The Di cult School-to-Work Transition of High School Dropouts: Evidence from a eld experiment

Transcription

1 The Di cult School-to-Work Transition of High School Dropouts: Evidence from a eld experiment Pierre Cahuc, Stéphane Carcillo, Andreea Minea June / 44

2 ntroduction Research suggests that private sector subsidies perform better than public sector ones (Heckman et al.1999, Kluve and Schmidt, 2002; Card et al. 2010, 2015; Kluve et al. 2016) n France, subsidies in the non-market sector is an important lever of employment policy. Despite the substantial cost, little is known on their impact on school-to-work transitions. Here we test programs by measuring the chances of getting a callback from employers for youth with various pathways. The method involves sending résumés of youth who, over a 3-year period after their leaving high-school, have been: unemployed unemployed with some temporary work experience employed in non-subsidized or subsidized jobs, in the market or the non-market sector, with or without skill certi cation. 2 / 44

3 Presentation plan 1. Background 2. Experimental design 3. Results 3 / 44

4 1. Background n 2016, in France, the unemployment rate of youth reached 24.3% from 18,5% at the beginning of the crisis. High school dropouts are 3 times more likely than university graduates to be NEET and 1 in 6 youth aged never nished school (OECD, 2016). 4 / 44

5 Share of non subsidized jobs Share of subsidized jobs Share of employed Unemployment rate 1. Background Employment rate Unemployment rate Year since leaving education Year since leaving education Share of non subsidized employment Share of subsidized employment Year since leaving education Year since leaving education Source: Enquête emploi (NSEE) Figure: Labor market situation of high-school dropouts in France 5 / 44

6 1. Background Emploi d Avenir (EdA) created in 2012: targets youth aged 16-25, who are unemployed, have a low education level and have been searching for a job for at least 6 months during the previous year non-market sector employers bene t from a state subsidy of 75% of gross minimum wage for up to 3 years; the subsidy amounts to 35% for market sector ones the program is associated with a mandatory training component that can lead to a certi cation of skills. 6 / 44

7 1. Background One year after their entry into an EdA, 3 in 4 young people had already bene tted from training. However, only 52% of youth with a non-market sector contract had acquired a skill certi cation; they were 35% among those with a market sector contract. n 2014, 80% of entries into EdA were in the non-market sector. 7 / 44

8 2.1 The ctitious applicants Applicants are identical in all points, with the exception of their employment status and type of contract in the past 3 years: they are young men, who graduated from middle school with a delay of 2 years and then went to a vocational high school they attended high school for 2 years without acquiring a degree after leaving school when they were 19, they had one year of unemployment following that year, they either bene tted from an Emploi d Avenir or from a non-subsidized contract for 3 years, or they were again unemployed for 3 years they are almost 24 at the time of the experiment, have an elementary level of English and basic T skills. 8 / 44

9 2.1 The ctitious applicants Applicant pro les correspond to the two types of occupations they apply for: gardener: youth attended a vocational high school preparing a diploma in the area of landscaping (Travaux paysagers) receptionist: youth attended a vocational high school preparing a diploma in the area of services for the management of places open to the public (Services de proximité et vie locale, Spécialité: Gestion des espaces ouverts au public). 9 / 44

10 Share of subsidized jobs 2.1 The ctitious applicants Production Maintenance Gardening Logistics Reception Profession Administration Trade Studies/R&D Teaching/Care Other Figure: Share of subsidized jobs by profession among youth ( ) 10 / 44

11 2.1 The ctitious applicants The employed youth pro les Youth with an employment experience pro le are unemployed since January Market Non-Market Subsidized job Yes No Yes No Vocational degree Yes No Yes No Yes No Yes No Note: This table displays the employment experience of applicants who could have worked either in the market or in the non-market sector, on a job either subsidized or non-subsidized and who could get either a vocational degree or no vocational degree. 11 / 44

12 2.1 The ctitious applicants The unemployed youth pro les We create four types of unemployed pro les: unemployed with no short xed-term contract experience in the previous 3 years unemployed with 1 short xed-term contract experience unemployed with 2 short xed-term contracts experiences unemployed with 3 short xed-term contracts experiences. Short-term xed contracts are in occupations other than those youth apply for (gardener and receptionist): sales promoter (animateur de vente) in a home furnishings retail chain crew member (équipier polyvalent) in a fast-food chain. 12 / 44

13 2.1 The ctitious applicants The variation in the previous employment statuses, contract type, sector and skill certi cation for the employed youth and that in the number of short xed-term contracts performed by unemployed youth results in 12 applicant pro les: (2 contract types) x (2 sectors) x (2 training trajectories= vocational title obtained or not) + (4 unemployment paths) 13 / 44

