RETIREMENT RESEARCH April 2017, Number 17-7 DO HEALTH INSURANCE REFORMS BOOST DEMAND FOR OLDER WORKERS BY SES? By Matthew S. Rutledge and Caroline V. Crawford* Introduction Working longer is an effective way for individuals to cost of older workers. This brief examines whether improve their retirement security, but a critical ques- the regulations improved labor market outcomes for tion is whether employers will hire and retain them. older workers, by education, particularly at the small This concern is especially acute for less-educated firms directly affected by the regulations. workers, who often hold middle-skill jobs that are at The discussion proceeds as follows. The first greater risk of disappearing.1 section provides background on the premium restric- One potential way to boost the prospects of older tions. The second section describes the data and workers is to reduce the cost to companies of employ- methodology used for the analysis. The third section ing them. For example, some policy experts have pro- presents the results. The final section concludes that posed making Medicare the primary payer for older while the earnings gap between workers in large and workers’ health care costs or eliminating their Social small firms did shrink, especially for workers with Security payroll taxes.2 And European countries have only a high school degree, the premium restrictions tried wage subsidies.3 did little to increase employment for older workers. This brief, based on the results of a recent pa- per, considers a policy that has already been tried in the United States.4 In the 1990s, nearly every Health Insurance Premium state gradually imposed restrictions on how much employer-sponsored health insurance premiums can Regulations vary across small firms based on the characteristics of While insurers are generally unable to vary employer- their workers. Prior to these regulations, a small firm sponsored health insurance premiums among the that hired even one additional older worker ran the workers at any single firm, premiums historically risk of higher premiums for all of its workers. The could vary substantially from firm to firm, with the regulations made premiums less – or, in some states, highest premiums for companies employing older not at all – dependent on the age or health composi- and less healthy workers. This pricing regime put tion of a firm’s employee pool, thereby reducing the small firms at a particular disadvantage. At a large * Matthew S. Rutledge is a research economist at the Center for Retirement Research at Boston College (CRR). Caroline V. Crawford is a research associate at the CRR. 2 Center for Retirement Research employer, the cost of hiring an additional “high-risk” the worker is with a small firm (fewer than 100 work- worker would be spread among many employees. ers);8 and 3) real annual earnings. The data include But at a small firm, the cost of that high-risk worker other important personal characteristics, notably would raise every worker’s premium, burdening the educational attainment, which is a proxy for socioeco- employer and/or the other employees. nomic status. The education measure is defined as Following the recommendations of the National less than high school, high school, some college, and Association of Insurance Commissioners, states college graduate. The results are presented separately began to pass legislation in the early 1990s restrict- by these categories. ing the ability of health insurers in the small-group The study also compiled a comprehensive dataset market, in which smaller employers bought coverage, on small-group premium restrictions by state and to vary their premiums by the characteristics of the over time. The variable of interest is the strength of firms’ employees. Between 1990 and 1994, 45 states a state’s rate bands, which proscribe the maximum passed some form of restriction on premium setting. ratio of premiums for firms with unhealthy policy- The degree of restriction varied across states and holders relative to firms with healthy policyholders. A over time. Some states adopted “community rating” rate-band ratio of 1 (1:1), i.e. community rating, is the policies that prohibit insurers from varying premium most restrictive policy because premiums do not vary costs for different firms due to their employees’ by the health composition of a firm’s employees. A characteristics. Only a few states passed “pure” com- rate-band ratio of 0 indicates that the state has no rate munity rating policies that forced insurers to charge band law, and is therefore the least restrictive. States every small employer in the state the same average with rate bands in between are assigned the reciprocal per-enrollee cost. Most states with community rating of their ratio; that is, a state where premiums can be 4 adopted “modified” policies, which prohibit the use of times higher in unhealthier firms (a 4:1 ratio) is given underwriting based on health status but permit pre- the value of 0.25. mium variation for other group characteristics such No state had any restriction as of 1989, but the as age and gender. regulations came quickly thereafter: by 1991, 17 states Instead of