![]() |
![]() |
![]() ![]() ![]() ![]() ![]() |
||
![]() |
![]() ![]() ![]() ![]() ![]() |
![]() |
![]() |
The purpose of this section is to examine the mechanisms by which the apparent adjustments to the change in the entrance requirement take place. We begin by examining the hazard plots for employment spells that end in quits as opposed to layoffs. Figures 11 and 12 provide these plots for 1989 and 1990, respectively.37 The plots are derived using a competing risks framework where for the layoff "hazard", spells ending in quits are treated as right-censored. In both years, the hazard rates for spells ending in quits are substantially lower than for those ending in layoffs, at least for durations below 30 weeks. Perhaps the most striking finding is that spells ending in quits show little or no evidence of entrance requirement effects. Rather, the spells ending in layoffs display the patterns observed in Figure 7. Indeed, those patterns appear even stronger when layoffs alone are examined. These results are more consistent with the implicit contract model than with the search model, in which job terminations would be employee-initiated. The dramatic difference between the two figures may seem surprising given that under the Canadian UI system during this period, one could quit a job and still qualify for benefits. However, the potential penalty for quitting a job was an extension of up to six weeks of the initial waiting period during which no benefits are paid. This penalty, together with the absence of experience rating, may have been sufficient to induce firms to initiate separations. Indeed, implicit in the "community pressure" model is the notion that the employer is best placed to ensure that as many members as possible qualify for UI. The employer's role is particularly important if some employees would prefer to work more weeks than the minimum entrance requirement.38 In Table 5 we present the covariate estimates from the duration model associated with the likelihood function specified in equation 7 above. In this estimation the estimated coefficients on the covariates are permitted to take different values at the 10-week point in 1989 (column 2) and the 14 week point in 1990 (column 3). A comparison of these estimates with the covariate coefficient values relevant for the rest of the range of weeks (column 1) provides information on the characteristics of individuals who worked just enough to qualify for UI in each year. The column 1 estimates of the covariate values for the main range of weeks are similar to the estimates in Table 3a, though effects such as education and age are stronger. The key difference is the change in the unemployment rate effect from positive to significantly negative. These suggest that the earlier positive coefficient may have been capturing a combination of a positive effect at the spike points and negative effects everywhere else. A more positive unemployment rate effect at the entrance requirement is consistent with the implicit contract model where poor demand shocks will cause more firms to bunch their workers at H = Hmin. ![]() View Figure 11 ![]() View Figure 12 In general, the estimated coefficients for the 10-week point in 1989 fit with expectations about individuals who might work just enough to qualify for UI. Individuals with post-secondary or greater education are significantly less likely to leave an employment spell at this point than those who have completed high school. In fact, our sample contained so few university educated individuals separating from employment at this point that we had to combine that group with individuals with lower levels of post-secondary education for purposes of estimation. Perhaps not surprisingly, individuals in primary and food processing industries show a much stronger propensity to fail out at this point than at other weeks. Interestingly, construction does not seem to be an industry that generates employment spells of exactly Hmin weeks. Firm size is another strong indicator for leaving employment at Hmin weeks, with individuals working for larger firms being much less likely to fail out at this point. This result could arise if the more personal relations in small firms lead to collusion between the firm and worker in the timing of employment spells.39 Alternatively, small firms may be more likely to be based in the region and to be more susceptible to community pressure. The SAMEMP variable, which takes a value of one for individuals who have worked for the current employer in a previous year, is an attempt to capture implicit contracts between workers and firms on recurring work patterns, but the evidence is at best weak that individuals who have worked for a firm before are any more likely to leave a job at 10 weeks. On the other hand, unionized workers are much less likely to fail out at 10 weeks. This could reflect union organization of more stable jobs or more formal work arrangements making firm-worker collusion more difficult. Finally, the regional unemployment rate does not have a significant impact on the likelihood of a separation occurring at 10 weeks.
