We use a gradient-boosting machine learning model trained on CPS microdata to estimate individual unemployment risk based on demographic, economic, and geographic characteristics. The resulting predictions of unemployment risk are then aggregated to assess the distribution of unemployment risk across time, industry, and other dimensions.

We show that much of the national shortfall in leisure & hospitality employment since the pandemic can be attributed to shortfalls in labor demand in dense urban cores. This is likely the result of work-from-home policies reducing foot traffic in these areas. Given the composition of the pre-pandemic workforce, reduced immigration during the pandemic likely also suppressed the labor supply for the industry.

This thesis studies the impact of long-term changes in labor market opportunities on suicide and overdose mortality rates. I use a Bartik instrument to quantify the exposure of commuting zones to exogenous labor demand shocks and confidential CDC mortality data to measure mortality rates. While I find no consistent effect of labor market shocks on suicide, I find that negative shocks lead to increases in overdose mortality.