The Economic Impact of Proposition 72 on California Employers

Dr. Aaron S. Yelowitz

University of Kentucky

September 6, 2004

Abstract: This study provides a comprehensive evaluation of the economic impact of the 2003 Health Insurance Act (HIA) using the most recent data available. A number of important results emerge. This “pay or play” mandate results in 1.98 million previously uninsured Californians receiving employer-provided health insurance, nearly double the number cited by most advocates. This represents 31 percent of the uninsured in California. Despite its modest impact on reducing the number of uninsured, the HIA is nonetheless much more expensive than previous studies have indicated. The cost for employers is between $12.8 and $13.2 billion. The costs for the uninsured, approximately $4.4 billion, represent about one-third of the total employer cost. Between $1.5 and $1.7 billion represents a cost shift from government health insurance to employer-provided health insurance. By far the largest single group cost is for those who already have employer-provided health insurance. The premium sharing and dependent requirements of HIA entail costs of between $5.8 and $6.0 billion for employers.

            The degree that employers ultimately bear these costs depends on the types of behavioral responses that may occur, on which the existing literature on HIA is largely silent. One likely possibility is that employers will attempt to shift the cost of the HIA onto employees in the form of lower wages. If employers can shift the entire cost of the HIA onto employees in the form of lower wages, tax revenue falls substantially. The loss in tax revenue could be as high as $3.6 to $4.9 billion from such wage shifting. Between $665 and $860 million of this loss is from reduced income tax collections for California. For many low-wage workers, it will prove impossible for employers to shift wages, because of the California minimum wage. This study finds that there are more than 680,000 California workers earning less than $9.31 per hour in large employers who are either uninsured or on government insurance. The HIA could cause a substantial number of these workers to lose their jobs since employers cannot fully shift the cost of the mandate through wages. This study estimates that more than 67,000 employees are expected to lose their jobs when wage shifting is possible, and more than 151,000 employees are expected to lose their jobs when wage shifting is not possible as a result of the HIA.

Contact information: Dr. Aaron Yelowitz, Department of Economics, University of Kentucky, 335 Business and Economics Building, Lexington, KY 40506-0034. Phone: (859) 257-7634. Fax: (859) 323-1920. Email: aaron@uky.edu. URL: http://gatton.uky.edu/faculty/yelowitz. The author acknowledges the helpful comments of David Neumark, Chris Bollinger, Chris Clark, Attila Cseh, Larry Yelowitz, and Jason Yelowitz.

            Dr. Aaron S. Yelowitz is currently an Associate Professor in the Department of Economics at the University of Kentucky. He also is a joint faculty member in the Martin School of Public Policy and Administration at the University of Kentucky. He is also a Research Associate at the National Bureau of Economic Research, a Faculty Affiliate at the Joint Center for Poverty Research, and a Research Associate at the Institute for Research on Poverty, and the economics department liaison for the UK Center for Poverty Research. He serves as an associate editor for the Journal of Public Economics.

            Dr. Yelowitz received his doctorate in economics from the Massachusetts Institute of Technology in 1994, and his bachelor’s degree in business and economics from the University of California, Santa Barbara in 1990. Prior to his arrival at the University of Kentucky in 2001, he was an Assistant Professor in the Department of Economics at the University of California, Los Angeles. He has received funding from the Association for Public Policy and Management, the Economic Research Initiative for the Uninsured, the National Academy of Science, The Employment Policies Institute, Social Security Administration, the Joint Center for Poverty Research, and the U.S. Department of Agriculture.

            Dr. Yelowitz’s publications focus on the economic consequences of Medicaid, public housing, SSI, WIC, Food Stamps, and AFDC/TANF. Many of the papers focus on the linkages between different poverty alleviation programs, such as the disincentives to leave AFDC because of the loss of Medicaid health insurance. He has published articles in the The Journal of Political Economy, The Quarterly Journal of Economics, Journal of Health Economics, Journal of Public Economics, The Journal of Human Resources, Economic Inquiry and Pediatric Neurology. He has refereed for more than 25 peer-reviewed publications, including The American Economic Review, The Journal of Political Economy, and The Quarterly Journal of Economics. He has presented his research in more than 50 academic settings, including MIT, Harvard, Yale, University of Chicago, and the National Bureau of Economic Research.

            Dr. Yelowitz has taught graduate classes about public economics and health economics, and undergraduate classes about labor economics, public economics, and poverty and welfare programs. Many of his research papers appear on the graduate syllabi for labor economics, public finance, and health economics at major economics departments throughout the country.

            Dr. Yelowitz’s most recent work focuses on the impacts of living wage mandates and health insurance mandates. He has presented his living wage research at the American Economic Association meetings, to the City of Atlanta Living Wage Commission, and to the U.S. Chamber of Commerce. He also has testified as an expert witness in a recent Santa Fe living wage trial. His most recent living wage study is forthcoming in the peer-reviewed journal Economic Development Quarterly. His scholarly work on health economics has appeared in some of the most prominent general-interest economics journals, including The Quarterly Journal of Economics and The Journal of Political Economy. He has also published chapters concerning health insurance in edited volumes for Russell Sage Foundation and the W.E. Upjohn Institute for Employment Research.

1. Introduction

            The Health Insurance Act (“HIA”) of 2003, or Senate Bill 2 (“SB 2”), is a “pay or play” mandate that requires California employers to pay a fee to the state to provide health insurance unless the employer provides health insurance coverage directly, in which case the fee is waived. The bill was passed by the legislature on September 12, 2003, and signed by then-Governor Gray Davis on October 5, 2003. A referendum on the November ballot, Proposition 72, will allow voters to decide whether the legislation he signed will stand, where a “yes” vote on Proposition 72 will approve the HIA, while a “no” vote will repeal it.

            In the event that an employer chooses to “play” (and offer health insurance), it must offer the minimum benefit as specified by the Knox-Keene act of 1975 or a variety of other regulations. All Knox-Keene licensed plans must provide a set of basic minimum benefits: inpatient and outpatient care, physician services, preventive services, lab and radiology, home health, hospice, and emergency services. This coverage must also include a prescription drug benefit plan. In the event that an employer chooses to “pay”, the employer is responsible for paying 80 percent of a yet undetermined fee established by the Managed Risk Medical Insurance Board (“MRMIB”).

            The mandate imposes different requirements based on firm size. For employers with 200 or more California employees, the mandate begins January 1, 2006 and requires coverage for both the worker and his or her dependents (including spouses and same-sex domestic partners). For employers with 50 to 199 California employees, the mandate begins January 1, 2007, and requires coverage for workers but not dependents. Employers with 20 to 49 employees are exempt unless the state of California provides a tax credit equal to 20 percent of the employer’s net cost of the fee, in which case they face the same requirements as those with 50 to 199 employees. Smaller employers are entirely exempt.

            The HIA is more than a mandate to cover uninsured workers. It also introduces a new “premium sharing” mandate for currently covered workers. Employers are required to contribute at least 80 percent of premium costs for all eligible workers. This provision was presumably intended to minimize the financial impact of HIA on newly covered employees, many of whom may have difficulty paying for their own health premiums. The provision also reflects the sentiment of HIA’s authors to “level the playing field” between employers who provide and pay for a significant amount of health insurance premiums, and those who do not. For many low-wage workers, the employer cost is even higher than 80 percent of premium costs. In firms required to pay for individual (family) care, enrollee contributions for workers whose wages are less than 200 percent of the poverty line for a household of one (three) are capped at 5 percent of wages, further raising the employer’s cost. As it is written, HIA counts only individual earnings, as opposed to family income for this poverty determination.

            The mandate specifies that employees qualify if they work 100 hours a month for three months. Those who meet this work requirement, and are employed at firms of an appropriate size, are required to pay (at most) 20 percent of the cost of coverage. HIA allows employers to deduct this payment directly from their employees’ paychecks.

            In addition to providing new coverage to some uninsured workers and imposing a “premium sharing” mandate, HIA also entails substantial shifting of costs onto employers because it “crowds out” other forms of health insurance. In addition to the costs of the uninsured, which has been the main focus of almost all existing studies, there is a substantial shifting of responsibility for paying for health insurance from the government to employers (akin to a tax increase), from currently covered employees to employers, and across employers (although this cost is often not counted as a new one to employers since it is an inter-employer transfer). ^,

            HIA mandates that employers provide coverage for workers already receiving benefits through government sponsored insurance programs such as Medicaid, Medicare, and Champus/Tricare, programs targeted to the poor, elderly, and those in the Armed Forces. In doing so, the bill vastly expands its reach to the more than one million individuals in California who currently receive health insurance coverage through only these programs. In addition, the bill creates a massive cost shift from these government programs to private businesses. This is particularly damaging due to the fact that these programs are at least partially, and often fully, funded by federal dollars. Currently, many employers only offer a supplement to employees that qualify for government insurance. Under HIA, the provision of this supplement does not count as “playing” and employers must provide full health coverage to recipients regardless of their current insurance status or employee desire.

            Employees who qualify for full medical insurance through government programs and are currently working at least 100 hours per month are classified as enrollees and are mandated to pay for coverage under HIA. Enrollees may voluntarily provide, to MRMIB, the information necessary to determine eligibility for Medi-Cal (the name of Medicaid in California) or the Healthy Families Program. In the event that an enrollee is determined to be eligible for these benefits, or is currently receiving the benefit, their enrollee contribution is refunded. The employer’s contribution, however, is not refunded and is used to pay the state’s contribution under the matching funds portion of Medicaid. In the event that the enrollee is receiving Medicare or Champus/Tricare coverage, the enrollee is additionally provided coverage either through the employer’s private plan or a contribution to the state fund. Any wraparound plan to supplement the government insurance programs would have to be offered in addition to this mandated coverage.

            It is difficult to arrive at a precise calculation of the impact of the HIA because a number of key, unknown variables will come into play if the law takes effect. Some of the provisions of the law leave a great deal of room for interpretation or are simply unknown. Perhaps what is most important, it is not known what the annual “fee” per employee will be from the “pay” part of the mandate. Moreover, the legislation gives the state the power to determine what is covered by the state plan against which private coverage is measured. If employers provide “inadequate” health coverage, they will be forced to find other coverage or have their workers covered by the state plan.

            Despite these difficulties, the goal of this study is to evaluate the economic impact of HIA on employers in California. It is clear that the uninsured present a real problem for policy makers to try to solve, but the fundamental question is whether the HIA is a good way to solve it relative to other policy options. The key reason to focus on employers is that passing the health care obligation onto them has the potential for many labor market distortions including job loss and wage shifting. These sorts of rational employer reactions to the HIA mandate create economic inefficiency (known as “deadweight loss”) for three critical reasons. First, some low-skill and less-experienced workers who would otherwise be able to find jobs and add value to the economy instead will become unemployed. Second, HIA has far-reaching effects on many currently insured workers by placing greater government regulation on premium cost sharing and benefits offered. To the extent that employers and employees already have agreed on an acceptable compensation package, this sort of intervention makes both parties worse off. Third, by reinforcing the link between employment and health insurance, the HIA potentially exacerbates issues like “job lock”.

            In the current analysis on employer costs, I assume that the generosity of the state’s plan (and the fee) is equivalent to the expense of the median health insurance plan. With the current ambiguity in the law, many researchers have taken an approach like this. With that in mind, a number of important results emerge from this study. The fully phased-in mandate (covering employers with 20 or more employees) results in 1,983,420 previously uninsured Californians receiving employer-provided health insurance. This represents 31 percent of the 6.4 million uninsured in California. Despite its modest impact on reducing the number of uninsured, the HIA is nonetheless much more expensive than previous studies have indicated. The cost for employers from the full mandate is between $12.8 and $13.2 billion. The costs for the uninsured, approximately $4.4 billion, represent about one-third of the total employer cost. Between $1.5 and $1.7 billion represents a cost shift from government health insurance to employer-provided health insurance. By far the largest single group cost is for those who currently have employer-provided insurance. The premium sharing and dependent requirements of HIA entail costs of between $5.8 and $6.0 billion for employers.

            The degree that employers ultimately bear these costs depends on the behavioral responses that may occur, on which the existing literature is largely silent. Based on existing economic studies, one likely possibility is that employers will attempt to shift the cost of the mandate onto employees in the form of lower wages (Gruber, 1994). When employers can shift the entire cost of the mandate onto employees in the form of lower wages, tax revenue falls substantially. The overall loss in tax revenue could be as high as $3.6 to $4.9 billion. Between $665 and $860 million of this loss is from reduced income tax collections on the part of California. For many low-wage workers, it will prove impossible for employers to shift wages, because the California minimum wage is $6.75 per hour. This study finds, for example, that there are more than 680,000 workers earning less than $9.31 per hour in large employers who are either uninsured or on government insurance. The family mandate could cause a substantial number of these workers to lose their jobs since employers cannot fully shift the cost of the mandate onto them. Using reputable employment elasticity estimates from the economics literature, 67,521 employees are expected to lose their jobs when wage shifting is possible, and 151,482 employees are expected to lose their jobs when wage shifting is not possible as a result of the HIA.

            The remainder of the paper is arranged as follows. Section 2 reviews previous cost estimates on the HIA and explains why such divergent results emerge from the different studies. Section 3 reviews the three microdata sets used in the economic analysis. The first two, the Current Population Survey (“CPS”) and County Business Patterns (“CBP”) are both published by the Census Bureau. The third, the California Employer Health Benefits Survey (“CEHBS”) is published by the Kaiser Family Foundation and Health Research and Educational Trust (“KFF/HRET”). This section also contrasts features of the CPS with the California Health Interview Survey (“CHIS”), a data set promoted by some researchers. Sections 4 and 5 analyze the coverage and cost effects of the legislation, assuming no behavioral responses on the part of the employers. Section 6 then presents evidence on two likely avenues through which employers will adjust their behavior – wage shifting and layoffs. Since layoffs result in fewer workers being covered under the HIA, this section also revises the cost estimates in light of these responses. Finally, it studies the impact of HIA on different socioeconomic groups. Section 7 concludes with remaining open questions and policy alternatives to HIA.

2. Why are the Proposition 72 cost estimates so divergent?

2a. Existing estimates

            Broadly, there are five groups that have produced widely cited and wildly different estimates of the impact of SB 2 on employers in various reports or “fact sheets.” These groups include The Employment Policies Institute, the California Chamber of Commerce, the UCLA Center for Health Policy Research, The Institute for Industrial Relations, and the California Medical Association.

            The range of costs to employers varies by an order of magnitude. The estimates from California Chamber of Commerce (and to a lesser extent The Employment Policies Institute) are often used by opponents of Proposition 72. This current study is the second report I have written for The Employment Policies Institute. In Yelowitz (2003), I found that the cost to employers for the fully phased-in mandate was $11.4 billion based on data from 2001 (and expressed in 2001 nominal dollars). In this current study, which uses newer cost data, the cost to employers of the full mandate is between $12.8 and $13.2 billion (and expressed in 2003 nominal dollars).

            Two reports have been produced by the Los Angeles County Economic Development Corporation (“LAEDC”) and are associated with the California Chamber of Commerce. In Kyser, et al. (2003), the authors find that HIA mandate would cost employers $5.7 billion (expressed in 2003 nominal dollars). In a more recent paper, Baker, et al. (2004) find a sharply lower “bottom line” cost to employers of $3.4 billion (again, expressed in 2003 nominal dollars). By inflating the cost estimate to nominal 2007 dollars using double-digit growth rates, their employer cost of $5.3 billion appears similar to the previous Kyser, et al. (2003) study.