community rating, the majority of had adopted some restriction. By 1995, 46 states had states adopted rate-band policies, which allow the a restriction in place, and most were strong restric- insurer to vary premiums within an acceptable range; tions – that is, 1.86:1 or lower. Subsequently, some for example, a rate-band ratio of 3:1 restricts the high- states relaxed their restrictions, though many strong est group premium to three times the rate of the least regulations remained in place just before rate bands expensive policy. The rate-band ratios varied across were standardized by the ACA (see Figure 1 for each states and were adjusted periodically by state legisla- state’s status as of 2013). tures, until the Affordable Care Act (ACA) standard- ized the difference between firms with older and younger employee pools at 3:1 starting in 2014.5 Figure 1. Premium Restrictions by State, 2013 The state-by-state adoption of premium-setting restrictions provides a good setting for a natural experiment. Most previous analyses have focused on whether these regulations reduced health insur- ance coverage for younger, healthier workers.6 The few studies that have examined their impact on labor market outcomes did not focus on older workers, who are most likely to be affected.7 Data and Methodology Weak The analysis uses data on individuals ages 25-61 from Strong the Current Population Survey for 1989-2013. The out- Community rating comes of interest are: 1) an indicator for whether the Data not available individual is employed; 2) an indicator for whether Source: Rutledge and Crawford (2016). Issue in Brief 3 The analysis estimates regressions for each labor represents the predicted probability of working in a market outcome as a function of the rate-band ratio, small firm. The gray portion is the predicted proba- an indicator for workers ages 50 or older, and the bility of working in a large firm; though this outcome interaction of these two variables. The earnings re- is not expected to be influenced by the premium gression also includes interactions with a “small firm” restrictions, it helps to establish whether large-firm indicator. Finally, each model controls for standard employment was changing at the same time for older personal demographic characteristics, the state un- workers (ages 50-61) vs. prime-age workers (ages employment rate in the given year, and state and year 25-49). The full height of each bar is the predicted fixed effects. probability of being employed overall. Contrary to the expectation that small-firm employment would increase, all employment rates Results – both overall and in small firms – look virtually the same, no matter the strength of the premium By reducing the cost of providing health insurance restriction. The left panel of Figure 2 indicates that for older workers, the premium restrictions were the overall employment rate (as predicted from expected to increase their employment at the small the regression results) for older individuals barely firms affected by the law. In addition, the restrictions increases as the strength of the premium restrictions were also expected to increase older workers’ earnings increases: from 69 percent with no rate band to 72 at small firms, because lower health premiums could percent with community rating. Small-firm employ- allow employers to pass on the savings to workers ment among individuals 50 and older increases by in the form of higher wages.9 The restrictions were even less. Prime-age individuals effectively saw no not expected to increase employment or earnings in change in small-firm employment or overall employ- larger firms, whose insurance premiums were not ment, which means that older workers did not see affected by the regulations. any relative increase either. Thus, the reforms did not appear to improve older workers’ prospects of being Employment employed at all, even at the small firms that would be most concerned about how expensive these workers Figure 2 shows the employment rate by rate band are to insure. strength, age, and firm size, as predicted from the regression estimates.10 The red portion of the bars Earnings Unlike employment, both prime-age and older work- Figure 2. Predicted Employment Rate by Rate ers – especially those in small firms – do appear to be Band Strength, Age, and Firm Size paid more in states with stronger premium restric- 100% tions (see Figure 3 on the next page). But, contrary to Large firm Small firm expectations, older workers do not appear to benefit 80% much more than prime-age workers. In general, workers at large firms earn more on 60% average than workers at small firms, and the gap is especially large for older workers. But the premium 40% regulations made older workers, in concept, more at- tractive to small firms, who could then afford to offer 20% them higher wages. Indeed, the results indicate that the gap between earnings at large and small firms 0% None Weak Strong CR None Weak Strong CR closes by more in states with stricter premium restric- Ages 50-61 Age 50-61 Ages25-49 Age 25-49 tions. In states with no premium restrictions, work- ers ages 50-61 earn about $13,400 less in small firms Note: “Community rating” refers to a 1:1 rate-band ratio, than in large firms. That difference falls steadily “Strong” ratios are between 1.2:1 and 1.67:1, and “Weak” as insurance premiums become more restricted; in ratios are between 1.86:1 and 4:1. states with community rating, the large-small firm Source: Authors’ estimates from the U.S. Census Bureau, earnings gap is only $9,980, 26 percent less than in Current Population Survey (CPS), 1989-2013. unrestricted states. 