اثر مزایای بیکاری در نرخ اشتغال مجدد: شواهدی از اصلاح بیمه بیکاری فنلاندی

In January 2003, the unemployment benefits in Finland were increased for workers with long employment histories. The average benefit increase was 15% for the first 150 days of the unemployment spell. At the same time severance pay system was abolished. In this paper we evaluate the effect of the change in the benefit structure on the duration of unemployment by comparing the changes in the re-employment hazard profiles among the unemployed who were affected by the reform to the changes in a comparison group whose benefit structure remained unchanged. We find that the change in the benefit structure reduced the re-employment hazards by on average 17%. The effect is largest at the beginning of the unemployment spell and disappears after the eligibility period for the increased benefits expires.

The Finnish unemployment insurance (UI) system was reformed in January 2003. As an attempt to simplify the benefit rules, a severance pay system that had existed until the end of 2002 was abolished and daily UI allowances correspondingly increased for 150 days. The average increase in the daily UI allowance was 15%, but only the unemployed with long employment histories were eligible for this higher allowance. According to the search theory such a change from a lump-sum severance pay to a higher daily allowance should weaken the incentives of job search and increase the duration until re-employment. The effect on re-employment hazards should be largest at the beginning of the unemployment spell and decrease as the benefit expiry date approaches. In this paper we use the Finnish UI benefit reform to evaluate these predictions. The Finnish benefit reform provides a relatively clean policy experiment that can be used to evaluate the effects of benefit structure on the re-employment rates. Since only part of the unemployed were affected by the reform, we can compare the changes in re-employment rates in a group whose benefit structure changed to a comparison group whose benefits remained unchanged. The reform also took place at a time when the macroeconomic environment was stable and at a time when no other major policy reforms were implemented. These features minimise the risk that our results would be contaminated by macroeconomic cycles or other policy changes. A fundamental identification problem in analysing the effects of UI benefits on re-employment rates is that UI benefits depend on previous earnings. However, pre-unemployment earnings are likely to be correlated with several other factors that affect re-employment rates. Lack of independent variation in UI benefits in a typical cross-section data makes it very difficult to disentangle the effect of UI benefits from the effects of these other factors.1 A common strategy to overcome this problem and to identify causal effects of UI benefits on re-employment rates is to exploit policy reforms that have led to different changes in benefits in different groups of unemployed workers. This approach has been used previously in evaluating policy reforms in Germany (Hunt, 1995), Sweden (Carling et al., 2001 and Bennmarker et al., 2007), Austria (Lalive et al., 2006) and the New York State (Meyer and Mok, 2007).2 Our basic approach is similar to the approach used in these previous papers but our set up also differs in several ways. First, in most previous papers the identification is based on comparing the changes in re-employment rates across groups that face different changes in benefits because they differ in the pre-unemployment wage. We identify the effect of UI benefits based on the differences in the benefit changes across groups that differ mainly in the length of previous work experience. This may be important if demand for workers with different skills changes differentially at the time of the reform. Second, we examine a more experienced group that becomes unemployed after having been displaced from a permanent job. Third, in our case the benefit changes involve only new entrants to unemployment, which makes it easier to account for possible anticipatory effects. Finally, our setting involves not only an increase in benefits but also a removal of severance pay and replacing it with higher UI allowance paid conditional on remaining unemployed. Since the expected value of the benefit increase was roughly equal to the severance pay, our estimates should be interpreted as the effects of (dis)incentives created by UI benefits rather than a combined effect on liquidity constraints and incentives. We have access to extremely detailed administrative data. Data contain the dates of entry into and exit out of unemployment. Our data also include detailed information on the benefits, reported by the UI funds themselves. We have information on the daily amounts of benefits, the dates when the benefits are paid out and, importantly, administrative information on the remaining benefit eligibility at the end of each quarter. The data also contain information on the key variables that determine the eligibility for the higher daily UI allowance and the eligibility for the severance pay. Unfortunately no information on the amount of severance payments is included in the data. In addition to estimating the average effect of the change in benefit structure on the re-employment hazard, we also estimate the effect of benefit structure on the entire re-employment hazard profile. In practise this is done by allowing the effect of the reform to vary across elapsed duration of unemployment as predicted by the search theory. A similar approach has been used in Lalive et al., 2006 and Bennmarker et al., 2007. In contrast to these papers, we find that the increase in the unemployment benefits had a large and statistically significant negative effect on the re-employment rates during the first months after entry into unemployment. As predicted by the search theory, this effect disappears after eligibility for the higher benefits expires. If we restrict the effect to be constant over elapsed duration, we find a relatively large, 17%, decline in the re-employment hazard. This corresponds to an 12% increase in duration until employment. The remaining part of this paper is organised as follows. In Section 2, we describe the details of the Finnish unemployment benefit system and the 2003 benefit reform. Section 3 describes the data and Section 4 the empirical methods. The main results are presented in Section 5 and further results regarding the effects of abolishing the severance pay system in Section 6. Section 7 concludes.

Concerns about the effect of job destruction on the most vulnerable groups increase the demand for social insurance provided by the unemployment benefits. While the benefits may cushion the effect of job loss in groups that have the greatest difficulty in finding new employment, they also have a side effect of decreasing the incentives to search for new jobs. Therefore, it is important to set up a benefit structure that minimises the disincentive effect of benefits. In this paper we have evaluated the effects of a reform that changed the benefit structure for a group of older workers. A lump-sum severance pay was replaced with a 15% higher daily allowance for the first 150 days of unemployment. Though the two benefit schemes had roughly similar expected value for the unemployed, the reform had a substantial effect on the re-employment rates. Based on our estimates we calculated that the expected time until re-employment increased by 33 days or about 12%. It should be noted that many unemployed individuals exit from the data for other reasons before finding work and this number cannot be directly interpreted as an effect on the duration of unemployment. Yet our results indicate that an increase in daily allowance decreases incentives substantially while a lump-sum transfer seems to have relative little effect. We also estimate the effect on the entire hazard profile and find that an increase in daily UI allowance decreases the re-employment hazard but the hazard rate returns to the pre-reform level once the period of increased benefits expires. We find no evidence that the unemployed anticipate the change in the benefit level by increasing their search effort before the benefits are decreased. In contrast, it seems that a decline in benefits increases re-employment rates only about one or two months after the benefits have been reduced. Taken at face value, this would imply that the unemployed are myopic and start searching more actively only after benefits have been reduced.