            The figures from the UCLA Center for Health Policy Research, the Institute for Industrial Relations at Berkeley, and the California Medical Association (“CMA”) are often used by advocates of Proposition 72. It is important to note that of these groups, only the CMA has a widely published cost estimate for employers. The researchers associated with the UCLA Center for Health Policy Research have not published a total cost estimate to employers, but have published a sheet estimating that 1,070,000 uninsured people will be covered by SB 2 (Brown, et al., 2003). Nonetheless, several researchers affiliated with the UCLA group apparently believe the cost of HIA will be much lower. In a recent court filing criticizing the Kyser, et al. (2003) study, E. Richard Brown stated “a better estimate of the cost of extending coverage to currently uninsured employees under Proposition 72 would be no higher than $2.7 billion, and likely far lower than that” (Brown, 2004). Another researcher affiliated with the center, Gerald F. Kominski, also stated in a court filing that using the UCLA study’s estimates on the number of uninsured, the total after-tax cost to employers and employees would be $1.8 billion, and with the Kyser, et al. (2003) estimates of the uninsured, no more than $2.7 billion (Kominski, 2004). Kominski believes that it is a “near impossibility” that Proposition 72 will cost $7 billion to employers and employees.

            Three reports have been produced by the researchers affiliated with the Institute for Industrial Relations at Berkeley. As with the UCLA group, the Berkeley affiliates have not published a “bottom line” estimate on the cost to employers. In Dube and Reich (2003), the authors estimate that the median covered California business will see annual increases in costs of $1,343 per worker it newly insures, after deducting corporate income tax deductions and expressed in 2002 dollars. In Dube (2003b), the author finds that 1.56 million people (1.16 million workers) would be covered by SB 2 based on 2001 data. Presumably, using the numbers from these two studies would produce an estimated employer cost for the uninsured of roughly the same magnitude as the UCLA researchers, perhaps $1.6 billion (the product of $1,343 per worker and 1.16 million workers, and expressed in nominal 2002 dollars). In a final study, which is no longer posted on the Berkeley web site, Dube (2003a) finds that 650,000 Medicaid recipients are eligible for SB 2, and that by shifting the responsibility for their coverage from the federal government and state government to employers, California would save $620 million annually. As the author notes, however, “Since the employer’s plan will become the primary insurer, the bulk of this cost will be shifted from the taxpayers to the employers” (Dube, 2003a, p. 3). Since the state only pays half of the Medicaid cost (the federal match rate is 50 percent), an implication of the Medicaid shifting would be that the pretax cost to employers is in the range of an additional $1.24 billion beyond the cost of covering the uninsured.

            Finally, the California Medical Association (“CMA”) has published a widely cited estimate that is usually thought of as the “lower bound” for the costs of HIA. In a one-page fact sheet (CMA, undated), the CMA estimates that the cost for extending the mandate to employers with 50 or more employees is $1.3 billion. They also estimate that the fully phased-in mandate for the uninsured would cost employers $1.7 billion. In CMA (2003c), which announced these estimates, the CMA also stated that “This legislation is also expected to provide $700 million in savings to the state’s Medi-Cal system and reduce inappropriate use of emergency rooms and the workers’ comp system by workers who lack health insurance.” As in the Dube (2003a) study, this Medi-Cal savings comes at the expense of employers but the CMA neglects to include this in the employer’s cost. Their estimate implies that the pretax cost of Medi-Cal shifting is in the range of an additional $1.4 billion to employers.

2b. How are costs counted?

            There are several key reasons why the published estimates of the cost to employers of SB 2 vary so much. First, the treatment of currently covered employees in firms that pay less than 80 percent of premiums varies. Related to this, the treatment of the “poverty subsidy” varies. Second, the treatment of insured workers who are not covered by their own employer plan varies, where “insured” can mean coverage through government insurance, private insurance, or a spouse’s plan. Third, the count of the number of uninsured workers (and dependents) varies. Fourth, the treatment of corporate tax deductions varies. Finally, the treatment of health care inflation varies.

            The first issue is how the studies deal with the “premium sharing” part of the SB 2 mandate. The HIA mandates that employers with 50 to 199 employees provide health insurance for single workers and pay for at least 80 percent of the premium cost. It also mandates that employers with 200 or more employees provide and pay for at least 80 percent of the cost of a single or family plan (as applicable for the employee). If a tax credit were implemented for employers with 20 to 49 employees, employers ultimately would be responsible for 64 percent of the premium cost of a single plan. This “premium sharing” part of the mandate is ignored by all of the studies on employer costs except the EPI studies. If the employer provides health insurance but pays less than the mandated percentage, the legislation entails redistribution from the employer to the employee, and is a true cost to the business. Even though California employers nearly meet the premium sharing part of the mandate requirements on average, there is a great deal of dispersion with some employers paying more than 80 percent and others paying far less. KFF/HRET used the CEHBS to estimate that in 2002, 20 percent of small or medium employers and 21 percent of large employers did not cover 80 percent of the premium costs of a single plan, and approximately 50 percent of large employers did not cover 80 percent of the premiums for a family plan. The results in Section 4 of my current study reveal that nearly 500,000 Californians who qualify for the HIA are covered by employer insurance where the employer pays for none of the costs. From the employer’s viewpoint, the additional cost of paying for these individuals is the same as for uninsured individuals under the mandate.

            In addition to this “premium sharing” mandate for all eligible employees, the HIA also mandates that low-wage workers pay no more than 5 percent of wages toward the cost of their health insurance. In this case, the employer would be responsible for more than 80 percent of the premium costs. In firms required to pay individual health care, enrollee contributions for individuals whose wages are less than 200 percent of the poverty line, or $18,620 in 2004, are capped at 5 percent of wages. For employers that are required to offer family coverage, enrollee contributions are capped at 5 percent of wages for employees earning less than 200 percent of the poverty line for a family of three, or $31,340 in 2004. For a full-time, full-year worker, these limits correspond to hourly wage rates of $8.95 and $15.06, respectively. As it is written, HIA counts only individual earnings, as opposed to family income or full-time equivalent earnings at various wage rates toward this poverty determination. In Yelowitz (2003, Table 7), I found that based on this narrow “individual earnings” measure, as many as 2.8 million Californians would qualify for the poverty subsidy and cost employers even more. Despite the potential for this poverty subsidy to raise costs, I have completely excluded this factor in my cost estimates because of its ambiguity. Ignoring it yields underestimates of the true cost to employers.

            The second factor for the widely varying employer costs is the treatment of people who are covered through Medicaid, Medicare, Champus/Tricare, privately-purchased health insurance, or a spouse’s plan. Crowding-out Medicaid, Medicare, and Champus/Tricare in favor of employer insurance represents a shifting of costs from the state and federal government to employers. Crowding-out privately-purchased health insurance represents a shifting of costs from the employee to the employer. The last group, crowding-out a spouse’s plan for a worker’s own plan, is a shifting of costs from one employer to another and thus not viewed as a new cost to employers as a whole (though the logic is not really as simple as this). From the summary above, it appears that many researchers are aware of these sorts of crowding-out, but only the EPI studies count this as a legitimate cost to business.

            Third, all researchers agree that uninsured workers and dependents who are covered by SB 2 are a cost to the employer. But there is disagreement on the count of uninsured workers (and dependents) covered by SB 2. Some studies have used insurance counts from the CHIS (for example, Brown, et al., 2003). Others (Dube 2003a, 2003b; Yelowitz, 2003, 2004) have primarily used the CPS. Finally, others (Dube and Reich, 2003; Baker, et al., 2004; Kyser, et al., 2003) have primarily relied on aggregate employment data or firm level data, supplemented with auxiliary data sources. The key difference here is that the CHIS data gives substantially lower estimates of the number of uninsured than the CPS, and some researchers seem to believe the CHIS is a better data set. A discussion of the merits of the data sources is postponed until Section 3. For all its discussion, however, the debate about the correct number of uninsured covered by SB 2 is not the main driver of total employer costs. For example, in my current study, if the costs of the uninsured were cut in half, the cost to employers would still be between $10.6 and $11.1 billion, rather than the estimated $12.8 to $13.2 billion.

            Fourth, the treatment of federal and state corporate income tax deductions varies. Every cost estimate touted by advocates of SB 2 discounts the cost of the mandate by the presumed “tax savings” from reduced corporate income tax payments. One recent study promoted by opponents of SB 2 (Baker, et al., 2004) also discounts employer costs for the corporate income tax, but several others do not (Yelowitz 2003, 2004; Kyser, et al., 2003). The studies that do include this tax saving often use a tax rate in the range of 40 percent, which inevitably leads to a substantial discount of employer costs. For example, in one study touted by opponents, the estimate of the employer’s cost falls from $5.775 billion to $3.423 billion by including this tax savings (Baker, et al., 2004, Table 8).

            It is important to understand that the corporate income tax is a tax on corporate profits. Thus, even some large corporations with many assets holdings and high market valuations will face low corporate tax rates because accounting profits can be volatile from year-to-year. Joel Friedman of the Center on Budget and Policy Priorities recently published an informative paper that illustrates the true reach (or lack of reach) of the corporate income tax. Friedman (2003, p. 4) states that of the 27 million businesses that filed tax returns in 2000, only 2.2 million (or about 8 percent) were subject to the corporate income tax. He also clears up the perception that the statutory federal rate of 35 percent is the most appropriate number. He reports on page 6 that:

“The corporate income tax rate is typically thought to be 35 percent. The reality is more complicated. The 35 percent rate is the highest statutory corporate rate; lower levels of corporate income are taxed at lower rates. The first $50,000 of taxable corporate income faces a 15 percent tax rate, and the next $25,000 is subject to a 25 percent rate. From $75,000 to $10 million of taxable profits, corporations pay a 34 percent rate. For taxable income above $10 million, the rate is 35 percent. These lower graduated rates phase out for corporations with larger incomes.”

            Friedman (2003) cites a Congressional Research Service (“CRS”) study that shows the corporate income tax rate has averaged 26.3 percent for non-financial, domestic corporations, or about one-quarter lower than the 35 percent statutory rate. Finally, the study notes that “C” corporations are subject to the corporate income tax, but the profits of businesses other than “C” corporations are subject to the individual rather than the corporate income tax. The number of “C” corporations peaked at 2.6 million in 1986, and has declined since then. At the same time, there has been rapid growth in another type of corporation, known as “S” corporations. “S” corporations do not pay corporate income tax, but rather pass through profits to their shareholders, who in turn include this business income on their individual income tax returns.

            Finally, a recent, revealing study by the General Accounting Office (GAO, 2004), covering the “boom years” of 1996 to 2000, offers other insights on the corporate income tax. The GAO compared tax liabilities of foreign and U.S. controlled corporations from 1996 to 2000 by examining the Internal Revenue Service’s Statistics of Income (“SOI”) samples of corporate tax returns, and found that a majority of all corporations reported no tax liability during these years. Overall, 71.3 percent of foreign controlled corporations (“FCCs”) and 61.3 percent of U.S. controlled corporations (“USCCs”) paid no tax (GAO, 2004, p. 15, Table 4). In addition, 94 percent of USCCs and 89 percent of FCCs reported tax liabilities of less than 5 percent of their total income (GAO, 2004, p. 7). Even among large corporations – those with assets of at least $250 million or gross receipts of at least $50 million in constant 2000 dollars – an estimated 82 percent of USCCs and 76 percent of FCCs reported a tax liability of less than 5 percent of their income (GAO, 2004, p. 16-17, Tables 5 and 6).

            These tax return findings from the GAO show the ability of corporations to minimize their tax liabilities. Corporations that pay zero federal corporate income tax clearly do not face a marginal federal income tax rate of 35 percent on average, but one that is much lower. Thus, the tax savings by assuming the highest marginal rate is overstated. The GAO report shows that an overwhelming percentage of corporations face an average tax rate of less than 5 percent, again indicating that they are not close to this part of the tax schedule.

            Based on the Friedman (2003) report, the CRS survey, and the GAO findings, one can reasonably conclude that the importance of discounting the mandate’s cost for the corporate income tax has been overstated and misapplied by the studies touted by advocates of SB 2. Given these findings, the burden of proof should be to demonstrate that businesses affected by SB 2 do, in fact, face the kinds of marginal tax rates that are assumed in some studies. To date, not one of those studies has cited any evidence on the effective tax rates that businesses really face.

            Finally, the treatment of health care inflation varies across studies. Most of the studies use outdated premium data, which tends to understate the true impact of the law on businesses. For example, Yelowitz (2003) relies on health coverage and health premium data from 2001. It is widely recognized that premiums have increased dramatically over the past three years, and will likely increase even more prior to the implementation date of the SB 2 mandate in 2006. According to the KFF/HRET, health insurance premiums in California grew by 15.8 percent in 2003, 13.4 percent in 2002, and 10 percent in 2001. Even advocates of Proposition 72 concede that premiums have grown dramatically. Most of the widely available, current studies do not account for the dramatic premium increases in recent years and therefore provide outdated underestimates of the cost impact of SB 2 to employers.

2c. Review of existing research on HIA

            With this background in mind, this subsection will review how each of the widely cited studies on the HIA accounts for the various cost issues. It will follow the ordering of the studies from Section 2a. Table 1 summarizes the key assumptions used by various studies that compute employer costs from HIA. Excluded from the table are studies that do not provide explicit cost estimates.

[INSERT TABLE 1 HERE]

            First, Yelowitz (2003) analyzed of data from a number of sources, the most important being the March 2002 CPS Annual Demographic Survey. The CPS is administered by the Bureau of Labor Statistics and the Census Bureau. This survey provides health insurance estimates for the 2001 calendar year. The study also relied on data from California’s Employment Development Department (“EDD”) and the 1977 National Medical Care Expenditure Survey (“NMCES”).

            The study revealed that the total cost to employers of the SB 2 mandate is $9.96 billion if it covers only employers with 50 or more employees, and is $11.36 billion if it covers employers with 20 or more employees (Yelowitz 2003, Table 5, p. 6). For the larger mandate, $4.56 billion is spent on the uninsured. This represents approximately 40 percent of the newly mandated cost. An additional $4.39 billion is spent on raising the employer’s premium-sharing percentage to 80 percent threshold for those with more modest employer coverage or covering dependents. The study also found that there are significant costs to employers from the state of California “shifting” those with government health insurance onto the employers, as intended by the HIA. The newly mandated cost for employers from those with government insurance (Medicaid, Medicare, or Champus) exceeded $1 billion.

            The estimates of coverage and costs contained in Yelowitz (2003) were significantly higher than other publicly released estimates because other estimates ignored several categories of individuals. These groups include employees currently receiving government insurance (such as Medicaid, Medicare, or Champus) or choosing to pay for their own private coverage – groups specifically covered under the HIA. Moreover, the HIA requires employers to pay for at least 80 percent of the premiums for a single or family plan (depending on firm size). Many employers pay far less than this percentage, and their costs will go up as a result of the mandate. Excluding these large categories of people, and assuming that all employer-provided insurance will meet the rich mandated benefits of the “play” portion of the HIA significantly understates the true cost and impact of this mandate.

            The cost estimates from Yelowitz (2003) ignored several issues that would modestly affect the conclusions. First, although the number of workers who would receive the “poverty subsidy” was computed, the increased cost from such workers to the employer was not calculated. Second, the cost estimates use information from 2001, while premiums have increased dramatically since that time. Third, the analysis relied solely on health insurance premium data contained in the CPS, even though the KFF/HRET publications generally suggested higher premium costs. Fourth, the cost estimates do not include any discount for corporate tax savings. As discussed in the previous section, my reading of the literature is that the importance of the corporate income tax, with regard to cost savings, has been massively overstated, and if the decision boils down to ignoring corporate income taxes or assuming the highest marginal tax rate, the evidence strongly suggests the first option is more realistic. Fifth, the study counted as an employer cost the crowding out of spouse’s plan for a worker’s own plan. As discussed earlier, although the inter-employer transfer is not neutral in terms of employer costs, there are admittedly some cost savings to small and medium employers. The current study correctly accounts for this issue, but the substantive conclusions remain unchanged.

