Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Micro-econometric evaluation of subsidized employment in morocco: the case of the "Idmaj" program

Micro-econometric evaluation of subsidized employment in morocco: the case of the "Idmaj" program This paper aims to assess the impact of the Moroccan wage subsidy program "Idmaj". It applies the propensity score matching method to the data from a survey conducted by the Ministry of Labour on a sample of eligible individu- als. Our results suggest that wage subsidies in Morocco have a positive but marginally significant effect on reducing unemployment and improving employment and a significant negative impact on wages. It also highlights some heterogeneous effects of the program, particularly on women. Finally, it appears that the program did not serve as a stepping stone to higher-paying, high-quality work and, in contrast, it had a stigmatizing effect on beneficiaries. Keywords: Propensity score matching, Impact evaluation, Active labour market policy, Wage subsidies JEL Classification: C14, C31, H43, J38, J68 unconditional average magnitude of these effects remains 1 Introduction low (Kluve et al. 2019).The assessment is even more pessi - Active labour market policy (ALMP) programs are usu- mistic according to other meta-analysis that reveal a neg- ally classified into three categories (Card et al. 2010). The ative impact, at best zero, on beneficiaries’ employment first one includes training programs that aim to improve and income (Card et  al. 2018).For developing countries, employee’s skills. The second encompasses job search the evidence on the effectiveness of ALMP programs still assistance and job-matching schemes. The last one relates sketchy (Grimm and Paffhausen 2014), due to the lim - to wage subsidies, both those provided to employees to ited number of rigorous impact evaluations (McKenzie reduce inactivity traps and those provided to employers 2017). For example, in the very influential meta-analysis to reduce the costs of hiring. These subsidies, which are of Betcherman et  al. (2004), which covers 159 studies, based on long-standing theoretical arguments (Kaldor there are only 39 evaluations from developing and tran- 1936), have been used extensively by developed countries sition countries, of which only 4 drew on randomized since the 1980s, before they were popularized in develop- experiments. Moreover, the findings of the few existing ing countries (Auer et al. 2005). The aftermath of the last evaluations should be taken with caution, as the found financial crisis, the persistence of mass unemployment effects vary across intervention’ type and depending on and the inevitable process of labour automation are all the group of beneficiaries and country context (Cho and recent trends that have given new impetus to wage subsi- Honorati 2013). Therefore, additional studies assessing dies as a primary tool of ALMP (Banerji et al. 2014). However, the available literature about their effective - ness provides mixed evidence. The recent meta-anal - ysis show positive effects (Yeyati et  al. 2019), but the 1 Among the main meta-analyzes in developing countries, we can cite: Cho and Honorati 2013; Grimm and Paffhausen 2014; Blattman and Ralston 2015; Mckenzie 2017; Escudero et al. 2019. *Correspondence: naima.samoudi1995@gmail.com 2 Only one of which (Galasso et  al. 2004) was published in an academic Laboratory of Applied Economics (LEA), Med V University, Rabat, Morocco journal (McKenzie 2017). © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. 17 Page 2 of 13 A. Chatri et al. the effectiveness of wage subsidies in developing coun - data used. Section “6” presents and discusses the empiri- tries are of crucial importance. cal results. The final section presents  the conclusions, The case of Morocco, where the available labour fac - policy implications and limitations of this paper. tor seems to be strongly under-used and inefficiently allocated, is an excellent example in this perspective. The 2 Literature review unemployment rate has exceeded 10% on average over The available literature shows that the effectiveness the last 20  years, with 70% are long-term unemployed, of wage subsidies is very heterogeneous and depends, 60% are first-time job seekers, and the employment rate among other factors, on their design arrangements, the is now only 40%. The high weight of graduate unemploy - modalities of their implementation, the targeting of their ment, which is five times higher than that of without beneficiaries, their expected outcomes, the market labour diploma people, as well as that of informal employment, functioning, as well as the positioning of countries in the unpaid employment and underemployment, which economic cycle (Card et  al. 2010; Almeida et  al. 2014; represent respectively 27%, 18% and 10% of the total Gautié 2015). Nevertheless, three consensual conclusions employment, reveal, moreover, the suboptimal use of deserve to be highlighted. the available labour force (Haut-commissariat au plan First, subsidies provided to firms for reducing their & World Bank 2017).This situation has occurred despite cost of hiring tend to produce positive results compared the implementation, particularly since the beginning of to those offered directly to the unemployed (Venetoklis the 1990s, of an ambitious ALMP, which includes, among 2004; De Mel et  al. 2010, 2016; Bruhn 2016). Secondly, other governmental programs, a wage subsidy scheme, the quasi-experimental methods produce higher effects called "Idmaj". Unfortunately, none of these public pro- than those of the experimental ones (Kluve 2010; Grimm grams has been evaluated, and our knowledge about and Paffhausen 2014). This may partly explain why these their effectiveness is still very limited. subsidies are relatively more effective in developing Therefore, this paper seeks to fill this gap by estimat - countries than in developed countries (Betcherman et al. ing the causal effect impact of the "Idmaj" Program using 2004; Kluve et  al. 2017). Indeed, as experiment evalua- the propensity score matching method. It contributes tions are rare in developing countries (Mckenzie 2017), to the available literature on subsidized employment in the few existing evaluations use mostly quasi-experimen- four ways. First, as counterfactual evaluations outside tal methods (Galasso and Ravallion 2004; Broecke 2013; the OECD are quite limited, it provides evidence on the Souag 2020). Thirdly, most authors have concluded that experience of developing countries. Second, while the the impact of these subsidies was limited (Escudero et al. available literature focuses more on schemes acting on 2019; Kluve et al. 2019). Betcherman et al. (2004) report, either the demand or the supply side of the labour market for example, that of the 159 evaluation studies reviewed, and targeting low-skilled individuals, this paper exam- more than two-thirds had a negative effect on employ - ines one particular program that affects the both sides of ment and only 5 studies had a positive effect on the the labour market and targets, as many countries in the income of beneficiaries. MENA region (Groh et al. 2012, 2014; Broecke 2013; Pre- The available literature has shed light on several collat - mand et  al. 2016; Souag 2020), higher degree graduates. eral effects of wage subsidies that can explain their lim - Third, it goes beyond the program’s direct impacts on ited effectiveness. Their potential stigmatizing effect has, employment to capture its effects on beneficiaries’ wages indeed, highlighted by the first experimental evaluations and their working conditions. Finally, it highlights the in the United States. Burtless (1985) has shown that these heterogeneous effects of the program. subsidies have been used by American employers as a The rest of the paper is organized as follows: Sect. “2” discrimination tool against beneficiaries, who were less presents a review of the literature. Section “3” presents likely to find a job than non-beneficiaries. Similarly, other the "Idmaj" program. Propensity score matching esti- authors have shown that beneficiaries have concealed mation methodology and empirical implementation are their eligible status because they considered it degrad- discussed in Sect. “4”. Section “5” provides details on the ing and stigmatizing (Woodbury and Spiegelman 1987; However, there are some recent signs that things are changing. For instance, Such evidence is also rare even for programs developed by civil society or in Morocco, the first and, to our knowledge, the only experimentation related international organizations, which are with limited scope and targeted gener- to the labour market is Bausch et  al. (2017). Experiments recently developed ally narrow local individuals. In this respect, the paper of Bausch et al. (2017) in the other countries of the MENA region (Groh et al. 2012; Groh et al. 2014; is an exception. This paper has examined the impact of a skills training pro - Premand et al. 2016) have also not focused on the mechanism discussed here, gram, called "100 Hours to Success", on the financial behaviour, employabil - namely wage subsidies. For the latter, the experiment of Galasso et al. (2004) ity and educational choices of the rural young living in Morocco’s Oriental on the argentine program “Proempleo” is one of the rare exceptions in devel- Region. oping countries. Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 3 of 13 17 Dubin and Rivers 1999).This stigma effect has also been the match between the profile of job seekers and the verified in other countries, both developed (Bell et  al. needs of the market, through contractualized training, 1999) and developing (Kluve et al. 1999). skills training or retraining. The second program, called Likewise, several studies have shown that these subsi- "Moukawalati" and renamed "self-employment", aims to dies are likely to produce some spillover effects that reduce promote entrepreneurship and business creation through their effectiveness (Martin and Grubb 2001; Kluve et  al. training and financial assistance. The third program, the 2017; Escudero et  al. 2019). When it comes to the wind- subject of this paper, is a subsidized employment pro- fall effect, it has been shown that wage subsidies mainly gram which supports the transition of youth from school help those who, even in the absence of the program, would to work through company placements. have found a job in both developed (Katz 1996; Bell et al. This program, called "Idmaj" or "insertion", has been 1999; Crépon et  al. 2013) and developing countries (Gal- designed to reach a double objective. Firstly, the improve- asso et  al. 2004; Broecke 2013). The substitution and dis - ment of the employability by offering to the targeted placement effects have been also verified by several works unemployed a professional experience through 24-month (Dahlberg and Forslund 2005; Crépon et al. 2013; Moczall internships. Secondly, the increase of the share of high- 2014). The size of these effects depends on the methods skilled and of qualified employees by giving incentives used. It seems that it is larger in econometric studies (60– for firms to hire the young graduates. Indeed, it is an 70%) than survey studies (15–40%) (Calmfors et al. 2002). employment subsidy targeting higher education gradu- Furthermore, many studies have focused on the perma- ates, as the individuals who are eligible for it are all job nence of the effects of wage subsidies. However, except seekers registered with ANAPEC and holders of a higher few papers that have validated the stepping stone effect to (tertiary) education diploma, vocational training diploma stable, better-paying jobs (Pallais 2014), the general trend or a baccalaureate at least. of these studies shows that the positive impact of these Two types of contracts are offered under the "Idmaj" subsidies declines over time (Hujer et al. 2004; Venetoklis program: the insertion contract (IC) and the professional 2004; Card and Hyslop’s, 2005; Kvasnicka 2009; David and integration contract (PIC). Both are a tripartite contract Houseman 2010; Groh et al. 2016; De Mel et al. 2016). between (i) a company interested in hiring an eligible job Finally, the existing literature has highlighted the het- seeker (ii) an eligible job seeker and (iii) the ANAPEC, erogeneous effects of wage subsidies by subgroups of par - which is responsible for validating and following up these ticipants. For example, Galasso et al. (2004) concluded that contracts. the "Proempleo" program was significant only for women, The IC, put in place from the start of the program, pro - youth, and highly educated workers. The positive impact vides to private companies involved in the program total on women is corroborated by many studies (Bergemann exemption from (i) Income Tax, ii) Social Charges and and Van den Berg 2008; Groh et al. 2016).Other works have (iii) Tax on Vocational Training for the internship allow- found heterogeneous effects depending on both gender and ance paid to eligible individuals up to a ceiling of 6000 areas. This is the case for example of that of Caliendo et al. MAD (Moroccan Dirham), for a period of 24  months. (2008), for the German scheme and that of Souag (2020) Two other incentive measures were added in 2016. This for the Algerian one. By sectors of activity, Stephan (2014) concerns, on the one hand, the payment by the Govern- detected a little heterogeneity of the German program. ment of employer’s and employee’s contributions for basic compulsory health insurance during the intern- 3 Description of the "Idmaj" program ship period and, on the other hand, the extension of the In 2000, the active labour market policy in Morocco aforementioned tax and social exemptions for an addi- took a qualitative step forward with the creation of the tional year, if the trainee is integrated under permanent National Agency for the Promotion of Employment employment (PE) contract. and Competencies (ANAPEC ). This public agency is For its part, the PIC, which was put in place in 2011, designed to (i) implement active labour market policies completes the IC by providing an integration training (ALMPs) (ii) to connect employers to job seeker, and (iii) subsidy (bonus) amounting to 25,000 MAD to companies to provide information and guidance to job seekers and recruiting eligible graduates under (PE) contract at the young entrepreneurs. Not far, three main programs were end of a 6 to 9 months integration training course as part set up in 2006. The first, called "Taehil ", aims to improve of the system of the ICs. However, it is worth noting that the PIC is reserved for specific job seekers who have great difficulties finding their first job, namely young graduates of some general higher education courses (baccalaure- Abbreviation used in Morocco in reference to the initials of the name of the ate + 3 and more) enrolled at ANAPEC for at least 1 year. agency in French: Agence Nationale de Promotion de l’Emploi et des Com- pétences. 17 Page 4 of 13 A. Chatri et al. That being said, the "Idmaj" program remains the cen - The hypothesis in (3), known as conditional independ - tral component of ALMP in Morocco. Actually, by the ence (Lechner 1999) or selection on observables (Heck- end of its start-up year (2006), the number of beneficiar - man et  al. 1985), assumes that there is no difference ies was 32,880 young graduates, before it rises steadily between treated and untreated groups, conditionally to the to over 102,773 in 2018. This increased number of ben - observables (noted X). The hypothesis in (4) implies the eficiaries merely reflects the budgetary cost of the pro - existence of an overlap (common support) between these gram and cannot be understood as a sign of effectiveness, two groups. Rosenbaum and Rubin (1983) call the com- which involves identifying the changes observed in ben- bination (3) and (4) the "strong ignorability" hypothesis, eficiaries that are attributable exclusively to the program. which is the identifying hypothesis of ATT, since it verifies: Such is the purpose of this paper. 