Get Complete Project Material File(s) Now! »
The determinants of the takeup
Because of data limitations, the study focuses on a particular unemployed popu-lation with very low daily benefit (lower than or equal to 20 euro). Yet, if particular, this population is non negligible, as it accounts for 12% of the flows to compensated unemployment as part of the main UI benefit.10. This population is also particularly in diﬃculty on the labour market, as their benefit and therefore their previous earnings are very low, which is often associated with low qualifications and low reemployment probability. Both their weight among all people receiving UI benefits as well as their situation on the labour market justify that we pay a particular attention to their outcomes on the labour market. Then, if the results obtained cannot necessarily be extrapolated to the whole unemployed population, they are of particular policy relevance if we think the State should provide a specific support to the population further away from the labour market.
An exploration of the observable characteristics of eligible workers choosing to exercise the option right brings useful insights on the profile of the takers. If not allowing to conclude on their preferences, it can proxy for their prospects on the labour market, the way they anticipate them and their impatience and risk aversion.
The profile of the taker is quite consistent across observable variables (Table 1.2). When compared to eligible non-takers, he is, on average, younger and more frequently male, which are characteristics generally associated with riskier behaviours or lower loss aversion (Falk et al., 2015; Albert and Duﬀy, 2012; Gächter, Johnson and Herrmann, 2007; Holt and Laury, 2002; Jianakoplos and Bernasek, 2006). He is also more skilled and more educated, which can explain that he may anticipate a quick return to the labour market, and may be less reluctant to give up on additional days of entitlement. He also works more hours, indicating a stronger attachment to the labour market and more stable jobs, which is also in line with the higher proportion of men. The lower tenure can be explained by the fact that takers are younger and then have less experience on the labour market. Concerning the right’s characteristics, the analysis is less straightforward: the initial potential benefit duration (PBD) associated to the former right is lower for takers, which can explain why they are less reluctant to give up on their former right. However, the remaining benefit duration at the moment they have to decide whether exercising the option right or not is higher for takers, which may seem surprising. Both indicators taken together mean that they have less consumed of their former right, and then that they have spend less time unemployed as part of their former right, which may explain why they anticipate a quick return to the labour market and choose to exercise their option right despite a high remaining PBD. This is in line with the fact that they have experienced less unemployment spells over the whole period (October, 2014 – May, 2017). We cannot exclude that the caseworkers also play a role in shaping such a profile: according to their beliefs about which type of people would benefit the most from the option right, they may themselves select the type of unemployed persons they will advice to exercise it, and for whom they will devote more time to explain its terms and conditions. Then, the observed diﬀerences between eligible takers and non-takers would not necessarily be due to diﬀerences in preferences or risk-aversion, but to caseworkers’ beliefs.
Table 1.2 also shows the diﬀerence between the population of takers based on the 20e condition with the whole population of interest. It indicates that, if takers exhibit large diﬀerences with eligible non-takers, they also diﬀer to a large extent from the whole population of non-eligible non-takers.11 Overall, the gap in some variables observed between takers and eligible non-takers are observed in the opposite direction between eligible non-takers and non-eligible non-takers: indeed, takers are much more frequently male as compared to eligible non-takers, but more frequently female as compared to non-eligible non-takers, as eligible non-takers are at 69% women. The pattern is less clear in terms of education and qualifications. Takers are much more numerous to have a vocational diploma, go less frequently to higher education, and are often classified as skilled employee or blue collar workers, then located in the middle of the qualification distribution. They also have less tenure, which is again in line with their young age, work less frequently full-time, and have a lower average potential benefit duration.
The empirical strategy to assess the impact of the option right on labour market outcomes consists in taking advantage of the existence of a threshold defining the eligibility condition, at 20e in the daily benefit distribution, as part of a regression discontinuity design. The idea is that people located very close to the threshold are likely to be similar, on average, in all respects but their eligibility status. Therefore, any systematic diﬀerence in their outcomes can be imputed to the fact that some are eligible to, and then may exercise the option right. This “quasi experimental design” is closely related to a local randomisation in the neighbourhood of the threshold as on which side any person will be located can be considered random, as long as some assumptions are verified. Empirical methodology – The estimated equation is the following: Y = α + τ1DBp≤c + δf f(DBp − c) + δgg((DBp − c)1DBp≤20) (1.1).
with Y being the outcome, such as unemployment duration in this case, 1DBp≤c an indicator equal to 1 when the previous daily benefit is lower or equal to c, the cutoﬀ value and f(.) and g(.) are flexible functions that we allow to diﬀer on each side of the cutoﬀ.
In this setting, the RD design is qualified as “fuzzy” in the sense that the probability to exercise the option right does not jump from 1 to 0 when crossing the 20e threshold, for two reasons: (i) all eligible persons below 20e will not exercise it; (ii) some people above 20e are eligible under the 30% ratio condition and will choose to exercise the option right.
It follows that: P r(OR = 1SDBp = 20 − ) < 1 and P r(OR = 1SDBp = 20 + ) > 0.
with OR being a dummy indicating if the person takes the option right.
