اثرات بلند مدت الزامات در جستجوی کار: شواهدی از اصلاح کمک هزینه افراد جویای کار انگلستان

This paper investigates long-term returns from unemployment compensation, exploiting variation from the UK JSA reform of 1996, which implied a major increase in job search requirements for eligibility and in the related administrative hurdle. Search theory predicts that such changes should raise the proportion of nonclaimant nonemployed, with consequences on search effort and labor market attachment, and lower the reservation wage of the unemployed, with negative effects on post-unemployment wages. I test these ideas on longitudinal data from social security records (LLMDB). Using a difference in differences approach, I find that individuals who start an unemployment spell soon after JSA introduction, as opposed to 6 months earlier, are 2.5–3% more likely to move from unemployment into Incapacity Benefits spells, and 4–5% less likely to have positive earnings in the following year. This latter employment effect only vanishes 4 years after the initial unemployment shock. Also, annual earnings for the treated individuals are lower than for the non-treated. These results suggest that while tighter search requirements were successful in moving individuals off unemployment benefits, they were not successful in moving them onto stable or better jobs, with fairly long-lasting unintended consequences on a number of labor market outcomes.

Despite a substantial literature on the impact of unemployment insurance (UI) on the duration of unemployment and re-employment rates,1 less is known on its long-term effects on work careers. But the channels through which UI affects the process of return to work, mainly job search effort and reservation wages, are clearly also likely to have an impact on the quality of post-unemployment jobs and in general on future work careers. For example it may be argued that more generous UI gives workers the opportunity of not simply accepting the first job offer that comes along, but of waiting for a good job, that provides the best match for their skills. Since early work by Diamond (1981), a number of papers in the theoretical literature have pointed out that UI may have beneficial effects, mainly by encouraging workers to wait for high-productivity jobs in an environment with search frictions and heterogeneous jobs (Acemoglu, 2001, Acemoglu and Shimer, 1999, Acemoglu and Shimer, 2000 and Marimon and Zilibotti, 1999). However, recent empirical work has generally found little evidence of beneficial effects of UI on post-unemployment earnings or job stability (see Card et al., 2007a, and references therein). This paper provides new evidence on the long-term returns to UI, exploiting variation from the UK Jobseekers' Allowance (JSA) reform of 1996. The JSA was introduced in October 1996 to replace the previous Unemployment Benefit/Income Support system, and represented a major reform to the existing UK system of welfare benefits for the unemployed. One of the most important changes with respect to the previous system was a substantial tightening of search requirements for eligibility and in the related administrative hurdle. There is now broad consensus on the strong positive effects of the JSA on the claimant outflow rate. In particular, the months following JSA introduction coincided with a record fall in the number of unemployment benefit claimants. In this paper I explore the link between tighter search requirements and a number of post-unemployment outcomes, including future employment rates, weeks worked, earnings and new benefit spells. The impact of higher search requirements on average search intensity is theoretically ambiguous, as some will search more intensively to meet the requirements, while others may consider the requirements too burdensome and give up search (see Manning, 2009), with an ambiguous impact on the exit rate into new jobs. But the introduction of stricter eligibility criteria unambiguously reduces utility during claimant job search, with negative effects on reservation wages and post-unemployment wages, and raises the share of nonclaimants in the nonemployment stock, thus possibly raising the take-up rate of other kinds of benefits and lowering future employment rates. Experimental evidence for the US indeed tends to show that tighter job search requirements have negative, albeit mild, effects on the time spent of benefits, while the effects on re-employment rates are less clear-cut (see Meyer, 1995), thus warning that not all those who are moved off benefits necessarily find new jobs. Related to this point, Card et al. (2007b) find that the spike in the re-employment hazard at benefit exhaustion is much smaller than the spike in the unemployment exit hazard, indicating that many workers leave the unemployment register without returning to work. I use a difference in differences approach to estimate the effects of unemployment compensation on subsequent careers, by comparing long-term outcomes for cohorts of unemployment entrants before and after JSA introduction in October 1996. As these two cohorts may differ in seasonal factors, I construct similar reference cohorts for 1997 and 1998, and then look at difference in differences across cohorts and years. There is an aspect of the JSA rules that makes this procedure non-standard, namely that when the JSA was introduced, the new eligibility requirements applied not only to the new claimant inflow, but to the existing stock of unemployed claimants as well, so there is no major discontinuity to expect between labor market outcomes of workers who became unemployed just before and just after JSA introduction. But the distance between the start date of an unemployment claimant spell and the date of JSA introduction is indicative of the spell's probability of being treated, and this will be the basis of identification. The key caveat to a causal interpretation of the resulting estimates is that unemployment entry cohorts may differ in unobservable characteristics that affect their post-unemployment outcomes, independently of JSA rules. A related concern is that seasonal factors may vary over time, and thus that differences in labor market outcomes across unemployment entry cohorts and years may not simply pick up the effect of JSA, but also the effect of potential interactions between, say, seasonality and business cycles. I investigate these possibilities with a number of robustness tests. First, for all entry cohorts in 1996–1998, I control for the vacancy to unemployment ratio in the month of unemployment entry, in order to capture the effect of changing aggregate labor market conditions. Second, I run a falsification test on a “placebo” JSA reform in 1997, using 1998 and 1999 entry cohorts to control for seasonal factors. Third, I estimate the effect of JSA on post-unemployment outcomes using an IV strategy, which consists in instrumenting the probability of being covered by the new JSA rules by the distance between the date of unemployment entry and October 1996, when the JSA rules kicked in. Again I control for seasonal factors exploiting information from later cohorts of unemployment entrants. None of these tests seem to invalidate the results of the main difference in differences estimates. My empirical analysis leads to three main findings. First, the JSA has had a strong, positive and significant impact on the outflow from claimant unemployment for the individuals affected, but a negative impact on weeks worked 1 year later. While the reform successfully managed to move claimants off benefits, it was not successful in getting them onto new, lasting jobs. Thus it seems that job search requirements mostly worked through raising the non-monetary search effort costs of remaining on UI, rather than enhancing job finding rates. Second, I find that JSA has had a negative and significant impact on post-unemployment annual earnings. My estimates indicate an initial impact of about 600£ on annual earnings, which persists for as much as 4–5 years after an unemployment shock. Log weekly earnings (conditional on working) are also somewhat reduced by the JSA, although the associated effects are too imprecise to exclude a wide variety of scenarios. Third, while JSA has moved individuals off unemployment-related benefits, it has increased the incidence of other benefits, most notably health-related benefits. Starting a spell soon after JSA introduction, as opposed to 6 months earlier, implies an increase of 2.5–3% in the probability of claiming Incapacity Benefits (the equivalent Disability Insurance in the US) 6 months after unemployment exit. My work complements existing evidence on the impact of UI on labor market careers by investigating long-run effects of tighter search requirements on post-unemployment outcomes. I use social security data containing complete labor market histories, which allow me to combine information on benefit spells and earnings, and to achieve a more long-term perspective on the impact of UI than previously addressed in the literature. Furthermore, UI systems have several institutional features, and I estimate the effects of major changes in job search requirements, while most of the previous literature focused on the effects of either changes in UI benefit levels or in their maximum duration. As it will be illustrated below, an increase in search requirements is predicted to lower reservation wages and raise exits into nonclaimant nonemployment, even when the actual level of benefits received remains unchanged. Finally, I consider a new potential dimension of the long-term effects of UI, namely the start of other benefit spells, with potential consequences on total benefit expenditure. This relates to a recent literature on higher take-up rates of health-related benefits in a number of countries (see Autor and Duggan, 2003 and Autor and Duggan, 2006, for the US; and Faggio and Nickell, 2004, for the UK). The paper is organized as follows. The next section briefly discusses related work. Section 3 describes the JSA features that are going to be relevant in my analysis. Section 4 proposes a simple job search model to represent the likely effects of JSA. Section 5 describes the data set used. Section 6 presents my methodology and some preliminary evidence. Section 7 presents my main findings on the effect of JSA on a number of post-unemployment outcomes and robustness tests. Section 8 concludes.