14 2.2 The applications Applications are made to job o ers from all French departments between February 15th and July 15th Job o ers are identi ed using in priority the website of Pôle Emploi. A few private job search websites were equally used. Applications are sent only when it is possible to contact the recruiter directly by . The same recruiter could be contacted only once. The typical application included a résumé and a cover letter. Two applications were sent for each job o er, but in consecutive days: one with an employed pro le and one with an unemployed pro le. The name of the applicant, the application pro le, layout type and application day were selected at random. 14 / 44

15 2.2 The applications n total 5,388 applications were sent, with, on average, 674 applications per unemployed pro le and 337 per employed pro le. Callbacks to job applications were received by and by phone, until September 5th When recruiters provided a positive answer to an application, an was sent to thank the recruiter and inform him that the applicant had accepted another o er. 15 / 44

16 3. Results Descriptive statistics Dependent variable: Received callback for interview or information request Mean Std. deviation Résumé attributes All applicants Unemployed, no job experience at all Unemployed, 1 temporary job Unemployed, 2 temporary jobs Unemployed, 3 temporary jobs Employed, market, subsidized, certi ed skills Employed, market, non-subsidized, certi ed skills Employed, market, subsidized, no certi ed skills Employed, market, non-subsidized, no certi ed skills Employed, non-market, subsidized, certi ed skills Employed, non-market, non-subsidized, certi ed skills Employed, non-market, subsidized, no certi ed skills Employed, non-market, non-subsidized, no certi ed skills Job characteristics Gardener position Receptionist position Job in the non-market sector Job in the market sector Job in the public sector Job in the private sector Note: The rst column of the table reports the mean value of the primary dependent variable which is equal to one if the résumé received a callback from the employer. The second column reports the standard deviation of this variable. 16 / 44

17 3. Results Employer descriptive statistics Mean Std. deviation Employment pool characteristics Unemployment rate (%) Unemployment rate, quantile Unemployment rate, quantile Unemployment rate, quantile Unemployment rate, quantile Unemployment rate, quantile Employer s characteristics Market.776 Non Market.224 Private.871 Public.129 Job characteristics Gardener position.495 Receptionist position.505 Open-ended contract.285 Fixed-term contract.715 Job characteristics in the market sector Gardener position.497 Receptionist position.503 Open-ended contract.333 Fixed-term contract.667 Job characteristics in the non-market sector Gardener position.535 Receptionist position.465 Open-ended contract.106 Fixed-term contract.894 Note: The table reports the unemployment rate of the commuting zone of the vacant job, the share of market and non-market vacant jobs, the required profession and the type of contracts of vacant jobs. 17 / 44

18 3. Results We estimate the following linear probability model: y ij = α + β i 1(i) + x j γ 0 j + ε ij y ij : indicator variable equal to one if the applicant of type i receives a callback from job j. A callback is de ned as a demand for an interview or a request for additional information. 1(i) : indicator function equal to one for applicant of type i x j : vector of characteristics of the job (profession, type of contract, size of rm, etc.) ε ij : error term. We also estimate a Probit model: estimated marginal e ects are very similar to OLS results. 18 / 44

19 3. Results The E ect of ndividual Pathways on Callback : Unemployment Dependent variable: Received a callback (1) (2) (3) (4) All applicants Gardener Receptionist (1) Unemployed, 1 temporary job (2) Unemployed, 2 temporary jobs (3) Unemployed, 3 temporary jobs (4) Employed, market, subsidized, certi ed skills (5) Employed, market, non-subsidized, certi ed skills (6) Employed, market, subsidized, no certi ed skills (7) Employed, market, non-subsidized, no certi ed skills (8) Employed, non-market, subsidized, certi ed skills (9) Employed, non-market, non-subsidized, certi ed skills (10) Employed, non-market, subsidized, no certi ed skills (11) Employed, non-market, non-subsidized, no certi ed skills (12) Constant (ref: unemployed, no job experience at all) N 5, 388 5, 388 2, 720 2, 668 Adj-R Department xed e ects no yes yes yes Month xed e ects yes yes yes yes Note: The dependent variable is a dummy variable equal to one if the application gets a callback. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 19 / 44

