Research

Unequal Recovery from Recessions: Skills Learning Among Young Workers (Job Market Paper)

• • Abstract: Numerous studies indicate that graduating and entering the job market during a recession leads to persistent wage losses, initially at 2% per one-point rise in the national unemployment rate and remaining significant over a decade. However, this statistic conceals significant heterogeneity in outcomes. I analyze the National Longitudinal Survey of Youth data set and use quantile regression to summarize the variation of young workers’ life-cycle wage profiles. I find workers in higher residual wage quantiles have faster recovery speeds. This heterogeneity can be attributed to some unobservable characteristics of workers. To understand these findings, I present a life-cycle model with two-dimensional Ben-Porath skills accumulation among workers differing in three dimensions. Dimensions related to wage determination that are not directly observed from data are true abilities, initial self-beliefs about true abilities, and initial stock of skills. This model extends the classical Ben-Porath model by accounting for workers’ uncertainties about their skills accumulation and introducing aggregate fluctuations in the demand for skills. The counterfactual experiment reveals that the endogenous change in workers’ investment in skills accumulation due to the recession explains up to 20% of wage losses every year throughout the recovery period and 90% from the recovery periods even after the recession is over. I also find workers with the lowest ability to accumulate human capital suffer the most of the losses, as supposed to those with low initial levels of skills or initial low self-belief.

  • Full Paper: Latest Version

Estimating Demand Shocks from Foot Traffic: A Big-Data Approach

with Marina Azzimonti  and David Wiczer

• • Abstract: Unlike the manufacturing industries, the demand shocks in customer-facing sectors are challenging to measure because the arrival of customers largely influences them. In this paper, we employ high-frequency foot traffic data to estimate demand dynamics in the Service, Retail Trade, and Health sectors from 2018 to 2019 in three New York boroughs: Manhattan, Brooklyn, and Bronx. We begin by eliminating the location-specific and trend biases in sampling, which is the technical contribution of our paper. We find substantial within-year variations in establishments with high foot traffic and large monthly fluctuations in demand, even among establishments with zero growth. Using an unsupervised machine learning approach, we categorize establishments into four groups and then estimate an AR(1) process of year-over-year demand growth patterns, allowing for group-specific effects. We find that the two-step grouped fixed-effects estimators are heterogeneous across clusters, and they are different than pooled regression estimates, indicating that the pooled regression results are not representative. The “Fast-Growth” and “Fast-Declining” groups of establishments also exhibit substantial monthly volatility in demand, suggesting that the growth in demand is not a smooth process. We uncover a negative relationship between persistence and the standard deviation of demand shocks within an establishment, implying similar unconditional variance across groups. Heterogeneity in demand dynamics is observed when restricting attention to a single geographical unit or production sector.

  • Full Paper: Latest Version