4 Center for Retirement Research Figure 3. Predicted Gap Between Average Large fect between the two age groups is not statistically sig- and Small Firm Earnings, by Rate-Band Strength nificant. The fact that prime-age workers – whose job and Worker Age, in 2013 Dollars prospects at small firms were less likely to have been reduced by concerns about their effect on the firms’ $15,000 health insurance costs – also benefited may indicate that firms decided to use any savings from older workers’ lower health premiums to raise salaries for $10,000 all of their employees. But it may also indicate that small-firm earnings increased in stricter-regulated states for other reasons. $5,000 Results by Education $0 None Weak Strong CR None Weak Strong CR The results in Figure 4 indicate that the socioeconom- Ages 50-61 Age 50-61 Ages25-49 Age 25-49 ic group that most benefits from premium-setting re- strictions is high school graduates. Older high school Note: “Community rating” refers to a 1:1 rate-band ratio, graduates working in small firms see a statistically “Strong” ratios are between 1.2:1 and 1.67:1, and “Weak” significant increase in their earnings (6.1 percent), all ratios are between 1.86:1 and 4:1. else equal, when their state moves from no rate band Source: Authors’ estimates from the 1989-2013 CPS. restriction to (pure or modified) community rating. Individuals with college experience also see increas- ing earnings – though of a lesser magnitude and The large-small firm earnings gap also falls for statistically insignificant – when community rating is prime-age workers, however: from $10,670 with no adopted. premium restrictions to $8,150 with community rat- Otherwise, the results do not provide much evi- ing. Because the initial gap is smaller for prime-age dence that the policies were effective. No group saw workers in all rate-band categories, the percentage the expected improvement in small-firm employment. change – 23 percent – is almost equal to the one seen High school dropouts actually saw the earnings gap by older workers, and the difference in the policy’s ef- between large- and small-firm workers increase, Figure 4. Estimated Effect of Adopting Community Rating on Labor Market Outcomes for Individuals Ages 50-61, by Education -5.9 Earnings at a smallfirm Earnings at a small firm 6.1 1.4 3.7 Less than HS -1.7 HS only Small firm employment 0.0 Small firm employment -2.3 Some college -0.7 College+ 3.3 Overall employment Overall employment 3.1 1.1 0.8 -8.0 -4.0 0.0 4.0 8.0 12.0 Estimated percent change Note: Solid bars are statistically significant at least at the 10-percent level. Source: Authors’ estimates from the 1989-2013 CPS. Issue in Brief 5 contrary to expectations. College graduates had no Endnotes statistically significant change in employment or the earnings gap. 1 Autor (2014). 2 Goda, Shoven, and Slavov (2007, 2009). Conclusion 3 The results of the subsidy policies have been The labor supply of older workers is increasing, but mixed; see Huttunen, Pirttilä, and Uusitalo (2010); they cannot work longer if jobs are not available to Boockmann et al. (2007); Schunemann, Lechner, and them. Tighter regulation of health insurance pre- Wunsch (2011); Garcia-Perez and Sanz (2009); and miums for less healthy workers in the small-group Eppel and Mahringer (2012). market should have benefited older workers, allow- ing them greater employment opportunities in small 4 Rutledge and Crawford (2016). firms, but the small-firm employment increase was minuscule. The earnings gap between large and 5 The American Health Care Act of 2017, proposed small firms did shrink in states with stronger premi- by Republicans in the House of Representatives, um restrictions, but older workers did not see greater would raise this ratio to 5:1. increases than younger workers, who stood to benefit, at most, indirectly. The earnings gap shrunk the most 6 Buchmueller and DiNardo (2002) and Adams for high school graduates, but no education group (2007) find some evidence of adverse selection be- saw statistically significant increases in small-firm cause the policies forced insurers to charge more to employment. younger, healthier workers, which made insurance These results suggest that indirectly reducing less attractive to this group. the cost of hiring older workers – by restricting their health insurance premiums – does not improve the 7 Kapur (2003, 2004); Kaestner and Simon (2002). labor market outcomes of any socioeconomic group. Instead, policymakers may want to consider trying 8 Though most premium restrictions define the more direct measures, perhaps by eliminating payroll small-group market as applying to firms with 50 em- taxes for older workers and their employers. ployees or fewer, the data do not allow for identifying   firms of that size consistently over time. 9 Younger workers who were unhealthy could also benefit, but the data lack health status information until 1995, by which time most states had already adopted the premium restrictions. 