In order to accurately measure the cost of job loss for workers, two issues must be addressed: the exogeneity of the observed separation to worker characteristics, and the representativeness of the sample used for the measurement. This project would advance knowledge along both dimensions. The difficulty of isolating exogenous individual separation events has led the literature to use separations that encompass multiple worker layoffs at the same time (mass layoffs) as the exogenous event upon which to base estimates of the earnings loss associated with an involuntary change of employers. Thus, the job losses associated with mass layoffs are a convenient natural experiment for investigating outcomes whose measurement would otherwise be biased by endogeneity of the separation. In previous research (McKinney and Vilhuber, 2006; Lengermann and Vilhuber, 2002; Abowd, McKinney and Vilhuber, 2008; Bowlus and Vilhuber, 2002), the investigators have found some evidence that the mass layoff event is related to characteristics of the workers employed at the firm. This project would continue that research, investigating whether the mass layoff event can be considered statistically exogenous for the worker and, if not in general, under what circumstances the exogeneity assumption can be maintained. Second, the previous literature has used survey-based person and/or household panels, or administrative data for individual US states to investigate these issues. While providing detailed demographic information, surveys are often limited by the number of workers actually observed as part of a mass layoff, restricting the analysis of specific sub-groups and limiting the extent to which geographic or firm-level information can be incorporated into the analysis. In the past, users of administrative data have been restricted to using a single geographic entity, not being able to follow workers across political boundaries, and have not had access to much information on worker or firm characteristics.
This project would address some of those shortcomings by using the Census Bureau Longitudinal Employer-Household Dynamics Program Infrastructure file system, covering more than 98% of private employment in 46 states as of January 2008 (with 31 states available for research purposes), which have been linked to select detailed demographic information on workers, and detailed firm-level information, using survey and census data available at the U.S. Census Bureau. This project explores and exploits detailed geographic variation, investigates and uses geographic mobility across state borders, and incorporates worker and firm level characteristics into the unique longitudinal aspect and quasi-universal coverage of administrative datasets. The project also provides newer and updated information on states that have been analyzed previously. It leverages the magnitude of the dataset to explore alternate measures of mass layoffs, highlighting the possible sensitivity of the results to the choice of particular measures of mass layoffs. Besides providing updated results, the wealth of information allows the investigator us to determine how one measures an event (or different events) ties directly into the question of the exogeneity of the layoff, as does the cause of the mass layoff event itself. The ability to define a number of new measures, to expand outcome measures beyond the original location of the worker, and to subset the analysis by detailed demographic and firm characteristics, including geography, helps to check the robustness of any inference made about earnings losses of workers.
Broader Impacts. The economic impact of job losses on workers, in particular those resulting from large mass layoffs, can be substantial. Workers may experience earnings 10 to 20 percent below normal earnings levels even five years or longer after losing a job. Accurately measuring the cost of job loss is policy relevant and has motivated regulations and laws to mitigate or alleviate these losses. Assessing the cost of such laws depends critically on both the estimated loss by each affected worker and the estimated incidence of exogenous job loss.