0 0 0 E Y |X, D = 1 = E Y |X, D = 0 = E Y |X 4 Methodology and empirical implementation (4) This paper uses the standard econometric evaluation However, when the number of X is large, a “Curse of model, Roy Rubin’s model of potential outcomes (Roy dimensionality” (Heckman et  al. 1997) could affect the 1951; Rubin 1974), to determine the causal effect of the width of the common support. To overcome this prob- "Idmaj" program on our outcome variables. In particular, lem, Rosenbaum and Rubin (1983) proposed to match it seeks to estimate the Average Treatment on Treated treated and untreated individuals not on the basis of X (ATT) parameter, which is given by: but based on the propensity score p (X), which defined as the probability of receiving the treatment giving X. Its 1 0 ATT = E Y |D = 1 − E Y |D = 1 (1) balancing property ensures that treated and untreated persons with the same value p (X) also have the same with D ∈ {0, 1} a binary variable that indicates the treat- distribution of X. These authors have also illustrated that ment status. It takes 1 in case of treatment and 0 other- while treatment assignment is strongly ignorable given X, 1 0 wise. Y is the outcome in case of treatment, and Y the it is also ignorable given p (X). Therefore, we have: outcome in case of non-treatment. 1 0 As E Y |D = 1 is not observable, and taking into Pr X|D = 1, p(X) (Y , Y )⊥D|p(X) (5) account the absence of an explicit rule of assignment to the "Idmaj" program and that there is no instrument or 0 < Pr(D = 1|p(X)) < 1 (6) discontinuity in treatment to exploit, we use in this paper the matching method, which permits us to identify our Once these conditions are satisfied, all biases due to the ATT under the following assumptions : observables may be eliminated, allowing us to estimate the ATT according to the following general formula : 1 0 (Y , Y )⊥D|X (2)   � � T C   ATT : Y − ω(i, j)Y (7) 0 < Pr(D = 1 X)< 1 i j (3) i∈T j∈C with: N The number of treated, Y : The observed out - come of the matched treated, Y : The observed outcome 6 of the matched untreated and ω i, j is the weight used to Certainly, the observable average outcome of the untreated could be used as a credible counterfactual for this non-observable outcome in randomized aggregate outcomes of the matched untreated. experiments, nevertheless in the presence of non-experimental data, as in this Estimating the ATT expressed in (7) above involves paper, this would lead to a selection bias. For this purpose, different estimat - first the prediction of the propensity scores with any dis - ing strategies are developed (Heckman et  al. 1999) to avoid this bias, which comes from both observable and unobservable factors. The feasibility of each crete choice model can be used. In this paper, we use the of these strategies is based on some assumptions, which in turn depend on the "Probit" model. The second step of our strategy consists nature of the program and its context. For example, the regression disconti- nuity (Imbens and Lemieux 2008) is suitable for programs using an index to decide who is eligible to enrol in the program and who is not. The instrumen - tal variables method (Angrist et al. 1996) which relies on some external source of variation to determine treatment status is suitable mainly for programs It is to highlight that there is another method, which is based directly on with imperfect compliance, or universal coverage. Unlike these two non- the propensity score. This is the inverse probability of treatment weighting experimental methods which produce estimates of the counterfactual through (IPTW) popularized by Hirano et al. (2003). explicit program assignment rules, other methods, such as the difference-in- differences and the matching methods, can be used for programs with vol - It is worth noting that both this model and "logit" model allow consist- untary enrolment, when the program assignment rules are less clear or when ent estimation of the conditional probability of participation in the program none of the former methods is feasible. in the interval [0–1]. These two models generally yield similar results: esti- mates from a "logit" represent π⁄√3 of those obtained with a "probit". Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 5 of 13 17 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 x Propensity Score kdensity Treatedkdensity Untreated Untreated Treated Fig. 1 Propensity score distribution and common support region of the choice of an appropriate matching estimator, being optimize the tradeoff between variance and bias (Garrido underlined that many popular matching estimators are et al. 2014; Heckman et al. 1997). suggested in the literature (nearest neighbor, caliper, ker- That being said, using this matching for estimating our nel and stratification…). Although they share the same ATT depends on its ability to verify both common sup- Eq.  (7) for calculating ATT, these estimators differ in port region and balancing properties. Figure 1 shows the both the definition of the neighbourhood for the treated existence of a sizeable overlap of [0.2–0.8]. Likewise, our individuals and in the way that weights are assigned to tests, in Table  2, confirm that the balancing property is those neighbors (Stuart 2010). When it comes to choos- also satisfied. ing the best estimator, the available literature offers very As a final step in our methodological approach, we little guidance (Huber et  al. 2013) and argues that there conduct a sensitivity analysis of the results obtained. As is no "winning" estimator (Caliendo and Kopeinig 2008), noted previously, the propensity score matching method so their performance depends above all on the data set relies on conditional independence, a strong assumption (Zhao 2003). Furthermore, it asserts that the choice of which assumes that program uptake is based entirely on the matching estimator requires evaluating tradeoffs observed characteristics (Khandker et al. 2009). So if any between efficiency and bias (Garrido et al. 2014). unobserved factors exist and affect participation status Therefore, the kernel matching algorithm developed by and the variable of interest, then a hidden bias is involved Heckman et al. (1997) seems to be the most appropriate (Rosenbaum 2002). In this respect, and after estimat- in the context of this paper due to the smallness of our ing the ATTs, we will check their sensitivity to devia- sample. This matching approach maximizes the estimates tion from the identification assumption, using the MH ‘efficiency and precision using more observations. Con - bounds program dedicated to binary variables accord- trary to other algorithms that use one (nearest neigh- ing to the bounding approach proposed by Robensuam bour) or at best a few observations from the comparison (2002) and explained in the paper of Becker and Caliendo group (calliper matching), Kernel matching estimator (2007). This allows highlighting the degree by which an constructs the counterfactual by using for each treated unmeasured variable influencing the odds of participa - unit, all the untreated units weighted by the distance that tion in "Idmaj" program will also modify its effects on our separates each of them from the treated (Morgan and outcome variables. Harding 2006). Thereby, weights are assigned so that less similar unites receive less weight when estimating ATT, 5 Database and vice versa (Handouyahia et al. 2013). The data used here are taken from a random survey con - This matching requires, moreover, the choice of a ker - ducted in 2010 by the Ministry of Labour in coordination nel function, and more importantly, the choice of a band- with the ANAPEC, as part of the strengthening of obser- width that controls the degree of smoothing applied vation and analysis capacities of the Moroccan labour to the data. In this paper, we used the Gaussian kernel market. Its objective was to provide an accurate descrip- function and the default bandwidth of 0.06 which may tion of the socio-economic situation of the interviewees 0 1 2 3 17 Page 6 of 13 A. Chatri et al. and appreciate, in a descriptive way, the effect of the only became eligible for the program, using this year as "Idmaj" program on their professional trajectories. While the baseline seems more reasonable. it was not completely designed for impact evaluation We also excluded variables that do not jointly cover purposes, this survey fulfills three characteristics that, beneficiaries and non-beneficiaries. Finally, for the out - according to the available literature, support the estima- lier and the missing values, and due to the small size of tion of the causal effect by the matching method (Heck - our sample, we created an additional modality "non- man et al. 1998). First, it uses the same data source and, response" for qualitative variables with too many missing therefore, measures the observed characteristics in the observations. Likewise, and considering the bias that it same way for treated and untreated groups. Second, it is can produce, the imputation by the mean has only been specific only to eligible individuals, which may alleviate used for the three variables which suffer from few miss - the problem of determining a «common reference date» ing data or outliers. These variables are: "Reservation for treated and untreated individuals (Lechner 1999). wage", "Mother’s level of education", and "Father’s level of Third, it contains a rich set of covariates, which sustains education". the plausibility of the conditional independence hypoth- All these choices have affected the size of our final sam - esis discussed above. ple, which ended up with 621 individuals, including 377 This survey focused on two samples constituted by sys - treated and 244 untreated. tematic random sampling with equal probability from Regarding the covariates included in our estimate, (i) a nominative list of beneficiaries of "Idmaj" insertion they were chosen on the basis of a set of methodological contracts in 2006 and 2007, and (ii) a list of job seekers considerations. First, the identification of ATT using the enrolled in ANAPEC, eligible for the program but never score propensity matching method implies, on the one having benefited from it since their enrollment until the hand, the inclusion of all covariates that influence simul - day of the survey. The retrospective questions of this taneously treatment and outcome variables and, on the survey aimed to collect information on activity situa- other hand, the exclusion of covariates affected by treat - tions of interviewees during the following four phases: (i) ment (Heckman et  al. 1998). For this reason, and since before registering with ANAPEC, (ii) between registra- our data are post-program, we have excluded variables tion and signature of the IC, (iii) during the IC period and that are not fixed overtime at the risk of being affected (iv) the post IC period. by treatment, yet the parents’ profession and their level of After the sampling, the survey was conducted on a education will be kept, both because of their importance sample of 2500 beneficiaries and 500 individuals for the in explaining participation in the program, as well as in control group. This paper uses a data of random subsam - labour market insertion (social capital) and because they ple of a total size of 879 eligible individuals for the pro- are difficult to change over time. Secondly, to choose the gram, which 627 beneficiaries in 2006 or 2007 and 252 most relevant variables, we have proceeded to an iterative non-beneficiaries. selection of our variables according to a process guided However, some choices seemed essential to improve by three references: (i) the significance of the statistical the coherence and quality of the data. u Th s, we chose, in tests, (ii) the quality of the matching, which involves veri- this paper, only one reference year for the program. This fying the existence of the common support and the bal- choice was necessary to construct some variables on the ancing condition between treated and untreated, and (iii) same basis and compare the two groups at similar time the empirical literature suggestions, according to which horizons. It was also more appropriate, as we had no way one should integrating variables that capture the four to decompose the control group, which is half the size following dimensions: eligibility criteria, socio-demo- of the treatment group, to evaluate both years simulta- graphic characteristics, qualifications and labour market neously. In addition, while data indicate that 21% of the history (Heckman et al. 1999). untreated were enrolled in ANAPEC until 2007 and thus Therefore, our final specification includes 23 covariates belonging to the four above-mentioned dimensions: eligibility criteria (diploma and years of 9 enrolment in ANAPEC), socio-demographic charac- The survey contains another sample of companies whose results are not used in this work. teristics (program entry age, gender, health problems, Not having been able to access the global sample, this sub-sample made available to us by the Ministry of Labour was, at our request, drawn ran- domly. The choice of 2006, instead, will reduce the size of the control group, which Since data are post-program, it was necessary to retain some variables at is already limited. program entry to predict propensity scores. For example, the calculation of age and seniority of enrollment in ANAPEC at the start of the program was The excluded variables are: marital status, number of children, number feasible for treated individuals, but for those untreated will require the set- of non-resourceful persons in the household, type of housing, region, and ting of a single reference year to allow their matching with the treated. household size. Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 7 of 13 17 Table 1 Variables construction Variable Description Years of enrolment in ANAPEC In years Program entry Age In years Health problems 1: has a health problem, 0 otherwise Reservation wage En MAD Internship 1: trainee before enrollment at ANAPEC; 0 otherwise Diploma Perception 1: positive perception, 0 otherwise Training days Number of training days from graduation to the enrollment at ANAPEC Gender 1: man; 0 woman Mother’s job 1: in employment or retirement, 0 otherwise Father’s job 1: in employment or retirement, 0 otherwise Extracurricular activity 1 if yes, 0 otherwise Mother’s level of education In Years Father’s level of education In Years Studies 1 in education before the enrollment at ANAPEC, 0 otherwise Employment 1 in employment before the enrollment at ANAPEC 0 otherwise Associative activities 1 if yes, 0 otherwise Duration of unemployment In days (from graduation to the enrollment at ANAPEC) Education sector 1: public, 0 otherwise Unemployment 1: unemployed before enrollment at ANAPEC, 0 otherwise Diploma In years Duration of other inactivity Number of days spent in inactivity (other than internship and studies) Internship days Number of internship days from graduation to the enrollment at ANAPEC Other forms of inactivity 1: inactive before ANAPEC enrollment, 0 otherwise mother’s and father’s level of education and their jobs), 6 Results and discussion qualifications (last individual’s professional situation As mentioned above, the aim of this paper is to assess the before enrolling in ANAPEC: employment, unemploy- effectiveness of the "Idmaj" program not only with regard ment, internship, study or other types of inactivity), to its direct expected outcomes, which is improving the and labour market history (duration of studies, unem- employability of its beneficiaries, but also with regard to ployment, internship and other types of inactivity the quality of the job held. Therefore, we have estimated experienced by the individuals from the time of their the ATT of wage subsidies granted under the "Idmaj" graduation until their registration with ANAPEC). We program on five types of outcomes: (i) status with regard have, moreover, integrated the following variables that to unemployment, (ii) status with regard to employment, are likely to capture some forms of unobservable indi- (iii) the level of salary, (iv) the benefit or not of social vidual heterogeneity: diploma perception, reservation cover, and (v) the number of hours worked per week. wage, participation in extracurricular and associative These outcome variables are all binary, apart from that activities. of the number of hours worked per week and the earned Tables 1 and 2, present respectively, our variables and wage. The latter is unconditional on wage employment. the tests of difference in means. These tests show in uTh s, the employment (respectively unemployment) particular that, before matching, treated and untreated result variable takes the value 1 if the individual finds a groups are statistically different only in terms of four job (does not find a job) and 0 otherwise. It should also variables, which are age, years of enrolment in ANA- be noted that unemployment outcome variable is condi- PEC, reservation wage, and health problems. After tional on labour market participation (we have taken only matching, these tests confirm, as mentioned above, the unemployed persons and excluded the inactive ones). that our specification satisfies the balancing property. Table  4 contains the results of our estimates and pro- For its part, Table  3 reports the results of the propen- vides ATTs for the whole sample and by subgroups con- sity score estimation. structed by gender and age. 17 Page 8 of 13 A. Chatri et al. Table 2 Balancing covariates test Variable Mean (1) Mean (2) t-test (1) t-test (2) Treated Control Treated Control t p > t t p > t Years of enrolment in ANAPEC 1.932 2.617 1.935 1.982 − 6.130 0.000 − 0.550 0.582 Program entry Age 25.395 26.633 25.344 25.263 − 3.640 0.000 0.290 0.775 Health problems 0.046 0.013 0.046 0.026 2.270 0.023 1.430 0.153 Reservation wage 3152.400 2979.900 3148.800 3231.900 2.220 0.027 − 1.010 0.312 Internship 0.078 0.046 0.079 0.086 1.590 0.113 − 0.350 0.725 Diploma Perception 0.676 0.625 0.678 0.683 1.290 0.199 − 0.150 0.878 Training days 135.150 97.842 117.700 108.870 1.150 0.250 0.410 0.685 Gender 0.551 0.504 0.550 0.530 1.140 0.255 0.540 0.591 Mother’s job 0.1 0.079 0.100 0.121 0.870 0.385 − 0.890 0.374 Father’s job 0.873 0.842 0.873 0.886 1.090 0.276 − 0.540 0.590 Extracurricular activity 0.349 0.321 0.347 0.332 0.710 0.479 0.420 0.673 Mother’s level of education 2.970 2.722 2.978 2.960 0.740 0.461 0.060 0.954 Father’s level of education 6.745 6.480 6.760 6.681 0.690 0.493 0.230 0.816 Studies 0.105 0.121 0.106 0.096 − 0.590 0.555 0.430 0.667 Employment 0.116 0.1 0.114 0.114 0.620 0.532 − 0.020 0.987 Associative activities 0.068 0.058 0.068 0.075 0.450 0.649 − 0.390 0.694 Duration of unemployment 318.090 295.090 308.390 287.810 0.470 0.640 0.500 0.619 Education sector 0.114 0.129 0.114 0.131 − 0.580 0.561 − 0.720 0.473 Unemployment 0.530 0.512 0.531 0.533 0.420 0.678 − 0.050 0.963 Diploma 14.130 14.133 14.136 14.143 − 0.040 0.968 − 0.090 0.925 Duration of other inactivity 24.803 25.512 24.870 23.289 − 0.070 0.945 0.190 0.851 Internship days 33.151 32.362 32.252 29.357 0.090 0.928 0.400 0.686 Other forms of inactivity 0.059 0.058 0.060 0.073 0.060 0.954 − 0.740 0.462 Bold means that the variables are significant at 1% , 5% or 10%  Mean (1) before matching Mean (2) after matching Overall, our results show that the "Idmaj" program Furthermore, our results show that the "Idmaj" program reduced the probability of being unemployed by 7.5% provides some heterogeneous effects. The impact on the reduction of unemployment is only observed for women points and increased the chances of finding a job by 8.2 and young people under 24  years with 9.5 percentage percentage points. However, this positive effect on ben - points and 12 percentage points respectively. Similarly, eficiaries is only significant at the 10% level. On the other only women have an improvement in their probability of hand, the program had a negative and statistically signifi - finding a job (+ 11 percentage points). This positive impact cant effect (at 1%) on wages of beneficiaries, which were on women is corroborated by the literature (Bergemann lower by 487 MAD than those of non-beneficiaries. These and Van den Berg 2008). Concerning the negative effect on results are consistent with those of Martin and Grubb wages, it was once more observed only among women, for (2001), who suggest that subsidies may produce positive effects but at the expense of wages. These findings may whose the participation in the program caused them a loss also reflect the stigmatizing effect of the program, and of 697 MAD compared to non-beneficiaries. its use by the employers as a filtering tool, leading them While the global results do not appear a statistically sig to consider the participants to be unproductive and have nificant effect of the program on the working conditions, to be paid lower wages than those otherwise hired. This the disaggregated analysis did reveal some valuable find - negative impact on salaries reflects the ineffectiveness of ings. For example, men were less likely to be covered by a the 24-month internship, which does not seem to have social security system than non-treated (− 13 percentage improved the skills and the productivity of the beneficiar - points). This probability also decreases by (− 11 percent - ies and, consequently, their chances to find good-paying age points) among young aged 25–34  years. In addition, and quality jobs. These conclusions imply that the "Idmaj" the latter beneficiaries had to work four hours longer program does not have the expected stepping stone effect. than the non-participants. This did not seem, moreover, Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 9 of 13 17 Table 3 Estimation of propensity scores (Probit Model: Marginal effects at the mean) variable dy/dx Std.Err Z P > z [95%Conf Interval] X Years of enrolment in ANAPEC − 0.071 0.017 − 4.240 0.000 − 0.104 − 0.038 2.202 Program entry Age − 0.020 0.007 − 3.030 0.002 − 0.033 − 0.007 25.882 Health problems 0.281 0.072 3.920 0.000 0.141 0.422 0.033 Reservation wage 0.000 0.000 1.590 0.113 − 0.000 0.000 3084.58 Internship 0.215 0.079 2.730 0.006 0.061 0.369 0.066 Diploma Perception 0.063 0.046 1.380 0.166 − 0.026 0.153 0.656 Training days 0.000 0.000 1.410 0.159 − 0.000 0.000 120.47 Gender 0.060 0.043 1.380 0.168 − 0.025 0.144 0.533 Mother’s job 0.054 0.078 0.700 0.487 − 0.099 0.208 0.092 Father’s job 0.067 0.062 1.090 0.277 − 0.054 0.188 0.861 Extracurricular activity 0.022 0.046 0.480 0.632 − 0.068 0.113 0.338 Mother’s level of education − 0.005 0.007 − 0.650 0.518 − 0.018 0.009 2.872 Father’s level of education 0.002 0.006 0.320 0.745 − 0.009 0.013 6.640 Studies 0.021 0.091 0.230 0.821 − 0.157 0.198 0.111 Employment 0.152 0.077 1.990 0.047 0.002 0.303 0.110 Associative activities − 0.045 0.090 − 0.500 0.614 − 0.221 0.131 0.064 Duration of unemployment 0.000 0.000 1.450 0.148 − 0.000 0.000 309.038 Education sector − 0.081 0.069 − 1.170 0.240 − 0.217 0.054 0.120 Unemployment 0.126 0.067 1.880 0.060 − 0.005 0.257 0.523 Diploma 0.025 0.022 1.160 0.247 − 0.017 0.067 14.131 Duration of other inactivity 0.000 0.000 0.530 0.597 − 0.000 0.001 25.082 Internship days − 0.000 0.000 − 0.660 0.509 − 0.001 0.000 32.841 Other forms of inactivity 0.099 0.099 1.000 0.318 − 0.095 0.293 0.059 Bold means that the variables are significant at 1% , 5% or 10% to be compensated by an increase in their chances of that the impact of wage subsidies are more beneficial finding a job compared to non-beneficiaries. This result when accompanied by complementary services (Kluve indicates that the stigmatizing effect of the "Idmaj" pro - et al. 2019). gram is particularly significant among older youth, who Other conditionalities may also be useful to ensure that must be long-term unemployed or with higher degrees, the program can produce lasting effects beyond the expi - for which the statistics are already alarming. In Morocco, ration of subsidies, such as the requirement to keep sub- 70% of the unemployed are long-term unemployed, and sidized employees for a specified period. The tendency of unemployment rates for holders of higher degrees are employers to pay low wages to the beneficiaries can, on almost five times higher than for those with no degree. the other hand, be attenuated by more meaningful differ - Our results suggest revisiting the program’s parameters entiation of the wage ceiling, which serves as the basis for and making certain adjustments to increase its effective - tax and social security exemption. ness. In this respect, particular attention should be given Finally, our results support a more refined targeting of to the professional internship to make it a real lever for women and youth for whom the program has proven to productivity gains, since only 2% of the beneficiaries in be more effective. Indeed, there is ample evidence that our sample report having received coaching from the careful targeting avoids the loss of funds and allows for host firm during the internship. It is, therefore, necessary rigorous monitoring of results, and that the effectiveness to supervise this internship with more rigorous specifica - of subsidies is better when targeted to a specific group tions aimed at ensuring real supervision of the trainees (Almeida et al. 2014). and the improvement of their skills. It must be accompa- Before generalizing these findings, we conducted a nied by a reduction in its duration, since 24  months of sensitivity test to check the quality of our estimates. internship is a relatively long period that can produce While the raised impacts of the "Idmaj" program are counterproductive effects. It would also be appropriate to significant (at the 10%) on unemployment (respectively empower beneficiaries by requiring them to participate employment), the sensitivity test will determine at what in employability training. Indeed, it is widely recognized degree of presence of unobserved selection (Gamma) 17 Page 10 of 13 A. Chatri et al. Table 4 The ATT estimates by Kernel Matching Table 5 Sensitivity test for unobserved heterogeneity Outcome n. treated n. control ATT T Gamma Unemployment Employment Q_mh− P_mh− Q_mh + P_mh + The whole sample Unemployment 370 235 − 0.075* − 1.832 1 1.49 0.067 1.54 0.060 Employment 370 235 0.082* 1.804 1.1 1.00 0.157 1.02 0.153 Wage 269 155 − 487.153*** − 2.480 1.3 0.154 0.438 0.097 0.461 Social protection 269 141 − 0.055 − 1.564 1.5 0.380 0.351 0.512 0.304 Worked hours per 269 155 1.350 1.138 Bold means that the variables are significant at 1% , 5% or 10% week Gamma: odds of differential assignment due to unobserved factors Women Q_mh+ : Mantel–Haenszel statistic (assumption: overestimation of treatment Unemployment 166 115 − 0.095* − 1.892 effect) Employment 166 115 0.112* 1.714 Q_mh−: Mantel–Haenszel statistic (assumption: underestimation of treatment effect) Wage 118 70 − 697.783* − 1.955 p_mh+ : significance level (assumption: overestimation of treatment effect) Social protection 118 68 0.013 0.163 p_mh−: significance level (assumption: underestimation of treatment effect) Worked hours per 118 70 1.044 0.931 week Men suggests a cautious interpretation of the obtained results Unemployment 204 116 − 0.037 − 0.666 when both effects on unemployment and employment Employment 204 116 0.054 0.981 would no longer be significant if an unobserved vari - Wage 151 82 − 345.330 − 1.234 able increases the chances of program participation Social protection 151 71 − 0.135** − 4.302 by 10% for two individuals with the same observable Worked hours per 151 82 2.440 1.638 characteristics. week Youth [18–24] Unemployment 170 71 − 0.120* − 1.746 7 Conclusion, policy implications and limitations Employment 170 71 0.100 1.290 This paper sought to assess the effectiveness of the sub - Wage 124 41 − 408.998 − 1.131 sidized employment program "Idmaj" not only in terms Social protection 124 36 − 0.028 − 0.401 of reducing unemployment and promoting employment, Worked hours per 124 41 0.433 0.242 but also in terms of beneficiaries’ wages, working hours, week and social security status. This evaluation is conducted Youth[25–34] using the propensity score-matching method based on Unemployment 190 142 − 0.042 − 0.645 the Ministry of Labor’s survey of a sample of eligible Employment 190 142 0.039 0.676 individuals for the "Idmaj" program. Our results have sig- Wage 138 101 − 416.596 − 1.371 nificantly improved our knowledge of the effectiveness Social protection 138 93 − 0.115** − 2.462 of the program; especially since none of Morocco’s pub- Worked hours per 138 101 4.266** 2.363 lic ALMP schemes has been subject to a counterfactual week impact evaluation.  t statistic; * p < 0.10, ** p < 0.05, *** p < 0.01 Our results reveal that the marginally positive effects of the "Idmaj" program on reducing unemployment and increasing employment were accompanied by some sig- these effects will become insignificant. Since the impact nificant collateral effects on beneficiaries. In particu - is negative on unemployment (positive on employment), lar, the program seems to have produced a stigmatizing the focus will be on the negative unobserved selec- effect. This were manifested through the negative and tion Q_mh- (positive unobserved selection Q_mh + for statistically significant impact on beneficiaries’ wages, employment). The results in Table  5 affirm that in the and, for some categories (notably young aged 25 to absence of unobserved selection (Gamma = 1), the 34  years), on their working conditions as an increase in "Idmaj" scheme does have a significant effect at the working hours and exclusion from social security cover- 10% level (p_mh = 0.06) on reducing unemployment/ age. Our findings also reflect the ineffectiveness of the increasing employment. Then, the Q_mh− /Q_mh+ sta- 24-month internship provided by the program on the tistic reveals that at relatively low levels of presence of beneficiaries’ skills and productivity and their chances of unobserved selection (Gamma = 1.1), these effects will finding gainful employment. As a result, the program did be non-significant (p_mh− = 0.157 for unemployment not serve as a stepping stone. and p_mh+ = 0.153 for employment). This analysis Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 11 of 13 17 Moreover, our results highlight some heterogeneous important limitation of this paper lies in the narrowness effects of the program. They showed a positive impact on of our sample, in particular, that of untreated individuals the professional insertion of the most-excluded individu- which is, and contrary to the usual evaluation conditions, als in the Moroccan labour market, particularly women narrower than that of treated individuals. and young people aged between 18 and 24  years. How- ever, the positive effect for women is obviated by their Abbreviations lower wages than they would have earned without the ALMP: Active Labour Market Policy; ANAPEC: Agence Nationale de Promo- program. As for young people aged 25–34, they have to tion de l’Emploi et des Compétences; ATT : Average treatment on treated; IC: Insertion contract; MAD: Moroccan dirham; PE: Permanent contract; PIC: work more and with less social security when they ben- Professional integration contract. efit from the program. These mixed and small effects suggest revisiting the Acknowledgements The authors thank the editor and the two anonymous reviewers for their valu- "Idmaj" program to improve its effectiveness. Refining able comments on a previous version of this manuscript. its targeting, rethinking its conditionalities and orienting them towards the sustainability of its impacts, strength- Authors’ contributions CA is the major contributor in writing and correcting the manuscript. The ening its complementarity with other levers of the ALMP, three authors decided the method used in the study and assisted in its appli- and enhancing its flexibility are all adjustments that seem cation. All authors read and approved the final manuscript. necessary to attenuate the negative effects of the program Authors’ information in question and increase its net benefit. Indeed, it would, CA: Professor-researcher at the University Mohammed V of Rabat. His research for example, be more appropriate to replace the current interests currently focuses on the evaluation of the impact of active labor single salary ceiling serving as the basis for tax and social market programs in developing countries. exemptions by differentiated limits to reduce the employ - HK: Laureate of the Master’s Degree in Economics and Evaluation of Public ers’ bias towards low wages. Likewise, for enhancing the Policies, Mohammed V University of Rabat. 