The cost of health care for state and local government employees is increasing rapidly, as it is for workers across the economy. Since state and local governments are large employers — one in seven people work for state and local governments — these cost increases are materially important. Estimates suggest that state and local governments spent $70 billion on health insurance in 2001 (in 2012 dollars), and $117 billion in 2010. The real increase was roughly $2,400 per state and local government employee, or $150 per U.S. resident.

Adjusting to these cost increases is more difficult for state and local governments than for private businesses. One strategy that businesses use to address rising costs is to pass those costs back to workers in the form of increased cost sharing for health insurance, less generous coverage, lower contributions to employee benefits, or smaller wage increases (Summers, 1989; Gruber, 1994; Kolstad and Kowalski, 2012). However, in a setting where wages and benefits are covered by union contracts — as is the case with a good share of state and local employees — the ability to effect these adjustments may be limited. When wages and benefit packages cannot be adjusted, increases in health care spending are equivalent to an increase in input costs, much like a price increase for electricity would be. In private businesses, some of this cost increase would show up in higher prices. Prices are not as flexible in the public sector, however, since the price for state and local services is the tax rate. Tax increases may be directly constrained by institutions, as with property tax limits in California, or may be politically difficult. Debt issuance by state and local governments similarly faces institutional and political constraints. If limitations to adjustment of taxes and debt are binding, that leaves reductions in inputs, and with them the quality or amount of public service provision, as the possible responses to increased benefit costs.

The exact impact of rising benefit costs therefore depends on which aspects of public budgets are constrained and which are relatively flexible. When compensation schemes, revenue, and debt issuance are fixed, cost increases may reduce the quality of public services (e.g., worse schools and more crime). Loose deficit-financing restrictions may allow burdens to be shifted onto future taxpayers. Cross-government transfer arrangements (e.g., revenue sharing across school districts) may similarly loosen the revenue-raising constraints faced by local governments. Finally, the strength of public-sector unions may drive the extent to which benefit costs can be shifted back onto government employees. The question of which margins will yield is ultimately empirical. Our research therefore focuses on examining data that can help identify which are the most important margins in practice.

We undertake two types of empirical analysis. First, we examine data describing how much state governments contribute to health insurance for their employees. We use these data to assess the extent to which state governments have shifted the costs of health insurance back to workers, in the form of less generous coverage. We find that in recent years, when fiscal conditions have been tight, health insurance premiums for state workers have grown materially less rapidly than premiums for comparable private- sector employers; this slower premium growth for state workers reflects, for example, changes from traditional comprehensive plans to networked plans, increases in deductibles, and/or non-transparent reductions in access due to reductions in payments to providers. Interestingly, the share of the premium paid by state workers has tended to rise in states with high rates of public-sector unionization, where the employee share started at a low base, while the share has fallen elsewhere.

Our interpretation of these outcomes is as follows. During the good times that preceded the financial crisis and recession, pension and health benefits were an effective, non-transparent way to increase public-worker compensation; state workers were able to build up such benefits without bearing significant costs in the form of wage offsets. When the crisis hit and budgets tightened, public budgets fell under greater scrutiny. Because state workers were initially paying less than dollar-for-dollar of benefits, their compensation was excessively tilted toward benefits. So when faced with compensation cuts, state workers were relatively willing to accept benefit cuts, which came in the form of stingier insurance and a higher worker-financed share of the premium.

We next turn to an analysis of rising benefit costs in the context of school districts, where workers’ health benefits have taken center stage in recent budget debates (Costrell and Dean, 2013). In this setting we can more fully assess the effects of benefits on total compensation costs; total spending; revenue raising; and a proxy, albeit a limited one, for student outcomes — the dropout rate. The analysis relies on differences in baseline district benefit levels and differences in regional growth in health expenditures; we use these to predict the benefit growth that would occur if school districts took no offsetting actions. Our initial finding, namely, that these “exogenous” factors predict actual benefit growth quite accurately, suggests that, at least on average, school districts did little to counteract benefit growth within the benefit package itself.

Looking both across districts and across employee groups within districts (e.g., across teachers, administrators, maintenance, and food-service workers), we find that only a small fraction of increases in benefit costs are offset through reductions in wages. Each dollar in benefit growth is associated with an 85-cent increase in total compensation. The results thus provide evidence that the market for public-sector workers deviates from the competitive, private-sector benchmark analyzed by Summers (1989), Gruber (1994), and Kolstad and Kowalski (2012). We next analyze how school districts finance these increases in benefits. To our initial surprise, we find that benefit-driven increases in employee compensation were financed by transfers from higher levels of government.

A detailed inspection of these revenues reveals that they come from sources subject to significant discretionary reporting (Cullen, 2003). One third of the relevant dollars are associated with “categorical aid” for students classified as having special needs or requiring remedial education. Recent work documenting troublingly high error rates in school lunch programs (Bass, 2010) emphasizes the flexibility of school reporting and the limitations of the systems through which eligibility claims are validated. We also find that the strength of teachers’ unions mediates school districts’ responses to benefit growth. The relationship between our projections of benefit growth and actual benefit growth is strongest in school districts with strong teachers’ unions. Districts with weak unions appear to have offset increases in health care costs much more through reductions in the generosity of benefits. Inflows of categorical aid also appear to be mediated by union strength. The same is true of inflows of general formula assistance, though this result is imprecisely estimated.

Finally, we find that benefit growth was associated with declines in student performance as measured by dropout rates. The reorganization of students required to increase flows of categorical aid may thus have worked to students’ detriment. Because we estimate this final result with moderate precision on a sample severely constrained by data limitations, it should be treated with caution.

This Research Brief is based on Clemens and Cutler (2014), available at ttp://