10 The estimates are from a multinomial logit regres- sion, where the outcomes are: 1) working at a large firm (the base outcome); 2) working at a small firm; and 3) not working. Results are similar in individual regressions examining employment (vs. not working) or small firm employment (vs. large firm employ- ment). 6 Center for Retirement Research References Adams, Scott. 2007. “Health Insurance Market Re- Huttunen, Kristiina, Jukka Pirttilä, and Roope form and Employee Compensation: The Case of Uusitalo. 2010. “The Employment Effects of Low- Pure Community Rating in New York.” Journal of Wage Subsidies.” Discussion Paper 4931. Bonn, Public Economics 91: 1119-1133. Germany: Institute for the Study of Labor (IZA). Autor, David H. 2014. “Polanyi’s Paradox and the Kaestner, Robert and Kosali Ilayperuma Simon. 2002. Shape of Employment Growth.” Working Paper “Labor Market Consequences of State Health In- 20485. Cambridge, MA: National Bureau of Eco- surance Regulation.” Industrial and Labor Relations nomic Research. Review 56(1): 136-159. Boockmann, Bernhard, Thomas Zwick, Andreas Am- Kapur, Kanika. 2003. “Labor Market Implications of mermüller, and Michael Maier. 2007. “Do Hiring State Small Group Health Insurance Reform.” Subsidies Reduce Unemployment Among the Public Finance Review 31(6): 571-600. Elderly? Evidence from Two Natural Experiments.” Discussion Paper 07-001. Mannheim, Germany: Kapur, Kanika. 2004. “The Impact of the Health Centre for European Economic Research (ZEW). Insurance Market on Small Firm Employment.” Journal of Risk and Insurance 71(1): 63-90. Buchmueller, Thomas and John DiNardo. 2002. “Did Community Rating Induce an Adverse Selection Rutledge, Matthew S. and Caroline V. Crawford. Death Spiral? Evidence from New York, Pennsyl- 2016. “Would Reducing the Price of Employing an vania, and Connecticut.” The American Economic Older Worker Improve Labor Market Outcomes Review 92(1): 280-294. by Socioeconomic Status? Evidence from Health Insurance Premium Restrictions.” Working Paper Eppel, Rainer and Helmut Mahringer. 2012. “Do 2016-17. Chestnut Hill, MA: Center for Retire- Wage Subsidies Work in Boosting Economic ment Research at Boston College. Inclusion? Evidence on Effect Heterogeneity in Austria.” Working Paper. Vienna, Austria: Austrian Schunemann, Benjamin, Michael Lechner, and Con- Institute of Economic Research (WIFO). ny Wunsch. 2011. “Do Long-term Unemployed Workers Benefit from Targeted Wage Subsidies.” Garcia-Perez, J. Ignacio and Yolanda Rebollo Sanz. Discussion Paper 2011-26. St. Gallen, Switzerland: 2009. “Do Wage Subsidies Affect the Subsequent University of St. Gallen. Employment Stability of Permanent Workers?: The Case of Spain.” Working Paper Econ 9.18. U.S. Census Bureau. Current Population Survey, 1989- Seville, Spain: Universidad Pablo de Olivide. 2013. Washington, DC. Goda, Gopi Shah, John B. Shoven, and Sita Nataraj Slavov. 2007. “A Tax on Work for the Elderly: Medicare as Second Payer.” Working Paper 13383. Cambridge, MA: National Bureau of Economic Research. Goda, Gopi Shah, John B. Shoven, and Sita Nataraj Slavov. 2009. “Removing the Disincentives in So- cial Security for Long Careers.” In Social Security Policy in a Changing Environment, edited by Jeffrey Brown, Jeffrey Liebman and David A. Wise, 21-38. Cambridge, MA: National Bureau of Economic Research. RETIREMENT RESEARCH About the Center Affiliated Institutions The mission of the Center for Retirement Research The Brookings Institution at Boston College is to produce first-class research Syracuse University and educational tools and forge a strong link between Urban Institute the academic community and decision-makers in the public and private sectors around an issue of criti- cal importance to the nation’s future. To achieve Contact Information Center for Retirement Research this mission, the Center sponsors a wide variety of Boston College research projects, transmits new findings to a broad Hovey House audience, trains new scholars, and broadens access to 140 Commonwealth Avenue valuable data sources. Since its inception in 1998, the Chestnut Hill, MA 02467-3808 Center has established a reputation as an authorita- Phone: (617) 552-1762 tive source of information on all major aspects of the Fax: (617) 552-0191 retirement income debate. E-mail: crr@bc.edu Website: http://crr.bc.edu © 2017, by Trustees of Boston College, Center for Retirement Research. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that the authors are identified and full credit, including copyright notice, is given to Trustees of Boston College, Center for Retirement Research. The CRR gratefully acknowledges the Alfred P. Sloan Foundation for its support of this research. The findings and conclu- sions expressed are solely those of the authors and do not represent the opinions or policy of the Alfred P. Sloan Foundation, Boston College, or the Center for Retirement Research.