24-month professional internship impact on productivity, SN: PhD student in economics, Mohammed V University of Rabat. Research clear specifications and regular monitoring and evalua - field: Employment economics. tions should be put in place. The flexibility of this intern - ship should also be increased, by allowing, for example, Funding Not applicable. the employee to change his initial employer more than once if this would prove useful for improving his employ- Availability of data and materials ability. More broadly, and as this wage subsidies program The data used in the present study are available from the Moroccan ministry of employment, direction of the national observatory of employment. But is used in Morocco not to address cyclical unemploy- restrictions apply to the availability of these data, which were used at our ment, but rather the structural one, employment-friendly request for this study. Data are however available from the authors upon rea- growth and structural reforms of the labour market will sonable request and with permission of direction of the national observatory of employment. be required to increase the employability of young gradu- ates sustainably. Declarations Finally, it should be mentioned that this paper is not without limitations and shortcomings. Indeed, as noted Ethics approval and consent to participate above, the design of the program whereby treatment Not applicable. exposure is not determined by some external selection Consent for publication criteria, such as a random allocation or an exogenous eli- Not applicable (the paper has used anonymized data, so no consent was gibility threshold, involved the use of the propensity score obtained). matching method based on the strong assumption of Competing interests selection on observables. That said the scope of our find - The authors declare that they have no competing interests. ings would have been much more robust and interest- Received: 10 December 2020 Accepted: 13 June 2021 ing if relevant pre-treatment data were available for our two groups. Indeed, the post-treatment survey led us to exclude certain co-variables that are not fixed over time but potentially crucial for participation and outcomes. References This has reduced the heterogeneity on the observables as Almeida, R., Orr, L., Robalino, D.: Wage subsidies in developing countries as long as program participation has been modelled based atool to build human capital: design and implementation issues. IZA J. Labor Policy 3, 1–24 (2014) on a few statistically significant co-variables. These ele - Angrist, J.D., Imbens, G.W., Rubin, D.: B: Identification of causal effects using ments could explain the relative weakness of the statisti- instrumental variables. J. Am. Stat. Assoc. 91(434), 444–455 (1996) cal significance of our results and their relative sensitivity Au Plan, H. C., & Mondiale, B: Le marché du travail au Maroc: défis et opportu- nités (2017). concerning unobservable factors. Another and no less 17 Page 12 of 13 A. Chatri et al. Auer, P., Efendioglu, L., Leschke, J.: Active labour market policies around the Garrido, M.M., Kelley, A.S., Paris, J., Roza, K., Meier, D.E., Morrison, R.S., Aldridge, world: Coping with the consequences of globalization, 2nd edn. Interna- M.: D: Methods for constructing and assessing propensity scores. Health tional Labour Office, Geneva (2005) Serv. Res. 49(5), 1701–1720 (2014) Banerji, A., Newhouse, D., Pacci, P., Robalino, D.: Working through the crisis: jobs Gautié, J. Le chômage. La découverte (2015). and policies in developing countries during the great recession. World Grimm, M., Paffhausen, A.: Interventions for employment creation in micro, Bank, Washington DC (2014) small and medium sized enterprises in low and middle income coun- Bausch, J., Dyer, P., Gardiner, D., Kluve, J., Kovacevic, S.: the impact of skills train- tries–a systematic review. University of Passau, Passau (2014) ing on the financial behaviour, employability, and educational choices Groh, M., Krishnan, N., McKenzie, D., Vishwanath, T.: Soft skills or hard cash? The of rural young people: findings from a randomized controlled trial in impact of training and wage subsidy programs on female youth employ- Morocco. Impact Report 6, 24 (2017) ment in Jordan. The World Bank, Washington (2012) Becker, S.O., Caliendo, M.: Sensitivity analysis for average treatment effects. Groh, M., McKenzie, D., Shammout, N., Vishwanath, T.: Testing the importance Stata. J. 7(1), 71–83 (2007) of search frictions, matching, and reservation prestige through rand- Bell, B., Blundell, R., Van Reenen, J.: Getting the unemployed back to work: the omized experiments in Jordan. The World Bank, Washington (2014) role of targeted wage subsidies. Int. Tax Public Fin. 6(3), 339–360 (1999) Groh, M., Krishnan, N., McKenzie, D., Vishwanath, T.: Do wage subsidies provide Bergemann, A., Van den Berg, G.: J: Active labor market policy effects for a stepping-stone to employment for recent college graduates? Evidence women in europe—a survey. Annales d’Economie et de Statistique 7, from a randomized experiment in Jordan. Rev. Econ. Stat. 98(3), 488–502 385–408 (2008) (2016) Betcherman, G., Dar, A., &Olivas, K: Impacts of active labor market programs: Handouyahia, A., Haddad, T., Eaton, F.: Kernel matching versus inverse prob- New evidence from evaluations with particular attention to developing ability weighting: a comparative study. Int. J. Math. Comput. Sci. 7(8), and transition countries, World Bank Social Protection Discussion Paper 1218–1233 (2013) Series no. 29142 (2004). Heckman, J.J., Robb, R.: Alternative methods for evaluating the impact of Blattman, C., & Ralston, L: Generating employment in poor and fragile states: interventions: An overview. J. Econ. 30(2), 239–267 (1985) evidence from labor market and entrepreneurship programs (2015). Heckman, J.J., Ichimura, H., Todd, P.: E: Matching as an econometric evaluation Broecke, S.: Tackling graduate unemployment in North Africa through employ- estimator: Evidence from evaluating a job training programme. Rev. Econ. ment subsidies: a look at the SIVP programme in Tunisia. IZA Journal of Stud. 64(4), 605–654 (1997) Labor Policy 2(1), 1–19 (2013) Heckman, J.J., Ichimura, H., Todd, P.: Matching as an econometric evaluation Bruhn, M: Can wage subsidies boost employment in the wake of an economic estimator. Rev. Econ. Stud. 65(2), 261–294 (1998) crisis? Evidence from Mexico. The World Bank (2016). Heckman, J.J., Lalonde, R.J., Smith, J.A.: The economics and econometrics of Burtless, G.: Are targeted wage subsdies harmful: evidence from a wage active labor market programs. In: Ashenfelter, O., Card, D. (eds.) Handbook voucher exeperement. Ind. Labor Relat. Rev. 39, 105–114 (1985) of labor economics. Springer, Amsterdam (1999) Caliendo, M., Kopeinig, S.: Some practical guidance for the implementation of Hirano, K., Imbens, G.W., Ridder, G.: Efficient estimation of average treatment propensity score matching. J. Econ. Surv. 22(1), 31–72 (2008) effects using the estimated propensity score. Econometrica 71(4), Caliendo, M., Hujer, R., Thomsen, S.: Identifying effect heterogeneity to improve 1161–1189 (2003) the efficiency of job creation schemes in Germany. Appl. Econ. 40(9), Huber, M., Lechner, M., Wunsch, C.: The performance of estimators based on 1101–1122 (2008) the propensity score. J. Econ. 175(1), 1–21 (2013) Calmfors, L., Forslund, A., & Hemstrom, M: Does active labour market policy Hujer, R., Caliendo, M., Thomsen, S.L.: New evidence on the effects of job work? Lessons from the Swedish experiences. Lessons from the Swedish creation schemes in Germany a matching approach with threefold Experiences (2002). heterogeneity. Res. Econ. 58, 257–302 (2004) Card, D., Hyslop, D.: Estimating the effects of a time-limited earnings subsidy Imbens, G.W., Lemieux, T.: Regression discontinuity designs: a guide to prac- for welfare-leavers. Econometrica 73(6), 1723–1770 (2005) tice. J Econ 142(2), 615–635 (2008) Card, D., Kluve, J.: What works? A meta-analysis of recent active labor market Kaldor, N.: Wage subsidies as a remedy for the unemployment. J. Polit. 5, 78 program evaluations. J. Eur. Econ. Assoc. 16(3), 894–931 (2018) (1936) Card, D., Kluve, J., Weber, A.: Active labour market policy evaluations: a meta- Katz, L: Wage subsidies for the disadvantaged. NBER Working Paper 5679 analysis. Econ. J. 120(548), F452–F477 (2010) (1996). Cho, Y., Honorati, M.: Entrepreneurship programs in developing countries: a Khandker, S., Koolwal, G., Samad, H.: Handbook on impact evaluation: quantita- meta regression analysis. The World Bank, Washington (2013) tive methods and practices. The World Bank, Washington (2009) Crépon, B., Duflo, E., Marc Gurgand, M., Rathelot, R., Zamoray, P.: Do labor Kluve, J.: The effectiveness of European active labor market programs. Labour market policies have displacement effects? Evidence from a clustered Econ. 17(6), 904–918 (2010) randomized experiment. Q. J. Econ. 128(2), 531–580 (2013) Kluve, J., Lehmann, H., Schmidt, C.: Active labor market policies in Poland: Dahlberg, M.: Forslund A: Direct displacement effects of labour market pro - human capital enhancement, stigmatization, or benefit churning? J. grammes. Scand. J. Econ. 107(3), 475–494 (2005) Comp. Econ. 27(1), 61–89 (1999) David, H., Houseman, S.: Do temporary-help jobs improvelabormarketout- Kluve, J., Puerto, S., Robalino, D., Romero, J.M., Rother, F., Stöterau, J., Weiden- comes for low-skilled workers? Evidence from" Work First". Am. Econ. J. kaff, F., Witte, M.: Interventions to improve the labour market outcomes Appl. Econ. 2(3), 96–128 (2010) of youth: a systematic review of training, entrepreneurship promotion, De Mel, S., McKenzie, D., Woodruff, C.: Wage subsidies for microenterprises. Am. employment services, and subsidized employment interventions. Camp. Econ. Rev. 100(2), 614–18 (2010) Syst. Rev. 12, 7 (2017) De Mel, S., McKenzie, D., Woodruff, C.: Labor drops: experimental evidence Kluve, J., Puerto, S., Robalino, D., Romero, J.M., Rother, F., Stöterau, J., Weiden- on the return to additional labor in microenterprises. The World Bank, kaff, F., Witte, M.: Do youth employment programs improve labor market Washington (2016) outcomes? A quantitative review. World Dev. 114, 237–253 (2019) Dubin, J.A., Rivers, D.: Experimental estimates of the impact of wage subsidies. Kvasnicka, M: Does Temporary Help Work Provide a Stepping Stone to Regular J. Econ. 56, 219–242 (1999) Employment? NBER, Working Paper N.w13843 (2009). Escudero, V., Kluve, J., Mourelo, E.L., Pignatti, C.: Active labour market Lechner, M.: Earnings and employment effects of continuous off-the-job train- programmes in Latin America and the Caribbean: evidence from a meta- ing in East Germany after uniication. J. Bus. Econ. Stat. 17(1), 74–90 (1999) analysis. J. Dev. Stud. 55(12), 2644–2661 (2019) Levy Yeyati, E., Montané, M., Sartorio, L: What works for active labor market Galasso, E., Ravallion, M.: Social protection in a crisis: Argentina’s Plan Jefes y policies? CID Working Paper Series (2019). Jefas. World Bank Econ. Rev. 18(3), 367–399 (2004) Martin, J.P., Grubb, D.: What works and for whom: a review of OECD Countries’ Galasso, E., Ravallion, M., Salvia, A.: Assisting the transition from workfare to experiences with active labour market policies. Swed. Econ. Pol. Rev. 8(2), work: a randomized experiment, industrial and labor relations review. Pol. 9–56 (2001) Res. Work. Pap. 57(5), 128–42 (2004) Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 13 of 13 17 McKenzie, D.: How effective are active labor market policies in developing Souag, A.: The heterogeneous effect of employment agencies program on countries? A critical review of recent evidence. World Bank Res. Observ. labor force behavior in Algeria: a dynamic approach. Springer, Berlin 32(2), 127–154 (2017) (2020) Moczall, A.: Eets d ff ’aubaine et de substitution d’un dispositif allemand de sub - Stephan, G.: Hétérogénéité sectorielle des effets d’un dispositif de subvention vention salariale pour demandeurs d’emploi difficilement employables. salariale sur les salaires et l’emploi en Allemagne. Travail et Emploi 139, Travail Et Emploi 139, 39–59 (2014) 60–74 (2014) Morgan, S.L., Harding, D.: J: Matching estimators of causal effects: prospects Stuart, E.A.: Matching methods for causal inference: a review and a look and pitfalls in theory and practice. Sociol. Methods Res. 35(1), 3–60 (2006) forward. Stat Sci 25(1), 1 (2010) Pallais, A.: Inefficient hiring in entry-level labor market. Am. Econ. Rev. 104(11), Venetoklis T: An evaluation of wage subsidy programs to SMEs utilising 3565–3599 (2014) propensity score matching, Research Reports 106, VATT Institute for Premand, P., Brodmann, S., Almeida, R., Grun, R., Barouni, M.: Entrepreneurship Economic Research (2004). education and entry into self-employment among university graduates. Woodbury, S.A., Spiegclman, R.G.: Bonuses to workers and employers to World Dev. 77, 311–327 (2016) reduce unemployment. Am. Econ. Rev. 77, 513–530 (1987) Rosenbaum, P.R.: Overt bias in observational studies. Observational studies, pp. Zhao, Z.: Data issues of using matching methods to estimate treatment 71–104. Springer, New York (2002) effects: an illustration with NSW data set. China Center Econ. Res. 15, Rosenbaum, P.R., Rubin, D.: The central role of the propensity score in observa- 1–37 (2003) tional studies for causal effects. Biometrika 70(1), 41–55 (1983) Roy, A.: Some thoughts on the distribution of earnings. Oxf. Econ. Pap. 3, Publisher’s Note 135–146 (1951) Springer Nature remains neutral with regard to jurisdictional claims in pub- Rubin, D.B.: Estimating causal effects of treatments in randomized and nonran- lished maps and institutional affiliations. domized studies. J. Educ. Psychol. 66(5), 688 (1974) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal for Labour Market Research Springer Journals

Micro-econometric evaluation of subsidized employment in morocco: the case of the "Idmaj" program

Loading next page...
