p < 0.05, (***) p < 0.01. Each firm size is
run in a separate OLS regression. Standard errors, clustered at the
state level, are in parentheses. Employment data are in levels,
establishment data in logs. (*) p< 0.1, (**) p < 0.05, (***) p< 0.01.


REFERENCES

Aghion, P., U. Akcigit, and P. Howitt, 2013, What Do We Learn From Schumpeterian Growth Theory? Working Paper, National Bureau of Economic Research, Cambridge, MA. World Wide Web: http://www.nber.org/papers/w18824

Aron-Dine, A., L. Einav, and A. Finkelstein, 2013, The RAND Health Insurance Experiment, Three Decades Later, Journal of Economic Perspectives, 27(1): 197-222.

Bailey, J., 2013, The Effect of Health Insurance Benefit Mandates on Premiums, Eastern Economic Journal, 40(1): 119-127.

Bailey, J., 2014, Who Pays the High Health Costs of Older Workers? Evidence From Prostate Cancer Screening Mandates, Applied Economics, 46(32): 3931-3941.

Bhattacharya, J., and W. Vogt, 2000, Could We Tell If Health Insurance Benefit Mandates Cause Unemployment? A Note on the Literature, Working Paper, University of Chicago Stigler Center. World Wide Web: www.chicagobooth.edu/assests/stigler/167.pdf

Cutler, D., and B. Madrian, 1998, Labor Market Responses to Rising Health Insurance Costs: Evidence on Hours Worked, RAND Journal of Economics, 29(3): 509-530.

Decker, R., J. Haltiwanger, R. Jarmin, and J. Miranda, 2014, The Role of Entrepreneurship in US Job Creation and Economic Dynamism, Journal of Economic Perspectives, 28(3): 3-24.

Feldman, R., 1993, Who Pays for Mandated Health Insurance Benefits? Journal of Health Economics, 12(3): 341-348.

Fronstin, P., 2012, Self-Insured Health Plans: State Variation and Recent Trends by Firm Size, Technical Report, Employee Benefit Research Institute, Washington, DC. World Wide Web: http://www.ebri.org/publications/notes/index.cfm?fa=notesDisp&content_id=5131

Garicano, L., C. Lelarge, and J. Van Reenen, 2012, Firm Size Distortions and the Productivity Distribution: Evidence From France, Working Paper, National Bureau for Economic Research, Cambridge, MA. World Wide Web: http://www.nber.org/papers/w18841

Glied, S., J. M. Lambrew, and S. Little, 2003, The Growing Share of Uninsured Workers Employed by Large Firms, Technical Report, The Commonwealth Fund, New York. World Wide Web: http://www.commonwealthfund.org/~/media/files/publications/fund-report/2003/oct/the-growing-share-of-uninsured-workers-employed-by-large-firms/glied_largefirms_672-pdf.pdf

Gohmann, S. F., and M. McCrickard, 2009, The Effect of State Mandates on Health Insurance Premiums, Journal of Private Enterprise, 24(2): 59-73.

Gruber, J., 1994a, The Incidence of Mandated Maternity Benefits, American Economic Review, 84(3): 622-641.

Gruber, J., 1994b, State-Mandated Benefits and Employer-Provided Health Insurance, Journal of Public Economics, 55(3): 433-464.

Gruber, J., and M. Lettau, 2004, How Elastic Is the Firm's Demand for Health Insurance? Journal of Public Economics, 88(7-8): 1273-1293.

Haltiwanger, J., R. Jarmin, and J. Miranda, 2013, Who Creates Jobs? Small vs. Large vs. Young, Review of Economics and Statistics, 95(2): 347-361.

Jensen, G. A., and J. R. Gabel, 1992, State Mandated Benefits and the Small Firm's Decision to Offer Insurance, Journal of Regulatory Economics, 4(4): 379-404.

Kaestner, R., and K. I. Simon, 2002, "Labor Market Consequences of State Health Insurance Regulation, Industrial and Labor Relations Review, 56(1): 136-159.

Kapur, K., P. Karaca-Mandic, S. M. Gates, and B. Fulton, 2012, "Do Small Group Health Insurance Regulations Influence Small Business Size? Journal of Risk and Insurance, 79(1): 231-259.