This paper has investigated the post-unemployment effects of tighter job search requirements, exploiting variation provided by the introduction of the UK JSA reform in October 1996. In a simple job search framework, one expects that tighter requirements for UI eligibility lower the reservation wage and thus the quality of post-unemployment jobs, and raises the fraction of nonclaimant nonemployed, with consequences on labor market attachment and job search effort. Using administrative longitudinal data on spells on unemployment benefits and earnings, I find that JSA has had a positive and significant impact on the claimant unemployment exit rate, as well as on exits into other benefits, and a negative and significant impact on the probability of working for up to 4 years following an unemployment spell. Starting a spell soon after JSA introduction, as opposed to 6 months earlier, raises the likelihood of a spell on Incapacity Benefits by about 2.5–3%, and lowers the likelihood of positive earnings by about 4–5%, together with the level of earnings and the number of weeks worked. Weekly earnings (conditional on work) also seem to be lower for the treated, but the confidence intervals around these estimated effects are too large to exclude a wider variety of scenarios. Overall, all the estimated effects tend to be stronger for the 16–24 than the 25–64 year old sample. A possible interpretation is that tighter search requirements implied by the JSA indeed moved claimants off unemployment benefits, without really raising job finding rates. Among claimants treated by the JSA, those who found jobs quickly did not see their fortunes much changed with respect to the previous regime, as implied by the absence of significant effects on weekly earnings in the year following job loss. But those who left the unemployment register without finding a job might have in general become detached from the labor market — for example by no longer perceiving themselves as “workers,” as they were not covered by labor force welfare — and in particular were more likely to start spells on benefits that were not search related, with detrimental effects on their search effort and fairly long-lasting effects on their employment rates. This was reflected in a net loss in (unconditional) weeks worked and earnings with respect to the previous system during about 3 years after a job loss.