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

Monthly time series of counts of separations by type from HRDC's administrative ROE data are employed in this study.⁹ The main advantages of the ROE data for this exercise are that it is collected as a 10% random sample of the complete inflow of job separations and that it categorises individuals according to the same reasons for separation that are used in administering the UI system. Over the period, the categories on the ROE form have been: shortage of work, strike or lockout, return to school, illness or injury, quit, pregnancy or adoption, retirement, work sharing, apprentice training, age 65, dismissal, leave of absence, and other. An ROE should be issued every time a worker separates from an employer, and it is a prerequisite for receiving UI benefits. Since dismissed workers were also penalised by Bills C-21 and C-113, it might also be of interest to consider this category of separations. We can do so here only in a very limited way, since dismissals were not listed as a distinct reason for separation until 1990.¹⁰ Previously they had been included in the "other" category. To account for some of the heterogeneity in the labour market, the counts are disaggregated by sex and into 15- to 24- and 25- to 54-year age groupings. Unfortunately, no distinctions can be made on the basis of temporary versus permanent layoffs, industry, or region in the data extract available to us. The ROE is also matched with the UI administrative Status Vector (SV) file to determine the fraction of separations that result in a UI claim. None of the series were deseasonalised prior to our analysis.

Basic descriptive statistics for the ROE data are presented in Tables 1 and 2 by UI policy period: (1) in the baseline period before Bill C-21, from January 1980 to October 1990; (2) during the partial disentitlement while C-21 was in effect, from November 1990 to March 1993; and (3) following the total disentitlement, under Bill C-113, starting in April 1993 and extending 9 and 15 months for the SV rate and ROE data, respectively.¹¹ Descriptive statistics are of the total separation rate (S/E), which includes all 13 reason-for-separation categories, the voluntary quit rate (VQ/E), a composite group of the other and dismissed categories (O-D), and the rate of layoffs (SW/E). Employment (E) is the age-specific count measured by Statistics Canada's Labour Force Survey (LFS).

From Table 1 it is clear that the separation rate of young male workers is almost twice that of prime-age men, with, on average, total separations equal to just over 7% of employment occurring each month. For females the gap is not quite as large, and the level is slightly lower, with the 15- to 24- and 25- to 54-year age groups' separation rates being 5.5 and 3.5%, respectively. It is difficult to draw inferences about how the levels change across the periods from the gross numbers in this table because the beginning of period 2 coincides quite closely with the onset of the 1991_92 recession, and the different periods cover different months of the year. These levels are therefore contaminated by cyclical and seasonal fluctuations which are quite large. They do, however, give us some sense of the magnitude of the flows in question. About 1.5 to 2% of employed 15- to 24-year-old workers quit their job each month, compared with 0.5 to 0.8% of the 25- to 54-year age group for both sexes. A similar, but not as pronounced, difference across the age groups is observed for the "other and dismissed" group; about 1.4 to 1.5% of the younger group is laid off each month, but the same number for the older groups is only about 0.7 to 0.8%. In contrast to the "quit" and the "other and dismissed" reasons for separation discussed thus far, where there have been much greater differences across age groups than between the sexes, the greatest differences in the layoff ("short work" or SW) rate is by sex. Men have much higher layoff rates than women. Those for women are in the 1.3 to 1.4% range, whereas men experience a rate of between 2 and 3%. This likely reflects differences in the industries in which men and women are employed. Standard errors of the series of 15- to 24-year-old are also uniformly larger than those for the 25- to 54-year age group for both sexes. This appears to be attributable to both the smaller underlying sample size and the greater intrinsic volatility of the labour force behaviour of the younger workers.

Matched to the ROEs are monthly counts of the UI claims by reason for separation, where the month associated with the claim is the month of the separation. Our definition of the UI claim rate is simply the fraction of ROEs that are used to establish claims.¹² Table 2 presents a summary of the VQ claim rates by period. The claim rate is seen to decline substantially across the periods, but as with the separation rates seasonal and cyclical factors confound inference. In contrast to the separation rates the largest difference in claim rates is by age, not sex, with the younger quitters having a lower claim rate.

Footnotes

9	The authors would like to thank Ging Wong, Carol Guest, and Anne Routhier of HRDC for producing the administrative data and generously making it available to us.
10	Looking at the data, starting in 1990, it appears that it took close to a year from the time dismissals were first reported until they reached a stable level.
11	While the data file actually extended into 1995 when our sample was taken, the files take a long time to "fill up" because of lags in collecting and inputting the ROE data. The matched UI claim data take even longer to become reliable. Since it would be easy for lags in the data collection process to be spuriously interpreted as drops in the count of separations or claims, we end the series at very conservative dates.
12	For a discussion of different definitions of the claim rate see Storer and Van Audenrode (1995). A related issue is that several ROEs may be combined to initiate a single claim: that is, a claim may be based on multiple distinct jobs. If a claim is based on, say, two ROEs, then the numerator and denominator of our claim rate are both incremented by two; our measure is the fraction of job separations (ROEs) that are used to establish any claim.