            In Kyser, et al. (2003), the authors find that HIA mandate would cost employers $5.7 billion (expressed in 2003 nominal dollars) and that the total cost would be $7.2 billion (Table 6, p. 14). The authors relied on publicly available aggregate employment data from Report 524 of the California EDD for the third quarter of 2001, as well as aggregate health insurance information from the 2002 CEHBS publication produced by the KFF/HRET. The authors compute the number of workers not offered health insurance by firm size (using information from Chart 2 in the KFF/HRET publication), and find that 1,031,858 such workers in firms with 20 or more employees (Table 4). They also find that 1,068,052 workers were offered coverage but refuse, using a take-up rate of offered insurance of 90 percent (Table 5). Thus, they find that 2,099,910 workers would be eligible for the full HIA mandate, in various firm size groupings. They compute an employer cost of approximately $3.7 billion for workers in medium employers (using 80 percent of the single premium from the KFF/HRET publication), and a cost of approximately $2.1 billion for workers in large employers (using 80 percent of a weighted average of the premiums for single and family plans). They argue that because of the poverty subsidy, their estimates of SB 2 costs to businesses are conservative.

            The Kyser, et al. (2003) study has received a great deal of attention and was prominently promoted by the California Chamber of Commerce. It also received a fair amount of criticism. One of the most germane is the number of uninsured workers covered by SB 2. In a recent court filing, Brown (2004) questions the “LAEDC estimate of 2,099,910 employees not currently receiving health care coverage at work and presumably eligible for Proposition 72 coverage. That estimate is almost double what it should be.” Brown, et al. (2003) find that 1,070,000 uninsured workers and dependents are covered by SB 2. A more careful reading reveals that these authors only estimate 698,000 workers eligible for SB 2, so the gap between their estimates is even larger. Based on my own work, and that of Dube (2003b), I believe that both the LAEDC and UCLA estimates of the number of covered workers are wrong. Dube (2003b) reports that 1.2 million workers are eligible through their own employer in 2001, and in unpublished tabulations of my own work, I found that 1.7 million uninsured workers are eligible. Both of these findings come from the March 2002 CPS. In my current study, I find that 1.5 million uninsured workers would be covered by their own employer using March 2003 data.

            One valid criticism that Brown (2004) offered is that a substantial number of workers who turn down coverage from their employer may have insurance from a spouse; he argues that nearly three-quarters of those who turn down coverage are, in fact, insured. To the extent that these workers are covered by a working spouse whose employer meets the mandate’s requirements, then the costs generated by these workers do represent a neutral inter-employer transfer. To the extent that those workers are covered by a spouse’s plan that does not meet all of the mandate’s requirements, or to the extent they are covered by government or privately-purchased health insurance, then some or all of the costs generated by these workers are indeed a new cost to employers.

            Baker, et al. (2004) provide revised estimates from the earlier study by many of the same authors. Although these authors continue to use the same California EDD data from 2001, they use an updated chartbook from the 2002 CEHBS produced by KFF/HRET that was prepared to shed light on the SB 2 mandate; the new chartbook grouped employers according to the mandate’s provisions. In addition, premium information was taken from 2003 rather than 2002. In the new study, the authors found that 1,773,394 uninsured workers (from 2001) would be eligible for the SB 2 mandate, a number much closer to my own tabulations for the 2001 calendar year from the March 2002 CPS.

            The authors use updated premium information (and then convert their estimates into nominal 2007 dollars), incorporate a rough measure of the impact of the poverty subsidy, and adjust their cost estimate downward by assuming that the aggregate federal and state corporate marginal income tax rate is 40.72 percent. In doing so, they find a cost to employers of $3.4 billion in 2003 dollars, and $5.3 billion in 2007 dollars.

            Although one can debate about the precise costs to employers for uninsured workers from the two studies promoted by the California Chamber of Commerce (and legitimate concerns about over-estimating the number of uninsured), my fundamental criticism is that they miss the costs from workers who already have insurance. Their “bottom-line” estimates for the cost to employers make no adjustment for workers and dependents currently receiving government insurance although this group is discussed in Baker, et al. (2004, Section VI). More surprisingly, they do not account for the effects of premium mandate on employers even though KFF/HRET has published estimates of the distribution of premiums. Although my own work would suggest lower employer costs for uninsured workers than either of the two reports (when taking the most similar comparisons), because the LAEDC/Chamber reports ignore these other legitimate employer costs, my results remain much higher. Because of these methodological concerns, the cost estimates in Kyser, et al. (2003) and Baker, et al. (2004) should be viewed with considerable skepticism.

            Another study that has received a great deal of attention is a fact sheet of Brown, et al. (2003). Although this fact sheet does not provide any estimates of costs, Brown (2004, p. 3) states that he believes it is “the one relied upon by the Legislative Analyst’s Office in its Fiscal Impact Statement for Proposition 72.” The authors find that 1,070,000 uninsured would be covered by full SB 2 mandate. They also estimate that 698,000 workers are eligible for SB 2. Virtually all advocates for Proposition 72 use the Brown, et al. (2003) estimate of the number of uninsured.

            The authors use the CHIS in conjunction with the March 2001 and 2002 CPS to provide estimates for relevant factors not included in CHIS. Their CHIS sample included uninsured people who were aged 18 to 64, worked for an employer for wages, worked at least 23 hours per week and had been employed in the same position for at least three months. They excluded workers with any insurance coverage. The CPS is used to impute dependent coverage for these uninsured workers.

            There are many serious criticisms of this approach. First, the authors provide no cost estimates for employers, so it is far less ambitious than some other studies. Second, the focus on uninsured workers misses many of the costs to employers, even for the uninsured. With this focus, the authors cannot assess the impact of the HIA on those with government and privately-purchased health insurance. Moreover, by excluding workers with insurance, they understate the number of uninsured dependents. Their method implies that an insured worker with single coverage at a large employer whose children have no insurance would not be included in counts of the uninsured or employer costs. Thus, even if one agrees with their use of the CHIS, the number of uninsured eligible for SB 2 must be understated. Brown, et al. (2003) also estimate far fewer workers eligible for SB 2 than other sources, including Dube (2003b), who is a coauthor of this study. Third, by focusing on the uninsured, the authors ignore the impact of workers who received employer-provided health insurance that pays for less than 80 percent of the premiums. Fourth, by screening exclusively on workers under the age of 65, Brown, et al. (2003) ignores the wasteful, redundant mandated employer costs for older workers who already have Medicare coverage.

            Another study that has received a great deal of attention is Dube and Reich (2003). They analyze the 2003 California Establishment Survey (“CES”), which surveyed business and nonprofit establishments with five or more employees (except for the agricultural industry). The CES was designed by Reich, funded by the UC Institute for Labor and Employment, and conducted during the summer of 2003 by the UC Berkeley Survey Research Center. The response rate is not reported in their study, but they construct weights that they claim solves any nonresponse issues.

            The CES finds that 64 percent of all “businesses respondents” support the health insurance mandate, as do 59 percent of “business respondents” at employers that do not offer health coverage. It is important to note that according to the authors, the overwhelming majority – 91 percent – of “business respondents” were not the owners whose profits would be lowered, but rather hired employees such as personnel department officials who very well could personally benefit from the mandate even if profitability falls. There is no explanation by these authors whether such employee responses are the official positions of the businesses or the respondent’s own personal position.

            Interestingly, one of the CES findings is that 90 percent of the employers that currently do not offer health benefits are in markets where their competitors do not provide such benefits either. This finding seems to invalidate one of the arguments put forth by advocates for Proposition 72, namely that the mandate levels the playing field for “responsible” companies. The argument – that companies that do not provide affordable health care to their employees have an advantage over companies that do – appears to not be terribly important, because the “race to the bottom” has already occurred.

            The most widely cited and used finding from Dube and Reich (2003), however, is their cost per newly covered worker. While many advocates for Proposition 72 use the Brown, et al. (2003) study for their estimate of the affected uninsured, they often use the Dube and Reich (2003) study for the per-employee cost. This study finds that the median covered California business will see annual increases in costs of $1,343 per worker it newly insures (after deducting corporate income tax deductions). To arrive at this number, they use average health premium costs from the 2002 CEHBS published by the KFF/HRET. The total premium was $2,845 for a single plan and $7,471 for a family plan from the 2002 survey. They also estimate a marginal cost per dependent of $2,085, using the aggregate number of dependents in family plans in the CPS. The employer is responsible for 80 percent of these costs, or $2,276 for a single plan, $5,976.80 for a family plan, and $1,668 per newly covered dependent. For employers with 20 to 49 workers, the mandated cost from the single plan would be even lower, $1820.80, because of the state tax credit. From there, they assume a combined federal/state marginal corporate income tax rate of 40 percent, which is very close to the maximum possible rate for most businesses once the federal deduction for taxes paid to the state is taken into account. After taking into account this tax savings, the annual per-employee cost for a newly insured worker would be $1,092.48 for employers between 20 and 49 employees, $1,365.60 for employers between 50 and 199 employees, $3,586.08 for family coverage in firms with 200 or more employees, and $1,000.80 per dependent.

            To arrive at their estimate of $1,343 per newly insured worker, they adjust these single plan baseline values for the fact that “some of these workers who are not insured through their own employer are dependents of spouses. These individuals do not represent added costs to employers but rather shifts in costs between employers.” For large employers, they also add the expected number of dependents per insured worker and the cost per dependent to the calculation.

            Their estimate of $1,343 fails to account for at least some of the issues discussed in Section 2b. First, through no fault of their own, the findings use outdated premium information. Analogous premium data for 2003 would be $3,102 for a single plan and $8,504 for a family plan (rather than $2,845 for a single plan and $7,471 for a family plan). Second, without presenting additional evidence on the marginal tax rates that corporations face in reality, it is not compelling to discount the cost estimates by the maximum rate. Third, and importantly, their number does not account for the premium increases to the employer for already covered employees. Fourth, they are not clear on how they treated workers and dependents with government health insurance or privately-purchased health insurance. All that we can be sure about is they downweighted the cost numbers for inter-employer transfers, which they assume to be neutral. Fifth, and most importantly, their finding is for the median business, yet the distribution of mandated costs across covered employers is clearly bimodal. HIA creates relatively low costs for uninsured workers at employers with 20 to 199 employees, and a distinctly higher set of costs for uninsured workers at employers with 200 or more employees. Rather computing the median cost for each distinct group, Dube and Reich group all of these firms together in coming up with their estimate. They state: “The median California covered business (i.e. with 20 or more employees) will see an annual increase in costs of $1,343 per worker it newly insures.” The key issue here is that the median business is not informative about the cost of the family mandate. Based on third quarter 2001 California EDD data, there were 1,075,523 businesses in California, of which 133,957 had 20 or more employees and a mere 6,664 had more than 250 employees. Clearly the median covered business does not face the family mandate, only the single mandate. If Dube and Reich (2003) had presented the median cost for newly insured workers at medium and large employers separately, those figures would have been more informative (but still subject to the other criticisms).

            Dube (2003a, 2003b) has also written two papers that have received relatively less attention. In fact, one of them (Dube 2003a) is no longer available on the Institute of Industrial Relations web site. In Dube (2003b), he uses the March 2002 CPS to estimate the impact of SB 2 on health insurance coverage. He estimated that there were 6.72 million uninsured Californians in 2001, and that by applying the provisions of the SB 2 legislation for employers with 20 or more employees, 1.56 million uninsured Californians and 1.16 million uninsured workers would be covered by the law. The total number of uninsured who are covered, which is based on the same data as in my earlier study, is somewhat lower than my estimate of 1.95 million uninsured in firms with 50 or more employees, and much lower than my estimate of 2.29 million uninsured Californians using 20 or more employees (Yelowitz, 2003, Table 4). It is also 45 percent higher than the Brown, et al. (2003) estimate of 1.07 million, however, and is only infrequently cited in public policy discussions.

            In a second piece, Dube (2003a), he again uses the March 2002 CPS, this time to estimate the impact of SB 2 on Medi-Cal costs. He finds that 650,000 Medi-Cal enrollees are SB 2 eligible working family members in firms with 20 or more employees. This is virtually identical to the count of 641,239 newly mandated Medicaid enrollees that I found (Yelowitz, 2003; Table 4) for employers with 50 or more employees, but considerably lower than the 693,160 enrollees I found in firms with 20 or more employees. His cost savings to the state is estimated to be $620 million annually, while the estimated cost to employers in Yelowitz (2003, Table 5) is $675 million. Overall, the differences in findings between this study and my own seem relatively modest.

            Dube (2003a) also raises the interesting point that federal Medicaid matching dollars could be lost if employers decide to “play” rather than “pay”. If the employer provides health insurance for the current Medi-Cal enrollee (and dependents), then “Medi-Cal will become at most a supplementary insurer” (Dube, 2003a, p. 2). In this situation, Medicaid will provide wraparound coverage, covering benefits like vision or dental that may not be provided by the employer. Since Medicaid enrollment falls, both state and federal expenditure fall at the expense of employers. If the employer “pays”, then the worker would still receive Medicaid, but the statewide employers’ pool would compensate the Department of Health Services for the state’s portion of covering Medicaid enrollees. Thus, the federal match is not lost for these recipients in this situation. Finally, he is candid about the shifting of costs to employers in this situation: “Since the employer’s plan will become the primary insurer, the bulk of this cost will be shifted from the taxpayers to the employers” (Dube, 2003a, p. 3).

            Overall, Dube (2003a, 2003b) provides only fragments of the information needed to assess the costs to employers. Nonetheless, his use of the CPS for health insurance statistics provides support to the notion that it is an appropriate data set to use for evaluations like these. Moreover, although Dube’s methodology for imputing SB 2 coverage clearly differs from my own, his numbers are only modestly smaller than my own.

            Finally, the CMA has produced a widely cited one-page sheet that estimates the cost of SB 2 to employers at $1.341 billion for employers with 50 or more employees, and $1.689 billion for employers with 20 or more employees (CMA, undated). They use a count of the uninsured (1.56 million) from the Dube (2003b) study. They assume an annual gross cost of $2,400 per worker or adult dependent and $1,100 per dependent minor. The actual marginal cost per dependent in their study is $1,616.40. Thus, the CMA is assuming 60.3 percent of dependents in large employers are children, and 39.7 percent are adults. They also assume a corporate tax rate of 40 percent, but acknowledge that it “may be less for some employers.”

            The CMA estimate is a straightforward, but naive, estimate of the costs to employers. First, they provide only weak justification for their premium numbers, and those numbers are considerably smaller than other sources. Second, they ignore the impact of the premium-sharing part of the HIA mandate, as well as the additional poverty subsidy for low-wage workers. Third, they ignore the costs to employers from workers who currently receive government health insurance or privately-purchased health insurance. Fourth, although they concede that some employers face lower marginal tax rates, they provide no evidence on the percentage of employers that do face such high tax rates. Fifth, their calculations do not account for the dramatic increase in health care premiums in the last three years. Because of these problems, the CMA numbers have very little credibility.

3. Data Description

3a. 2003 March Current Population Survey, Annual Social and Economic Survey

            The primary dataset used in the analysis is the 2003 March CPS Annual Social and Economic Survey (U.S. Department of Commerce, Bureau of the Census, 2003). The March CPS Annual Social and Economic Survey (“ASEC”) was formerly called the Annual Demographic Survey.