Kowalski, A. E., W. J. Congdon, and M. H. Showalter, 2008, State Health Insurance Regulations and the Price of High-Deductible Policies, Forum for Health Economics and Policy, 11(2): 1-26.

Lahey, J. N., 2012, The Efficiency of a Group-Specific Mandated Benefit Revisited: The Effect of Infertility Mandates, Journal of Policy Analysis and Management, 31(1): 63-92.

LaPierre, T. A., C. J. Conover, J. W. Henderson, J. Allen, and B. A. Taylor, 2009, Estimating the Impact of State Health Insurance Mandates on Premium Costs in the Individual Market, Journal of Insurance Regulation, 27(3): 3-36.

Laudicina, S., J. Gardner, and K. Holland, 2011, State Legislative Healthcare and Insurance Issues, Technical Report, Blue Cross and Blue Shield Association.

Lennon, C., 2015, The Individual-Specific Impact of Employer-Sponsored Insurance: Evidence From the Affordable Care Act, Technical Report, University of Pittsburgh, PA. World Wide Web: http://www.conorjlennon.com/uploads/3/9/6/0/39604893/aca_summer_2015.pdf

Mathur, A., 2010, Health Insurance and Job Creation by the Self-Employed, Small Business Economics, 35(3): 299-317.

Meer, J., and J. West, 2011, Identifying the Effects of Health Insurance Mandates on Small Business Employment and Pay, Technical Report, Texas A&M, College Station. World Wide Web: http://econweb.tamu.edu/jmeer/abs\s\do5(m)w\s\do5(m)andates.html

Sloan, F., and C. Conover, 1998, Effects of State Reforms on Health Insurance Coverage of Adults, Inquiry, 35(3): 280-293.

Summers, L. H., 1989, Some Simple Economics of Mandated Benefits, American Economic Review, 79: 177-183.

Syverson, C., 2011, What Determines Productivity, Journal of Economic Literature, 49(2): 326-365.

van der Goes, D. N., J. Wang, and K. C. Wolchik, 2011, Effect of State Health Insurance Mandates on Employer-Provided Health Insurance, Eastern Economic Journal, 37(4): 437-449.

Webber, D., and J. Bailey, 2014, The Political Roots of Health Insurance Benefit Mandates, Working Paper, Mercatus Center, Arlington, VA. World Wide Web: http://mercatus.org/publication/political-roots-health-insurance-benefit-mandates

Wolaver, A., T. McBride, and B. Wolfe, 2003, Mandating Insurance Offers for Low-Wage Workers: An Evaluation of Labor Market Effects, Journal of Health Politics, Policy, and Law, 28(5): 883-926.

James Bailey, PhD, is Assistant Professor in the Department of Economics and Finance at Creighton University, 2500 California Plaza, Omaha, NE 68178. James Bailey can be contacted via e-mail: jamesbailey@creighton.edu. Douglas Webber, PhD, is Assistant Professor in the Department of Economics at Temple University, Philadelphia, PA. We thank the Mercatus Center at George Mason University for funding and helpful comments.

(1) In an earlier study, Webber and Bailey (2014) find that states with more doctors per capita passed more mandates, and states with greater political contributions by health insurers passed fewer. In this study, we therefore use doctors per capita and political contributions by insurers as our two instruments for the number of mandates in each state. The 2014 study also finds that other variables that might be expected to affect the passage of mandates, such as the political party of the governor and state legislature or the proportion of individuals with private health insurance, were in fact insignificant.

(2) The QCEW is publicly available on the Bureau of Labor Statistics Web Site: http://www.bls.gov/cew/home.htm

(3) Summers intended for his model to apply to a range of mandated benefits, from workers' compensation to health insurance benefit mandates. A complication arises when applying the Summers model to health insurance benefit mandates--namely, that health insurance is a lump-sum benefit. Working 1 more hour at the margin rarely leads to a marginal increase in health insurance coverage. This difference led Feldman (1993) to propose his own model, in which mandating generous health insurance coverage leads employees to work more hours to bring their total monetary compensation back up, restoring the balance of nonhealth consumption with health insurance. In the Feldman model, the effect of health insurance mandates is to increase hours worked among the employed while decreasing the total number of employees. Cutler and Madrian (1998) make a similar argument. Whereas the Summers model differs from the two later models on how mandates should affect hours worked, all three models describe mechanisms by which mandates could reduce employment.