 
/lp/springer-journals/micro-econometric-evaluation-of-subsidized-employment-in-morocco-the-3M7v0zs5m0
Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2021
ISSN
1614-3485
eISSN
2510-5027
DOI
10.1186/s12651-021-00300-5
Publisher site
See Article on Publisher Site

Abstract

This paper aims to assess the impact of the Moroccan wage subsidy program "Idmaj". It applies the propensity score matching method to the data from a survey conducted by the Ministry of Labour on a sample of eligible individu- als. Our results suggest that wage subsidies in Morocco have a positive but marginally significant effect on reducing unemployment and improving employment and a significant negative impact on wages. It also highlights some heterogeneous effects of the program, particularly on women. Finally, it appears that the program did not serve as a stepping stone to higher-paying, high-quality work and, in contrast, it had a stigmatizing effect on beneficiaries. Keywords: Propensity score matching, Impact evaluation, Active labour market policy, Wage subsidies JEL Classification: C14, C31, H43, J38, J68 unconditional average magnitude of these effects remains 1 Introduction low (Kluve et al. 2019).The assessment is even more pessi - Active labour market policy (ALMP) programs are usu- mistic according to other meta-analysis that reveal a neg- ally classified into three categories (Card et al. 2010). The ative impact, at best zero, on beneficiaries’ employment first one includes training programs that aim to improve and income (Card et  al. 2018).For developing countries, employee’s skills. The second encompasses job search the evidence on the effectiveness of ALMP programs still assistance and job-matching schemes. The last one relates sketchy (Grimm and Paffhausen 2014), due to the lim - to wage subsidies, both those provided to employees to ited number of rigorous impact evaluations (McKenzie reduce inactivity traps and those provided to employers 2017). For example, in the very influential meta-analysis to reduce the costs of hiring. These subsidies, which are of Betcherman et  al. (2004), which covers 159 studies, based on long-standing theoretical arguments (Kaldor there are only 39 evaluations from developing and tran- 1936), have been used extensively by developed countries sition countries, of which only 4 drew on randomized since the 1980s, before they were popularized in develop- experiments. Moreover, the findings of the few existing ing countries (Auer et al. 2005). The aftermath of the last evaluations should be taken with caution, as the found financial crisis, the persistence of mass unemployment effects vary across intervention’ type and depending on and the inevitable process of labour automation are all the group of beneficiaries and country context (Cho and recent trends that have given new impetus to wage subsi- Honorati 2013). Therefore, additional studies assessing dies as a primary tool of ALMP (Banerji et al. 2014). However, the available literature about their effective - ness provides mixed evidence. The recent meta-anal - ysis show positive effects (Yeyati et  al. 2019), but the 1 Among the main meta-analyzes in developing countries, we can cite: Cho and Honorati 2013; Grimm and Paffhausen 2014; Blattman and Ralston 2015; Mckenzie 2017; Escudero et al. 2019. *Correspondence: naima.samoudi1995@gmail.com 2 Only one of which (Galasso et  al. 2004) was published in an academic Laboratory of Applied Economics (LEA), Med V University, Rabat, Morocco journal (McKenzie 2017). © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. 17 Page 2 of 13 A. Chatri et al. the effectiveness of wage subsidies in developing coun - data used. Section “6” presents and discusses the empiri- tries are of crucial importance. cal results. The final section presents  the conclusions, The case of Morocco, where the available labour fac - policy implications and limitations of this paper. tor seems to be strongly under-used and inefficiently allocated, is an excellent example in this perspective. The 2 Literature review unemployment rate has exceeded 10% on average over The available literature shows that the effectiveness the last 20  years, with 70% are long-term unemployed, of wage subsidies is very heterogeneous and depends, 60% are first-time job seekers, and the employment rate among other factors, on their design arrangements, the is now only 40%. The high weight of graduate unemploy - modalities of their implementation, the targeting of their ment, which is five times higher than that of without beneficiaries, their expected outcomes, the market labour diploma people, as well as that of informal employment, functioning, as well as the positioning of countries in the unpaid employment and underemployment, which economic cycle (Card et  al. 2010; Almeida et  al. 2014; represent respectively 27%, 18% and 10% of the total Gautié 2015). Nevertheless, three consensual conclusions employment, reveal, moreover, the suboptimal use of deserve to be highlighted. the available labour force (Haut-commissariat au plan First, subsidies provided to firms for reducing their & World Bank 2017).This situation has occurred despite cost of hiring tend to produce positive results compared the implementation, particularly since the beginning of to those offered directly to the unemployed (Venetoklis the 1990s, of an ambitious ALMP, which includes, among 2004; De Mel et  al. 2010, 2016; Bruhn 2016). Secondly, other governmental programs, a wage subsidy scheme, the quasi-experimental methods produce higher effects called "Idmaj". Unfortunately, none of these public pro- than those of the experimental ones (Kluve 2010; Grimm grams has been evaluated, and our knowledge about and Paffhausen 2014). This may partly explain why these their effectiveness is still very limited. subsidies are relatively more effective in developing Therefore, this paper seeks to fill this gap by estimat - countries than in developed countries (Betcherman et al. ing the causal effect impact of the "Idmaj" Program using 2004; Kluve et  al. 2017). Indeed, as experiment evalua- the propensity score matching method. It contributes tions are rare in developing countries (Mckenzie 2017), to the available literature on subsidized employment in the few existing evaluations use mostly quasi-experimen- four ways. First, as counterfactual evaluations outside tal methods (Galasso and Ravallion 2004; Broecke 2013; the OECD are quite limited, it provides evidence on the Souag 2020). Thirdly, most authors have concluded that experience of developing countries. Second, while the the impact of these subsidies was limited (Escudero et al. available literature focuses more on schemes acting on 2019; Kluve et al. 2019). Betcherman et al. (2004) report, either the demand or the supply side of the labour market for example, that of the 159 evaluation studies reviewed, and targeting low-skilled individuals, this paper exam- more than two-thirds had a negative effect on employ - ines one particular program that affects the both sides of ment and only 5 studies had a positive effect on the the labour market and targets, as many countries in the income of beneficiaries. MENA region (Groh et al. 2012, 2014; Broecke 2013; Pre- The available literature has shed light on several collat - mand et  al. 2016; Souag 2020), higher degree graduates. eral effects of wage subsidies that can explain their lim - Third, it goes beyond the program’s direct impacts on ited effectiveness. Their potential stigmatizing effect has, employment to capture its effects on beneficiaries’ wages indeed, highlighted by the first experimental evaluations and their working conditions. Finally, it highlights the in the United States. Burtless (1985) has shown that these heterogeneous effects of the program. subsidies have been used by American employers as a The rest of the paper is organized as follows: Sect. “2” discrimination tool against beneficiaries, who were less presents a review of the literature. Section “3” presents likely to find a job than non-beneficiaries. Similarly, other the "Idmaj" program. Propensity score matching esti- authors have shown that beneficiaries have concealed mation methodology and empirical implementation are their eligible status because they considered it degrad- discussed in Sect. “4”. Section “5” provides details on the ing and stigmatizing (Woodbury and Spiegelman 1987; However, there are some recent signs that things are changing. For instance, Such evidence is also rare even for programs developed by civil society or in Morocco, the first and, to our knowledge, the only experimentation related international organizations, which are with limited scope and targeted gener- to the labour market is Bausch et  al. (2017). Experiments recently developed ally narrow local individuals. In this respect, the paper of Bausch et al. (2017) in the other countries of the MENA region (Groh et al. 2012; Groh et al. 2014; is an exception. This paper has examined the impact of a skills training pro - Premand et al. 2016) have also not focused on the mechanism discussed here, gram, called "100 Hours to Success", on the financial behaviour, employabil - namely wage subsidies. For the latter, the experiment of Galasso et al. (2004) ity and educational choices of the rural young living in Morocco’s Oriental on the argentine program “Proempleo” is one of the rare exceptions in devel- Region. oping countries. Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 3 of 13 17 Dubin and Rivers 1999).This stigma effect has also been the match between the profile of job seekers and the verified in other countries, both developed (Bell et  al. needs of the market, through contractualized training, 1999) and developing (Kluve et al. 1999). skills training or retraining. The second program, called Likewise, several studies have shown that these subsi- "Moukawalati" and renamed "self-employment", aims to dies are likely to produce some spillover effects that reduce promote entrepreneurship and business creation through their effectiveness (Martin and Grubb 2001; Kluve et  al. training and financial assistance. The third program, the 2017; Escudero et  al. 2019). When it comes to the wind- subject of this paper, is a subsidized employment pro- fall effect, it has been shown that wage subsidies mainly gram which supports the transition of youth from school help those who, even in the absence of the program, would to work through company placements. have found a job in both developed (Katz 1996; Bell et al. This program, called "Idmaj" or "insertion", has been 1999; Crépon et  al. 2013) and developing countries (Gal- designed to reach a double objective. Firstly, the improve- asso et  al. 2004; Broecke 2013). The substitution and dis - ment of the employability by offering to the targeted placement effects have been also verified by several works unemployed a professional experience through 24-month (Dahlberg and Forslund 2005; Crépon et al. 2013; Moczall internships. Secondly, the increase of the share of high- 2014). The size of these effects depends on the methods skilled and of qualified employees by giving incentives used. It seems that it is larger in econometric studies (60– for firms to hire the young graduates. Indeed, it is an 70%) than survey studies (15–40%) (Calmfors et al. 2002). employment subsidy targeting higher education gradu- Furthermore, many studies have focused on the perma- ates, as the individuals who are eligible for it are all job nence of the effects of wage subsidies. However, except seekers registered with ANAPEC and holders of a higher few papers that have validated the stepping stone effect to (tertiary) education diploma, vocational training diploma stable, better-paying jobs (Pallais 2014), the general trend or a baccalaureate at least. of these studies shows that the positive impact of these Two types of contracts are offered under the "Idmaj" subsidies declines over time (Hujer et al. 2004; Venetoklis program: the insertion contract (IC) and the professional 2004; Card and Hyslop’s, 2005; Kvasnicka 2009; David and integration contract (PIC). Both are a tripartite contract Houseman 2010; Groh et al. 2016; De Mel et al. 2016). between (i) a company interested in hiring an eligible job Finally, the existing literature has highlighted the het- seeker (ii) an eligible job seeker and (iii) the ANAPEC, erogeneous effects of wage subsidies by subgroups of par - which is responsible for validating and following up these ticipants. For example, Galasso et al. (2004) concluded that contracts. the "Proempleo" program was significant only for women, The IC, put in place from the start of the program, pro - youth, and highly educated workers. The positive impact vides to private companies involved in the program total on women is corroborated by many studies (Bergemann exemption from (i) Income Tax, ii) Social Charges and and Van den Berg 2008; Groh et al. 2016).Other works have (iii) Tax on Vocational Training for the internship allow- found heterogeneous effects depending on both gender and ance paid to eligible individuals up to a ceiling of 6000 areas. This is the case for example of that of Caliendo et al. MAD (Moroccan Dirham), for a period of 24  months. (2008), for the German scheme and that of Souag (2020) Two other incentive measures were added in 2016. This for the Algerian one. By sectors of activity, Stephan (2014) concerns, on the one hand, the payment by the Govern- detected a little heterogeneity of the German program. ment of employer’s and employee’s contributions for basic compulsory health insurance during the intern- 3 Description of the "Idmaj" program ship period and, on the other hand, the extension of the In 2000, the active labour market policy in Morocco aforementioned tax and social exemptions for an addi- took a qualitative step forward with the creation of the tional year, if the trainee is integrated under permanent National Agency for the Promotion of Employment employment (PE) contract. and Competencies (ANAPEC ). This public agency is For its part, the PIC, which was put in place in 2011, designed to (i) implement active labour market policies completes the IC by providing an integration training (ALMPs) (ii) to connect employers to job seeker, and (iii) subsidy (bonus) amounting to 25,000 MAD to companies to provide information and guidance to job seekers and recruiting eligible graduates under (PE) contract at the young entrepreneurs. Not far, three main programs were end of a 6 to 9 months integration training course as part set up in 2006. The first, called "Taehil ", aims to improve of the system of the ICs. However, it is worth noting that the PIC is reserved for specific job seekers who have great difficulties finding their first job, namely young graduates of some general higher education courses (baccalaure- Abbreviation used in Morocco in reference to the initials of the name of the ate + 3 and more) enrolled at ANAPEC for at least 1 year. agency in French: Agence Nationale de Promotion de l’Emploi et des Com- pétences. 17 Page 4 of 13 A. Chatri et al. That being said, the "Idmaj" program remains the cen - The hypothesis in (3), known as conditional independ - tral component of ALMP in Morocco. Actually, by the ence (Lechner 1999) or selection on observables (Heck- end of its start-up year (2006), the number of beneficiar - man et  al. 1985), assumes that there is no difference ies was 32,880 young graduates, before it rises steadily between treated and untreated groups, conditionally to the to over 102,773 in 2018. This increased number of ben - observables (noted X). The hypothesis in (4) implies the eficiaries merely reflects the budgetary cost of the pro - existence of an overlap (common support) between these gram and cannot be understood as a sign of effectiveness, two groups. Rosenbaum and Rubin (1983) call the com- which involves identifying the changes observed in ben- bination (3) and (4) the "strong ignorability" hypothesis, eficiaries that are attributable exclusively to the program. which is the identifying hypothesis of ATT, since it verifies: Such is the purpose of this paper. 