            The CPS is recognized as a credible and widely respected survey. It currently surveys nearly 80,000 households for the March supplement, and asks questions that specifically address issues of health coverage and health insurance. It is administered by the Bureau of the Census for the Bureau of Labor Statistics and has been conducted for more than 50 years. The response rate for the March survey is exceptionally high for a voluntary, household-based survey. The sample is scientifically selected to represent the civilian noninstitutional population. The Census Bureau states that the CPS sample provides estimates for the nation as a whole and serves as part of model-based estimates for individual states and other geographic areas. The CPS is conducted by telephone and in-person (and thus includes residences without telephones).

            The March 2003 CPS ASEC surveys 216,424 people across the nation (78,310 households), and 16,779 people in California (5,600 households). When appropriately weighted, I derived a population count of 285,934,600 for the United States, and a population count for California of 35,159,001. The count for California exactly matches published Census tabulations, while the count for the United States appears to be subject to a trivial amount of rounding error. Unless otherwise noted, all estimates in the paper are based on the weighted data.

            The ASEC only identifies 32 of 58 counties in California (either individually or within an MSA). The other 26 counties are not identified in the ASEC. These counties are: Alpine, Amador, Calaveras, Colusa, Del Norte, Glenn, Humboldt, Imperial, Inyo, Kings, Lake, Lassen, Mariposa, Mendocino, Modoc, Mono, Nevada, Plumas, San Benito, Santa Cruz, Shasta, Sierra, Siskiyou, Tehama, Trinity, and Tuolumne. Fortunately, these unidentified counties are sparsely populated; the 32 sampled counties contain roughly 95 percent of California’s population. In total, the 26 unsampled counties had a population of 1,540,112 in the 2000 Census, out of 33,871,648 in the state as a whole.

            In general, using the CPS would be problematic for constructing state-level estimates. Fortunately, the California sample is sufficiently large to overcome these concerns. State-level estimates of the uninsured in the annual March CPS are typically unreliable due to small state sample sizes. Approximately 2,000 to 3,000 unweighted households are needed in a state sample to generate reliable estimates of subpopulations (such as uninsured children below 200 percent of the poverty line).

            The ASEC asks detailed questions about health insurance and work behavior for the entire previous calendar year. Health insurance status is asked for all household members; the survey includes questions about employer-provided health insurance, private health insurance, and government insurance. The CPS does not directly ask people whether they are uninsured. The survey asks about coverage of specific types of insurance and respondents who answer no to all of the categories are considered uninsured. The March 2003 CPS asks about health insurance coverage in 2002. The March CPS asks respondents about coverage at any time during the preceding calendar year so being uninsured reflects lack of health insurance throughout the calendar year. It is thought that the CPS misclassifies insurance status for some people. In the analysis that follows, I use health insurance definitions identical to those of the Census Bureau.

            The CPS is useful as a source of estimates of the insured and uninsured populations at the state level. According to the Census Bureau, the March CPS is perhaps the most widely used source of data on health insurance coverage in the United States. It is the official source of estimates used to allocate federal funding to states for the State Children's Health Insurance Program (“SCHIP”), which amounted to $3.7 billion in Federal Fiscal Year 2002. The March CPS provides reliable estimates of the net change in the number of uninsured people from one year to the next. Even critics of the CPS concede “Despite its limitations, the CPS provides a useful measure of changes over time in health insurance coverage and uninsurance.” (Brown, et al., 2002, p. 61).

            Employment information is elicited for all household members over the age of 15. The survey includes questions on usual hours worked per week, annual earnings, weeks worked per year, industry, and firm size for all adults. In contrast, almost all of the labor market information in the CHIS refers to the sample respondent only. Typically, a single CPS respondent reports for everyone in the household, although telephone callbacks to obtain particular items of information known only by someone else in the household is fairly common. Even researchers who have found “small” effects of SB 2 have used the CPS labor market questions in their research to construct estimates of the mandate’s impact (Dube and Reich 2003, Brown, et al. 2003).

            The CPS also provides state- and county-level geographic identifiers, as well as demographic information for all respondents on age, education, race, ethnicity, gender, marital status, and immigrant status. It also provides sufficient information to identify family relationships across household members, which is critical in forming the “health insurance units” defined by the HIA.

            The Census Bureau also provides imputations for a number of variables that are of policy interest, including the fungible value of Medicare and Medicaid, the employer’s contribution for health insurance, and state and federal income tax liabilities. The subsequent analysis will make use of the employer contribution imputations as well as the tax imputations.

3b. 2003 California Employer Health Benefits Survey

            The CEHBS is a joint product of KFF/HRET. The survey was designed and analyzed by researchers at the KFF and HRET, and administered by National Research LLC (“NR”). The findings are based on a random sample of 864 interviews with employee benefit managers in private employers in California. NR conducted interviews from May to August 2003. The sample of employers was drawn from the Dun & Bradstreet list of private employers with three or more workers. The survey is based on a national employer survey conducted annually by KFF/HRET.

            The survey asked questions about the following types of health plans: Conventional (fee-for-service) plans, Health Maintenance Organizations, Preferred Provider Organizations, and Point-of-Service plans. Conventional plans comprise a very small share of the California market, however.

            The CEHBS is the source of health premium information in numerous studies, including Dube and Reich (2003), Kyser, et al. (2003), Baker, et al. (2004), and the current study. The CEHBS will be used to merge premium and cost sharing information for single and family plan by industry and firm size to individual CPS respondents. By merging this CEHBS data to the CPS, I am able to derive an alternative set of cost estimates that do not rely on the CPS employer contribution imputations provided by the Census Bureau. I work with the 2003 CEHBS data rather than earlier years for several reasons. First, and most importantly, the premium information for 2003 better reflects current conditions in the health care market. Second, in years prior to 2003, the CEHBS sample of employers was poststratified using frequency distributions from Dun & Bradstreet. Concerns about the volatility of counts in recent years led KFF/HRET to use the Statistics of U.S. Businesses conducted by the U.S. Census as the basis for the post-stratification adjustment in 2003. Due to this change, KFF/HRET recalculated the weights for survey years 2000 to 2002. This change in weighting has little impact on worker-based estimates, but did have an impact on estimates expressed as a percentage of employers.

            Two different samples are drawn from the CEHBS. One sample consists of all 3,222 firms that provided answers about their firm size. The other sample consists of 864 completed interviews that provided detailed answers about their health plans; of these firms, 760 offered health insurance plans and when appropriately weighted, represent 7,863,192 covered employees at these firms.

3c. 2001 County Business Patterns

            While the CPS data provides a reasonable estimate of firm size in six intervals, I augment these responses with information from the 2001 CBP, the most recent year available. As will be shown in Section 4, one can make more accurate imputations of SB 2 eligibility by exploiting the fact that the CPS respondent reports firm size, county of residence, and industry.

            The CBP is an annual series that provides economic data by industry at various levels of geographic aggregation. The series is useful for studying the economic activity of small areas, analyzing economic changes over time, and as a benchmark for statistical series, surveys, and databases between economic censuses.

            The CBP covers most of the country’s economic activity. The series excludes data on self-employed individuals, employees of private households, railroad employees, agricultural production employees, and most government employees. Beginning in 1998, data are tabulated by industry as defined in the North American Industry Classification System (“NAICS”). This series has been published annually since 1964 and at irregular intervals dating back to 1946.

            The CBP provides data on the total number of establishments, mid-March employment, first quarter and annual payroll, and number of establishments by nine firm size groups by detailed industry for all counties in the United States and the District of Columbia.

            Most geographic codes are derived from the physical location address reported in Census Bureau programs. The Internal Revenue Service provides supplemental address information. Those employers without a fixed location within a state (or of unknown county location) are included under a “statewide” classification at the end of the county tables. This incomplete detail causes only slight understatement of county employment.

            An establishment in the CBP is a single physical location at which business is conducted or services or industrial operations are performed. It is not necessarily identical with a company or enterprise, which may consist of one or more establishments. When two or more activities are carried on at a single location under a single ownership, all activities generally are grouped together as a single establishment. The entire establishment is classified on the basis of its major activity and all data are included in that classification. Establishment-size designations are determined by paid employment in the mid-March pay period. The firm size group “1 to 4 employees” additionally includes establishments that did not report any paid employees in the mid-March pay period but paid wages to at least one employee at some time during the year.

            Establishment counts represent the number of locations with paid employees any time during the year. This series excludes governmental establishments except for wholesale liquor establishments, retail liquor stores, Federally-chartered savings institutions, Federally-chartered credit unions, and hospitals.

            Paid employment consists of full- and part-time employees, including salaried officers and executives of corporations, who are on the payroll in the pay period including March 12. Included are employees on paid sick leave, holidays, and vacations; not included are proprietors and partners of unincorporated businesses.

            Some data in the CBP is withheld from publication because of risk of disclosure of the operations of an individual employer. The number of establishments in an industry classification and the distribution of these establishments by firm size group are not considered to be disclosures, so this information may be released even though other information is withheld from publication.

            In total, 5,371 observations are available at various levels of aggregation in California; most of the analysis using the CBP will rely on the highest quality, most disaggregated data, resulting in 1,991 industry/county observations. In particular, the baseline models will rely on industries broken out by three-digit NAICS code, focusing on the same geographic coverage as the CPS, and including only observations with correctly coded employment data.

4. Coverage Effects

4a. Insurance coverage in California

            According to data from the CPS, California had a population of 35.2 million in 2002, up from 34.5 million in 2001. Nearly 82 percent of this population had health insurance either through their employer, private coverage, government coverage, or some combination of the three. Table 2 shows a breakdown of insurance coverage by provider in California.

[INSERT TABLE 2 HERE]

Over 57 percent of Californians had insurance solely through their employer or private coverage. The number of people with employer or private coverage increased by more than 615,000 between 2001 and 2002. Nearly 6 million Californians receive their medical insurance coverage solely through Medicare, Medicaid or Champus, increasing by about 43,000 between 2001 and 2002. The number of uninsured fell by more than 320,000, from 6.7 million in 2001 to 6.4 million in 2002.

4b. Impact of SB 2 mandate on coverage

            While advocates of Proposition 72 frequently state its intent is to provide coverage to employees whose employers provide no insurance, a more careful reading of the legislation shows the actual coverage is significantly broader.

[INSERT TABLE 3 HERE]

Table 3 shows that, in total, nearly 18 million Californians are covered by HIA requirements, including nearly 11 million workers. As shown in the more detailed Appendix Tables 1 and 2, only 3.6 million Californians (and 2.3 million eligible workers) receive fully funded health insurance from their employers. Assuming the employer is providing an acceptable level of coverage to be classified as “playing,” it can be said with certainty that this category, and only this category, is entirely unaffected by the HIA mandate.

            Over 10.6 million Californians (6.5 million eligible workers) currently receive health coverage from their employer but are forced to pay a portion of the premium, with over 488,000 people (over 316,000 eligible workers) paying the entire cost of coverage. In terms of the employer’s cost from HIA, these workers are just as expensive as uninsured workers. Many of the 6.5 million eligible workers will require additional coverage to meet the minimum standards under HIA in terms of both cost and quality of coverage.

            Previously uninsured individuals make up a minority of those employees who are affected by this legislation. Under HIA, only 1.98 million individuals who previously had no insurance receive new coverage. This means that approximately 69 percent of currently uninsured Californians will still have no health coverage as a result of HIA. Even if one excludes employees and dependents who previously received limited benefits (employer coverage below the level acceptable to be considered “playing”), those without any insurance account for only 57 percent of those affected by the legislation.

            Over 1.5 million individuals (nearly 700,000 workers) who currently have non-employer health insurance coverage will receive additional coverage as a result of the legislation. These individuals include the more than 360,000 Californians that choose to purchase private coverage and over one million Californians (more than 415,000 workers) currently receiving coverage through government programs.

5. Estimated Cost of Coverage

5a. Background on health insurance costs in California

            Before assessing the impact of the HIA on employer costs, it is useful to examine the employer health insurance market in California. KFF/HRET has published annual surveys about California’s health insurance market using the CEHBS from 1999 onward. These surveys form the basis for many of the cost estimates of SB 2 (Dube and Reich, 2003; Kyser, et al., 2003; Baker, et al., 2004; Kominski, 2004).

            The 2003 CEHBS completed interviews with benefits managers for each of 864 firms; 760 of the firms offered health insurance plans. When these 760 firms are appropriately weighted, they represent 7,863,192 covered employees with health insurance at these firms. Questions were asked about four types of health plans: conventional plans, HMOs, PPOs, and POS plans. The questionnaire specifically asked about premiums for each of the four health plans; the premiums used here are a weighted average based on actual participation rates in the plans. Premium questions are asked separately for single plans and family plans (covering a family of four). The benefits manager at a firm with 20 or more employees was asked both about the employee’s premium contribution to the most popular plan in each plan type and the employee’s total payment for COBRA continuation coverage (which is then adjusted to reflect total premiums). In firms with 19 or less employees, similar questions are asked about the employee’s contribution and the total payment (COBRA continuation coverage is not mandated for such firms). From these responses, the often cited premium numbers emerge.

[INSERT TABLE 4 HERE]

            Table 4 shows the results on total health premiums. The averages in the first two cells of the first column of data, $3,102 and $8,504, exactly replicate the widely cited results used in other studies, and are identical to figures produced in Chart 10 of KFF/HRET (2004). The remaining numbers in the table have not been previously disseminated. As expected, the median cost is modestly lower than the mean because of some outliers; the median cost of a single plan was $3,001 and the median cost of a family plan was $8,345. The lower median cost will be used as a benchmark throughout the cost analysis. The first two rows, based on all industries and firm sizes, also show substantial variation across plans – the 75/25 percentile range is more than $900 for a single plan, and more than $2,500 for a family plan.

            These averages mask variation across industry and firm size. The remaining rows show the results from the CEHBS data. There is substantial variation in total health premiums across industry. For single plans, the median cost for a plan varies from $2,513 in mining to $3,094 in the service sector. When grouped by firm size (according to the CPS categories, e.g., “1 to 9 employees”, “10 to 24 employees”, “25 to 99 employees”, “100 to 499 employees”, “500 to 999 employees”, and “1000 or more employees”), it is interesting to note that the premium costs for employees in the largest firms are only trivially smaller than covered employees in firms with 10 to 24 workers, and much more expensive than for employees in firms with 1 to 9 workers. This could call into question whether the firm size cutoffs in HIA are justified on equity grounds. For family plans, the median cost for a plan again varies substantially across industry – from $6,179 in mining to $9,815 in transportation/utilities/communications. A fairly clear pattern emerges with family premiums by firm size: premiums tend to increase with firm size. The lowest family premiums are for employees in firms with 1 to 9 workers (at $7,800) and highest for employees in firms with 1000 or more workers (at $8,508).

            This table also reveals a potential problem in the cost estimates of others who have relied on the aggregated premium costs for all firms. Based on the CEHBS data, the mean cost of a family plan is higher in large firms (“100 to 499 employees”, “500 to 999 employees”, and “1000 or more employees”) than the published aggregate estimate would indicate. For larger firms, the cost estimates used in other studies may be understated by as much as 3.2 percent (for firms with 500 to 999 employees).

            Finally, Table 4 could provide guidance on the industries that are most likely to “pay” rather than “play” in response to the HIA mandate. Employees in the transportation/utilities/communications, financial, and service industries have the highest health insurance premiums for both single and family plans. If the MRMIB sets one fee for the single plan for all participants, and one fee for the family plan for all participants, then the employees in these industries could see their employers opt-in to government run health care if the fee is lower than their premiums.