Both imperfect takeup and the existence of other eligibility criteria take us away from the standard “sharp” RD design. Yet, the identification remains possible as long as we have a jump in the probability of treatment at the cutoﬀ, though lower than one: P r(OR = 1SDBp = 20 − ) ≠ P r(OR = 1SDBp = 20 + ).
Labour market impact of the option right
Option right impact on unemployment spell duration – Before examining the eﬀect of the option right on labour market outcomes, we first have to determine what are the most relevant outcomes. A first natural indicator to look at is the duration of the unemployment spell. Previous literature has shown that the elasticities of unemployment duration with respect to the level of unemployment benefits as well as the potential benefit duration were positive (Schmieder and Von Wachter (2016) for a review). Then, in this setting, we expect the eﬀect of the option right to go in two opposite directions: the eﬀect of receiving higher benefits for a shorter potential duration is a priori unclear. In practice, the unemployment spell duration can be defined in several ways, considering the data at hand. A first measure – the paid unemployment spell duration – corresponds to the addition of all subperiods during which benefits were paid, within the same spell.18 The full unemployment spell duration corresponds to all registered subperiods within the same spell, including the unpaid ones, which, by definition of the spell, last less than 4 months. However, restricting the analysis to the unemployment spell duration may some-times not be relevant as, if the person keeps going back and forth on the labour market, the unemployment spell may be short without necessarily corresponding to a stable exit to the labour market.19 That is why this measure will be presented with other complementary outcome variables meant to capture a medium to long-term eﬀect.
Tables 1.12 and 1.13 show, making the polynomial order vary, that taking the option right has a strong and significant eﬀect on unemployment spell duration – both paid and unpaid. If we focus on the quadratic specification, without any controls, of Table 1.12, the option right leads to an increase in the paid unemployment duration of about 161 days. The eﬀect is markedly large, and the option right seems to have a very detrimental impact on the employment outcomes of a population already in a precarious situation. In particular, if we consider that the average duration of a spell at the cutoﬀ is around 100 days, the eﬀect is equivalent to multiplying the spell duration by 2.6.20 At first sight, evidence would lead to the conclusion that letting the unemployed choose the terms and conditions of their compensation is a very ineﬃcient way of ensuring satisfactory coverage and a quick return to the labour market.
The strong and positive eﬀect on unemployment duration is confirmed by Fig-ures 1.13 and 1.14. The addition of covariates does not change the order of magnitude of the results for any specification, which is reassuring on the validity of the RDD. The results are also very consistent across the local linear regression and the higher-order polynomials.
Placebo tests are performed at diﬀerent cutoﬀ values randomly chosen along the distribution of daily benefit (Table 1.14), and display no significant estimates.
These diﬀerent findings indicate that benefiting from a shorter potential duration with a higher level of benefits makes the duration of the unemployment spell increase. In this specific context, the elasticity of unemployment duration to the level of unemployment benefits outweighs the elasticity of unemployment duration with respect to the potential benefit duration. This result is in line with the literature, as elasticities of non employment duration or benefit duration with respect to the benefit level are usually higher than the same elasticities measured with respect to potential benefit duration (see Schmieder and Von Wachter (2016) for a recent review).
The optimisation ability of the unemployed
If the impact of the option right on unemployment duration is markedly negative in the short run, and less clear in the long run, the conclusion in terms of unemployed people optimisation ability is not straightforward. Assuming that the the welfare objective is to maximise the total capital of benefits earned, we can look at the eﬀect of the option right on total benefits. The theoretical benefits capital is always higher in case of non-takeup, as it is equal to the theoretical capital in case of takeup, plus the capital from the remainder of the former right. However, since the unemployed person will not necessarily use his full potential benefit period if he finds a job before the end of his entitlement, it is likely that the actual benefit capital received during the unemployment spell does not correspond to the theoretical capital. Indeed, Table 1.27 shows that exercising the option right makes the total amount of benefits received over the subsequent spell increase. The diﬀerence is lower in the long run (Table 1.28) though substantial and significant, meaning that those not taking the option right catch up to some extent. Yet, maximising the amount of benefits received does not necessarily makes the worker better-oﬀ, as this can result in more days of unemployment.
To assess the extent to which workers are able to make the right insurance decision, one important parameter to take into account is whether they reach the exhaustion point of their entitlements. Indeed, if by exercising the option right they loose in terms of benefit duration, they could still be entitled to a long coverage. If they are not at risk of running out of benefits, taking the option right would simply mean having higher benefits, and would be a risk-less way to maximise the amount of benefits collected. Table 1.20 shows that indeed, only a small share of takers, 17%, do exhaust their UI entitlement. However, the strong eﬀect of the option right on the probability to exhaust benefits (Table 1.20) suggests that at least part of the unemployed choosing the option right because they anticipate a quick return to the labour market fail in their prediction.