With a few exceptions, the estimates for coefficients at 14 weeks in 1990 are more poorly defined than their counterparts in columns 1 and 2 in Table 5. These estimates are also closer in both magnitude and sign to the estimates for the main range of weeks than to the estimates for the 10-week point in 1989. Given the relatively small increase in the spike in the hazard at 14 weeks between 1989 and 1990, the estimates in column 3 may not accurately reflect the characteristics of individuals who plan on working exactly Hmin weeks. As discussed earlier, individuals may not be able to quickly target their employment spells to exactly Hmin weeks when the entrance requirement is changed. In that case, the estimates in column 2 deserve more attention. Those estimates suggest that individuals working just the minimum number of weeks tend to be disproportionately poorly educated, non-unionized, and in the primary and related processing industries. We next examine individuals who worked just long enough in 1989 to qualify for UI benefits in order to see how they adjusted to the increase in the entrance requirement. In this analysis we allow for the possibility that some individuals meet the entrance requirement by working several jobs separated by brief non-employed searches. Thus, we create a variable, TOTEND, which equals the total duration of job spells not separated by periods of UI receipt for an individual.40 We select a sample of individuals who have employment spells in 1989 such that 10 < TOTEND < 13, and whose last employment spell in 1989 ends before November 18. The latter restriction is imposed to make examination of work experience in the experimental period in 1990 more straightforward. Given our goal of selecting a sample of minimum employment UI users, length biased sampling issues are not a concern. In drawing this sample we include employment spells that begin in 1988, providing they end in 1989. Once this sample is selected, we compare their labour force experience in the first 46 weeks of 1989 and 1990. Table 6 provides mean characteristics for 1989 and 1990 for the following: all workers in Canada who were not full-time students and were under age 65 in the given year; all workers living in the maximum entitlement regions who were not full-time students and who were under age 65; and the sample of individuals who worked for from 10 to 13 weeks in 1989. The characteristics recorded for the 10 to 13 weeks sample accord well with the pattern for the estimated covariates at the 10-week spike in 1989. These individuals are substantially less educated than the average worker in their regions and, especially, when compared to Canadian workers overall. They are much more likely to work for establishments of under 20 people and are about half as likely to be unionized as other Canadians. They are much more concentrated in seasonal industries with 33.2 per cent of them being in agriculture, forestry or fishing compared to 9.3 per cent for the maximum entitlement regions as a whole and 4.2 per cent for Canada. They are also disproportionately found in the food and beverage industry and the service industry and are significantly under-represented in other manufacturing, sales, and education and health. The seasonal nature of their jobs is reinforced by their reason for leaving the last job in the year (conditional on having left a job in the sample year).
Only 6.9 per cent of the 10 to 13-weeks workers quit their job as compared to 32.1 per cent for the maximum entitlement regions overall and 50.6 per cent for Canada as a whole. This accords with the evidence presented earlier on the quits vs. layoffs breakdown and with the idea that if one is going to work just enough weeks to qualify for UI it is best to do so in a job from which a layoff is expected in order to avoid the penalties associated with quitting. Fully 58 per cent of the 10 to 13-weeks sample gave as their reason for leaving their last job that the job had ended because of its seasonal nature. This compares to 22 per cent for Canada as a whole. The 10 to 13-weeks workers also take much lower paying jobs and tend to work much longer hours when they do work. The long hours are consistent with the structure of UI benefits which are calculated based on weekly earnings. Workers trying to generate high UI benefits-especially those with little education facing a low hourly market wage-can best do so by working long hours. Finally, the 10 to13-weeks workers are much more likely to have two jobs in the year than other workers. Overall, one forms a picture of individuals with low education who work for low pay and long hours in predominantly seasonal industries. How did these workers adjust to the increase in the entrance requirement? First, almost all of them continue to work in 1990; only 3 out of 170 individuals who worked between 10 and 13 weeks in 1989 did not work in 1990. 