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

Unemployment insurance (UI) systems, as is well known, are complex social programs with a wide variety of potential effects on economic behaviour. Aspects of UI that have been examined include the effects of: benefit replacement rates on unemployment durations (Meyer 1990), UI qualifying periods on employment durations (Green and Riddell 1993), experience rating on the amplitude of seasonal employment fluctuations (Anderson 1993), overall benefit generosity on employers' choices between layoffs and hours reductions (Burdett and Wright 1989), and differential generosity on interregional labour mobility (Day 1992). Interestingly, however, one aspect of UI that has received almost no attention to date is that in many systems, including Canada's, entitlement to benefits depends critically on the reason why one left one's last job. Most systems, for example, impose substantial penalties on workers who quit or were dismissed from their last job, relative to workers separating for other reasons.¹

In this paper we study the effect of non-neutralities in UI related to separation reasons on the separation behaviour of Canadian workers, using two recent legislative changes (Bills C-21 and C-113) that dramatically increased the UI penalties for quitting one's job or being dismissed for cause. These changes to the UI legislation first partially and then completely disentitled those who quit without just cause and those dismissed for cause from receiving UI benefits. We argue that the responses of Canadian workers to these changes are of interest for two distinct reasons. First, in the period preceding the legislative changes being studied, voluntary quits accounted for between 19 % (prime age males) and 37 % (young females) of all separations. To the extent that quit penalties are actually effective in reducing overall UI claim rates, they have the potential to generate substantial cost savings for the UI system as a whole. The size of the savings are, of course, of direct interest to policymakers.

A second reason for examining the natural experiments provided by the disentitlement of Canadian job quitters is that they may provide useful information on the empirical relevance of some simple theoretical models of job separations. For example, in the extreme case of a pure "efficient separations" model (e.g. Becker et al. 1977; McLaughlin 1991) labels attached to separations have no market consequences apart from their implications for UI eligibility. As a result, any distinction between the UI eligibility of quitters and other job losers should lead firms and workers to "relabel" all UI-eligible quits as layoffs, with no associated change in either the overall separation rate or the UI claim rate. Under such a "pure relabelling" scenario, UI quit penalties would be totally ineffective in generating cost savings for the UI system. This is not the case for some other models of separations, including ones in which wage rigidities (Hashimoto and Yu 1980) or asymmetric information (Gibbons and Katz 1991) play a role. In these models, firms may be reluctant to acquiesce in the relabelling of separations, or workers may not wish to be relabelled because of the signalling value of labels. In such models, the introduction of UI quit penalties should reduce the total number of separations and impose real costs on workers, leading to the potential for substantial cost savings from the introduction of UI quit penalties. An examination of Canadians' responses to these legislative changes is, therefore, also of broader relevance to our conceptualisation of how labour markets function.

The empirical analysis in this paper is based on Human Resources Development Canada's (HRDC) administrative files, together with background variables from Statistics Canada. The administrative data provide monthly counts of job separations by reason, but because of a reporting change for dismissals in the midst of the period in question, our focus is primarily on voluntary quitters. In our analysis we first establish that both the legislative changes examined here did indeed make it harder for workers who were reported to have quit a job to claim UI: the fraction of voluntary quitters and those dismissed for cause who actually receive UI benefits does fall substantially following the legislative changes, particularly the complete disentitlement. Despite this, however, we find no evidence of a change in the reported frequency of quits in response to the first, less severe legislated increase in quit penalties. While this may reflect our inability to separate signal from noise, it also indicates that any changes in quit rates resulting from the legislation were not very large, since the standard errors of the estimates are sufficiently small that any economically large change would be observed. With respect to the second, more drastic policy change, we do find a significant decrease in the measured quit rate for all demographic groups except prime-age males. Finally, and perhaps most importantly, in no case do we find evidence that penalty-induced decreases in the quit rate are accompanied by concomitant increases in the separation rate for non-penalised reasons, such as layoffs. Indeed, most of the appropriate point estimates are negative, indicating that the observed reductions in quits correspond to reductions in total separations, rather than a relabelling thereof.

Although the analysis in this paper could be improved by access to more detailed microdata on separations over time, and while our aggregate analysis might as a result fail to identify certain population subgroups (e.g. frequent users) who might have been able to respond to the quit penalties by relabelling their separations, we conclude from our analysis here that simple relabelling of separations in response to the UI disentitlement of quitters was not an important phenomenon for the Canadian labour market as a whole. Instead, it appears that the legislation either had no effect on workers' separation behaviour (the case of prime-age men) or worked not by relabelling separations but by "inhibiting" workers from quitting. As a consequence, we are led to question the empirical relevance of simple "efficient separations" models of the dissolution of firm-worker matches: labels do seem to matter for reasons apart from pure UI eligibility. Another consequence is that the savings to the UI system resulting from UI quit penalties, rather than being negligible, might in fact have been larger than what is implied by the drop in the claim rate of those who quit. This is because the decreased claim rate of quitters does not incorporate the unclaimed UI benefits by those who were inhibited from leaving their jobs, which are also a cost savings to the system.

Section 2 sketches the relevant institutional and policy background. Section 3 contrasts some alternative theories in the literature and what they imply for the expected effects of these policy changes. Section 4 discusses important identification issues, while the data and some descriptive statistics are presented in Section 5. The statistical methodology is outlined in Section 6, and the results are presented and discussed in Section 7. Section 8 concludes.

Footnotes

1

The one exception we are aware of is Ragan (1984), who examines the effects of a general tightening up of the eligibility requirements for those who quit or were dismissed for cause in the United States in the 1970s. Ragan used aggregate manufacturing data for a subset of states in the four even-numbered years from 1972 to 1980, excluding 1976. Like our findings for prime-age males, but unlike our findings for other groups, he did not find a significant effect of UI quit penalties on the quit rate.