[INSERT TABLE 5 HERE]

            Table 5 shows the percentage of premiums that employees pay for; recall that one important aspect of the HIA is the “premium sharing” mandate that limits employee contributions to 20 percent. The percentages in the first two cells of the first column of data, 14 percent for single plans and 30 percent for family plans, exactly replicate the results produced in Chart 12 of KFF/HRET (2004). The most striking finding from the aggregate numbers is that many currently covered employees in firms that already provide health insurance to families will be affected by the HIA provisions. The employee’s share of premiums for a family plan for the median employee is 25 percent; under HIA, this would fall to 20 percent. Indeed, at least one-quarter of all covered employees pay at least 44 percent of the family plan premiums. Those covered employees affected by the HIA mandate would have their contributions reduced by more than 50 percent; the employees would have to absorb these additional costs under HIA. Table 5 shows that because of the premium sharing mandate, HIA has a far more expansive reach than advocates often claim.

            The disaggregated results again call into question the justifications for the firm size cutoffs in the HIA. For single plans, the median employee cost share percentage is the lowest for small firms and even the 75^th percentile is under the HIA requirement. With several exceptions, it appears that the premium sharing part of the HIA single mandate is not terribly binding. The table makes clear, however, that the premium sharing part of the family mandate is extremely important for costs. With the exception of employees in transportation/utilities/communications, the “premium sharing” part of the family mandate would be binding for at least one-quarter of the covered employees in all other industries. Employees in construction, retail, and wholesale industries would be most affected by the premium sharing part of the family mandate. In construction, for example, nearly half of the covered employees would see their contribution for family plans fall by approximately 50 percent; the employer would be responsible for the additional cost.

5b. Estimated costs of HIA mandate

            Two different methods are used to estimate the costs of HIA; one relies on imputed CPS employer contributions and the other relies on the CEHBS premium and cost sharing information. The technical discussion of how the costs are calculated is postponed to Appendix 2.

[INSERT TABLE 6 HERE]

            Table 6 presents the results using these different methods. It is estimated that the fully phased in HIA will cost employers between $12.8 and $13.2 billion. The difference between these two figures is based on the assumptions about how health insurance premiums are assigned to CPS respondents. In the event that the mandate was restricted only to firms with 50 or more employees, the costs are estimated to be between $11.3 and $11.9 billion. By comparing across columns, the estimated cost for businesses with 20 to 49 employees is approximately $1.3 to $1.5 billion. This cost already nets out the savings from the 20 percent tax credit that these employers would receive.

            In the case when the largest number of uninsured receive new coverage (when firms with 20 or more employees are included) this coverage breaks down to nearly $6,500 per newly covered individual. As the percentages show, the reason for this relatively high cost is that nearly two-thirds of the costs associated with this bill involve providing insurance to individuals who already have it. This cost number recognizes the full impact of the mandate on the currently insured, which is neglected by other authors. For every dollar spent under HIA, only 30 to 35 cents benefit the previously uninsured.

            The estimates of coverage above are significantly higher than other publicly released estimates. The costs for the uninsured, approximately $4.4 billion, represent about one-third of the total employer cost. This pretax estimate for the uninsured is substantially lower than Kyser, et al., (2003) and moderately lower than Baker, et al. (2004), but is also substantially higher than CMA (undated). Neglected in these studies, however, and by far the largest single group cost is for those who currently have employer-provided insurance. The premium sharing and dependent requirements of HIA entail costs of between $5.8 and $6.0 billion for employers. Those with “Employer-based coverage only” and “Employer-based coverage and government coverage” represent roughly one-half of the cost of the mandate.

            Between $1.5 and $1.7 billion represents a cost shift from government health insurance to employer-provided health insurance. This is roughly consistent with the claims of Dube (2003b) and CMA (2003c) who emphasize the flip-side – the savings to the state of California. Currently, more than 8.7 million Californians receive insurance through Medicare, Medicaid and Champus/Tricare. Of these, over 2 million are affected by the HIA mandate due to their work effort, tenure, or the size of the firm where they work.

            Enrollees are given the option of providing the necessary information to MRMIB in order to determine eligibility of either Medi-Cal or HFP. Eligible enrollees will be enrolled in these Medicaid wraparound coverage and refunded their enrollee contribution. Employer contributions will be used to pay the state’s portion of the matching funds for Medicaid. In this way, HIA amounts to a tax of at least $791 million (using CPS premiums) on employers to fund Medicaid, or the tax could be as high as $887 million (using CEHBS premiums).

            In addition to Medicaid recipients, enrollees who qualify for Medicare and Champus/Tricare will now receive additional, possibly redundant coverage from their employer. The reason the coverage is redundant is that HIA prohibits employers from providing wraparound coverage for these people – rather they must receive the full menu of health benefits, even though many of those benefits were already provided by government insurance. Currently, many of these employees receive supplemental coverage that provides vital benefits such as vision care, dental services, and prescription drugs. Under HIA, the provision of the supplemental benefits alone will not qualify an employer as “playing”. Instead, employers will be forced to provide upwards of $800 million of coverage to individuals who already have basic insurance. As a result, few employers will retain the incentive to continue to provide this supplemental coverage. Due to the fact that Medicare and Champus/Tricare are funded solely through federal dollars, the state government will see no cost savings. Private businesses, however, will supplement Medicare and Champus/Tricare with $643 million (using CPS premiums) or $812 million (using CEHBS), with around 85 percent of this cost going to Champus/Tricare.

            There are several issues in estimating the HIA costs that were ignored. First, there is no adjustment for the poverty subsidy. This would increase the employer’s cost. The key problem in incorporating the poverty subsidy is that it is unclear how far-reaching it actually is. The provisions of HIA note that the subsidy is based on individual earnings, not family income. But the provisions do not state clearly whether the subsidy is triggered by actual annual earnings or full-time, full-year equivalent earnings. It is possible for a part-time worker earning a higher wage to be eligible for the poverty subsidy, while a full-time worker earning a lower wage to be ineligible. It is highly unlikely that this is the intent of HIA, but rather poor wording on the part of the bill’s sponsors. Thus, I am reluctant to include the “poverty subsidy” until there is additional clarification about this. Second, I do not account for the corporate income tax deduction, for the reasons explained in Section 2. Third, I use an hours cut off of 25 hours per week (rather than 23, which is closer to the mandate’s actual provision). This higher cutoff will tend to understate costs. Fourth, even though HIA intends to extend its reach to seasonal workers and those with multiple jobs, I do not explicitly try to classify them as eligible for HIA. HIA (2003, p. 4) states “It is the further intent of the Legislature that workers who work on a seasonal basis, for multiple employers, or who work multiple jobs for the same employer should be afforded the opportunity to have health coverage in the same manner as those who work full-time for a single employer.” Fifth, even though the data is as up-to-date as possible, the latest premium data comes from 2003. Even within the past year, there has been a substantial increase in health care costs for employers and employees.

6. Labor market responses

            My reading of the literature indicates that none of the existing studies account for potential behavioral responses on the part of employers in a serious fashion. Whenever possible, profit-maximizing employers will react to HIA by shifting costs onto employees in the form of lower wages, and a number of credible, peer-reviewed studies suggest this is likely. In this case, the employee rather than the employer bears the cost of the mandate. In the case of the least skilled workers, however, wage shifting is simply not an option. In my study, the CPS data show that employers will be unable to shift the cost of the mandate onto a substantial number of employees. These employees are at risk of losing their jobs, either through labor force cuts or competition from more experienced workers attracted by the new benefits.

6a. Theory and evidence on mandated benefits

            As the HIA mandate has not yet been implemented, there are no studies of the actual effects of the law. In addition, there are very few states (and certainly none comparable to California) that have enacted such a sweeping mandate, so there is little direct evidence on the mandate’s potential impact.

            Summers (1989) presents theoretical arguments for mandated benefits relative to public provision of a good. He notes that “if employers and employees can negotiate freely over the terms of the compensation package, they will reach a mutually efficient outcome.” Yet, Summers argues that there are potential market failures that could lead to the case for public provision or mandated benefits. These market failures include “merit goods”, irrational consumers, externalities, and adverse selection.

[INSERT FIGURE 1 HERE]

            Figure 1 shows the typical supply and demand framework used to analyze a tax or mandated benefit. Ignoring the presence of the minimum wage, the efficient labor market equilibrium occurs at the intersection of the worker’s labor supply (S⁰) and employer’s labor demand curves (D⁰). This give the equilibrium quantity of labor (L⁰) and equilibrium wage rate (w⁰). Assuming that the labor market is competitive and there are no market failures or government distortions, the employment level and wage rate are economically efficient.

            Under a typical tax imposed on employers (the demanders of labor), the demand curve shifts down to D¹, and the new labor market equilibrium is (L¹,w¹). This is shown in Figure 2.

[INSERT FIGURE 2 HERE]

The tax imposes economic inefficiency, known as deadweight loss, represented by the yellow triangle. Government intervention lowers the employment level and wage rate.

            As Summers (1989, p. 180) notes, “mandated benefits do not give rise to deadweight losses as large as those that arise from government tax corrections.” The reason is that because the mandated benefit is potentially valuable to the employee, the labor supply curve shifts downward as well. The new employment level is given in Figure 3 by the intersection of the new labor demand curve, D¹, and the new labor supply curve, S¹.

[INSERT FIGURE 3 HERE]

            This equilibrium represents a situation with lower employment than without any government interference, but higher employment than with tax-financed provision of a benefit. The new labor market allocation, (L²,w²), also has lower wages for workers than either tax-financed provision or no government intervention. The inefficiency from such a mandated benefit is given by the smaller yellow triangle. The allocation with mandated benefits could therefore involve substantial wage reductions for employees.

            This straightforward framework ignores several important features, however. First, as Summers (1989, p. 180) points out, the “mandated benefits represent a tax at a rate equal to the difference between the employers cost of providing the benefit and the employee’s valuation of it, not a rate equal to the cost of the employer of providing the benefit.” One critical issue then becomes the employee’s valuation of the benefit. Under HIA, more than 1.08 million current recipients of government health care will have that insurance crowded out by employer-provided health insurance. For these enrollees, the additional value of the benefits from the HIA mandate are surely quite small, so the economic inefficiency and employment losses look more like Figure 2. The same is also probably true for the uninsured who are eligible, but not participating, in Medicaid. Brown, et al., (2002, p. 48) estimate that 1.12 million adults and children are eligible for Medi-Cal or Healthy Families but not participating. It is also likely that some of the uninsured – especially younger, healthier adults – do not put a very large valuation on health insurance.

            Second, as Summers (1989, p. 181) notes, if there is a binding minimum wage, then “wages cannot fall to offset employers’ cost of providing a mandated benefit, so it is likely to create unemployment.” As will be shown below, there are 4.3 million Californian workers with wages below $9.31 per hour, including more than 680,000 workers in large firms who are either uninsured or on government insurance. It is likely that wages would not be able to fully adjust downward for such workers. More generally, when wages are rigid and do not move downward in response to the mandate (which is especially likely in the short-run), then the larger economic inefficiencies illustrated in Figure 2 become more likely.

            Finally, the framework above shows that mandated benefits are still a government tax, even if they are not explicitly called a tax. Summers (1989, p. 182) cautions about the government’s use of mandated benefits. He says “There is no sense in which benefits become ‘free’ just because the government mandates that employers offer them to workers.” Reinhardt (1987, p. 124) notes that “the fiscal flows triggered by mandate would not flow directly through the public budgets does not detract from the measure’s status of a bona fide tax.”

            Although there are a number of empirical studies that examine the impact of employer mandates for family leave and workers compensation, the study that provides the closest (albeit not that close) analog to the HIA mandate is Gruber (1994). Gruber (1994) studied several state and federal laws that mandated comprehensive childbirth benefits in health insurance policies, and therefore substantially raised the cost of insuring women of childbearing age. Between 1975 and 1978, some states passed laws that prohibited treating pregnancy differently from “comparable illnesses.” In October 1978, the Federal government passed the Pregnancy Discrimination Act, which prohibited any differential treatment of pregnancy in the employment relationship. Using the CPS, he finds shifting of the costs of the mandates from the employer to the employee in the form of lower wages on the order of 100 percent. In fact, some of his specifications suggest overshifting of wages.

            Thus, Gruber (1994) provides strong evidence that firms will lower wages (when possible) to “pay” for the mandate. Even those who have higher wages are affected by the mandate if their employer does not provide coverage. The 1.08 million current recipients of government health care who will have that insurance crowded out by employer-provided health insurance will likely see lower wages but no commensurate increase in health benefits.

6b. Employer responses to HIA

            In the context of HIA, there are a number of responses that the simple supply and demand framework captures, and other responses that it is less suited for. The “pay or play” mandate should lead to wage shifting and employment losses, similar to Figure 2 or Figure 3 (depending on the additional valuation from employer-provided coverage). But HIA also creates “employment notches” by dramatically raising the cost of hiring the 20^th, 50^th, or 200^th worker. The cutoffs in the law mean that for a firm that was not previously providing health insurance, it would have to provide and/or pay more for health insurance for all of the inframarginal employees by hiring the marginal 20^th, 50^th or 200^th worker. Thus, in addition to the typical disemployment effect from the tax, the “employment notches” of HIA present important barriers to growth for certain firms and also create incentives for firms to downsize or consolidate part-time jobs. HIA also creates “hours notches” by defining eligible employees as those who work 100 or more hours per month for three months. This potentially creates incentives for employers to limit weekly hours to 23 or less and to increase turnover for short-term employees.

            In principle, the HIA tries to legislate away economic responses. HIA (2003, p. 11) states “It shall be unlawful for an employer to designate an employee as an independent contractor or temporary employee, reduce an employee’s hours of work, or terminate and rehire an employee if a purpose of which is to avoid the employer’s obligations under this part.” To the extent that employers are unable to minimize HIA’s cost impact by limiting hours, the likelihood of other economic adjustments (like wage shifting and job loss) becomes more likely.

6c. Wage shifting and unemployment

[INSERT TABLE 7 HERE]

            As shown in Table 7, the California labor market consisted of 17,883,738 workers during the 2002 calendar year. This includes CPS respondents who answered affirmatively to the questions “Did ... work at a job or business at any time during 2002?” or “Did ... do any temporary, part-time, or seasonal work even for a few days during 2002?”. Slightly more than two-thirds of these workers had employer health insurance. Approximately 9.4 percent of workers had some form of government insurance. Overall, nearly 3.6 million workers, or 20 percent, were uninsured in 2002. This estimate is slightly higher than the Baker, et al. (2004) estimate, who find 3.4 million workers were uninsured (based on 2001 California EDD data).

            The final two columns of Table 7 show insurance characteristics for low-wage workers. Two cutoffs are used: $9.31 per hour and $10.36 per hour. For the first group, nearly 38 percent are uninsured, and nearly 16 percent have government insurance. Only 41 percent currently have some form of employer health insurance. For the second group, the results are fairly similar – nearly 36 percent of workers lack insurance, nearly 15 percent have government insurance, and nearly 45 percent have employer coverage.

[INSERT TABLE 8 HERE]

            Table 8 explores these low-wage workers further. Even though there are 4.3 million workers under $9.31 per hour (and 5.4 million under $10.36 per hour), many will not work at large or medium firms, and others do have some sort of employer coverage. For workers earning under $9.31 per hour, more than 1.6 million are uninsured, and more than 500,000 have government coverage. A significant number of these workers are employed in large firms that would be potentially responsible for paying for 80 percent (or more) of the costs of a family plan. Nearly 480,000 uninsured workers earning under $9.31 per hour work at large firms, and more than 567,000 uninsured workers earn under $10.36 per hour at such firms. Approximately 200,000 additional workers under $9.31 per hour have government health insurance at such firms, as do nearly 230,000 workers under $10.36 per hour.