41 The great majority of them manage to increase their weeks of work beyond the new entrance requirement. Of those who do not have a job spell ongoing at the end of the year (85 per cent of the original sample), 9.7 per cent work between 1 and 9 weeks in 1990, 18.4 per cent work between 10 and 13 weeks as they had the year before, 18.4 per cent work exactly 14 weeks, 36.9 per cent work between 15 and 20 weeks, and the remaining 16.6 per cent work 21 weeks or more. These results accord with the evidence from the hazard rate analysis that few individuals who had worked from 10 to 13 weeks in 1989 continued to do so and that many moved not just to jobs that ended exactly at the new Hmin but in the 15 to 20-week range as well. The 1990 job characteristics for the sample of individuals who worked for 10 to 13 weeks in 1989 indicate offsetting patterns of response. On the one hand, these workers show dramatic movements out of the primary and service sectors. The proportion in fishing alone falls from 21 per cent in 1989 to 12.6 per cent in 1990. These movements are offset by increases in a variety of industries including Food and Beverage and Other Manufacturing. Further, there is a slight increase in their likelihood of being unionized and a fall from 58.1 per cent in 1989 to 51.8 per cent in 1990 in the percentage who say their last job terminated because of its seasonal nature. Thus, there appears to be a movement towards more stable job patterns. On the other hand, the probability that these individuals work part time rises. The percentage who leave jobs through quitting also increases with over 5 per cent of the sample claiming they quit because of low pay in 1990, an answer that is not given by anyone in 1989. The percentage holding two jobs in the year decreases but this is offset both by increases in the percentage holding exactly one job and the percentage holding three or more. Similarly, a lower proportion of individuals work for establishments of under 100 workers but the proportion working for the largest category of establishments also drops. Thus, there is evidence that some workers ended up in less stable work situations. Some of the hours of work changes could be associated with the onset of the recession; however, the decline in hours of work in these regions is much larger than that experienced in Canada as a whole. In an attempt to understand the role of firm-worker interaction in adjusting to the new requirement, we also selected a subsample of the 10-13 weeks workers who worked for exactly one employer in 1989 and who returned to that employer in 1990. Because there are only 68 such individuals, we do not report complete tabulations; however, the breakdown of weeks worked in 1990 is instructive. Of the portion of this subsample who do not have a right-censored spell at the end of 1990 (48 individuals), 4 per cent worked less than 10 weeks in 1990 for the employer they worked for in 1989, 27 per cent worked for between 10 and 13 weeks in 1990, 19 per cent worked 14 weeks, 50 per cent worked for 15 to 20 weeks, and the rest worked for 21 or more weeks in 1990. Thus, the results are mixed. Returning to the same employer made it more likely that an arrangement could be made yielding 14 to 20 weeks or work (though no more likely than for the sample as a whole that exactly 14 weeks would be worked) but also more likely that one would only be offered employment for the same 10 to 13-week spell as in 1989.
Overall, an examination of the subsample of individuals who worked just enough to qualify for benefits in 1989 reveals that a significant number managed to increase their weeks of work to the 14 to 20-week range in 1990. They appear to have done so largely by shifting toward more stable industries. Some of the adjustment also took the form of a decline in hours of work per week. Finally, it is noteworthy that virtually the same proportion of the sample received UI payments in 1990 as in 1989. Our final piece of evidence is presented in Table 7. It compares the aggregate labour market behaviour in the maximum entitlement regions to that for Canada as a whole. Between 1989 and 1990, the employment/population ratio increased in the maximum entitlement regions but declined in Canada as a whole. The increase in total employment was also somewhat greater in the maximum entitlement regions. The behaviour of unemployment is dramatically different. While unemployment increased by about nine percent in Canada as a whole, unemployment declined by almost one percent in the maximum entitlement regions. The 0.2 per cent decline in the unemployment rate in these regions (versus an increase of 0.6 per cent for Canada as a whole) is consistent with the estimates of the impact of the change in the entrance requirement made earlier in the paper. Generally speaking, these aggregate statistics, while admittedly crude, do support the conclusions of the micro-analysis that the change in the entrance requirement resulted in increased employment, reduced unemployment and little (if any) change in labour force participation.
|