Working Paper - July 2020
The economic questions that can be addressed using many of the large administrative employer-employee linked data sets with workers' earnings have been limited by the absence of information on workers' base wages, variable compensation, hours or weeks worked, and other factors determining workers' earnings. This paper presents a set of machine learning methods that identify each worker's unobserved persistent base wages, paydays weeks, and annual bonuses from the worker's quarterly earnings. I then implement and evaluate the quality of these methods using quarterly earnings data in the U.S. Census Bureau's Longitudinal Employer-Household Dynamics (LEHD) dataset, an employer-employee linked dataset for the United States. Using the estimated nominal wages of workers in 30 U.S. states, I document four patterns of nominal wage adjustment: i) estimated persistent wage changes exhibit downward nominal wage rigidity, ii) optimal real wage cuts are suppressed by downward nominal wage rigidity, iii) workers' nominal raises follow a Taylor-like pattern, with the probability of a wage raise spiking every four quarters, and iv) the timing of workers' annual raises are synchronized within the firm.
Working paper with Edward Olivares - June 2020
Six weeks after the national emergency declaration for the COVID-19 pandemic, unemployment insurance (UI) claims as a share of total state employment differed by as much as 24 percentage points across states. This paper examines three explanations for these dramatic differences: statewide stay-at-home orders, the industry composition of states' employment, and the historical utilization of states' UI systems. We find that i) the surge in UI claims began and often peaked before states enacted stay-at-home orders; ii) relative to other states, states that would eventually enact stay-at-home orders had high UI claim rates before the orders were enacted; iii) six weeks after the national emergency declaration, statewide stay-at-home orders accounted for less than 30% of the difference in cumulative initial UI claims between states that ever-versus-never enacted statewide stay-at-home orders; iv) industry-related exposure to the COVID-19 pandemic can account for twice as much of the state-level variation in UI claims relative to stay-at-home orders; and v) states with greater historical utilization of their UI systems by unemployed workers tended to both receive more UI claims and process these claims more quickly at the onset of the COVID-19 pandemic.
Job Market Paper - November 2019
Whether downward nominal wage rigidity causes firms to destroy jobs affects both optimal monetary policy and the asymmetry of employment fluctuations over the business cycle. This paper provides quasi-experimental evidence that downward nominal wage rigidity causes firms to destroy jobs and that this effect is empirically relevant for the macroeconomy. Given the unanticipated nature of the financial collapse in Q3 of 2008, differences across firms in their patterns of seasonal nominal wage adjustment generated heterogeneity in firms' exposure to downward nominal wage rigidity in Q4 of 2008. To identify these seasonal patterns, I develop a set of machine learning tools that I apply to longitudinal data on individual U.S. firms. I find that exposure to downward nominal wage rigidity generated by firms' seasonal wage adjustment patterns accounts for 23% of the spike in aggregate job destruction that occurred in Q4 of 2008. Since this empirical finding runs counter to the assumption in many macro models that downward wage rigidity does not cause job destruction, I present a model wherein downward nominal wage rigidity causes inefficient job destruction, while ensuring, à la Barro (1977), that workers and firms do not forgo mutually beneficial nominal wage cuts.
We consider sorting in the labor market, that is, whether high or low productivity workers and firms tend to match with each other, and how this varies over time using U.S. linked employer-employee data. Composition changes of workers and firms move in opposite directions over the business cycle. During and after recessions, low-rank workers are less likely to work, while the employment share of low-rank firms increases. The agreement between worker and firm ranks increases in the early stages of labor market downturns. We consider these results in the context of a model of cyclical labor market sorting.
Working Paper - November 2019
The pace of worker reallocation in the U.S., measured by employer-to-employer (EE) transitions, has been declining since the early 1990’s. This paper explores the impact of older workers’ decisions to delay their retirement from the labor force on the pace of worker reallocation. Using variation in the age composition of firms’ workforces and shifts in workers’ retirement rates caused by two Social Security rule changes, I measure the effect of a worker’s voluntary separation on a firm’s replacement hiring rate. After documenting that the average worker retires almost two years later for cohorts reaching early retirement age in 2008 versus 1990, I use a vacancy chain framework to evaluate the effect of workers’ delayed retirements on the overall pace of worker reallocation. A retiring worker may generate a chain of vacancies (and EE transitions) if the employer replaces the retiring worker with an already employed worker. This new hire must then quit their old job, thus creating a new vacancies and perpetuating the vacancy chain. I find that approximately 30% of the secular decline in the rate of EE transitions from 1990 to 2015 can be explained by the delaying of retirement by older workers.
Working Paper with Henry Hyatt and Kristin Sandusky - July 2020
Studies on the decline in U.S. labor market fluidity have focused on employees and excluded most of the rapidly growing population of the self-employed and business owners. In this paper, we use administrative records data to construct labor reallocation rates that include business income recipients. In our more inclusive definitions, declines in the annual hire and separation rates from 1994 to 2014 were smaller by 1.3 to 1.4 percentage points (8.3% to 8.7%), mainly among jobs that were secondary sources of income or short in duration. We present evidence that workers' transitions between wage and salary jobs and self-employment largely represent actual labor reallocation (as opposed to reclassification of employees as independent contractors). We also find that secondary business income displaced fewer wage and salary jobs in the 2010s than it did in the 1990s, suggesting that self-employment is an increasingly important part of labor market fluidity.
Work In Progress
A Microfoundation for Taylor-style Annual Nominal Wage Adjustment Patterns
Work in Progress
This paper presents a model of optimizing firm behavior that endogenously generates Taylor-style (1980) annual nominal wage raise schedules that are synchronized across workers within the firm. Using employer-employee linked data, I show that approximately 80% of U.S. workers are employed by firms that follow a pattern of synchronized annual nominal wage raises. The model combines downward nominal wage rigidity with costly evaluation of employees’ time-varying productivity to endogenously generate three related phenomenon. One, workers' nominal wages follow a regular annual raise schedule. Two, these raise schedules are coordinated among workers within the same firm. And three, workers' nominal wages are rarely indexed to inflation.