Health State Risk Categorization: A Machine Learning Clustering Approach Using Health and Retirement Study Data, The Journal of Financial Data Science May 2022 (forthcoming)
with Dr. Dhagash Mehta
Upward Intergenerational Insurance for Long-Term Care (Job Market Paper)
Abstract: This paper assesses the value of Medicaid for recent retirees in insuring against long-term care risks while taking into account child-to-parent transfers. Parent retirees receive substantial transfers in the forms of informal care and financial transfers from children. To understand the role of upward intergenerational insurance for old-age health risks, I develop a dynamic model for parent-child pairs and childless retirees. A vital feature of the model is that a parent and her child interact strategically to make decisions about transfers in a non-cooperative game. I estimate and calibrate the model to match the Health and Retirement Study (HRS) data. Using the calibrated model, I calculate the insurance values of Medicaid relative to its cost for retirees and child households. Compensating variation calculations suggest that childless retirees value every dollar of Medicaid insurance at $2.20, which is twice the value for parent retirees ($1.10). Furthermore, I find that middle-income parent retirees value Medicaid insurance less than poor and wealthy parent retirees. An additional result of the paper is that child households also value Medicaid. This decomposition provides a new consideration for the efficient design of Medicaid benefits, particularly in light of a growing population aging without children.
Information Technological Change and Sagging Non-Routine Cognitive Employment Growth in the 2000s
Abstract: The growth rate of non-routine cognitive occupations (in terms of employment share) declined since 2000 after strong growth in 1980s and 1990s. I propose and test the hypothesis that tasks related to information gathering and processing have been substituted by recent information technology after 2000. I nd that 57 percent of the slowing growth of employment share of non-routine cognitive occupations can be attributed to within-industry changes, and the rest (43 percent) is due to between-industry changes. The estimated model of conditional labor demand and product demand explains 95 percent of such a decline for the non-routine cognitive occupations after 2000. In a counterfactual exercise, I nd that the employment share of non-routine cognitive occupations would have grown by 1 percentage point instead of 0.26 percentage points annually after 2000, if recent technological change had complemented information-related tasks in the same way as before 2000. My nding depicts the intricacy between technology and employment under the broad context of skill-biased technology change.