There is increasing recognition that retirement decisions among married persons are influenced by health and financial incentives faced by both members of the couple. Coile examined the relative importance of one’s own and one’s spouse’s financial incentives in the retirement decision of couples (CC Coile, 2004b). In a separate piece, Coile examined how health shocks affected couples’ labor force participation, concluding that health shocks “result in real financial losses for families and are an important source of financial risk for older households” (CC Coile, 2004a). Coile and Gruber (CC Coile and J Gruber, 2007) and Coile (CC Coile, 2004b) argue that including forward-looking pension incentives in retirement models is essential—and they carefully show that peak value incentive calculations perform better than the option value approach first suggested by Stock and Wise (1990). Surprisingly, Coile’s paper on health shocks only controls for net wealth and liquid assets of the couple to examine how health shocks affect individual retirement decisions (CC Coile, 2004a).

We aim to combine the approaches in the above series of papers by controlling for current wealth (including pension accrual), as well as forward-looking financial incentives (peak value pension wealth) of both household members to understand the role of health shocks and financial incentives in an integrated model of a married individual’s retirement decision. This effort should clarify whether previously omitted forward-looking financial incentives can explain the strong role attributed to health shocks in the retirement decisions of coupled individuals, or vice versa. Secondly, it will help inform the debate on which determinant of retirement behavior is the most important empirically.

Our reduced form model will include own and spousal health characteristics and financial incentives, using 8 waves (1992 to 2006) of the Health and Retirement Survey (HRS) linked to Social Security Administration (SSA) data. Using linked HRS-SSA data provides the distinct advantages of minimizing measurement error on work behavior and pension earnings while affording the use of a rich set of initial health controls for both members of the couple. Also important for this paper, the HRS has a rich number of health shock measures: we will examine new acute events such as heart attacks and new cancer diagnoses (M McClellan, 1998) as well as broader measures of shocks such as new functional limitations (CC Coile, 2004a) and inpatient hospitalization of one’s spouse (N. A. Christakis and P. D. Allison, 2006). For context, we will also examine the extreme case of how the death of one’s spouse influences one’s own retirement decision.