20 3. Results The E ect of ndividual Pathways on Callback : Employment but no certi cation Dependent variable: Received a callback (1) (2) (3) (4) All applicants Gardener Receptionist (1) Unemployed, 1 temporary job (2) Unemployed, 2 temporary jobs (3) Unemployed, 3 temporary jobs (4) Employed, market, subsidized, certi ed skills (5) Employed, market, non-subsidized, certi ed skills (6) Employed, market, subsidized, no certi ed skills (7) Employed, market, non-subsidized, no certi ed skills (8) Employed, non-market, subsidized, certi ed skills (9) Employed, non-market, non-subsidized, certi ed skills (10) Employed, non-market, subsidized, no certi ed skills (11) Employed, non-market, non-subsidized, no certi ed skills (12) Constant (ref: unemployed, no job experience at all) N 5, 388 5, 388 2, 720 2, 668 Adj-R Department xed e ects no yes yes yes Month xed e ects yes yes yes yes Note: The dependent variable is a dummy variable equal to one if the application gets a callback. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 20 / 44

21 3. Results The E ect of ndividual Pathways on Callback : Employment with certi cation Dependent variable: Received a callback (1) (2) (3) (4) All applicants Gardener Receptionist (1) Unemployed, 1 temporary job (2) Unemployed, 2 temporary jobs (3) Unemployed, 3 temporary jobs (4) Employed, market, subsidized, certi ed skills (5) Employed, market, non-subsidized, certi ed skills (6) Employed, market, subsidized, no certi ed skills (7) Employed, market, non-subsidized, no certi ed skills (8) Employed, non-market, subsidized, certi ed skills (9) Employed, non-market, non-subsidized, certi ed skills (10) Employed, non-market, subsidized, no certi ed skills (11) Employed, non-market, non-subsidized, no certi ed skills (12) Constant (ref: unemployed, no job experience at all) N 5, 388 5, 388 2, 720 2, 668 Adj-R Department xed e ects no yes yes yes Month xed e ects yes yes yes yes Note: The dependent variable is a dummy variable equal to one if the application gets a callback. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 21 / 44

22 3. Results The E ect of Skill Certi cation on Probability of Callback (1) (2) (3) All applicants Gardener Receptionist Employment without certi ed skills Employment with certi ed skills Constant (ref: unemployed) N 5, 388 2, 720 2, 668 Adj-R Note: The dependent variable is a dummy variable equal to one if the application gets a callback. Employment with certi ed skills comprises all applications of applicants who were employed on jobs in the market or non-market sector, with or without subsidy, which o ered skill certi cation. Employment without certi ed skills comprises all applications of applicants who were employed on jobs in the market or non-market sector, with or without subsidy, which did not o er skill certi cation. Certi ed skills is an indicator variable equal to 1 if the applicant has a skill certi cation. All columns include department and month xed e ects. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 22 / 44

23 3. Results The E ect of ndividual Pathway on Probability of Callback Short work experience gives no advantage to unemployed who were in these jobs compared with unemployed with no work experience during their unemployment spell. Result consistent with Farber et al. (2016) and Nunley et al. (2016) who nd that low-level temporary jobs do not increase the probability of callback. Also, being employed rather than unemployed does not improve the likelihood of callback in the absence of training. Result consistent with Fremigacci et al. (2016) who nd that men unemployed for one year applying to waiter jobs in France do not have lower callback rates than those employed on xed-term contracts over the year. 23 / 44

24 3. Results The E ect of Skill Certi cation on Probability of Callback The impact of skill certi cation is large (+3.4 percentage points). Skill certi cation has a stronger impact when youth occupied a subsidized job rather than a non-subsidized one. The impact is even stronger if the job on which the experience was certi ed was in the non-market sector. Results that use a stricter de nition of the callback rate (=explicit request for an interview), con rm the hypothesis that training acquired in the market sector is perceived as less serious by employers. 24 / 44

25 3. Results The E ect of Subsidized and Non-Market Job Experience on Probability of Callback (1) (2) (3) (4) All applicants Gardener Receptionist Panel A (All job o ers) Non-subsidized market sector experience Non-subsidized non-market sector experience Subsidized non-market sector experience Subsidized market sector experience Constant (ref:unemployment) Certi ed skills no yes yes yes N 5, 388 5, 388 2, 720 2, 668 Adj-R Panel B (Market sector job o ers) Non-subsidized market sector experience Non-subsidized non-market sector experience Subsidized non-market sector experience Subsidized market sector experience Constant (ref:unemployment) Certi ed skills no yes yes yes N 4, 236 4, 236 2, 104 2, 132 Adj-R / 44

