Christian Belzil, "Unemployment Insurance and Subsequent Job Duration: Job Matching vs. Unobserved Heterogeneity", Journal of Applied Econometrics, Vol. 16, No. 5, 2001, pp. 619-636. 1. The Data Set The analysis presented in the paper is based on confidential administrative files from the former Canadian Ministry of Employment and Immigration (now called Human Resources and Development Canada). The data are constructed as an event history data set and cover a period going from January 1972 until December 1984. The data set contains several pieces of information about employment and unemployment spells of a random sample of Canadian labor force participants. The data are actually based on a merge of several administrative files such as Records of Employment (ROE) and the Unemployment Insurance administrative files, and they enable the researcher to recreate the sequence of labor market states occupied by a given individual. As is usually the case with administrative data, information on insured spells of unemployment (such as benefit durations and the weekly benefit level) are relatively accurate. However, the data are much less reliable when it comes to evaluating the labor market status of those unemployed for a relatively long period, especially those unemployed beyond benefit termination. The Records of Employment (ROE) identify the reason for separation and provide information about job tenure, age, experience and industry. In Canada, firms are legally required to issue a ROE for every job separation that takes place. The measure of experience available is the total number of weeks of employment from 1972 until 1984. It is therefore reliable for younger workers. The Unemployment Insurance file, along with some partial income tax records file, give information about potential benefit duration for the unemployed, weekly insurable earnings, unemployment duration, UI benefit level and the number of weeks of benefit entitlement left when a new job is accepted. The employer code available is used to identify individuals who have been laid off and returned with the same employer subsequently. The key UI variables are defined as follows: (maximum) Benefit Duration: The maximum number of weeks of UI benefits that an individual can draw as recorded in the Unemployment Insurance File by the date of the formal application. It is calculated based on the number of weeks worked during the previous year, up to a certain maximum (around 45) which may depend itself on the local rate of unemployment. When an individual experiences a subsequent spell (under the same claim), initial benefit duration is reduced by the number of weeks used during the previous spell. Potential benefit duration: The difference between maximum benefit duration and elapsed unemployment duration. When elapsed duration exceeds maximum benefit duration, potential benefit duration is set to 0. Benefit Level: The weekly benefits (measured in Canadian dollars) calculated from information on weekly insurable earnings. 2. Sampling Method The data set used in this paper contains 2610 individual records of labor market histories for young males who have suffered a job separation between January 1976 and February 1978. Each individual was between 18 and 25 at the time of the job separation and was followed until 1984 through administrative records. Consequently, the number of spells attached to each record varies considerably across individuals. Although the original file contains information on multiple unemployment and employment spells, for the current study, I only had access to a single unemployment and job spell per individual. Out of these 2610 records, 1910 are coded as layoffs while 700 individuals quit to become unemployed. As a first step, I retain the 1910 individuals who have experienced a layoff. As a second step, I exclude the 1001 cases where unemployment was followed by a job with the previous employer as well as the 700 individuals who quit their job. This is because those individuals who returned to their previous employer are most likely to be on temporary layoffs and are likely to have a distinct search behavior from those who were displaced permanently. Those who quit their job are more likely to have a weaker attachment to the labor market. The resulting sample contains 909 individuals who have experienced an involuntary job separation. Most of these individuals (889 cases) have found a new job with a different employer while 20 of them have been lost by the UI authorities, which means that I observe no subsequent job duration for these individuals. Given that reported unemployment duration is likely to be unreliable for these individuals, I censor unemployment duration at 50 weeks. In terms of the subsequent jobs, these 909 individual records are classified as follows: 464 subsequent job durations which were later terminated by a layoff; 289 subsequent job durations terminated by a quit; 136 subsequent job spells still in progress as of 1984; 20 cases where the unemployment spell is the last recorded state. Christian Belzil