            The additional cost to employers to provide the mandated coverage to these workers is likely to be high. Below each raw count of low-wage workers is the average mandated cost per worker (the employer cost presented here assumes the employer’s share is 80 percent of the total mandated cost). As can be seen by these figures, the employer cost per low-wage worker of the mandate is highest for large firms. One pattern that clearly emerges from these numbers is that uninsured workers and government insured workers entail the highest cost to business. In large firms, the average cost of providing coverage to low-wage uninsured workers is between $2,438 and $2,805. The cost to a worker who currently has government insurance is between $2,637 and $3,765. Workers covered by private coverage can be quite expensive, too, with costs greater than $2,000 in large firms. One other finding that comes out of this table is that workers with current employer-provided health insurance coverage are not “free” – the average cost at a large firm for such a worker is in the range of $1,200, while the average cost from HIA in a medium firm is in the range of $600 to $1000. These costs suggest that both the premium sharing provisions and the dependent provisions in the HIA mandate are quite costly.

            As suggested by the theoretical framework and existing empirical evidence, one strong possibility is that employers will try to pass the HIA costs onto employees in the form of lower wages. Table 9 examines the implications for tax revenue from such wage shifting, assuming that wages fully adjust for all workers (which is unrealistic, especially in the short-run).

[INSERT TABLE 9 HERE]

As in previous tables, results are presented using both the CPS health insurance premiums and the CEHBS premiums. Under these assumptions, between $4.6 and $4.9 billion of tax revenue is lost to the federal and state government. Of this, around half is a loss in federal tax revenue. The loss to the state of California is between $800 and $860 million per year from these wage reductions. The loss in payroll tax collections is more than $1 billion. The effects on the earned income tax credit are a more modest $136 to $147 million, but these losses mask the subsidy/tax nature of the credit. To the extent that EITC recipients have their wages reduced in the phase-out range, government tax revenue goes down. But to the extent that the recipients have their wages reduced in the 40 percent phase-in range, the government pays less in subsidy so government revenue goes up. The negative numbers in the last column for some groups reflect this second effect being stronger.

[INSERT TABLE 10 HERE]

            Although the previous literature on mandated benefits does provide guidance on the possibility of wage shifting, such wage shifting is unlikely for those near the minimum wage. Table 10 recalculates the tax loss numbers, excluding approximately 1.4 million workers for whom full wage shifting is not possible. These workers have actual mandated costs from HIA that would preclude such shifting. The results from Table 10 suggest much the same story in terms of revenue loss, with the state of California losing anywhere between $665 and $696 million in tax revenue, while the total tax revenue loss is around $3.6 billion.

            Employers of low-wage workers, however, will be unable to shift the full burden of HIA costs onto their employees. The estimates above show that wage shifting is constrained for approximately 1.4 million employees due to the current California minimum wage of $6.75 per hour. Operating under this constraint, employers are faced with a similar situation to an increase in the wage floor. They must accept lower profits, raise prices, or alter employment levels and skill levels to respond to the increased costs.

            In studying the effect on increases in mandated wage levels, Neumark (1995) found that current employees were often displaced by higher skill individuals attracted by higher wages. Lang (1995) found wage hikes shift “employment towards teenagers and students… [T]he competition from [these] higher quality workers makes low-skill workers worse off.” Neumark and Wascher (2000) convincingly reevaluate Card and Krueger’s (1994) study of minimum wages in New Jersey, and using payroll data find an employment elasticity of -0.22. In a prominent survey of labor economists, Fuchs, Krueger, and Poterba (1998) find that the mean estimate of the employment elasticity for teenagers is -0.21, and the median is -0.10. All of these findings suggest that in the absence of full wage shifting, there is a strong possibility of layoffs as a result of HIA, especially for low-skill workers.

            I next examine the possibility of job loss, using the Neumark and Wascher (2000) elasticity estimate of -0.22. To compute the employment loss, I considered several different scenarios. The first scenario assumes no wage shifting for any worker, while the second assumes full wage shifting until the minimum wage and disemployment effects thereafter. That is, the second scenario shifts as much of HIA onto the worker as possible in the form of lower wages, and only to the extent that wages would have to be shifted below the California minimum wage of $6.75 would employment losses ensue. For example, if the total mandated HIA cost were $2,080 for a worker, then this would translate into a $1.00 per hour shift in wages. If the worker earned, $7.50 per hour, only 75 cents of this mandate could be passed along in the form of lower wage; the remaining 25 cents is analogous to a minimum wage increase (where the percent change in wages is 25 cents divided by $6.75, or 3.7 percent). In the second scenario, I apply the Neumark and Wascher (2000) employment elasticity to this percentage to compute the employment losses. In all cases, the HIA mandate was converted into an hourly wage rate increase based on full-time/full-year work, which leads to the smallest possible disemployment effect. In other words, the percent increase in the wage floor would be larger using actual hours of work.

            Focusing on the 1.4 million workers where the wage floor would increase after wage shifting occurred, the wage floor increase for the mean (and median) worker is approximately 21 percent using CPS premiums. Ten percent of workers would experience an increase in the wage floor of 42 percent or more. The results are similar using CEHBS premiums.

[INSERT TABLE 11 HERE]

            Table 11 shows the disemployment results, using both sets of health premium data. When wage shifting is possible, approximately 70,000 workers will lose their jobs as a result of HIA, nearly 25 percent of whom already had employer-provided health insurance. More than 32,000 of these workers were uninsured meaning that in addition to not receiving health insurance, now they also lose their jobs. Around 11,500 workers with government insurance lose their job, meaning they continue to keep this insurance instead of being transferred to employer insurance. When wage shifting is not possible (as is likely in short-run), the results are even more dramatic. Around 150,000 workers lose their jobs, with roughly equal numbers coming from the uninsured and covered by employers.

6d. Other labor market adjustments

            The employment loss analysis using the supply and demand framework above likely understates the job loss. The HIA creates “employment notches” for hiring the 20^th, 50^th, and 200^th employee. Using the median CEHBS premiums for 2003, for example, a firm that was previously offering single coverage could face a marginal cost to its health care bill of approximately $850,000 for hiring the 200^th employee, because it would have to offer all of its employees family coverage if they qualified. Similarly, hiring the 20^th employee entails a marginal cost to its health care bill of nearly $40,000, while hiring the 50^th employee entails a marginal cost of nearly $26,000 due to the loss of the tax credit. Understanding how employers respond to such notches is clearly important in gauging the employment losses, but there is no convincing evidence on this. The HIA also creates the potential for an “hours notch” where employers limit workers to less than 23 hours per week. Although the legislation specifically outlaws reducing hours, it is likely that through attrition and new hiring, the same sort of outcome could be achieved. Evidence from Thurston (1997) finds that a similar hours notch in Hawaii did affect the work patterns of employees. The legal uncertainty that surrounds such adjustments potentially raises the marginal cost of doing such actions, shifting employer behavior to the legal wage-shifting and layoff behavior discussed earlier.

6e. Revised cost estimates and characteristics of workers

            The disemployment effects computed in Table 11 potentially lead to the worker being worse off, but the employer saving money. This cost savings is substantial, but only affects the overall employer costs by a modest amount. For example, when the largest number of workers lose their jobs (when wage shifting is not possible) the cost savings varies between

$483,634,933 and $649,412,060, thus the range of HIA costs varies between $12.4 and $12.6 billion.

            One question that arises in light of these employer responses is what kinds of characteristics do these newly unemployed workers possess. Table 12 compares the laid off workers to all workers in California.

[INSERT TABLE 12 HERE]

            First, these workers have much lower family income than the typical worker. Average family income for a California worker in 2002 was approximately $73,000; for workers displaced when there is no wage shifting, family income is around $50,000. For the smaller group of workers who lose their jobs when wage shifting is possible, family income is much lower, approximately $38,000. This group of roughly 70,000 displaced workers has much lower wages and is considerably younger than the typical California worker. They are modestly less likely to be married, male or a veteran. In terms of educational attainment, this group is far more likely to be a high school dropout or high school graduate. They are also much more likely to be of Hispanic origin. The basic conclusion is those who suffer the most under HIA – workers who lose their jobs – are already extremely disadvantaged in the labor market.

7. Conclusions and extensions

            This study finds a much larger estimate than other existing studies, including studies touted by opponents of Proposition 72. All of these studies have failed to take account of the full extent of the HIA on employers’ costs. The cost to employers of the legislation is expected to be in the neighborhood of $13 billion; much of this represents shifting of responsibility of paying for health care from other groups to employers. The cost of HIA is nearly $6,500 per newly insured individual, because nearly two-thirds of the costs associated with this bill involve individuals who already have health insurance. The likely employer responses include wage-shifting and layoffs.

            A number of factors were not addressed in this study, in part, because of the considerable ambiguity still surrounding the implementation of the HIA. The most important is the question of whether firms will “pay or play.” Baker, et al. (2004) provide an interesting discussion about some of the issues surrounding this decision. One key issue is to the extent that firms with higher expected health costs (e.g., those with older or unhealthier workers) opt-in to the state’s plan, and whether the fees associated with participation in the plan are experience-rated, as are taxes for various social insurance programs. Ultimately, the decision to opt-in to state-run health care would boil down to comparing the difference in costs and benefits from the state-run and private health care plans, along any other adjustments to the employee’s compensation package. In cases when the firm’s profitability and employee’s total compensation could both be increased (e.g., for firms with high expected health care costs), then the decision to “pay” rather than “play” seems quite likely.

            A final question is, if HIA is the wrong direction for health care reform, what is the right direction? The goal of any health care reform should be to reduce the number of uninsured in a cost efficient way. One possibility, though certainly not the only one, is an aggressive Medi-Cal outreach program in California. In Yelowitz (2003), I estimated using CPS data that virtually all uninsured children in California were eligible for some sort of government health insurance program. Along the same lines, Brown, et al., 2002 (p. 48) estimate using CHIS data that 355,000 uninsured children are eligible but not participating in Medi-Cal, and another 301,000 are eligible for Healthy Families. In addition they find that 413,000 uninsured parents are eligible but not participating in Medi-Cal, and another 52,000 non-elderly adults are eligible for Medi-Cal. Thus, more than 1.1 million Californians would be eligible for these government programs. The number of nonparticipating eligibles for government programs in Brown, et al. (2002) is larger than the number of Californians that Brown, et al. (2003) estimate would gain coverage under the fully phased-in HIA. If the state of California had an aggressive outreach campaign, this campaign could produce much more bang-per-buck than the HIA, because the federal government would absorb half of the Medicaid costs through the match rate and the employment loss that would take place would be minimal.

Selected References

Baker, Wally, Stephen P. Erie, Jack Kyser, Gregory Freeman, Michael D. Lloyd, Scott MacKenzie, “The Economic Impact of Mandatory Health Insurance Coverage on Californians,” Prepared for Californians Against Government Run Healthcare, June 22, 2004, http://www.stopthehealthtax.org/pdf/LAEDCReport.pdf.

Brown, E. Richard, “Declaration of E. Richard Brown, Ph.D., In Support of Petition for Writ of Mandate and Ex Parte Application for Order Shortening Time For Response to Petitioners’ Request for Production Documents,” filed in the Superior Court of the State of California, County of Sacramento, July 28, 2004.

Brown, E. Richard, Ninez Ponce, Thomas Rice, and Shana Alex Lavarreda. “The State of Health Insurance in California: Findings from the 2001 California Health Interview Survey,” Los Angeles, CA: UCLA Center for Health Policy Research, 2002, http://www.healthpolicy.ucla.edu/pubs/files/shic062002.pdf.

Brown, E. Richard, Hongjian Yu, Shana Alex Lavarreda, Lida Becerra, Arindrajit Dube, and Richard Kronick. “SB 2 Will Extend Coverage to 1 Million Uninsured Workers and Dependents,” Los Angeles, CA: UCLA Center for Health Policy Research, September 2003, http://www.healthpolicy.ucla.edu/pubs/files/SB2_FactSheet.pdf

California Healthcare Foundation. “The Health Insurance Act of 2003: An Overview of SB 2,” November 2003, http://www.chcf.org/documents/insurance/SB2FactSheet2.pdf.

California Health Interview Survey. CHIS 2001 Adult Public Use File, Release 3 [computer file]. Los Angeles, CA: UCLA Center for Health Policy Research, April 2004, http://www.chis.ucla.edu/main/PUF/DICTIONARY_ADULT_PUFA3_041404.pdf.

California Medical Association, “The Faces of Medi-Cal: A White Paper,” April 2003a, http://www.calphys.org/assets/applets/budget_white_paper.pdf.

California Medical Association, “SEIU-SB 2 Health Access Weighted Questionnaire,” September 12, 2003b, http://www.calphys.org/assets/applets/sb2_questionnaire.pdf

California Medical Association, “CMA Campaigns for Gov. Davis' Signature on Health Care Measure,” September 16, 2003c, http://www.calphys.org/html/bb377.asp.

California Medical Association, “Estimated Cost Impact and Savings SB 2 (Burton),” undated, http://www.calphys.org/assets/applets/sb2_cost.pdf.

Card, David, and Alan B. Krueger, “Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania,” The American Economic Review, September 1994, 772-793.

Cutler, David M., and Jonathan Gruber, “Does Public Insurance Crowd Out Private Insurance?,” The Quarterly Journal of Economics, May 1996, 391-430.

Dube, Arindrajit. “Impact of SB2 on Medi-Cal Costs,” Mimeo, Institute of Industrial Relations, UC Berkeley, September 8, 2003, 2003a.

Dube, Arindrajit. “Impact of SB2 on Health Coverage,” Mimeo, Institute of Industrial Relations, UC Berkeley, September 9, 2003, 2003b, http://www.iir.berkeley.edu/research/healthcoverage.pdf.

Dube, Arindrajit, and Michael Reich. “2003 California Establishment Survey: Preliminary Results on Employer Based Healthcare Reform,” Mimeo, Institute of Industrial Relations, UC Berkeley, September 18, 2003, http://www.iir.berkeley.edu/research/ces.pdf.

Friedman, Joel, “The Decline of Corporate Income Tax Revenues,” Mimeo, Center on Budget and Policy Priorities, October 24, 2003, http://www.cbpp.org/10-16-03tax.pdf.

Fuchs, Victor R., Alan B. Krueger, and James M. Poterba, “Economists’ Views about Parameters, Values, and Policies: Survey Results in Labor and Public Economics,” Journal of Economic Literature, September 1998, pp. 1387-1425.

General Accounting Office, Tax Administration: Comparison of the Reported Tax Liabilities of Foreign- and U.S.-Controlled Corporations, 1996-2000, GAO-04-358, February 27, 2004, http://www.gao.gov/new.items/d04358.pdf.

Gruber, Jonathan, “The Incidence of Mandated Maternity Benefits,” The American Economic Review, June 1994, 622-641.

Health Access California, “Fact Sheet: Health Coverage Impacts of SB 2 (Burton)” September 2, 2003a, http://www.health-access.org/docs/SB2HealthImpactFactsSept-2-03.doc.

Health Access California, “Some Responses about SB 2 (Burton)” September 18, 2003b, http://www.health-access.org/docs/SB2Responses.doc.

Health Access California, “Talking Points: SB 2 (Burton), Health Insurance for Working Californians,” undated document, accessed August 17, 2004a, http://www.health-access.org/docs/SB2TalkingPointsHAC.doc.

Health Access California, “The Future of National Health Care Reform is on the November 2004 California Ballot” January 27, 2004b, http://www.health-access.org/docs/SB2BallotCampaignFactSheet.doc.