To better understand if workers are able to optimise their compensation, I perform a heterogeneity analysis, intersecting the propensity to take the option right with the outcome in terms of unemployment spell duration. The idea is to analyse which subpopulations are more likely to take the option right, and whether they are right in doing so. Using age, gender and education categories and comparing the first and second-stage for each category, I observe that younger unemployed people have a higher jump in the probability of taking the option right at the 20e threshold (Tables 1.29 and 1.30), which is compatible with the less stable professional status that is often experienced in the early years of the career. Very young workers (under 35 years-old) are not perfectly optimising, as both their takeup and the negative impact on unemployment duration are of high magnitude. Older workers have insignificant first and second-stage estimates. Some of them do exercise the option right but they do not seem to resort to it under the 20e criterion. Similarly, Tables 1.31 and 1.32 point to a high takeup jump for people with a higher level of education, with a limited negative impact. In particular, when comparing the highest two categories, we observe that people with higher education display a significant and substantial jump in takeup while the impact on unemployment duration is not significant, whereas it is negative and of higher magnitude for people who only completed high school despite a higher first-stage. In line with this result on education, we observe that unskilled employees and workers often exercise the option right whereas the impact is markedly negative for them (Tables 1.33 and 1.34). On the reverse, skilled workers is the category where the jump in takeup is the highest with a limited and insignificant eﬀect on unemployment duration. It means that, although they are not at the top of the skill distribution, their better qualifications do play a role in protecting them from the adverse impact of the option right on labour market performance. The option right is also more detrimental to females, as they exhibit a higher discontinuity in takeup, coupled with a stronger negative impact on labour market outcomes (Tables 1.35 and 1.36).
All in all, it seems that male workers in the middle and at the top of the age and skills distribution and highly educated are the ones taking the most advantage of the option right, with a limited impact on the subsequent unemployment spell duration. One interpretation is that they are better at predicting their reemployment probability, and that they have at the same time better objective labour market prospects. More educated workers are more likely to take the risk of exercising the option right without making their unemployment spell duration increase too much, as they are better equipped to find a job rapidly and they may have a more stable professional status.
Adverse selection and moral hazard
The choice feature of the option right is a unique opportunity to try to measure the extent of adverse selection in the UI market. Indeed, the main rationale for the implementation of a UI mandated at the national level comes from the Rotschild-Stiglitz demonstration (Rothschild and Stiglitz, 1992) that, because of heterogeneity in risk types and asymmetry of information, there is no equilibrium supporting the provision of insurance. Empirically, most of the papers have taken this result as granted without questioning the actual presence of adverse selection, and rather focusing on moral hazard. In this subsection, I try to disentangle the adverse selection from the moral hazard by looking at both the predicted and realised unemployment duration. Unemployment duration is predicted using a sample of similar job-seekers during the two years preceding the implementation of the option right. I use a large set of covariates associated to the worker and to his last employer to capture as accurately as possible all the information that is available to the worker when he has to decide whether to take or not the option right. An alternative group of workers is used for the prediction in Tables 1.41 to 1.
Table 1.37 shows that the predicted unemployment duration is higher for eligible non-takers than for takers. The fifteen-day diﬀerence, representing a 10% increase relative to the predicted unemployment duration of takers, is indicative of significant adverse selection. Job-seekers with higher predicted unemployment duration are more likely to choose the longest UI coverage. This is confirmed by Tables 1.38 and 1.39 where the population is divided into quintiles of predicted unemployment duration. We observe that the takeup rate is a decreasing function of the predicted unemployment duration. It means that the higher the unemployment risk, the more likely it is that the worker will choose the longest coverage. Consistently, Table 1.39 shows that the jump in takeup at the threshold is the highest in the lowest quintiles. This positive correlation test with the predicted unemployment duration, although it may not capture the role of unobservables, is indicative of adverse selection.
Table of contents :
1 Generosity versus Duration Trade-Off and the Optimisation Ability of the Unemployed
2 Legislation and data
3 The determinants of the takeup
4 Empirical implementation
5.1 Labour market impact of the option right
5.2 The optimisation ability of the unemployed
5.3 Adverse selection and moral hazard
6 Concluding remarks
2 Unemployment Benefits and the Timing of Redundancies: Evidence from Bunching
2 Institutional Background
3 Empirical Evidence of Bunching
3.2 Documentation of the Bunching
3.3 Underlying Mechanisms: Exploration of the Bargaining Process126
4 Theoretical Framework of Negotiated Layoff
5 Heterogeneity in Bunching
5.1 Empirical Bunching Estimation
5.1.1 Baseline methodology
5.1.2 The Difference-in-Bunching Strategy
5.2 Characterisation of the buncher
5.3 Interaction Between Ability and Incentives
5.3.1 Individual characteristics
5.3.2 Firms’ characteristics
6 Robustness Checks and Extensions
6.1 Round-Number Fixed Effects
6.2 Response at the two-years cutoff
6.3 Extensive margin response
7 Welfare implications
8 Concluding Remarks
1 Institutional context
2.1 Other optimisation mechanisms
2.2 Bunching at the two-year threshold
3 Alternative theoretical framework
4 Methodological discussion
3 Entitled to Leave: the Impact of Unemployment Insurance Eligibility on Employment Duration and Job Quality
1 Institutional background
2 Data and Descriptive Statistics
3 Empirical evidence of a separation response
4 Impact of UI eligibility criterion on contract duration
5 Extensive margin effect of UI benefits