1. Employer-Sponsored Health Insurance for Early Retirees: Impacts on Retirement, Health and Health Care Erin Strumpf, Ph.D. McGill University AcademyHealth Health Economics Interest Group June 7, 2008 Funding from the National Institute on Aging, Grant Number T32-AG00186, is gratefully acknowledged.

2. Background Employer-sponsored health insurance is an important source of coverage for older Americans Rates of employer offer of retiree health insurance (RHI) have declined by 50%, from 66% of large firms in 1988 to 33% in 2005 Based on these declining rates, we can expect that future cohorts of retirees will have much lower rates of RHI coverage

3. Research Question What implications can we expect among Americans ages 45-64? Measure the effect of RHI offer on: –Retirement –Health –Health care spending

4. RHI Offer Retirement RHI Coverage Health How Does RHI Offer Affect Health? Medical Care Use and Spending

5. Existing Literature Effect of health insurance on retirement –Strong evidence that health insurance affects retirement decisions, but generalizability is often limited Effect of health insurance on health –Elderly (Medicare): no impact on mortality, some increase in utilization and improvement in self- reported health –Non-elderly: some evidence of small positive effects for marginal populations, mostly no measurable effects Madrian 1994, Gruber and Madrian 1995, 1996, Rust and Phelan 1997, Blau and Gilleskie 2001; McWilliams, et al. 2003, Levy and Meltzer 2004, Meara, et al. 2005, Finkelstein and McKnight 2005, Cutler and Vigdor 2005.

6. Data Health and Retirement Survey (HRS) 1992-2002 A longitudinal study of older Americans with interviews every two years Sample restrictions: –respondents aged 47-63 and report having employer- sponsored health insurance in 1992 –years when respondents are still under age 65 RHI Offer: can continue current employer- sponsored coverage in retirement

7. Identification Employer-sponsored coverage RHI OfferNo RHI Offer Don’t RetireRetire Don’t Retire Health and Medical Spending Need to show: RHI offer is conditionally exogenous. Conditional on offer status, there is no differential selection into retirement with respect to health.

8. Identification Is RHI offer conditionally exogenous? –Summary statistics for two groups –Robustness checks: subsamples and propensity score weighting Is there differential selection into retirement? –Interact health status with RHI offer in retirement model –Scale total estimates by percent retired –Estimates from retired, placebo tests on not retired

9. Summary Statistics 1992 Total ESIRHI offerNo RHI offer Age 55.2 0.051 55.4 0.061 54.8* 0.098 Education 12.9 0.033 13.0 0.039 12.7* 0.068 Fair/Poor Health 13% 0.004 13% 0.005 14% 0.008 OOP Health Spending 1,311 62 1,292 75 1,381 130 Mother Alive 45% 0.006 45% 0.008 44% 0.013 Married 83% 0.005 85% 0.006 77%* 0.011 Works Full-Time 65% 0.006 61% 0.008 73%* 0.011 Ages 47-63. Means and standard errors (adjusted for survey design and clustering at the individual level). * significantly different from RHI offer group at p<0.01.

10. Full-Time Retirement Pr(Retirement it ) = α + β1 RHIoffer i1 + X it β2 + Year t + ε Covariates: –sex, race, education level, age, marital status, self- reported health –spouse’s demographics –household income and assets –pension characteristics, vesting age, and industry and occupation Conditional on ESI and not retired in 1992 RHI offer increases probability of early retirement by 7 percentage points, or 35 percent

11. Differential Effects by Health Status Pr(Retirement it ) = α + β1 RHIoffer i1 + β2 HealthShock it + β3 RHIoffer i1 *HealthShock it + X it β4 + Year t + ε New health shock occurs before retirement –Chronic: congestive heart failure, high blood pressure, diabetes, lung disease, arthritis or a psychiatric illness (51%) –Acute: heart attack, angina, stroke or cancer (13%)

12. Differential Retirement by Health Status Full-Time Retirement Chronic Health ShockAcute Health Shock RHI offer0.065***0.084***0.067***0.082*** [0.010][0.014][0.010][0.013] Health Shock-0.037**-0.0390.0560.057 [0.019][0.023][0.044][0.055] Offer*Shock0.026-0.018-0.039-0.080 [0.025][0.029][0.040][0.046] Lagged Shock (2 yrs)0.0320.156* [0.025][0.062] Offer*Lagged Shock-0.037-0.067 [0.026][0.050] N12,3669,51612,0859,387 Marginal effects from probit models. Std errors adjusted for survey design and clustering at the individual level. *significant at 5%, ** 1%, *** 0.1%

13. Health Outcomes Y it = α + β1 RHIoffer i1 + X it β2 + Year t + ε Fair/poor health based on self-reported health measure (1=excellent, 5=poor) Change in self-reported health ranges from -4 to 4 Change in ADLs performed with difficulty ranges from -5 to 5 Covariates are sex, race, education level, age, and self- reported health in wave 1 Scaled estimates, use not retired group as a placebo test

14. Estimated Effect of RHI Offer on Health Status TotalScaledNot RetiredRetired Fair/Poor Self-Reported Health RHI offer-0.0083-0.0366-0.0089-0.0308 [0.007] [0.018] Change in Self-Reported Health RHI offer0.00010.0002-0.0038 [0.008][0.009][0.026] Change in ADLs RHI offer-0.0030-0.01240.0026-0.0396 [0.005] [0.022] The fair/poor health regression is conditional on not being in fair/poor health at baseline. Std errors adjusted for survey design and clustering at the individual level. *significant at 5%, ** 1%, *** 0.1%

15. Out-of-Pocket Medical Care Spending Distribution of medical care spending significantly right-skewed Calculate residual out-of-pocket spending after controlling for age, sex, race, education, baseline health status and year Centile treatment effect: Δ p = {resid spend p (offer = 1) – resid spend p (offer = 0)}

16. $

17. $

18. $

19. Insurance Value of RHI U (household income – out-of-pocket medical spending) Subject each individual to random draws from the empirical distribution of spending in the offered and not offered groups Calculate risk premia based on expected utility

20. Utility Analysis Results Out-of-Pocket Spending*Mean Risk Premium MeanSD Retired Men 60-64 No offer$2,288$4,674$8,929 RHI offer$1,824$4,020$5,101 Difference$465$3,828 Net$3,363 Retired Women 60-64 No offer$2,762$5,080$9,810 RHI offer$2,089$4,051$6,013 Difference$673$3,797 Net$3,124 These estimates use a CRRA utility function and a coefficient of risk aversion equal to 3. * Spending draws are capped at 90% of income.

21. Summary of Findings RHI offer increases the probability of early retirement by 35% RHI offer has no significant effects on health status RHI offer provides significant risk protection, decreasing out-of-pocket medical spending by 20% in the top 40% of the spending distribution among retirees Retired men aged 60-64 value RHI at about $3,400; women $3,100

22. Policy Implications Lower early retirement rates and delayed retirement Decreased financial risk protection: changes to individual insurance market and/or public programs Decline of employer-sponsored health insurance more broadly