Health Insurance Act of 2003, October 6, 2003, http://www.leginfo.ca.gov/pub/bill/sen/sb_0001-0050/sb_2_bill_20031006_chaptered.pdf.

Kominski, Gerald F., “Declaration of Gerald F. Kominski, Ph.D., In Support of Petition for Writ of Mandate,” filed in the Superior Court of the State of California, County of Sacramento, July 28, 2004.

Kyser, Jack, Wally Baker, Steven P. Erie, Harold Brackman, and Michael D. Lloyd. “Analysis of the Economic Impacts of Mandatory Health Coverage in California,” September 8, 2003, http://www.calchamber.com/pdf/sb2_analysis.pdf.

Lang, Kevin, “Minimum Wage Laws and the Distribution of Employment,” The Employment Policies Institute, Washington, DC, 1995.

Madrian, Brigitte C., “Employment-Based Health Insurance and Job Mobility: Is There Evidence of Job-Lock?” The Quarterly Journal of Economics, February 1994, 27-54.

Neumark, David, “Effects of Minimum Wages on Teenage Employment, Enrollment, and Idleness,” Mimeo, The Employment Policies Institute, 1995.

Neumark, David, and William Wascher, “Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment,” The American Economic Review, December 2000, pp. 1362-1396.

Reinhardt, Uwe, “Should all Employers be Required by Law To Provide Basic Health Insurance Coverage for Their Employees and Dependents?” in Government Mandating of Employee Benefits, Washington, D.C.: Employment Benefit Research Institute, 1987.

Summers, Lawrence H., “Some Simple Economics of Mandated Benefits,” The American Economic Review, May 1989, 177-183.

The Henry J. Kaiser Family Foundation and Health Research and Educational Trust, “California Employer Health Benefits Survey, 2003,” March 2004, http://www.kff.org/statepolicy/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=32778

Thurston, Norman K., “Labor Market Effects of Hawaii's Mandatory Employer-Provided Health Insurance,” Industrial and Labor Relations Review, October 1997, 117-135.

U.S. Department of Commerce, Bureau of Census, 1993, Measuring the Effects of Benefits and Taxes on Income and Poverty: 1992, Current Population Reports, Consumer Income, Series P-60, No. 186-RD, Washington, D.C., : U.S. Department of Commerce.

U.S. Department of Commerce, Bureau of the Census. Current Population Survey: Annual Social and Economic (ASEC) Supplement, 2003 [Computer file]. Washington, DC: U.S. Dept. of Commerce, Bureau of the Census [producer], 2003. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2004.

Yelowitz, Aaron S., “The Medicaid Notch, Labor Supply and Welfare Participation: Evidence From Eligibility Expansions,” The Quarterly Journal of Economics, November 1995, 909-939.

Yelowitz, Aaron S., “Public Policy and Health Care Choices of the Elderly: Evidence From the Medicare Buy In Program,” Journal of Public Economics, November 2000, 301-324.

Yelowitz, Aaron, “The Cost of California’s Health Insurance Act of 2003,” Mimeo, The Employment Policies Institute, October 2003, http://www.epionline.org/studies/epi_Yelowitz_10-2003.pdf

Yelowitz, Aaron, “Declaration of Aaron Yelowitz, Ph.D., In Support of Petition for Writ of Mandate,” filed in the Superior Court of the State of California, County of Sacramento, dated July 30, 2004, http://gatton.uky.edu/faculty/yelowitz/yelowitzprop72declaration.pdf.

Appendix 1: Technical assumptions used to impute HIA eligibility

            The HIA clearly delineates different requirements based on whether the firm has 20 to 49 California employees, 50 to 199 employees, or 200 or more California employees. The CPS, like most household surveys, only asks respondents about firm size in ranges, not the exact firm size. The CPS question is worded as: “Counting all locations where this employer operates, what is the total number of persons who work for ...'s employer?”, where firm size can range from 1 to 9 employees, 10 to 24 employees, 25 to 99 employees, 100 to 499 employees, 500 to 999 employees, and 1000 or more employees. These groupings are shown in the first column of Appendix Table 3.

[INSERT APPENDIX TABLE 3 HERE]

It is clear that workers who answer “1 to 9” can be classified as ineligible for HIA, and that those who answer “500 to 999” or “1000 or more” can be classified as eligible for the HIA family mandate. The other three groups include a mix of workers who qualify for different entitlements under HIA. Some in the “10 to 24” category will qualify for the single mandate if the state of California offers a 20 percent tax credit to businesses with between 20 and 49 employees. The same is true for some of the workers in the “25 to 99” category, while other workers in this group will qualify for the HIA single mandate. In the “100 to 499” grouping, some workers will qualify for the HIA single mandate and others for the family mandate. These classifications are shown in the second column of Appendix Table 3.

            Without any additional information, a researcher would have several options for these other three groupings. One option would be to be extremely conservative (liberal) by assuming that all firms in a certain grouping, such as the 25 to 99 group, are at the lowest (highest) end of the scale, but this would clearly understate (overstate) the costs of HIA. This conservative (liberal) assumption would say, for example, that all firms had 25 (99) employees and the employees were therefore eligible for the single coverage with a state tax credit (single coverage alone), and would systematically understate (overstate) the cost of the mandate. A second option would be to assume that either firms or employees are distributed uniformly across the distribution within a firm size grouping, an assumption listed in the third column of Appendix Table 3. For example, there are 75 separate groupings in the “25 to 99” category. If firms are distributed uniformly across this category, then the likelihood of an employee falling into the 25 to 49 group (and thus, eligible for single coverage with the state tax credit) would be 19.9 percent. This is because large firms have more employees and therefore contribute more to the probability. If employees are distributed uniformly across this category, then the likelihood of an employee falling into the 25 to 49 group would be much higher, 33.3 percent. For the distribution to be uniform in employment, there must be, for example, three times as many 25-employee firms as there are 75-employee firms.

            The options listed above appear to be the most relevant ones without additional data sources. The intuition for using the 2001 CBP data is to avoid ad hoc assumptions about the distribution of employers or employees. The CBP, for example, provides information by county and industry for the number of firms with 1 to 4 employees, 5 to 9 employees, 10 to 19 employees, 20 to 49 employees, 50 to 99 employees, 100 to 249 employees, 250 to 499 employees, 500 to 999 employees, and 1000 or more employees. Assume that a CPS respondent reports working at a firm with “25 to 99” workers, living in Los Angeles county, and being employed in the “Food Services and Drinking Places” industry (NAICS code 722). In this case, the CBP groupings of “20 to 49 employees” and “50 to 99 employees” can shed additional light on the HIA mandate. For purposes of illustration, assume that the CBP reported zero firms in Los Angeles County in this industry in the “20 to 49” group, and a positive number in the “50 to 99” group. In this case, using the CBP data in conjunction with the CPS data would generate a 100 percent probability of being affected by the HIA mandate with “single coverage”, and a 0 percent probability of being affected by the “single coverage with subsidy”. On the other extreme (again, for purposes of illustration), assume that the CBP reported zero firms in the “50 to 99 employees” category and a positive number in the “20 to 49 employees” category. Now using the CBP in conjunction with the CPS data would generate a 100 percent probability of “single coverage with credit” and a 0 percent probability of “single coverage.”

            Clearly, both of those extremes are not likely to occur very often in the CBP data, so how does one incorporate the CBP data with the CPS data in other situations? The probability will depend on the number of firms in the CBP categories (e.g., “20 to 49 employees” and “50 to 99 employees”) and average firm size for firms in each of the two categories. These numbers are used to compute covered employment and total employment. In addition, because the CBP actually groups firms into “20 to 49” employees rather than using the CPS cutoff of 25 employees, another adjustment will need to be made. Otherwise, some smaller firms with 20 to 24 employees will inappropriately downweight the probability of single coverage from HIA.

            The goal is to use this information to figure out the expected number of employees in the “50 to 99” grouping, and the expected number of employees in the “25 to 49” grouping. By doing so, dividing the employment count in the “50 to 99” grouping by the total employment count in the “25 to 99” grouping gives us the probability of a CPS respondent will receive single coverage from HIA. This expression is computed as follows:

where Pr(SING_25-99,c,i) is the estimated probability that a CPS respondent who reports being employed in a firm with 25 to 99 employees in county c and industry i is actually in a firm large enough to receive single coverage under the HIA mandate (as opposed to single coverage with the tax credit). The variables FIRM_20-49,c,i and FIRM_50-99,c,i are the number of firms in each grouping (“20 to 49”, and “50 to 99”) in county c and industry i and come directly from the CBP data. The variables EMP_20-49 and EMP_50-99 are estimates of firm size within each CBP grouping (based on all industries and counties; the computation is discussed below). Finally, the product (EMP_20-49*FIRM_20-49,c,i) gives the total employment in firms with 20 to 49 employees, but the denominator needs employment in firms with 25 to 49 employees. The variable PR_25-49|20-49 (also discussed below) is the probability an employee who works in a firm with a firm size of 20 to 49 employees is actually working in a firm with 25 to 49 employees.

            Return to the example of a CPS respondent who reports working at a firm with “25 to 99” workers, living in Los Angeles county, and being employed in the “Food Services and Drinking Places” industry. The 2001 CBP for Los Angeles County reveals a total of 15,187 “Food Services & Drinking Places.” Of those, 3,000 establishments have between 20 and 49 employees, and 883 had between 50 and 99 employees. Moreover, for purposes of illustration, assume that the average firm size is 30 employees for firms with 20 to 49 employees, and is 70 employees for firms with 50 and 99 employees. Finally, assume the probability of an employee being in a firm with 25 to 49 employees, conditional on being in a firm with 20 to 49 employees, is 88 percent. The total number of people employed in firms with 50 to 99 employees is therefore 61,810 workers (=883 establishments*70 employees). The total number employed in firms with 20 to 49 employees is 90,000 workers (=3000 establishments*30 employees). We derive the number who work in firms with 25 to 49 employees as 79,200 workers (=88 percent*3000 establishments*30 employees). As a consequence, 10,800 workers are in firms sized 20 to 24. The total number of workers in firms sized 25 to 99 is therefore 141,010 (=61,810+79,200). The probability of receiving single coverage for the CPS respondent is therefore 43.8 percent (=61,810/141,010). Thus, the probability of such a CPS respondent receiving single coverage with a credit is 56.2 percent.

            The above illustration provides general guidance on how the HIA coverage was imputed to CPS workers, but there are many additional details of the calculations that will be filled in below. Before filling in those details, I first show the methodology used to compute HIA eligibility for all of the groups in the analysis. Returning to Appendix Table 3, only three groups offer any sort of ambiguity in the CPS.

            The following probabilities (analogous to the equation above) are used in the analysis. For CPS respondents reporting a firm size with “10 to 24 employees” the following expressions are used to compute the probability of receiving no coverage under HIA or single credit coverage:

(1a)

(1b)

            For CPS respondents reporting a firm size with “25 to 99 employees” the following expressions are used to compute the probability of receiving single credit coverage or single coverage under HIA:

(2a)

(2b)

            Finally, for CPS respondents reporting a firm size with “100 to 499 employees” the following expressions are used to compute the probability of receiving family coverage or single coverage under HIA:

(3a)

(3b)

            In equations (1a), (2a), and (3a), EMP_10-19, EMP_20-49, EMP_50-99, EMP_100-249, and EMP_250-499 are the median estimate of employment size within that category, based on a model derived from the 2001 CBP and discussed later. The variables FIRM_10-19, FIRM_20-49, FIRM_50-99, FIRM_100-249, and FIRM_250-499 are raw CBP counts of the number of firms in each employee size grouping, based on the CPS respondent’s county c and industry i. Finally, PR_25-49|20-49 and PR_{200-249|100-249} are the estimated probabilities that an employee in the 20 to 49 (100 to 249) grouping is actually in the 25 to 49 (200 to 249) category. Note that in equation (1a), by using 1-PR_25-49|20-49=PR_20-24|20-49, we are computing the number of employees in the CBP firms size 20 to 49 who are actually in firms sized 20 to 24 employees. As discussed later, PR_25-49|20-49 and PR_{200-249|100-249} are weighted by employment.

            The general idea of all of the above equations is, within each CPS firm size category, to compute the expected employment above and below the HIA cutoff, in order to assign accurate eligibility to CPS respondents. The two remaining details are related to EMP_10-19, EMP_20-49, EMP_50-99, EMP_100-249, and EMP_250-499 – the estimated employment within a firm size grouping, and PR_25-49|20-49 and PR_{200-249|100-249} – the estimated probabilities of falling into certain groupings in the CBP data.

            First, in order to impute HIA eligibility to CPS respondents, equations (1a)-(3b) need estimates of firm size in the CBP. One naive, but simple, assumption, would be to take the midpoint of each firm size interval and use that as the estimate. In the example before, this would imply assuming that all firms in the “20 to 49” category had a firm size of 34.5, while all firms in the “50 to 99” category had a firm size of 74.5. This is problematic, however, because within any given category, firms tend to be more concentrated on the small end of the interval. Fortunately, the CBP data offers a way to estimate firm size in each category. For each county/industry observation in the CBP, the CBP gives total industry employment as well as number of firms in each category. I estimate the following median regression model:

(4)

where TOTEMP_c,i is the total industry employment in the 2001 CBP in a county/industry cell, and FIRM_f,c,i are the nine firm size groupings (f) in the CBP (“1 to 4 employees”,“5 to 9 employees”,“10 to 19 employees”,“20 to 49 employees”,“50 to 99 employees”,“100 to 249 employees”,“250 to 499 employees”,“500 to 999 employees”, and “1000 or more employees”). The estimated coefficients from this model minimizes ε, the sum of the absolute value of the residuals. The baseline results are presented in Appendix Table 4, column 1.

[INSERT APPENDIX TABLE 4 HERE]

            The results in the first column are estimated on the “cleanest” CBP sample – 1,991 county/industry observations, where industry is defined at the three-digit NAICS level. These are the results that will be used to construct EMP_10-19, EMP_20-49, EMP_50-99, EMP_100-249, and EMP_250-499 in equations (1a)-(3b). In order to maintain as much comparability to the CPS as possible, the geographic coverage is restricted to the 32 identified counties in the CPS. The sample in first column removes observations with miscoded employment. A CBP observation is considered a “miscode” if the total employment is smaller than the minimum that could conceivably exist based on the number of firms in each category. For example, if there was one firm in the “5 to 9 employee” category for a given county-industry observation, the observation would be considered a miscode if employment was reported as less than 5 employees. The coefficient estimates, β, reveal an employment level of 1.5 employees in firms with “1 to 4 employees”, a level of 7.2 for firms with “5 to 9 employees”, a level of 12.5 for firms with “10 to 19 employees”, a level of 30.0 for firms with “20 to 49 employees”, a level of 70.4 for firms with “50 to 99 employees”, a level of 146.9 for firms with “100 to 249 employees”, a level of 349.6 for firms with “250 to 499 employees”, a level of 664.6 for firms with “500 to 999 employees”, and a level of 2046.6 for firms with “1000 or more employees”. All of the coefficient estimates are very precisely estimated and each falls within the correct employment range. One can strongly reject the hypothesis that the coefficient estimates equal the midpoint value of the range. Instead, what clearly emerges is that in virtually every case, the employment level is below the midpoint value. By using the midpoint, we would tend to overstate the cost of HIA, because we would be more likely to assign CPS respondents to larger firms.

            The final six columns of Appendix Table 4 demonstrate the robustness of these results. Column (2) reestimates the sample with 2,774 observations in all 58 California counties. Columns (3) and (4) show the results by including employment miscodes, both using the CPS geographic coverage and using the entire state. Columns (5) through (7) return to the original sample, but successively include county fixed effects, industry fixed effects, and county and industry fixed effects. The most striking finding in these robustness checks is that they are all extremely similar to the baseline estimate.