26 3. Results The E ect of Subsidized and Non-Market Job Experience on Probability of Callback (1) (2) (3) (4) All applicants Gardener Receptionist Panel C (Non-market sector job o ers) Non-subsidized market sector experience Non-subsidized non-market sector experience Subsidized non-market sector experience Subsidized market sector experience Constant (ref:unemployment) Certi ed skills no yes yes yes N 1, 152 1, Adj-R Note: The dependent variable is a dummy variable equal to one if the application gets a callback. Non-subsidized market sector experience comprises all applications of applicants who were employed in the market sector, on non-subsidized jobs, with or without skill certi cation. Non-subsidized non-market sector experience comprises all applications of applicants who were employed in the non-market sector, on non-subsidized jobs, with or without skill certi cation. Subsidized non-market sector experience comprises all applications of applicants who were employed in the non-market sector, on subsidized jobs, with or without skill certi cation. Subsidized market sector experience comprises all applications of applicants who were employed in the market sector, on subsidized jobs, with or without skill certi cation. Certi ed skills is an indicator variable equal to 1 if the applicant has a skill certi cation. All columns include department and month xed e ects. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 26 / 44

27 3. Results The E ects of Being Employed on Subsidized Jobs and on Non-Market Jobs There seems to be no "stigma" e ect associated with subsidized employment for low skilled youth in France. Professional experience in the non-market sector increases the chances of callback more than employment in the market sector. This is again related to the credibility of training. Card et al. (2010), Kluve (2010), Benoteau (2015) found that private sector incentive schemes are more e ective than other programs for youth. Our paper! n absence of training, market sector experience, whether subsidized or non-subsidized, has no e ect on the probability of callback. Both market sector and non-market sector recruiters react in a similar way to candidates previous employment experience. 27 / 44

28 3. Results The mpact of Local Labor Market Conditions So far, we have found that only training leading to skill certi cation signi cantly raises callback rate among all employed applicant types. However, skill certi cation corresponds to the lowest level in the National Classi cation of Levels of Training (titre professionnel). The advantage provided by the degree should decline when the number of other candidates increases (higher unemployment rate). 28 / 44

29 3. Results The E ect of Certi ed Skills on Probability of Callback by Quintile of Unemployment Rate of the Commuting Zone of the Job O er (1) (2) (3) (4) (5) (6) All Q1 Q2 Q3 Q4 Q5 Certi ed skills Constant Unemployment rate Mean Min Max N 5, 144 1, , 020 1, Adj-R Note: The dependent variable is a dummy variable equal to one if the application gets a callback. Certi ed skills is an indicator variable equal to 1 if the applicant has a skill certi cation. QX stands for the number of the quintile of the unemployment rate of the commuting zone of the job o er. Mean, Min and Max denote the mean, the minimum and the maximum value of the quintile of the unemployment rate of the commuting zone respectively. The total number of observations is slightly lower compared with the previous tables because the employment zone was not identi ed for some job o ers. The regressions include department and month xed e ects. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 29 / 44

30 3. Results The E ect of Certi ed Skills on Probability of Callback by Quintile of Unemployment Rate of the Commuting Zone of the Job O er Training accompanied by skill certi cation signi cantly improves the callback rate only when the local unemployment rate is su ciently low. As a robustness check, we also use the average callback rate of our applicants by commuting zone as a measure of local labor market conditions. Results corroborate those that rely on the unemployment rate. 30 / 44

31 4. Conclusion This paper reports results from a eld experiment studying the impact of individual pathways with various forms of labor market experience for young high school dropouts. Results indicate that the likelihood of receiving a callback from employers sharply increases when youth get a skill certi cation. Other pathways in the labor market seem unable to improve youth s employment outlook. This result is consistent with previous research showing that accruing work experience, even in the market sector, is not always su cient to get more callbacks. t suggests that employment support measures should be conditional on getting a skill certi cation at the end of the employment period, at least for previously unskilled youth. 31 / 44

32 4. Conclusion The e ect of skill certi cation is more pronounced in tight labor markets. This highlights the importance of additional measures supporting the geographical mobility of youth. Our test cannot measure, however, the full potential e ects of training on employment (e.g. job stability or quality). Similarly, our results cannot capture the role of social networks of young people. This experiment suggests additional areas for future research: generalization of occupation types, focus on youth with higher levels of initial education, e ciency of other training programs, etc. 32 / 44

33 5. Appendix 33 / 44

34 5.1 Literature review Literature on eld experiments Eriksson and Rooth, 2014! subsequent work experience eliminates any potential negative e ect associated with past long-term unemployment spells. But contemporary spells are di erent: L Horty et al. (2016), Eriksson and Rooth (2014), Kroft et al. (2013)! it is long spells that have a negative impact on callback rates. Farber et al. (2016)! no relationship between callback rates and the duration of unemployment for mature and older female workers in the US. Also, holding temporary jobs may negatively a ect callbacks. Our paper! past unemployment experience has no e ect on callback rates. There is a negative in uence of local market conditions on the impact of training. Also, employment periods on xed-term contracts do not improve callbacks. 34 / 44