0 0 0 E Y |X, D = 1 = E Y |X, D = 0 = E Y |X 4 Methodology and empirical implementation (4) This paper uses the standard econometric evaluation However, when the number of X is large, a “Curse of model, Roy Rubin’s model of potential outcomes (Roy dimensionality” (Heckman et  al. 1997) could affect the 1951; Rubin 1974), to determine the causal effect of the width of the common support. To overcome this prob- "Idmaj" program on our outcome variables. In particular, lem, Rosenbaum and Rubin (1983) proposed to match it seeks to estimate the Average Treatment on Treated treated and untreated individuals not on the basis of X (ATT) parameter, which is given by: but based on the propensity score p (X), which defined as the probability of receiving the treatment giving X. Its 1 0 ATT = E Y |D = 1 − E Y |D = 1 (1) balancing property ensures that treated and untreated persons with the same value p (X) also have the same with D ∈ {0, 1} a binary variable that indicates the treat- distribution of X. These authors have also illustrated that ment status. It takes 1 in case of treatment and 0 other- while treatment assignment is strongly ignorable given X, 1 0 wise. Y is the outcome in case of treatment, and Y the it is also ignorable given p (X). Therefore, we have: outcome in case of non-treatment. 1 0 As E Y |D = 1 is not observable, and taking into Pr X|D = 1, p(X) (Y , Y )⊥D|p(X) (5) account the absence of an explicit rule of assignment to the "Idmaj" program and that there is no instrument or 0 < Pr(D = 1|p(X)) < 1 (6) discontinuity in treatment to exploit, we use in this paper the matching method, which permits us to identify our Once these conditions are satisfied, all biases due to the ATT under the following assumptions : observables may be eliminated, allowing us to estimate the ATT according to the following general formula : 1 0 (Y , Y )⊥D|X (2)   � � T C   ATT : Y − ω(i, j)Y (7) 0 < Pr(D = 1 X)< 1 i j (3) i∈T j∈C with: N The number of treated, Y : The observed out - come of the matched treated, Y : The observed outcome 6 of the matched untreated and ω i, j is the weight used to Certainly, the observable average outcome of the untreated could be used as a credible counterfactual for this non-observable outcome in randomized aggregate outcomes of the matched untreated. experiments, nevertheless in the presence of non-experimental data, as in this Estimating the ATT expressed in (7) above involves paper, this would lead to a selection bias. For this purpose, different estimat - first the prediction of the propensity scores with any dis - ing strategies are developed (Heckman et  al. 1999) to avoid this bias, which comes from both observable and unobservable factors. The feasibility of each crete choice model can be used. In this paper, we use the of these strategies is based on some assumptions, which in turn depend on the "Probit" model. The second step of our strategy consists nature of the program and its context. For example, the regression disconti- nuity (Imbens and Lemieux 2008) is suitable for programs using an index to decide who is eligible to enrol in the program and who is not. The instrumen - tal variables method (Angrist et al. 1996) which relies on some external source of variation to determine treatment status is suitable mainly for programs It is to highlight that there is another method, which is based directly on with imperfect compliance, or universal coverage. Unlike these two non- the propensity score. This is the inverse probability of treatment weighting experimental methods which produce estimates of the counterfactual through (IPTW) popularized by Hirano et al. (2003). explicit program assignment rules, other methods, such as the difference-in- differences and the matching methods, can be used for programs with vol - It is worth noting that both this model and "logit" model allow consist- untary enrolment, when the program assignment rules are less clear or when ent estimation of the conditional probability of participation in the program none of the former methods is feasible. in the interval [0–1]. These two models generally yield similar results: esti- mates from a "logit" represent π⁄√3 of those obtained with a "probit". Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 5 of 13 17 0 .2 .4 .6 .8 1 0 .2 .4 .6 .8 1 x Propensity Score kdensity Treatedkdensity Untreated Untreated Treated Fig. 1 Propensity score distribution and common support region of the choice of an appropriate matching estimator, being optimize the tradeoff between variance and bias (Garrido underlined that many popular matching estimators are et al. 2014; Heckman et al. 1997). suggested in the literature (nearest neighbor, caliper, ker- That being said, using this matching for estimating our nel and stratification…). Although they share the same ATT depends on its ability to verify both common sup- Eq.  (7) for calculating ATT, these estimators differ in port region and balancing properties. Figure 1 shows the both the definition of the neighbourhood for the treated existence of a sizeable overlap of [0.2–0.8]. Likewise, our individuals and in the way that weights are assigned to tests, in Table  2, confirm that the balancing property is those neighbors (Stuart 2010). When it comes to choos- also satisfied. ing the best estimator, the available literature offers very As a final step in our methodological approach, we little guidance (Huber et  al. 2013) and argues that there conduct a sensitivity analysis of the results obtained. As is no "winning" estimator (Caliendo and Kopeinig 2008), noted previously, the propensity score matching method so their performance depends above all on the data set relies on conditional independence, a strong assumption (Zhao 2003). Furthermore, it asserts that the choice of which assumes that program uptake is based entirely on the matching estimator requires evaluating tradeoffs observed characteristics (Khandker et al. 2009). So if any between efficiency and bias (Garrido et al. 2014). unobserved factors exist and affect participation status Therefore, the kernel matching algorithm developed by and the variable of interest, then a hidden bias is involved Heckman et al. (1997) seems to be the most appropriate (Rosenbaum 2002). In this respect, and after estimat- in the context of this paper due to the smallness of our ing the ATTs, we will check their sensitivity to devia- sample. This matching approach maximizes the estimates tion from the identification assumption, using the MH ‘efficiency and precision using more observations. Con - bounds program dedicated to binary variables accord- trary to other algorithms that use one (nearest neigh- ing to the bounding approach proposed by Robensuam bour) or at best a few observations from the comparison (2002) and explained in the paper of Becker and Caliendo group (calliper matching), Kernel matching estimator (2007). This allows highlighting the degree by which an constructs the counterfactual by using for each treated unmeasured variable influencing the odds of participa - unit, all the untreated units weighted by the distance that tion in "Idmaj" program will also modify its effects on our separates each of them from the treated (Morgan and outcome variables. Harding 2006). Thereby, weights are assigned so that less similar unites receive less weight when estimating ATT, 5 Database and vice versa (Handouyahia et al. 2013). The data used here are taken from a random survey con - This matching requires, moreover, the choice of a ker - ducted in 2010 by the Ministry of Labour in coordination nel function, and more importantly, the choice of a band- with the ANAPEC, as part of the strengthening of obser- width that controls the degree of smoothing applied vation and analysis capacities of the Moroccan labour to the data. In this paper, we used the Gaussian kernel market. Its objective was to provide an accurate descrip- function and the default bandwidth of 0.06 which may tion of the socio-economic situation of the interviewees 0 1 2 3 17 Page 6 of 13 A. Chatri et al. and appreciate, in a descriptive way, the effect of the only became eligible for the program, using this year as "Idmaj" program on their professional trajectories. While the baseline seems more reasonable. it was not completely designed for impact evaluation We also excluded variables that do not jointly cover purposes, this survey fulfills three characteristics that, beneficiaries and non-beneficiaries. Finally, for the out - according to the available literature, support the estima- lier and the missing values, and due to the small size of tion of the causal effect by the matching method (Heck - our sample, we created an additional modality "non- man et al. 1998). First, it uses the same data source and, response" for qualitative variables with too many missing therefore, measures the observed characteristics in the observations. Likewise, and considering the bias that it same way for treated and untreated groups. Second, it is can produce, the imputation by the mean has only been specific only to eligible individuals, which may alleviate used for the three variables which suffer from few miss - the problem of determining a «common reference date» ing data or outliers. These variables are: "Reservation for treated and untreated individuals (Lechner 1999). wage", "Mother’s level of education", and "Father’s level of Third, it contains a rich set of covariates, which sustains education". the plausibility of the conditional independence hypoth- All these choices have affected the size of our final sam - esis discussed above. ple, which ended up with 621 individuals, including 377 This survey focused on two samples constituted by sys - treated and 244 untreated. tematic random sampling with equal probability from Regarding the covariates included in our estimate, (i) a nominative list of beneficiaries of "Idmaj" insertion they were chosen on the basis of a set of methodological contracts in 2006 and 2007, and (ii) a list of job seekers considerations. First, the identification of ATT using the enrolled in ANAPEC, eligible for the program but never score propensity matching method implies, on the one having benefited from it since their enrollment until the hand, the inclusion of all covariates that influence simul - day of the survey. The retrospective questions of this taneously treatment and outcome variables and, on the survey aimed to collect information on activity situa- other hand, the exclusion of covariates affected by treat - tions of interviewees during the following four phases: (i) ment (Heckman et  al. 1998). For this reason, and since before registering with ANAPEC, (ii) between registra- our data are post-program, we have excluded variables tion and signature of the IC, (iii) during the IC period and that are not fixed overtime at the risk of being affected (iv) the post IC period. by treatment, yet the parents’ profession and their level of After the sampling, the survey was conducted on a education will be kept, both because of their importance sample of 2500 beneficiaries and 500 individuals for the in explaining participation in the program, as well as in control group. This paper uses a data of random subsam - labour market insertion (social capital) and because they ple of a total size of 879 eligible individuals for the pro- are difficult to change over time. Secondly, to choose the gram, which 627 beneficiaries in 2006 or 2007 and 252 most relevant variables, we have proceeded to an iterative non-beneficiaries. selection of our variables according to a process guided However, some choices seemed essential to improve by three references: (i) the significance of the statistical the coherence and quality of the data. u Th s, we chose, in tests, (ii) the quality of the matching, which involves veri- this paper, only one reference year for the program. This fying the existence of the common support and the bal- choice was necessary to construct some variables on the ancing condition between treated and untreated, and (iii) same basis and compare the two groups at similar time the empirical literature suggestions, according to which horizons. It was also more appropriate, as we had no way one should integrating variables that capture the four to decompose the control group, which is half the size following dimensions: eligibility criteria, socio-demo- of the treatment group, to evaluate both years simulta- graphic characteristics, qualifications and labour market neously. In addition, while data indicate that 21% of the history (Heckman et al. 1999). untreated were enrolled in ANAPEC until 2007 and thus Therefore, our final specification includes 23 covariates belonging to the four above-mentioned dimensions: eligibility criteria (diploma and years of 9 enrolment in ANAPEC), socio-demographic charac- The survey contains another sample of companies whose results are not used in this work. teristics (program entry age, gender, health problems, Not having been able to access the global sample, this sub-sample made available to us by the Ministry of Labour was, at our request, drawn ran- domly. The choice of 2006, instead, will reduce the size of the control group, which Since data are post-program, it was necessary to retain some variables at is already limited. program entry to predict propensity scores. For example, the calculation of age and seniority of enrollment in ANAPEC at the start of the program was The excluded variables are: marital status, number of children, number feasible for treated individuals, but for those untreated will require the set- of non-resourceful persons in the household, type of housing, region, and ting of a single reference year to allow their matching with the treated. household size. Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 7 of 13 17 Table 1 Variables construction Variable Description Years of enrolment in ANAPEC In years Program entry Age In years Health problems 1: has a health problem, 0 otherwise Reservation wage En MAD Internship 1: trainee before enrollment at ANAPEC; 0 otherwise Diploma Perception 1: positive perception, 0 otherwise Training days Number of training days from graduation to the enrollment at ANAPEC Gender 1: man; 0 woman Mother’s job 1: in employment or retirement, 0 otherwise Father’s job 1: in employment or retirement, 0 otherwise Extracurricular activity 1 if yes, 0 otherwise Mother’s level of education In Years Father’s level of education In Years Studies 1 in education before the enrollment at ANAPEC, 0 otherwise Employment 1 in employment before the enrollment at ANAPEC 0 otherwise Associative activities 1 if yes, 0 otherwise Duration of unemployment In days (from graduation to the enrollment at ANAPEC) Education sector 1: public, 0 otherwise Unemployment 1: unemployed before enrollment at ANAPEC, 0 otherwise Diploma In years Duration of other inactivity Number of days spent in inactivity (other than internship and studies) Internship days Number of internship days from graduation to the enrollment at ANAPEC Other forms of inactivity 1: inactive before ANAPEC enrollment, 0 otherwise mother’s and father’s level of education and their jobs), 6 Results and discussion qualifications (last individual’s professional situation As mentioned above, the aim of this paper is to assess the before enrolling in ANAPEC: employment, unemploy- effectiveness of the "Idmaj" program not only with regard ment, internship, study or other types of inactivity), to its direct expected outcomes, which is improving the and labour market history (duration of studies, unem- employability of its beneficiaries, but also with regard to ployment, internship and other types of inactivity the quality of the job held. Therefore, we have estimated experienced by the individuals from the time of their the ATT of wage subsidies granted under the "Idmaj" graduation until their registration with ANAPEC). We program on five types of outcomes: (i) status with regard have, moreover, integrated the following variables that to unemployment, (ii) status with regard to employment, are likely to capture some forms of unobservable indi- (iii) the level of salary, (iv) the benefit or not of social vidual heterogeneity: diploma perception, reservation cover, and (v) the number of hours worked per week. wage, participation in extracurricular and associative These outcome variables are all binary, apart from that activities. of the number of hours worked per week and the earned Tables 1 and 2, present respectively, our variables and wage. The latter is unconditional on wage employment. the tests of difference in means. These tests show in uTh s, the employment (respectively unemployment) particular that, before matching, treated and untreated result variable takes the value 1 if the individual finds a groups are statistically different only in terms of four job (does not find a job) and 0 otherwise. It should also variables, which are age, years of enrolment in ANA- be noted that unemployment outcome variable is condi- PEC, reservation wage, and health problems. After tional on labour market participation (we have taken only matching, these tests confirm, as mentioned above, the unemployed persons and excluded the inactive ones). that our specification satisfies the balancing property. Table  4 contains the results of our estimates and pro- For its part, Table  3 reports the results of the propen- vides ATTs for the whole sample and by subgroups con- sity score estimation. structed by gender and age. 17 Page 8 of 13 A. Chatri et al. Table 2 Balancing covariates test Variable Mean (1) Mean (2) t-test (1) t-test (2) Treated Control Treated Control t p > t t p > t Years of enrolment in ANAPEC 1.932 2.617 1.935 1.982 − 6.130 0.000 − 0.550 0.582 Program entry Age 25.395 26.633 25.344 25.263 − 3.640 0.000 0.290 0.775 Health problems 0.046 0.013 0.046 0.026 2.270 0.023 1.430 0.153 Reservation wage 3152.400 2979.900 3148.800 3231.900 2.220 0.027 − 1.010 0.312 Internship 0.078 0.046 0.079 0.086 1.590 0.113 − 0.350 0.725 Diploma Perception 0.676 0.625 0.678 0.683 1.290 0.199 − 0.150 0.878 Training days 135.150 97.842 117.700 108.870 1.150 0.250 0.410 0.685 Gender 0.551 0.504 0.550 0.530 1.140 0.255 0.540 0.591 Mother’s job 0.1 0.079 0.100 0.121 0.870 0.385 − 0.890 0.374 Father’s job 0.873 0.842 0.873 0.886 1.090 0.276 − 0.540 0.590 Extracurricular activity 0.349 0.321 0.347 0.332 0.710 0.479 0.420 0.673 Mother’s level of education 2.970 2.722 2.978 2.960 0.740 0.461 0.060 0.954 Father’s level of education 6.745 6.480 6.760 6.681 0.690 0.493 0.230 0.816 Studies 0.105 0.121 0.106 0.096 − 0.590 0.555 0.430 0.667 Employment 0.116 0.1 0.114 0.114 0.620 0.532 − 0.020 0.987 Associative activities 0.068 0.058 0.068 0.075 0.450 0.649 − 0.390 0.694 Duration of unemployment 318.090 295.090 308.390 287.810 0.470 0.640 0.500 0.619 Education sector 0.114 0.129 0.114 0.131 − 0.580 0.561 − 0.720 0.473 Unemployment 0.530 0.512 0.531 0.533 0.420 0.678 − 0.050 0.963 Diploma 14.130 14.133 14.136 14.143 − 0.040 0.968 − 0.090 0.925 Duration of other inactivity 24.803 25.512 24.870 23.289 − 0.070 0.945 0.190 0.851 Internship days 33.151 32.362 32.252 29.357 0.090 0.928 0.400 0.686 Other forms of inactivity 0.059 0.058 0.060 0.073 0.060 0.954 − 0.740 0.462 Bold means that the variables are significant at 1% , 5% or 10%  Mean (1) before matching Mean (2) after matching Overall, our results show that the "Idmaj" program Furthermore, our results show that the "Idmaj" program reduced the probability of being unemployed by 7.5% provides some heterogeneous effects. The impact on the reduction of unemployment is only observed for women points and increased the chances of finding a job by 8.2 and young people under 24  years with 9.5 percentage percentage points. However, this positive effect on ben - points and 12 percentage points respectively. Similarly, eficiaries is only significant at the 10% level. On the other only women have an improvement in their probability of hand, the program had a negative and statistically signifi - finding a job (+ 11 percentage points). This positive impact cant effect (at 1%) on wages of beneficiaries, which were on women is corroborated by the literature (Bergemann lower by 487 MAD than those of non-beneficiaries. These and Van den Berg 2008). Concerning the negative effect on results are consistent with those of Martin and Grubb wages, it was once more observed only among women, for (2001), who suggest that subsidies may produce positive effects but at the expense of wages. These findings may whose the participation in the program caused them a loss also reflect the stigmatizing effect of the program, and of 697 MAD compared to non-beneficiaries. its use by the employers as a filtering tool, leading them While the global results do not appear a statistically sig to consider the participants to be unproductive and have nificant effect of the program on the working conditions, to be paid lower wages than those otherwise hired. This the disaggregated analysis did reveal some valuable find - negative impact on salaries reflects the ineffectiveness of ings. For example, men were less likely to be covered by a the 24-month internship, which does not seem to have social security system than non-treated (− 13 percentage improved the skills and the productivity of the beneficiar - points). This probability also decreases by (− 11 percent - ies and, consequently, their chances to find good-paying age points) among young aged 25–34  years. In addition, and quality jobs. These conclusions imply that the "Idmaj" the latter beneficiaries had to work four hours longer program does not have the expected stepping stone effect. than the non-participants. This did not seem, moreover, Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 9 of 13 17 Table 3 Estimation of propensity scores (Probit Model: Marginal effects at the mean) variable dy/dx Std.Err Z P > z [95%Conf Interval] X Years of enrolment in ANAPEC − 0.071 0.017 − 4.240 0.000 − 0.104 − 0.038 2.202 Program entry Age − 0.020 0.007 − 3.030 0.002 − 0.033 − 0.007 25.882 Health problems 0.281 0.072 3.920 0.000 0.141 0.422 0.033 Reservation wage 0.000 0.000 1.590 0.113 − 0.000 0.000 3084.58 Internship 0.215 0.079 2.730 0.006 0.061 0.369 0.066 Diploma Perception 0.063 0.046 1.380 0.166 − 0.026 0.153 0.656 Training days 0.000 0.000 1.410 0.159 − 0.000 0.000 120.47 Gender 0.060 0.043 1.380 0.168 − 0.025 0.144 0.533 Mother’s job 0.054 0.078 0.700 0.487 − 0.099 0.208 0.092 Father’s job 0.067 0.062 1.090 0.277 − 0.054 0.188 0.861 Extracurricular activity 0.022 0.046 0.480 0.632 − 0.068 0.113 0.338 Mother’s level of education − 0.005 0.007 − 0.650 0.518 − 0.018 0.009 2.872 Father’s level of education 0.002 0.006 0.320 0.745 − 0.009 0.013 6.640 Studies 0.021 0.091 0.230 0.821 − 0.157 0.198 0.111 Employment 0.152 0.077 1.990 0.047 0.002 0.303 0.110 Associative activities − 0.045 0.090 − 0.500 0.614 − 0.221 0.131 0.064 Duration of unemployment 0.000 0.000 1.450 0.148 − 0.000 0.000 309.038 Education sector − 0.081 0.069 − 1.170 0.240 − 0.217 0.054 0.120 Unemployment 0.126 0.067 1.880 0.060 − 0.005 0.257 0.523 Diploma 0.025 0.022 1.160 0.247 − 0.017 0.067 14.131 Duration of other inactivity 0.000 0.000 0.530 0.597 − 0.000 0.001 25.082 Internship days − 0.000 0.000 − 0.660 0.509 − 0.001 0.000 32.841 Other forms of inactivity 0.099 0.099 1.000 0.318 − 0.095 0.293 0.059 Bold means that the variables are significant at 1% , 5% or 10% to be compensated by an increase in their chances of that the impact of wage subsidies are more beneficial finding a job compared to non-beneficiaries. This result when accompanied by complementary services (Kluve indicates that the stigmatizing effect of the "Idmaj" pro - et al. 2019). gram is particularly significant among older youth, who Other conditionalities may also be useful to ensure that must be long-term unemployed or with higher degrees, the program can produce lasting effects beyond the expi - for which the statistics are already alarming. In Morocco, ration of subsidies, such as the requirement to keep sub- 70% of the unemployed are long-term unemployed, and sidized employees for a specified period. The tendency of unemployment rates for holders of higher degrees are employers to pay low wages to the beneficiaries can, on almost five times higher than for those with no degree. the other hand, be attenuated by more meaningful differ - Our results suggest revisiting the program’s parameters entiation of the wage ceiling, which serves as the basis for and making certain adjustments to increase its effective - tax and social security exemption. ness. In this respect, particular attention should be given Finally, our results support a more refined targeting of to the professional internship to make it a real lever for women and youth for whom the program has proven to productivity gains, since only 2% of the beneficiaries in be more effective. Indeed, there is ample evidence that our sample report having received coaching from the careful targeting avoids the loss of funds and allows for host firm during the internship. It is, therefore, necessary rigorous monitoring of results, and that the effectiveness to supervise this internship with more rigorous specifica - of subsidies is better when targeted to a specific group tions aimed at ensuring real supervision of the trainees (Almeida et al. 2014). and the improvement of their skills. It must be accompa- Before generalizing these findings, we conducted a nied by a reduction in its duration, since 24  months of sensitivity test to check the quality of our estimates. internship is a relatively long period that can produce While the raised impacts of the "Idmaj" program are counterproductive effects. It would also be appropriate to significant (at the 10%) on unemployment (respectively empower beneficiaries by requiring them to participate employment), the sensitivity test will determine at what in employability training. Indeed, it is widely recognized degree of presence of unobserved selection (Gamma) 17 Page 10 of 13 A. Chatri et al. Table 4 The ATT estimates by Kernel Matching Table 5 Sensitivity test for unobserved heterogeneity Outcome n. treated n. control ATT T Gamma Unemployment Employment Q_mh− P_mh− Q_mh + P_mh + The whole sample Unemployment 370 235 − 0.075* − 1.832 1 1.49 0.067 1.54 0.060 Employment 370 235 0.082* 1.804 1.1 1.00 0.157 1.02 0.153 Wage 269 155 − 487.153*** − 2.480 1.3 0.154 0.438 0.097 0.461 Social protection 269 141 − 0.055 − 1.564 1.5 0.380 0.351 0.512 0.304 Worked hours per 269 155 1.350 1.138 Bold means that the variables are significant at 1% , 5% or 10% week Gamma: odds of differential assignment due to unobserved factors Women Q_mh+ : Mantel–Haenszel statistic (assumption: overestimation of treatment Unemployment 166 115 − 0.095* − 1.892 effect) Employment 166 115 0.112* 1.714 Q_mh−: Mantel–Haenszel statistic (assumption: underestimation of treatment effect) Wage 118 70 − 697.783* − 1.955 p_mh+ : significance level (assumption: overestimation of treatment effect) Social protection 118 68 0.013 0.163 p_mh−: significance level (assumption: underestimation of treatment effect) Worked hours per 118 70 1.044 0.931 week Men suggests a cautious interpretation of the obtained results Unemployment 204 116 − 0.037 − 0.666 when both effects on unemployment and employment Employment 204 116 0.054 0.981 would no longer be significant if an unobserved vari - Wage 151 82 − 345.330 − 1.234 able increases the chances of program participation Social protection 151 71 − 0.135** − 4.302 by 10% for two individuals with the same observable Worked hours per 151 82 2.440 1.638 characteristics. week Youth [18–24] Unemployment 170 71 − 0.120* − 1.746 7 Conclusion, policy implications and limitations Employment 170 71 0.100 1.290 This paper sought to assess the effectiveness of the sub - Wage 124 41 − 408.998 − 1.131 sidized employment program "Idmaj" not only in terms Social protection 124 36 − 0.028 − 0.401 of reducing unemployment and promoting employment, Worked hours per 124 41 0.433 0.242 but also in terms of beneficiaries’ wages, working hours, week and social security status. This evaluation is conducted Youth[25–34] using the propensity score-matching method based on Unemployment 190 142 − 0.042 − 0.645 the Ministry of Labor’s survey of a sample of eligible Employment 190 142 0.039 0.676 individuals for the "Idmaj" program. Our results have sig- Wage 138 101 − 416.596 − 1.371 nificantly improved our knowledge of the effectiveness Social protection 138 93 − 0.115** − 2.462 of the program; especially since none of Morocco’s pub- Worked hours per 138 101 4.266** 2.363 lic ALMP schemes has been subject to a counterfactual week impact evaluation.  t statistic; * p < 0.10, ** p < 0.05, *** p < 0.01 Our results reveal that the marginally positive effects of the "Idmaj" program on reducing unemployment and increasing employment were accompanied by some sig- these effects will become insignificant. Since the impact nificant collateral effects on beneficiaries. In particu - is negative on unemployment (positive on employment), lar, the program seems to have produced a stigmatizing the focus will be on the negative unobserved selec- effect. This were manifested through the negative and tion Q_mh- (positive unobserved selection Q_mh + for statistically significant impact on beneficiaries’ wages, employment). The results in Table  5 affirm that in the and, for some categories (notably young aged 25 to absence of unobserved selection (Gamma = 1), the 34  years), on their working conditions as an increase in "Idmaj" scheme does have a significant effect at the working hours and exclusion from social security cover- 10% level (p_mh = 0.06) on reducing unemployment/ age. Our findings also reflect the ineffectiveness of the increasing employment. Then, the Q_mh− /Q_mh+ sta- 24-month internship provided by the program on the tistic reveals that at relatively low levels of presence of beneficiaries’ skills and productivity and their chances of unobserved selection (Gamma = 1.1), these effects will finding gainful employment. As a result, the program did be non-significant (p_mh− = 0.157 for unemployment not serve as a stepping stone. and p_mh+ = 0.153 for employment). This analysis Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 11 of 13 17 Moreover, our results highlight some heterogeneous important limitation of this paper lies in the narrowness effects of the program. They showed a positive impact on of our sample, in particular, that of untreated individuals the professional insertion of the most-excluded individu- which is, and contrary to the usual evaluation conditions, als in the Moroccan labour market, particularly women narrower than that of treated individuals. and young people aged between 18 and 24  years. How- ever, the positive effect for women is obviated by their Abbreviations lower wages than they would have earned without the ALMP: Active Labour Market Policy; ANAPEC: Agence Nationale de Promo- program. As for young people aged 25–34, they have to tion de l’Emploi et des Compétences; ATT : Average treatment on treated; IC: Insertion contract; MAD: Moroccan dirham; PE: Permanent contract; PIC: work more and with less social security when they ben- Professional integration contract. efit from the program. These mixed and small effects suggest revisiting the Acknowledgements The authors thank the editor and the two anonymous reviewers for their valu- "Idmaj" program to improve its effectiveness. Refining able comments on a previous version of this manuscript. its targeting, rethinking its conditionalities and orienting them towards the sustainability of its impacts, strength- Authors’ contributions CA is the major contributor in writing and correcting the manuscript. The ening its complementarity with other levers of the ALMP, three authors decided the method used in the study and assisted in its appli- and enhancing its flexibility are all adjustments that seem cation. All authors read and approved the final manuscript. necessary to attenuate the negative effects of the program Authors’ information in question and increase its net benefit. Indeed, it would, CA: Professor-researcher at the University Mohammed V of Rabat. His research for example, be more appropriate to replace the current interests currently focuses on the evaluation of the impact of active labor single salary ceiling serving as the basis for tax and social market programs in developing countries. exemptions by differentiated limits to reduce the employ - HK: Laureate of the Master’s Degree in Economics and Evaluation of Public ers’ bias towards low wages. Likewise, for enhancing the Policies, Mohammed V University of Rabat. 