Working Papers
"Happy Together or Home Alone: A Structural Model of the Role of Health Insurance in Household Joint Retirement"
(Dissertation) (Now work with Leora Friedberg and Dajun Lin to get paper published) Download Appendix Here
Abstract: The baby boomers are approaching retirement, and the majority of them are married. Simultaneously, employers are less likely to provide retiree or spousal health insurance, making it important to understand how health insurance affects couples’ joint retirement decisions. I develop a dynamic model in which married couples jointly decide when to retire, how to use available insurance, and how much to save. Insurance plans vary by plan characteristics (including premium, deductible, and coinsurance rates). I estimate my model with Maximum Simulated Likelihood estimation using data from the Health and Retirement Study (HRS), the Medical Expenditure Panel Survey (MEPS) and the Panel Study of Income Dynamics (PSID). I find that, for workers who lose employer-provided coverage in retirement, gaining employer-provided retiree coverage delays retirement by 1.1 and 0.5 years for husbands and wives, respectively. Similarly, raising the Medicare eligibility age is predicted to delay retirement, while the Affordable Care Act (ACA) is predicted to accelerate it. The effects of Medicare are bigger than the effects of the ACA but smaller than the effects of EPHI due to the differences in plan quality. In addition, people value health insurance mostly (80%) because it smooths consumption. Furthermore, I find that spousal coverage motivates simultaneous retirement by delaying husbands' retirement and accelerating wives' retirement.
(Dissertation) (Now work with Leora Friedberg and Dajun Lin to get paper published) Download Appendix Here
Abstract: The baby boomers are approaching retirement, and the majority of them are married. Simultaneously, employers are less likely to provide retiree or spousal health insurance, making it important to understand how health insurance affects couples’ joint retirement decisions. I develop a dynamic model in which married couples jointly decide when to retire, how to use available insurance, and how much to save. Insurance plans vary by plan characteristics (including premium, deductible, and coinsurance rates). I estimate my model with Maximum Simulated Likelihood estimation using data from the Health and Retirement Study (HRS), the Medical Expenditure Panel Survey (MEPS) and the Panel Study of Income Dynamics (PSID). I find that, for workers who lose employer-provided coverage in retirement, gaining employer-provided retiree coverage delays retirement by 1.1 and 0.5 years for husbands and wives, respectively. Similarly, raising the Medicare eligibility age is predicted to delay retirement, while the Affordable Care Act (ACA) is predicted to accelerate it. The effects of Medicare are bigger than the effects of the ACA but smaller than the effects of EPHI due to the differences in plan quality. In addition, people value health insurance mostly (80%) because it smooths consumption. Furthermore, I find that spousal coverage motivates simultaneous retirement by delaying husbands' retirement and accelerating wives' retirement.
"Effect of Medicaid on Mental Health Care Use and Cost" with Caruso Brown (Deputy Executive Director of Region Ten, Virginia Community Services Board) and Steven Stern (Stony Brook University)
Abstract: We use a zero-inflated Poisson and linear regression models and data from the Medical Expenditure Panel Survey to examine the effects of demographic variables and health insurance coverage on the usage and unit cost of mental health related treatments. We find that individuals with Medicaid use more mental health services than those with private insurance but at significantly lower unit costs. Overall, our results suggest that Medicaid expansion will increase total mental healthcare spending by 28%. We suggest several reasons for this increase, and we discuss what those reasons imply about reducing Medicaid expenditures across states.
Abstract: We use a zero-inflated Poisson and linear regression models and data from the Medical Expenditure Panel Survey to examine the effects of demographic variables and health insurance coverage on the usage and unit cost of mental health related treatments. We find that individuals with Medicaid use more mental health services than those with private insurance but at significantly lower unit costs. Overall, our results suggest that Medicaid expansion will increase total mental healthcare spending by 28%. We suggest several reasons for this increase, and we discuss what those reasons imply about reducing Medicaid expenditures across states.
"Cost and Service Variation across Community Service Boards for Mental Health Services in Virginia" with Caruso Brown (Deputy Executive Director of Region Ten, Virginia Community Services Board) and Steven Stern (Stony Brook University)
Abstract: We estimate the effects of various factors on service provision and unit cost by public mental health providers using administrative data from Virginia and collected data from the Current Population Survey (CPS). We find that while there is significant variation in service provision levels and unit costs across providers, it is difficult to explain much of the variation. Especially with respect to cost results, this suggests some potential for cost savings, but also a need for better data on cost determinants.
Abstract: We estimate the effects of various factors on service provision and unit cost by public mental health providers using administrative data from Virginia and collected data from the Current Population Survey (CPS). We find that while there is significant variation in service provision levels and unit costs across providers, it is difficult to explain much of the variation. Especially with respect to cost results, this suggests some potential for cost savings, but also a need for better data on cost determinants.