[INSERT APPENDIX TABLES 5 and 6]

            Appendix Tables 5 and 6 show the results from estimating equation (4) using robust regression and ordinary least squares. Robust regression uses iteratively reweighted least squares to estimate both the regression coefficients and the standard errors. The procedure assigns weights to each of the observations. Those observations with high leverage or influence receive lower weights. As with median regression, this procedure reduces the impact of outlier observations. Appendix Table 5 reveals very similar (and sensible) findings using robust regression. The conclusions are not substantively changed in any way compared to using median regression. Appendix Table 6 estimates the models using ordinary least squares (“OLS”), and the results show the influence of outlier observations. The OLS results reveal an employment level of 6.8 employees in firms with “1 to 4 employees”, a level of -11.2 for firms with “5 to 9 employees”, a level of 28.7 for firms with “10 to 19 employees”, a level of 19.9 for firms with “20 to 49 employees”, a level of 71.8 for firms with “50 to 99 employees”, a level of 185.2 for firms with “100 to 249 employees”, a level of 197.2 for firms with “250 to 499 employees”, a level of 226.8 for firms with “500 to 999 employees”, and a level of 3086.3 for firms with “1000 or more employees”. Clearly these estimates, which include a negative value, are uninformative for imputing firm size.

            The estimates from the “baseline” median regression were used to impute firm size. It is reassuring to know that these coefficient estimates from this model using the 2001 CBP are nearly identical to averages published by the state of California’s EDD for the third quarter of 2001. The average employment in the EDD report was 1.3 for firms sized “0 to 4 employees”, 6.6 for firms sized “5 to 9 employees”, 13.6 for firms sized “10 to 19 employees”, 30.6 for firms sized “20 to 49 employees”, 69.0 for firms sized “50 to 99 employees”, 148.8 for firms sized “100 to 249 employees”, 341.0 for firms sized “250 to 499 employees”, 679.9 for firms sized “500 to 999 employees”, and 2377.3 for firms with “1000 or employees”.

            The final step in estimating the probabilities in equations (1a)-(3b) is computing PR_25-49|20-49 and PR_{200-249|100-249}. These are the probabilities that an employee in the 20 to 49 (100 to 249) grouping is actually in the 25 to 49 (200 to 249) category. To do this, Appendix Table 7 regresses the logarithm of number of employers on firm size.

[INSERT APPENDIX TABLE 7 HERE]

Firm size comes from the model estimated in Appendix Table 4, column (1), for five firm size groupings (“10 to 19 employees”,“20 to 49 employees”,“50 to 99 employees”,“100 to 249 employees”, and “250 to 499 employees”). The aggregate number of establishments in each category comes from the 2001 CBP. According to the CBP, in California, there were 394,771 establishments with “1 to 4 employees”, 138,517 with “5 to 9 employees”, 95,576 with “10 to 19 employees”, 70,768 with “20 to 49 employees”, 25,126 with “50 to 99 employees”, 13,898 with “100 to 249 employees”, 3,306 with “250 to 499 employees”, 1,201 with “500 to 999 employees”, and 647 with “1000 or more employees”. These aggregates are based on the same screens as in the model estimated in Appendix Table 4, column (1), that is, data available at the 3-digit NAICS level, without miscodes for employment, and in the counties covered by the 2003 March CPS.

            From the model estimated in Appendix Table 7, column (1), we obtain the predicted values and form a density function for all firms with 20 to 249 employees. From this density function, which tells us the probability of firms being in different firm size intervals, we compute the employee-weighted probability of being in firms of different sizes (using firm size as the weight). The results of this exercise are shown toward the bottom of Appendix Table 7. Unsurprisingly, the probability of being in a firm sized 25 to 49 workers, conditional on being in a firm sized 20 to 49 workers, is quite high, 87.8 percent. Being in a firm sized 200 to 249 workers, conditional on being in a firm sized 100 to 249 workers, is estimated to be 27.2 percent. These estimates form the basis for PR_25-49|20-49 and PR_{200-249|100-249} in assigning HIA eligibility.

            The final three columns of Appendix Table 7 show the results from various other specifications. Column (2) estimates the model for all nine firm size groupings (e.g., “1 to 4 employees”,“5 to 9 employees”,“10 to 19 employees”,“20 to 49 employees”,“50 to 99 employees”,“100 to 249 employees”,“250 to 499 employees”,“500 to 999 employees”, and “1000 or more employees”), and obtains a similar estimated probability for PR_25-49|20-49 but a considerably higher one for PR_{200-249|100-249} (38.4 percent instead of 27.2 percent). Using these probabilities would lead to a higher estimate from HIA than the baseline specification. The final two columns estimate the model using median regression and including the square of firm size. Both lead to extremely similar estimates of the probability as the baseline estimate.

[INSERT APPENDIX TABLE 8 HERE]

            Appendix Table 8 replicates this exercise using the CEHBS. One advantage of the CEHBS over the CBP is that the exact number of California employees is directly reported for each firm in the CEHBS. In total, there were 3,222 establishments in the CEHBS data, which includes firms that did not answer the health insurance questions. These 3,222 firms are grouped by firm size and the aggregate number of establishments is derived by adding up the employer weights for each firm size. Thus, each “observation” in the models in Appendix Table 8 may represent multiple firms of a given size from the CEHBS. As with the models in Appendix Table 7, the logarithm of the number of establishments is used in the estimation. The “baseline” estimate follows the same restrictions as in Appendix Table 7 (e.g., restricts the sample groups with firm sizes between 10 and 499), and uses 251 observations. Both the coefficient estimates, and the derived probabilities are similar to the CBP results. The probability of being in a firm with 25 to 49 employees, conditional on being in a firm with 20 to 49 employers is estimated at 87.6 percent rather than 87.8 percent. The probability of being in a firm with 200 to 249 employees, conditional on being in one with 100 to 249 employees, is estimated at 25.4 percent rather than 27.2 percent. Thus, the results using the CEHBS are quantitatively similar to those of the CBP. The final three columns of Appendix Table 8, using different samples and estimation techniques, show similar patterns to Appendix Table 7.

            With the information on average firm size from the CBP, as well as the probability of being in firms of different sizes, one can effectively use equations (1a)-(3b) to arrive at more accurate estimates of HIA eligibility. This information was applied to all CBP observations in California – the 5,371 observations for all counties and all levels of industry aggregation (zero-digit, two-digit, and three-digit). In instances when the CBP data was missing or miscoded (e.g., because of confidentiality concerns), then the following steps were taken to impute the probability by looking “upstream.” First, for industry/county observations that were missing the number of employers or total employment at the three-digit NAICS level, I substituted the probability from the two-digit industry code at the county-level. Second, if that was not possible, I substituted the probability from the three-digit code at the state-level. Third, if neither of those were possible, the next step was to substitute the probability from the two-digit code at the state-level. Fourth, if none of those steps worked, the next step was to substitute the probability for all industries at the county-level. Finally, for any remaining observations, I substituted the probability for all industries at the state-level.

            A similar procedure was done at other industry levels. For industry/county observations that were missing data at the two-digit NAICS level, I first substituted the probability from the two-digit industry code at the state-level. If that proved impossible, I substituted the probability for all industries at the county-level. Finally, for any remaining observations, I substituted the probability for all industries at the state-level. For county-level observations that were miscoded, I substituted the probability derived from the state-level.

            The final step in assigning HIA coverage for individuals is merging these probabilities of coverage from the CBP to the CPS. The key issue in doing this is that the March 2003 CPS does not directly provide NAICS codes, but rather, the Census coding of industry. Of the 16,779 CPS respondents in the unweighted sample, 8,155 (or 48.6 percent) with work in the previous calendar year had an assigned industry code. Of those 8,155 respondents with a Census industry code, 7,157 respondents (87.8 percent) could be matched to a three-digit NAICS code using a “roadmap” in the appendix of the CPS users manual. Another 8,068 (98.9 percent) could be matched to a two-digit NAICS code, and 8,106 (99.4 percent) could be matched to a one-digit NAICS code . Of the 8,155 respondents, 6,316 (77.4 percent) could be matched by county (or MSA) and two- or three-digit NAICS code for imputing the probability of HIA coverage. The remaining 1,839 respondents could not be matched by both geography and industry, so the county-level probability was instead used. There were 63 individuals who did not have a county (or MSA) code; they were assigned the state-level probability. Overall, 99.2 percent of the working respondents could be matched in one way or another to the CBP data.

            For each worker, a probability was assigned using equations (1a)-(3b) using their response to the firm size question, as well as their county of residence and industry. Using the geographic and industry variation leads to large variations in the probabilities for different workers in the same firm size category. For example, the likelihood that a CPS worker in the “100 to 499” firm size group is eligible for family coverage is 51.4 percent, with a standard deviation of 11.9 percent. The probability of a CPS worker in the “25 to 99” group being eligible for single coverage is 48.6 percent, with a standard deviation of 10.3 percent. Finally, the probability of a CPS worker in the “10 to 24” group being eligible for single coverage with a tax credit is 19.4 percent, with a standard deviation of 7.3 percent. Actual HIA eligibility was computed for each worker by drawing from the uniform distribution, and using the probability cutoffs generated by equations (1a)-(3b) to assign groupings, when appropriate.

            It is also necessary to create health insurance units (“HIUs”) to assign dependent coverage for workers in large firms. I define HIUs in the conventional way – by including the head of household, spouse, minor children under 18, unmarried children between 19 and 22 who are full time students, and disabled children. The HIU definition does not include the head’s parents, grandchildren, foster children, or unrelated individuals. Within a household, separate HIUs are created for related and unrelated subfamilies. The analysis does not create HIUs for domestic partners, so the impact of HIA is potentially understated.

Appendix 2: Technical assumptions used in forming cost estimates of HIA

            The cost analysis relied on two different sources of premium and cost sharing information: the imputed employer contributions provided in the CPS and the premium/cost sharing information from the CEHBS. Both sets of data revealed largely the same story in terms of costs.

            First, the imputed value of employer contributions, contained in the March 2003 CPS, was utilized to estimate the cost of providing coverage in California. The median contribution, for employers paying the entire premium amount in 2002, was $5,914 for a family coverage and $3,621 for individual coverage. The 2003 CPS figures for family coverage are substantially lower than the median cost in the CEHBS ($5,914 versus $8,345), while the cost for single coverage is somewhat higher ($3,621 versus $3,001). For one set of cost estimates, these medians are taken as the “full cost” of an acceptable plan and the total “fee” charged by MRMIB. Of this amount, the employer would be responsible for 80 percent, or $4731.20 for a family plan or $2,896.8 for a single plan.

            These CPS imputations appear very reasonable in light of other estimates. Brown, et al. (2002, footnote 7) find that in the private market the least expensive HMO that would provide comprehensive benefits and minimal cost sharing would cost a 26-year-old single person $1,456 annually. For a family of three the least expensive HMO providing comprehensive benefits would cost $5,486 annually. Since the CPS estimate is for the median cost across many types of plans and many types of individuals or families (rather than for the minimum cost for healthy, young individuals in HMOs), the numbers are quite consistent.

            Analogous figures from the March 2002 CPS show that the median contribution was $5,101 for a family coverage and $3,154 for individual coverage (Yelowitz, 2003). Thus, the total costs increased by 15.9 percent for a family plan in the CPS data, and 14.8 percent for a single plan. The premium increase from the 2002 to 2003 March CPS is very similar to the 15.8 percent increase in health insurance premiums published in KFF/HRET (2004, Chart 7).

            In the case where employees currently receive employer-based coverage and pay none of the premium, the estimated change in cost is $0 with the CPS premiums. In the case, however, where employees pay a portion of their premium, it would be unrealistic to assume that the employer pays at least 80 percent of the cost. Doing so dramatically underestimates the actual cost of meeting the “play” portion of the HIA mandate. To determine the cost of coverage when employers only pay a portion of the premium, the difference between the CPS imputed value of the employer contribution and 80 percent of the estimated cost of the median plan was calculated (in cases where the imputed value of the employer contribution was higher than the 80 percent fee, the estimated increased cost was $0). In cases where eligible workers were provided no current employer based coverage (either uninsured or government insurance only) or cases where the employer does not contribute to their employees premium, 80 percent of the cost of the plan is used to estimate the cost of these employers either “paying” or “playing” under the HIA mandate.

            Finally, these HIA costs were lowered to account for the possibility of “shifting” payment responsibilities across workers in a family. In cases where the HIA family mandate binds, the greater of the worker’s or spouse’s employer contribution was used for the employer’s contribution, thus lowering the additional cost of HIA. For example, if a spouse who worked at a small firm had a health plan with an employer contribution of $4,000 toward family coverage, and the worker declined coverage from a large firm, then the cost of HIA mandate is computed as the difference between the $4731.20 “fee” and the $4,000 contribution, or $731.20. If both spouses worked at firms with 200 or more employees, the mandated cost counts family coverage at one of the two firms only.

            The motivation for using a second source of data, namely the CEHBS, is that the CPS imputation for the employer’s contribution might be viewed as unsatisfactory because health care market has changed greatly since the 1977 NMCES. The CEHBS provides timely, detailed data about both single and family health insurance plans for California employers. In addition to this health insurance data, it also provides information about firm size in California and industry (broken out into mining, construction, manufacturing, transportation/utilities/communications, wholesale, retail, financial, service, and healthcare) . The CEHBS provides sample weights for covered California employees that allowed me to construct the empirical distribution of health care premiums and cost sharing in each relevant firm size/industry grouping for workers. For example, if there were two firms in the 100 to 499 grouping for retail, and one firm represented 1,000 covered California workers (when weighted) and the other represented 9,000 covered California workers, a CPS respondent in that firm size/industry grouping would have a 10 percent chance of being assigned the health insurance characteristics from the first firm and a 90 percent chance of being assigned health insurance characteristics from the second firm.

            One of the primary reasons for drawing from the empirical distribution in the CEHBS, rather than using the mean cost sharing information, is that the “premium sharing” part of the mandate provides a floor on the costs to the employer; even if the mean or median cost sharing is close to satisfying the “premium sharing” HIA requirements (which is largely true for single coverage but not family coverage), there are still distributional issues that will tend to understate costs. For example, if there were three equally sized employers, one who paid 95 percent of premiums, the second who paid 80 percent, and the third who paid 65 percent, then both the mean and median employer cost sharing would be 80 percent. Yet, both measures misstate the cost of HIA since the third employer must increase the cost sharing percentage.

            Once the CEHBS premiums and cost sharing information was assigned to the CPS worker, the cost of the mandate was computed. Because the CEHBS data are in many respects better, the calculation was somewhat different. First, there was no need to extrapolate from a single plan to a family plan (or vice-versa) – both are provided in the CEHBS. Second, some firms may pay for greater (less) than 80 percent of the premium costs, but offer less (more) generous benefits. I assume that if HIA mandates more generous (less) benefits than the employer currently provides, it could decrease (increase) its cost sharing to 80 percent and increase (decrease) its benefits with no increase in costs. Thus, from the cost sharing percentage and the total cost, I compute the employer’s current contribution, and compare that to 80 percent of the median CEHBS premium. The employer’s part of the fee for a family plan in the CEHBS would be $6,676, and either $1920.64 or $2400.80 for a single plan (depending on whether the firm was in the 20 to 49 employee range or above). From these fees, the employer’s actual contribution was subtracted to compute the marginal cost of HIA.

            Finally, as with the CPS premium imputations, these HIA costs were lowered to account for the possibility of “shifting” payment responsibilities across workers in a family. The procedure was identical to the one used on the CPS premium imputations.