200 likes | 307 Views
Employment RCTs in France. Bruno Crépon. Focus on employment issues. Starting 2006 we have launched many RCT’s in France Focus on Employment RCT’s Strong debates and pressures about reinforced counseling services 2005 end of the PES monopoly: entry of new firms
E N D
EmploymentRCTs in France Bruno Crépon
Focus on employment issues • Starting 2006 we have launched many RCT’s in France • Focus on Employment RCT’s • Strong debates and pressures about reinforced counseling services • 2005 end of the PES monopoly: entry of new firms • Counseling big issue: large changes in last years. Reinforced services becoming the corner stone of employment policies. Size of portfolios going from 120 to 30! • But: is it useful to invest so much in these policies • Some frequent questions: cost benefit, displacements
Example 1: Reinforced counseling • Our first randomized experiment. • Quite large : more than 50.000 individuals • Target is unemployed at risk of long term unemployment • Individual characteristics and caseworker’s appreciation • With a large provision of unemployment benefits : 120 days • Regular services: 120 unemployed by caseworker • Reinforced services : 30 unemployed by caseworker • Part of the idea is to “invest” in unemployed at the beginning of the spell and save unemployed benefits • Program lasts 6 months: hope was to have unemployed soon back to durable employment • Not a mandatory program
Example 1: Whoshouldprovide the services • Reinforced counseling program was provided by both the private and the public sector. First entry of private operators • Same program for private and public sectors • Private operators selected by bids to provide services in specific regions at the local level • Several bids organized • Private operators had contracts : • Maximum fee for one unemployed was between 3000 to 4000 euros depending on the bid • Paid in three parts • 30% at entry, 35% if “durable” employment within 6 months, 35% if still in employment 6 months later
Example 1: Private vs public counseling • Plan was to have 40,000 unemployed in each of the reinforced counseling programs • Target population: more than 250,000 unemployed • Three arms • Assigned to regular services: no participation in the reinforced counseling program • Assigned to reinforced counseling services provided by the public sector • Assigned to reinforced counseling services provided by the private sector
Example 1: Implementation • Assignment of unemployed to one of three groups • Directly by caseworkers at the first meeting interview • Identification and registration of the unemployed • Randomization. Due to constraints the probabilities were quite different • 85% for private, 7.5% for public reinforced, 7.5% for control group • Due to large number of individuals very rough data mostly administrative • Know if people left unemployment and the type of exit • Also know if people went back to unemployment and when
Example 1 : Comparisons • Can compare groups assigned • Assigned to the private program and assigned to the control group • Can also measure the effect on these who entered tje program • Entry into the program is also important • Part of incentives in the private operator contract NT T T NT NT Control Private Public
Example 1: Private vs public counseling • Effect of program participation: employment after 6 months • Exit rate at 6 months in the control group 23% • Private scheme : +4.5% • Public scheme : +10.2% • Same result for different durations: 3 months, 9 months and 12 months after random assignment • Effect only appears at 6 months and remain lower for the private sector
Example 1: Understanding the findings • One way to understand these results is in the incentives provided by the contract structure of private operators • Private operators can enroll as many unemployed as possible in the program and just park some of them • Entry into the program • Private scheme 46% • Public scheme 39% • If we look at the effect assigned individuals we obtain very similar results for public and private programs: +4% and +3% respectively
Example 1: Main lessons • Reinforced services increase the chances to go back to employment • However no cost benefit analysis • Idea that private performs better is wrong • Important to use randomized experiments to address such issue • Contract features plays a key role: strong incentives for private operators to include unemploymed and park some of them
Example 1: Were these results useful? • Who should provide the services is a real important issue • It was useful to show that private operators are not a cheaper substitute to public services • The study is a strong reference though not the only one • No direct causal relation between results and decision • The relationship of private operators with the Public employment service has turned to • Capacity potential : manage fluctuation in inflow through private operators • Specialization ask private operators to develop an expertise on some specific public • Increased monitoring • Change the contract structure (reduced fee at entry)
Example 2: Displacementeffects • Common idea that counseling help beneficiaries but can be at the expense of non beneficiaries • To catch the idea assume the number of available jobs is given helping some will not reduce the overall unemployment rate • Key issue to address the effect of policy • Should it be considered as a redistributive policy or a global solution to solve the unemployment problem? but little evidence • Program considered was a reinforced counseling scheme • Same idea to have 30 unemployed by caseworker instead of 120
Example 2: Specificdesign • It was possible to work with the PES to implement a specific design so as to be able to answer this question • Considers « areas » which are small independent labor market (Local Employment Agency) • Assign randomly areas to be control or treatment areas • In treatment areas assign individuals to be in a treatment group or in a control group • Compare the two controls: • Controls in treatment areas, exposed to treatment • « super controls » coming from control areas
Example 2: Specific design Large number of such paires of indpendant areas Comparing PC and LC Allows to measuredisplacement Comparing PC and T Allows to measure the trueeffect T LC PC Treated area 50% randomlyassigned to T 50% randomlyassigned to a Local Control Pure Control Area Local areas without the program
Example 2: Displacementeffects • Around 12,000 individuals in the experiment • Around 230 local areas • Quite simple to implement but need coordination • Get the list of eligible people from the PES • Send back the list of youths assigned to the program to PES and to the operator in charge of contacting the youth • Measurement was performed through surveys • 8 months, 12 months, 16 months and 18 months after random assignment • Short questionnaire based on main employment outcomes. Non response limited to 20%
Example 2: Displacementeffects • Find the program had an effect on the treated • Increased chance to find a job when in the program compared to the pure control • Also find significant displacement effects • Reduced chance to find a job when in the control compared to the pure control • Especially for men • The size of the displacement effect is quite large • Net effect of the policy is almost zero when accounting for displacement effects
Example 2: Displacementeffects • One other finding is that both the effect of the program and displacement effects are stronger in weak labor market • An important result as ALMP are countercyclical • This experiment was specific as it was entirely designed to answer a question • Not as in the previous example does the program had an effect or not • Promising area both for research and for cooperation with the PES. • PES has question it wants to answer
Example 2: Were these results useful? • The results have to be confirmed however one lesson from this RCT is that reinforced counseling scheme can be seen here as redistributive policy rather than a global solution suitable for large number of unemployed • Has to be kept for specific populations • We are currently examining a specific program designed to address the needs of youth in deprived suburbs • Recent changes in the Public Employment Service consistent with the findings: • Increased targeting • Direct Public Employment Service effort toward firm services so as to increase the demand for labor
Conclusion • We are in the process of acquiring a lot of information about the effect of employment policies • Example : Provide youth in deprived suburbs with a cash transfer conditional on attendance to their counseling program • Help them to find a job • Help them to build human capital • Is this appealing intervention the right answer?
Conclusion • Experiments are a way to get more precise and relevant answers to our questions • The progressin the understanding of relevant issues weare making are collective • The resultsweobtain are sharedwith the administration • They are part of the decision process • It is not “Test and if efficient scale-up” • Decision process in France is more complicated • Experiments are also a way to become more precise in the question wewant to answer