24-month professional internship impact on productivity, SN: PhD student in economics, Mohammed V University of Rabat. Research clear specifications and regular monitoring and evalua - field: Employment economics. tions should be put in place. The flexibility of this intern - ship should also be increased, by allowing, for example, Funding Not applicable. the employee to change his initial employer more than once if this would prove useful for improving his employ- Availability of data and materials ability. More broadly, and as this wage subsidies program The data used in the present study are available from the Moroccan ministry of employment, direction of the national observatory of employment. But is used in Morocco not to address cyclical unemploy- restrictions apply to the availability of these data, which were used at our ment, but rather the structural one, employment-friendly request for this study. Data are however available from the authors upon rea- growth and structural reforms of the labour market will sonable request and with permission of direction of the national observatory of employment. be required to increase the employability of young gradu- ates sustainably. Declarations Finally, it should be mentioned that this paper is not without limitations and shortcomings. Indeed, as noted Ethics approval and consent to participate above, the design of the program whereby treatment Not applicable. exposure is not determined by some external selection Consent for publication criteria, such as a random allocation or an exogenous eli- Not applicable (the paper has used anonymized data, so no consent was gibility threshold, involved the use of the propensity score obtained). matching method based on the strong assumption of Competing interests selection on observables. That said the scope of our find - The authors declare that they have no competing interests. ings would have been much more robust and interest- Received: 10 December 2020 Accepted: 13 June 2021 ing if relevant pre-treatment data were available for our two groups. Indeed, the post-treatment survey led us to exclude certain co-variables that are not fixed over time but potentially crucial for participation and outcomes. References This has reduced the heterogeneity on the observables as Almeida, R., Orr, L., Robalino, D.: Wage subsidies in developing countries as long as program participation has been modelled based atool to build human capital: design and implementation issues. IZA J. Labor Policy 3, 1–24 (2014) on a few statistically significant co-variables. These ele - Angrist, J.D., Imbens, G.W., Rubin, D.: B: Identification of causal effects using ments could explain the relative weakness of the statisti- instrumental variables. J. Am. Stat. Assoc. 91(434), 444–455 (1996) cal significance of our results and their relative sensitivity Au Plan, H. C., & Mondiale, B: Le marché du travail au Maroc: défis et opportu- nités (2017). concerning unobservable factors. Another and no less 17 Page 12 of 13 A. Chatri et al. Auer, P., Efendioglu, L., Leschke, J.: Active labour market policies around the Garrido, M.M., Kelley, A.S., Paris, J., Roza, K., Meier, D.E., Morrison, R.S., Aldridge, world: Coping with the consequences of globalization, 2nd edn. Interna- M.: D: Methods for constructing and assessing propensity scores. Health tional Labour Office, Geneva (2005) Serv. Res. 49(5), 1701–1720 (2014) Banerji, A., Newhouse, D., Pacci, P., Robalino, D.: Working through the crisis: jobs Gautié, J. Le chômage. La découverte (2015). and policies in developing countries during the great recession. World Grimm, M., Paffhausen, A.: Interventions for employment creation in micro, Bank, Washington DC (2014) small and medium sized enterprises in low and middle income coun- Bausch, J., Dyer, P., Gardiner, D., Kluve, J., Kovacevic, S.: the impact of skills train- tries–a systematic review. University of Passau, Passau (2014) ing on the financial behaviour, employability, and educational choices Groh, M., Krishnan, N., McKenzie, D., Vishwanath, T.: Soft skills or hard cash? The of rural young people: findings from a randomized controlled trial in impact of training and wage subsidy programs on female youth employ- Morocco. Impact Report 6, 24 (2017) ment in Jordan. The World Bank, Washington (2012) Becker, S.O., Caliendo, M.: Sensitivity analysis for average treatment effects. Groh, M., McKenzie, D., Shammout, N., Vishwanath, T.: Testing the importance Stata. J. 7(1), 71–83 (2007) of search frictions, matching, and reservation prestige through rand- Bell, B., Blundell, R., Van Reenen, J.: Getting the unemployed back to work: the omized experiments in Jordan. The World Bank, Washington (2014) role of targeted wage subsidies. Int. Tax Public Fin. 6(3), 339–360 (1999) Groh, M., Krishnan, N., McKenzie, D., Vishwanath, T.: Do wage subsidies provide Bergemann, A., Van den Berg, G.: J: Active labor market policy effects for a stepping-stone to employment for recent college graduates? Evidence women in europe—a survey. Annales d’Economie et de Statistique 7, from a randomized experiment in Jordan. Rev. Econ. Stat. 98(3), 488–502 385–408 (2008) (2016) Betcherman, G., Dar, A., &Olivas, K: Impacts of active labor market programs: Handouyahia, A., Haddad, T., Eaton, F.: Kernel matching versus inverse prob- New evidence from evaluations with particular attention to developing ability weighting: a comparative study. Int. J. Math. Comput. Sci. 7(8), and transition countries, World Bank Social Protection Discussion Paper 1218–1233 (2013) Series no. 29142 (2004). Heckman, J.J., Robb, R.: Alternative methods for evaluating the impact of Blattman, C., & Ralston, L: Generating employment in poor and fragile states: interventions: An overview. J. Econ. 30(2), 239–267 (1985) evidence from labor market and entrepreneurship programs (2015). Heckman, J.J., Ichimura, H., Todd, P.: E: Matching as an econometric evaluation Broecke, S.: Tackling graduate unemployment in North Africa through employ- estimator: Evidence from evaluating a job training programme. Rev. Econ. ment subsidies: a look at the SIVP programme in Tunisia. IZA Journal of Stud. 64(4), 605–654 (1997) Labor Policy 2(1), 1–19 (2013) Heckman, J.J., Ichimura, H., Todd, P.: Matching as an econometric evaluation Bruhn, M: Can wage subsidies boost employment in the wake of an economic estimator. Rev. Econ. Stud. 65(2), 261–294 (1998) crisis? Evidence from Mexico. The World Bank (2016). Heckman, J.J., Lalonde, R.J., Smith, J.A.: The economics and econometrics of Burtless, G.: Are targeted wage subsdies harmful: evidence from a wage active labor market programs. In: Ashenfelter, O., Card, D. (eds.) Handbook voucher exeperement. Ind. Labor Relat. Rev. 39, 105–114 (1985) of labor economics. Springer, Amsterdam (1999) Caliendo, M., Kopeinig, S.: Some practical guidance for the implementation of Hirano, K., Imbens, G.W., Ridder, G.: Efficient estimation of average treatment propensity score matching. J. Econ. Surv. 22(1), 31–72 (2008) effects using the estimated propensity score. Econometrica 71(4), Caliendo, M., Hujer, R., Thomsen, S.: Identifying effect heterogeneity to improve 1161–1189 (2003) the efficiency of job creation schemes in Germany. Appl. Econ. 40(9), Huber, M., Lechner, M., Wunsch, C.: The performance of estimators based on 1101–1122 (2008) the propensity score. J. Econ. 175(1), 1–21 (2013) Calmfors, L., Forslund, A., & Hemstrom, M: Does active labour market policy Hujer, R., Caliendo, M., Thomsen, S.L.: New evidence on the effects of job work? Lessons from the Swedish experiences. Lessons from the Swedish creation schemes in Germany a matching approach with threefold Experiences (2002). heterogeneity. Res. Econ. 58, 257–302 (2004) Card, D., Hyslop, D.: Estimating the effects of a time-limited earnings subsidy Imbens, G.W., Lemieux, T.: Regression discontinuity designs: a guide to prac- for welfare-leavers. Econometrica 73(6), 1723–1770 (2005) tice. J Econ 142(2), 615–635 (2008) Card, D., Kluve, J.: What works? A meta-analysis of recent active labor market Kaldor, N.: Wage subsidies as a remedy for the unemployment. J. Polit. 5, 78 program evaluations. J. Eur. Econ. Assoc. 16(3), 894–931 (2018) (1936) Card, D., Kluve, J., Weber, A.: Active labour market policy evaluations: a meta- Katz, L: Wage subsidies for the disadvantaged. NBER Working Paper 5679 analysis. Econ. J. 120(548), F452–F477 (2010) (1996). Cho, Y., Honorati, M.: Entrepreneurship programs in developing countries: a Khandker, S., Koolwal, G., Samad, H.: Handbook on impact evaluation: quantita- meta regression analysis. The World Bank, Washington (2013) tive methods and practices. The World Bank, Washington (2009) Crépon, B., Duflo, E., Marc Gurgand, M., Rathelot, R., Zamoray, P.: Do labor Kluve, J.: The effectiveness of European active labor market programs. Labour market policies have displacement effects? Evidence from a clustered Econ. 17(6), 904–918 (2010) randomized experiment. Q. J. Econ. 128(2), 531–580 (2013) Kluve, J., Lehmann, H., Schmidt, C.: Active labor market policies in Poland: Dahlberg, M.: Forslund A: Direct displacement effects of labour market pro - human capital enhancement, stigmatization, or benefit churning? J. grammes. Scand. J. Econ. 107(3), 475–494 (2005) Comp. Econ. 27(1), 61–89 (1999) David, H., Houseman, S.: Do temporary-help jobs improvelabormarketout- Kluve, J., Puerto, S., Robalino, D., Romero, J.M., Rother, F., Stöterau, J., Weiden- comes for low-skilled workers? Evidence from" Work First". Am. Econ. J. kaff, F., Witte, M.: Interventions to improve the labour market outcomes Appl. Econ. 2(3), 96–128 (2010) of youth: a systematic review of training, entrepreneurship promotion, De Mel, S., McKenzie, D., Woodruff, C.: Wage subsidies for microenterprises. Am. employment services, and subsidized employment interventions. Camp. Econ. Rev. 100(2), 614–18 (2010) Syst. Rev. 12, 7 (2017) De Mel, S., McKenzie, D., Woodruff, C.: Labor drops: experimental evidence Kluve, J., Puerto, S., Robalino, D., Romero, J.M., Rother, F., Stöterau, J., Weiden- on the return to additional labor in microenterprises. The World Bank, kaff, F., Witte, M.: Do youth employment programs improve labor market Washington (2016) outcomes? A quantitative review. World Dev. 114, 237–253 (2019) Dubin, J.A., Rivers, D.: Experimental estimates of the impact of wage subsidies. Kvasnicka, M: Does Temporary Help Work Provide a Stepping Stone to Regular J. Econ. 56, 219–242 (1999) Employment? NBER, Working Paper N.w13843 (2009). Escudero, V., Kluve, J., Mourelo, E.L., Pignatti, C.: Active labour market Lechner, M.: Earnings and employment effects of continuous off-the-job train- programmes in Latin America and the Caribbean: evidence from a meta- ing in East Germany after uniication. J. Bus. Econ. Stat. 17(1), 74–90 (1999) analysis. J. Dev. Stud. 55(12), 2644–2661 (2019) Levy Yeyati, E., Montané, M., Sartorio, L: What works for active labor market Galasso, E., Ravallion, M.: Social protection in a crisis: Argentina’s Plan Jefes y policies? CID Working Paper Series (2019). Jefas. World Bank Econ. Rev. 18(3), 367–399 (2004) Martin, J.P., Grubb, D.: What works and for whom: a review of OECD Countries’ Galasso, E., Ravallion, M., Salvia, A.: Assisting the transition from workfare to experiences with active labour market policies. Swed. Econ. Pol. Rev. 8(2), work: a randomized experiment, industrial and labor relations review. Pol. 9–56 (2001) Res. Work. Pap. 57(5), 128–42 (2004) Micro‑econometric evaluation of subsidized employment in morocco: the case of the "Idmaj"… Page 13 of 13 17 McKenzie, D.: How effective are active labor market policies in developing Souag, A.: The heterogeneous effect of employment agencies program on countries? A critical review of recent evidence. World Bank Res. Observ. labor force behavior in Algeria: a dynamic approach. Springer, Berlin 32(2), 127–154 (2017) (2020) Moczall, A.: Eets d ff ’aubaine et de substitution d’un dispositif allemand de sub - Stephan, G.: Hétérogénéité sectorielle des effets d’un dispositif de subvention vention salariale pour demandeurs d’emploi difficilement employables. salariale sur les salaires et l’emploi en Allemagne. Travail et Emploi 139, Travail Et Emploi 139, 39–59 (2014) 60–74 (2014) Morgan, S.L., Harding, D.: J: Matching estimators of causal effects: prospects Stuart, E.A.: Matching methods for causal inference: a review and a look and pitfalls in theory and practice. Sociol. Methods Res. 35(1), 3–60 (2006) forward. Stat Sci 25(1), 1 (2010) Pallais, A.: Inefficient hiring in entry-level labor market. Am. Econ. Rev. 104(11), Venetoklis T: An evaluation of wage subsidy programs to SMEs utilising 3565–3599 (2014) propensity score matching, Research Reports 106, VATT Institute for Premand, P., Brodmann, S., Almeida, R., Grun, R., Barouni, M.: Entrepreneurship Economic Research (2004). education and entry into self-employment among university graduates. Woodbury, S.A., Spiegclman, R.G.: Bonuses to workers and employers to World Dev. 77, 311–327 (2016) reduce unemployment. Am. Econ. Rev. 77, 513–530 (1987) Rosenbaum, P.R.: Overt bias in observational studies. Observational studies, pp. Zhao, Z.: Data issues of using matching methods to estimate treatment 71–104. Springer, New York (2002) effects: an illustration with NSW data set. China Center Econ. Res. 15, Rosenbaum, P.R., Rubin, D.: The central role of the propensity score in observa- 1–37 (2003) tional studies for causal effects. Biometrika 70(1), 41–55 (1983) Roy, A.: Some thoughts on the distribution of earnings. Oxf. Econ. Pap. 3, Publisher’s Note 135–146 (1951) Springer Nature remains neutral with regard to jurisdictional claims in pub- Rubin, D.B.: Estimating causal effects of treatments in randomized and nonran- lished maps and institutional affiliations. domized studies. J. Educ. Psychol. 66(5), 688 (1974)

Journal

Journal for Labour Market ResearchSpringer Journals

Published: Jun 25, 2021

Keywords: Propensity score matching; Impact evaluation; Active labour market policy; Wage subsidies; C14; C31; H43; J38; J68

References