Works in Progress
"Joint Health Transitions of Spouses" with Leora Friedberg (University of Virginia) and Dajun Lin (University of Virginia)
Abstract: We explore how health of spouses co-evolves in middle- and old-aged households. Evolution of health at later ages affects not only well-being but also optimal household saving behavior and government spending through Medicare and Medicaid. Much of the existing literature on these topics treats health of spouses as independent, though they are strongly correlated in the crosssection. If this correlation reflects selection into marriage exclusively (due to assortative matching, for example), then health might evolve conditionally independently after accounting for such a correlation.
We estimate a bivariate probit model using self-reported health of spouses in the first 12 waves of the Health and Retirement Study, and we condition on a rich set of variables associated with individual health, including age, education, race, chronic disease and experience with SSI or SSDI. We investigate three types of interdependencies among couples’ health. Because the long panel allows us to distinguish both individual health shocks and individual random effects, we allow both to be correlated between spouses. We further test whether the health of each spouse evolves conditionally independently.
We find that the evolution of spouses’ health is interdependent and, moreover, asymmetrically so between husbands and wives. Husbands and wives appear to help each other stay in good health, but only wives appear to help their husbands' health improve when either of them is in bad health. We argue that this asymmetric health interdependency may be causal, given our comprehensive controls for unobservable correlations between spouses and for individual determinants of health. While idiosyncratic shocks to health are correlated among spouses, individual random effects are not significantly correlated and have a substantially smaller variance than idiosyncratic shocks; this absence of persistent health differences across households reduces concerns that, for examples, assortative matching by health type might explain the health interdependency that we estimate. We seek corroborating evidence from the American Time Use Survey by investigating whether wives and husbands spend systematically different amounts of time caring for each other in particular health states.
Abstract: We explore how health of spouses co-evolves in middle- and old-aged households. Evolution of health at later ages affects not only well-being but also optimal household saving behavior and government spending through Medicare and Medicaid. Much of the existing literature on these topics treats health of spouses as independent, though they are strongly correlated in the crosssection. If this correlation reflects selection into marriage exclusively (due to assortative matching, for example), then health might evolve conditionally independently after accounting for such a correlation.
We estimate a bivariate probit model using self-reported health of spouses in the first 12 waves of the Health and Retirement Study, and we condition on a rich set of variables associated with individual health, including age, education, race, chronic disease and experience with SSI or SSDI. We investigate three types of interdependencies among couples’ health. Because the long panel allows us to distinguish both individual health shocks and individual random effects, we allow both to be correlated between spouses. We further test whether the health of each spouse evolves conditionally independently.
We find that the evolution of spouses’ health is interdependent and, moreover, asymmetrically so between husbands and wives. Husbands and wives appear to help each other stay in good health, but only wives appear to help their husbands' health improve when either of them is in bad health. We argue that this asymmetric health interdependency may be causal, given our comprehensive controls for unobservable correlations between spouses and for individual determinants of health. While idiosyncratic shocks to health are correlated among spouses, individual random effects are not significantly correlated and have a substantially smaller variance than idiosyncratic shocks; this absence of persistent health differences across households reduces concerns that, for examples, assortative matching by health type might explain the health interdependency that we estimate. We seek corroborating evidence from the American Time Use Survey by investigating whether wives and husbands spend systematically different amounts of time caring for each other in particular health states.
"Shared Extended-Family Caregiving in China" with Eric Fesselmeyer (National University of Singapore) and Steven Stern (Stony Brook University)
Abstract: We use different models and data from the China Health and Retirement Longitudinal Study (CHRLS) to examine how families make decisions about who cares for the aging parent, which adult children the aging parent provides childcare for, and financial transfers between the aging parent and adult children. The baby boomer generation in China is approaching their 70s, and the vast majority of them are parents of the first generation of children born under the one-child policy. Understanding how families coordinate childcare and long-term care decisions is important in developing appropriate public policies concerning long-term care arrangements for the elderly.
Abstract: We use different models and data from the China Health and Retirement Longitudinal Study (CHRLS) to examine how families make decisions about who cares for the aging parent, which adult children the aging parent provides childcare for, and financial transfers between the aging parent and adult children. The baby boomer generation in China is approaching their 70s, and the vast majority of them are parents of the first generation of children born under the one-child policy. Understanding how families coordinate childcare and long-term care decisions is important in developing appropriate public policies concerning long-term care arrangements for the elderly.
"Analysis of Commercial Structure and Urbanization in China" with Juan Wang (Hunan University of Commerce), supported by NSF of China ($30,000)
Abstract: We use a Multinomial Logit model and administrative data from Hunan, China to examine the effects of urbanization policies in China on the commercial structure of newly developed urban areas, and the health and welfare of people there.
Abstract: We use a Multinomial Logit model and administrative data from Hunan, China to examine the effects of urbanization policies in China on the commercial structure of newly developed urban areas, and the health and welfare of people there.