35 5.1 Literature review Literature on active labor market policies Heckman et al. (1999), Kluve and Schmid (2002), Sianesi (2002), Hujer (2004)! public sector job programs perform poorly in comparison to other interventions. Card et al. (2010, 2015)! private sector employment programs have large e ects in medium-long run, while public sector ones have no e ect. Autor et al. (2016)! employment programs for low skilled workers, notably temporary held jobs, have no e ect on earnings. Kluve et al. (2016)! youth programs that integrate multiple interventions are more likely to succeed. Programs focusing only on job search activities or subsidized employment have no impact. Our paper! neither public sector nor private sector subsidized employment makes a di erence for low skilled youth. Work experience has an impact only when accompanied with training leading to certi cation. 35 / 44

36 5.2 Robustness check Descriptive statistics Dependent variable: Received callback for interview Mean Std. deviation Résumé attributes All applicants Unemployed, no job experience at all Unemployed, 1 temporary job Unemployed, 2 temporary jobs Unemployed, 3 temporary jobs Employed, market, subsidized, certi ed skills Employed, market, non-subsidized, certi ed skills Employed, market, subsidized, no certi ed skills Employed, market, non-subsidized, no certi ed skills Employed, non-market, subsidized, certi ed skills Employed, non-market, non-subsidized, certi ed skills Employed, non-market, subsidized, no certi ed skills Employed, non-market, non-subsidized, no certi ed skills Job characteristics Gardener position Receptionist position Job in the non-market sector Job in the market sector Job in the public sector Job in the private sector Note: n this table we consider that the callback variable equals one if an interview is requested and zero in all other cases. Thus, the rst column of the upper part of the table reports the mean value of the primary dependent variable which is equal to one if the résumé received a callback from the employer explicitly asking to set up an interview and to zero otherwise. The second column reports the standard deviation of this variable. 36 / 44

37 5.2 Robustness check The E ect of ndividual Pathway on Probability of Callback Dependent variable: Received a callback for an interview (1) (2) (3) (4) All applicants Gardener Receptionist (1) Unemployed, 1 temporary job (2) Unemployed, 2 temporary jobs (3) Unemployed, 3 temporary jobs (4) Employed, market, subsidized, certi ed skills (5) Employed, market, non-subsidized, certi ed skills (6) Employed, market, subsidized, no certi ed skills (7) Employed, market, non-subsidized, no certi ed skills (8) Employed, non-market, subsidized, certi ed skills (9) Employed, non-market, non-subsidized, certi ed skills (10) Employed, non-market, subsidized, no certi ed skills (11) Employed, non-market, non-subsidized, no certi ed skills (12) Constant (ref: unemployed, no job experience at all) N 5, 388 5, 388 2, 720 2, 668 Adj-R Department xed e ects no yes yes yes Month xed e ects yes yes yes yes Note: n this table we consider that the callback variable equals one if an interview is requested and zero in all other cases. Thus, the dependent variable is a dummy variable equal to one if the application gets a callback asking for an interview. Robust standard errors are clustered at the job level and reported below the coe cients. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 37 / 44

38 5.3 Probit model The E ect of ndividual Pathway on Probability of Callback (Marginal E ects at the Mean) Dependent variable: Received a callback (1) (2) (3) All applicants Gardener Receptionist (1) Unemployed, 1 temporary job (2) Unemployed, 2 temporary jobs (3) Unemployed, 3 temporary jobs (4) Employed, market, subsidized, certi ed skills (5) Employed, market, non-subsidized, certi ed skills (6) Employed, market, subsidized, no certi ed skills (7) Employed, market, non-subsidized, no certi ed skills (8) Employed, non-market, subsidized, certi ed skills (9) Employed, non-market, non-subsidized, certi ed skills (10) Employed, non-market, subsidized, no certi ed skills (11) Employed, non-market, non-subsidized, no certi ed skills N 5, 388 2, 720 2, 668 Pseudo R Note: The table reports marginal e ects for the probability of receiving a callback based on probit regressions. The dependent variable is a dummy variable equal to one if the application gets a callback. Robust standard errors are clustered at the job level and reported below the marginal e ects. * signi cant at 10 percent, ** signi cant at 5 percent, *** signi cant at 1 percent. 38 / 44

39 5.4 Application examples 39 / 44

40 5.4 Application examples 40 / 44

41 5.4 Application examples 41 / 44

42 5.4 Application examples 42 / 44

43 5.4 Application examples 43 / 44

44 5.4 Application examples 44 / 44