Directorates
·	Director General's Office
·	Data Development
·	Strategic Evaluations
·	Partnership and Continuous Evaluations
Publications
·	Evaluation Reports
·	Evaluation Briefs
·	Labour Market Development Agreement Series
·	Lessons Learned Series
·	Evaluation Tool Kit Series
·	Evaluation Forum Newsletter
·	From Theory To Practice

2.1 Data

The data we use to investigate the impact of the switch to an hours-based system on employment outcomes and entitlement is the Canadian Out of Employment Panel (COEP). The COEP consists of administrative data linked to a survey. In particular, a sample of Records of Employment (ROEs) marking the end of jobs is drawn from all jobs ending in a given quarter. Workers associated with the sampled ROEs are then contacted in order to be surveyed. The survey provides extensive data on individual characteristics such as education and age, on employer characteristics such as firm size, and on household characteristics such as expenditures. This makes COEP a unique dataset that is very well suited to studying the issue at hand. Indeed, this study would be far less complete without the COEP. Before the advent of the COEP, researchers had to make do either with administrative data alone, which lack crucial information on individual characteristics such as education and family resources, or with survey data alone, the use of which necessitates considerable guessing about the exact job spells used to establish claims.⁴

At our time of study, data were available up to and including the sample of ROEs drawn in the third quarter of 1997. In order to maintain comparability, we used data corresponding to data ending in the first three quarters of 1996 and the first three quarters of 1997. Because self-employment spells and some jobs ending in quits are not eligible for UI, we restricted our attention to paid employment spells ending in layoff. In addition, following Green and Sargent (1998) we split our sample into seasonal and non-seasonal jobs. The seasonality split is created based on a question asking workers the nature of their work arrangement. We defined as seasonal those jobs listed by the worker in the survey as being of a seasonal or contract nature. All other layoffs we defined as non-seasonal.

Our examination focuses on employment duration rather than job duration since several jobs may be strung together to create Unemployment Insurance (UI) eligibility. We define an employment spell as consisting of weeks of consecutive employment. In particular, we begin with the job corresponding to the originally drawn ROE. We then consider the immediately preceding job, adding it to the ROE sample job if the end date of the preceding job is within two weeks of the start of the ROE sample job. If the preceding job is added to the ROE job for the purpose of generating an employment spell, then we move further back and consider the next previous job using the same criterion.⁵ While previous jobs may not be close enough in time to count as part of the same employment spell by our criteria, they may still be used by the worker to generate UI entitlement. They are eligible to be used in this way if at least some of their associated weeks fall in the qualifying period. Recall that the qualifying period equals the lesser of 52 weeks or the time since the last UI claim ended. To establish the qualifying period we link the ROE to UI claim records included as part of the COEP. We note the length of jobs that occur in the 52 weeks preceding the start of the employment spell or the period since the end of the last UI claim, whichever is less. We use this to calculate the number of weeks of employment suitable for establishing entitlement that the individual carries into the employment spell being studied. This is particularly important for examining changes in entitlement discussed in Section 3. Our sample sizes are: 5,348 seasonal employment spells and 4,143 non-seasonal employment spells ending in the first three quarters of 1996, and 4,316 seasonal employment spells and 3,287 non-seasonal employment spells ending in the first three quarters of 1997.

Footnotes

4	We are greatly indebted both to the individuals who spent innumerable hours of work creating the COEP and to those who currently administer it (in particular, Lars Vilhuber). They have made a truly complex dataset as easy to work with as one could possibly hope.
5	If there is a gap between the preceding job and the ROE sample job of, for example, two weeks, we do not count that two week gap as part of the total length of the employment spell.