(4) Available at http://kff.org/state-category/health-coverage-uninsured/private-sector-coverage/

(5) Small firms are also the most willing to drop coverage in the face of higher prices; Gruber and Lettau (2004) find the price elasticity of demand for insurance to be -0.5 for firms with under 100 employees, but only -0.128 for medium-sized firms, and impossible to calculate for large (1,000+ employee) firms because they essentially all purchase insurance.

(6) See below section for further discussion of self-insurance.

(7) Jensen and Gabel (1992) find that mandates do decrease the chance that employers offer health insurance. Gruber (1994b) and Kaestner and Simon (2002) find that mandates do not significantly reduce the number of individuals with employer-provided health insurance, and Sloan and Conover (1998) and van der Goes, Wang, and Wolchik (2011) find that they do. Kaestner and Simon (2002) find some evidence of wage decreases from mandates but conclude those decreases are likely spurious.

(8) Lennon (2015) provides a search-theoretic model in which insurance mandates cause workers with high expected health costs to experience lower earnings, higher unemployment, and longer unemployment spells.

(9) Kaestner and Simon (2002) and Meer and West (2011) study firms with fewer than 100 employees; Mathur (2010) studies firms with fewer than 50 employees, with a special focus on firms with 5 or fewer.

(10) We use an updated version of these data in this article. See Laudicina, Gardner, and Holland (2011) for a description.

(11) Quote is from the U.S. Bureau of Labor Statistics Web Site: http://www.bls.gov/cew/cewfaq.htm

(12) The literature has found robust evidence ex post moral hazard in health insurance (i.e., more generous health insurance coverage means higher health spending); see, for instance, the Aron-Dine, Einav, and Finkelstein (2013) discussion of the RAND health insurance experiment.

(13) The data are publicly available at www.followthemoney.org

(14) See Appendix Table A1 for the first-stage IV results. Our instruments pass the F-test for instrument strength (F = 41) and Hansen's J-test for overidentification (J = 1.87).

(15) Our results are also robust to using the raw number of firms rather than the logged value. For instance, the IV coefficient for firms with 1,000 or more employees is 2.173 in this case (statistically significant at the 5 percent level).

(16) Results available upon request. We do not include this measure in our main specifications because income is most certainly endogenous to employment, particularly firm size.

DOI: 10.1111/jori.12164
TABLE 1 Summary Statistics: State-Level Independent Variables, 1996-2010

Variable                          Mean      Standard Deviation  Minimum

Total mandates                       30.70            10.49        6
Health political spending     1,901,878        5,324,597         279
Insurance political spending    714,697        1,793,088         147
Doctors per 1,000 people              2.64             0.97        1.48

Variable                        Maximum

Total mandates                        62
Health political spending     92,347,416
Insurance political spending  20,680,858
Doctors per 1,000 people               8.73

Notes: All variables used are from state-level data from 1996 to 2010,
and have 620 observations. Spending variables are measured in real 2011
dollars. Data on physicians per capita come from the U.S. Department of
Health and Human Services' Area Resource File. Data on political
contributions come from the National Institute on Money in State
Politics. Data on mandates are from Laudicina, Gardner, and Holland
(2011).

TABLE 2 Summary Statistics: Employment by Establishment Size (State
Level)

Firm Size    Mean     Standard Deviation   Minimum   Maximum

   <5        156,506        178,297        16,408    1,163,052
    5-9      176,020        178,305        19,726    1,006,772
   10-19     234,964        239,531        25,394    1,412,658
   20-49     363,013        385,646        33,913    2,278,757
   50-99     284,196        310,586        21,691    1,830,533
  100-249    355,522        383,637        20,820    2,158,862
  250-499    203,077        218,189             0    1,188,226
  500-999    147,883        159,183             0      815,361
1,000+       238,420        263,512             0    1,211,509
All firms  2,159,602      2,295,256       169,536   12,900,000

Notes: All variables used are from state-level data from 1996 to 2010,
and have 620 observations. Data are from the Quarterly Census of
Employment and Wages.

TABLE 3 Summary Statistics: Number of Establishments by Employment Size
(State Level)

Firm Size   Mean    Standard Deviation  Minimum   Maximum

   <5       97,430        123,515         9,942    911,784
    5-9     27,195         27,620         3,121    155,412
   10-19    17,799         18,121         1,983    105,987
   20-49    12,263         12,993         1,186     76,914
   50-99     4,206          4,591           330     26,893
  100-249    2,407          2,602           142     14,660
  250-499      604            650             0      3,543
  500-999      221            239             0      1,237
1,000+         114            124             0        612
All firms  162,239        188,675        17,101  1,267,235

Note: All variables used are from state-level data from 1996 to 2010,
and have 620 observations. Data are from the Quarterly Census of
Employment and Wages.

TABLE 4 Effect of Mandates on Employment at Firms of Various Sizes

Firm Size          OLS      Fixed Effects  Instrumental Variables Fixed
                                                      Effects

   <5           6,503 (**)      207.6                  1,650
               (2,605)         (382.5)                (1,318)
    5-9         6,383 (**)     -174.1                   -148.1
               (2,450)         (266.6)                  (554.5)
   10-19        8,847 (**)       18.17                     6.84
               (3,363)         (409.5)                  (795.5)
   20-49       14,208 (**)      660.8                   -694
               (5,469)         (780.4)                (1,566)
   50-99       11,185 (**)      116.7                  -2050
               (4,387)         (556.7)                (1,440)
  100-249      13,631 (**)      851.6                   -361.7
               (5,260)         (554.4)                (1,807)
  250-499       7,259 (**)      343.5                 -1,256
               (2,992)         (414.1)                (1,652)
  500-999       4,802 (**)      349.1                   -611.8
               (2,135)         (600)                  (1,343)
1,000+          7,554 (**)    1,460                    2,767
               (3,180)         (991.5)                (2,543)
Year dummies      No              Yes                    Yes
State dummies     No              Yes                    Yes
Observations      620           620                      620

Notes: Each firm size is run in a separate regression. Standard errors
are in parentheses; the standard errors for the OLS and fixed effects
regressions are robust and clustered at the state level. Fixed effects
and IV regressions control for state population. The instrumental
variables column shows the second stage of IV regressions; the first
stage uses doctors per capita and political contributions by insurers
as instruments for mandates. (**) p < 0.05.

TABLE 5 Effect of Mandates on the Number of Establishments of Various
Sizes

Firm Size            OLS      Fixed Effects  Instrumental Variables
                                                Fixed Effects

   <5            0.039 (***)     -0.002            -0.001
                (0.010)          (0.002)           (0.006)
    5-9          0.034 (***)     -0.001            -0.010 (***)
                (0.009)          (0.001)           (0.003)
   10-19         0.036 (***)     -0.001            -0.013 (***)
                (0.009)          (0.001)           (0.004)
   20-49         0.038 (***)      0.000            -0.008 (**)
                (0.010)          (0.001)           (0.004)
   50-99         0.039 (***)     -0.003            -0.023 (***)
                (0.010)          (0.002)           (0.006)
  100-249        0.040 (***)      0.000            -0.005
                (0.010)          (0.001)           (0.004)
  250-499        0.038 (***)     -0.005            -0.005
                (0.011)          (0.004)           (0.003)
  500-999        0.033 (***)     -0.003            -0.012
                (0.011)          (0.003)           (0.010)
1,000+           0.036 (***)      0.004             0.043 (***)
                (0.011)          (0.004)           (0.013)
Year dummies    No               Yes               Yes
State dummies   No               Yes               Yes
Observations   620              620               620

Notes: Each firm size is run in a separate regression. Standard errors
are in parentheses; the standard errors for the OLS and fixed effects
regressions are robust and clustered at the state level. Fixed effects
and IV regressions control for state population. The instrumental
variables column shows the second stage of IV regressions; the first
stage uses doctors per capita and political contributions by insurers
as instruments for mandates. (**) p < 0.05, (***) p < 0.01.
COPYRIGHT 2018 American Risk and Insurance Association, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Bailey, James; Webber, Douglas
Publication:Journal of Risk and Insurance
Geographic Code:1USA
Date:Jun 1, 2018
Words:8620
Previous Article:ADVANTAGEOUS SELECTION, MORAL HAZARD, AND INSURER SORTING ON RISK IN THE U.S. AUTOMOBILE INSURANCE MARKET.
Next Article:DYNAMIC MORAL HAZARD: A LONGITUDINAL EXAMINATION OF AUTOMOBILE INSURANCE IN CANADA.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters