Alternative models of activation policies: the experience of public oriented services

Abstract

Purpose

The purpose of this paper is to understand the impact of activation policies in contexts of public-oriented employment services.

Design/methodology/approach

The fieldwork was conducted in Italy, using the regional case of Emilia-Romagna as a representation of public-oriented models. The empirical research relies primarily on quantitative research methods by means of impact evaluations based on very rich and recent administrative data that includes 20,014 observations. These are integrated with some interesting insights from qualitative research tools by means of semi-structured interviews with stakeholders in the fieldwork.

Findings

The overall level of effectiveness of activation policies does not seem to be particularly high. However, the results show that, to some extent, there are more advantages for job seekers furthest from the labor market. The stronger efforts made to assist those disadvantaged groups are probably related to a loss in the overall effectiveness of the system.

Research limitations/implications

This analysis has focused on a substantially homogeneous and economically developed region. Indeed, Italian activation policy reforms have been implemented differently in different areas of the country. Similar to other Southern European countries, they are characterized by regional fragmentation and implementation gaps in activation policies.

Practical implications

The findings of the present study are relevant to policymakers who deal with activation policies and to both public and nonpublic organizations involved in this field. It seems plausible to support the possibility that public-oriented models could represent a proficient alternative to proposals relying heavily on market-based interventions. Such evidence becomes particularly interesting in the aftermath of the Great Recession when this model is confronted with the consequences of the economic crisis.

Originality/value

The paper adds to the literature on public-oriented models, while overcoming some of the limitations of earlier research, which has been restricted mainly to cases with early traditions of marketized services.

Citation

Scarano, G. (2020), "Alternative models of activation policies: the experience of public oriented services", International Journal of Sociology and Social Policy, Vol. 40 No. 3/4, pp. 382-408. https://doi.org/10.1108/IJSSP-01-2020-0011

1. Introduction

The legacy of the Great Recession continues to be an issue in many countries. Although, in most cases, pre-crisis employment levels have been restored, employment rates have differed markedly across population groups. Some groups in society continue to be affected by weak labor market attachment. In this context, activation policies play an important role in the public debate as they have the potential to help all groups in society, especially those who face barriers and disincentives to finding a good job. Of the various aspects of activation policies, marketization has probably been discussed and investigated most (Van Berkel et al., 2011). For example, contracting out public employment services (PES) to private providers is deemed to help overcome a lack of staff, skills, and efficient processes. However, the literature to date is focused mainly on cases with an early tradition of marketized services, while less attention has been paid to experiences that still rely on public-oriented services. In these respects, Italy can be used as an interesting field of study because of its poor implementation of market-based strategies.

The aim of this paper is to investigate the effects of participating in activation policies in contexts where there is less reliance on market-based interventions, and in particular, to examine what consequences such policies can have for groups affected by weak labor market attachment. This interest comes from the consideration of the ambiguous effects that market-based interventions may have on those job seekers furthest from the labor market due to the risk of private providers’ discrimination.

The data and information analyzed in this study were sourced from the analysis of very rich and recent administrative data set collected in an Italian region, Emilia-Romagna. The activation policies model considered may be investigated as an alternative to those relying on large scale marketization. The emerging considerations should open the debate to the point that some may claim that a more private-oriented provision of services should be reduced in favor of more public-oriented designs, despite the fact that policymakers in very different contexts continue to support the marketization strategy (Larsen and Wright, 2014).

The paper is structured as follows: section 2 summarizes the theoretical background on activation policies and marketization; section 3 outlines the different marketization trends and governance regimes as they pertain to activation across European countries; section 4 presents the Italian landscape in this field, and the specific model is analyzed; section 5 is dedicated to the data and methods. The main findings and conclusions are presented in the second part of the paper.

2. Activation policies and marketization

Most work on welfare state reforms emphasizes the role of activation and activation policies, becoming part of the widely accepted political debate in Europe. These policies have been considered as an increasingly important part of the “modernization” of welfare states in Europe (Barbier, 2004; Bonoli, 2010; Aurich, 2011; Van Berkel et al., 2012). The activation paradigm emerged between the Nineties and early Two-Thousands, based on the premise that long-term unemployment can have detrimental effects on individual employability, thereby triggering structural unemployment (Jackman and Layard, 1991; Zimmermann et al., 2014). For making sense of this concept, activation should be understood as depending on a framework favoring a combination of supportive and demanding policies (Eichhorst et al., 2008). The aim of these policies when they were introduced was to integrate the broader parts of the population into the labor market and to mobilize groups with significant barriers to the labor market integration. This required a broader array of new forms of intervention, as well as new structures of policy implementation. Consequently, different policy reforms have increasingly put emphasis on the governance of activation policies, that is, how the provision and delivery of such policies are organized (Borghi and Van Berkel, 2007; Van Berkel et al., 2012).

A growing body of literature has been devoted to various aspects of the governance of the welfare state and activation policies (Considine, 2001; Considine and Lewis, 2003; Van Berkel et al., 2011, 2012). Some of the aspects most frequently discussed include decentralization, marketization, collaboration/network, and performance management (Zimmermann et al., 2014). Undoubtedly, the introduction of markets for the provision of social services is the most salient and contested form of new governance that European welfare states have been confronted with, although to very different extents in different countries. Traditionally, PES had a monopoly-status in most countries, being the only institution that provided job-placement (Freedland et al., 2007). European states progressively started to allow private employment agencies to operate only during the Nineties, making PES operate in a competitive market. This process allowed for the entry of new forms of governing the provision of services by means of outsourcing the selection of services from public to private providers (Zimmermann et al., 2014).

Marketization can be considered one of the main governance aspects of activation policies mainly because it is supported by the assumption that contestability increases the efficiency and effectiveness of provision. Nevertheless, marketized provision can also be accompanied by negative side effects. As Crouch (2014, p. 8) specifies, « marketization will normally be expected to have two consequences: gains in efficiency that flow from the rational cost-effectiveness of the market; and losses in the damage done by the negative externalities that marketization almost necessarily produces». Unintended consequences can occur because public goals are not necessarily in line with private providers' interests. The private providers’ goal is to maximize their short-term profits, whereas the government tries to manage costs and benefits in a difficult trade-off between unemployment benefits and reintegration costs (Bruttel, 2005). In particular, the high costs involved with serving difficult clients may encourage private providers to assist only those who are easiest to place (“the creaming effect”) (Le Grand and Bartlett, 1993; Koning and Heinrich, 2013; Greer et al., 2018). In such cases, there can be very important social implications in terms of the equitability of performances.

The risk of private providers’ discrimination highlights an apparent contradiction in marketization. Market-based interventions may worsen precarity without achieving the stated goal of increasing labor-market participation (Greer et al., 2017). The consideration of such ambiguous consequences should open up the debate concerning the role of marketization for the governance of activation policies. Concerns, such as this one, may induce policymakers to focus more strongly on alternative forms of provision characterized by more public-oriented designs.

3. Marketization trends and governance regimes across Europe

Arrangements for the involvement of private employment agencies may include different contracting forms, all of them generating very different outcomes (Van Berkel et al., 2012). From the perspective of comparative welfare state research, the crucial question is whether there is a general and converging or a context-specific pattern of marketization strategies. Several studies come from the analyses of the countries that have proceeded furthest with such models on a large scale (Sol and Westerveld, 2005; Struyven; Steurs, 2005; Bredgaard and Larsen, 2008; Finn, 2011; Greer et al., 2017). They use, for the large part, qualitative means to explore patterns of the new trends of contracting-out logic and the implementation of market-based instruments in the provision of services. All these studies agree that marketization trends and their effects vary across countries due to a wide range of labor market and welfare state institutions. Future research on the issue should devote more attention to the specific national circumstances and institutions to yield a more in-depth understanding. Any attempt to study marketization should take into account two challenges: (1) the timing of the diffusion of the marketization across countries’ employment services landscapes, and (2) the kind of marketization, which entails quick and large scale contracting out of public services (Greer et al., 2017). Countries, in these respects, vary with regard to the relative market shares of the labor market intermediaries, ranging from a strongly public to mainly private sector driven system of employment services. Relying on the categories proposed by Van Berkel et al. (2012), it is possible to identify three main families of European countries characterized by different models of activation policies in consideration of their degree of marketization. The first one is that of the “committed marketizers” (UK and Netherlands), who were among the first to implement reforms characterized by the strong market provision. The second group corresponds to the “modernizers” (France, Germany, and Sweden), where marketization has been implemented more gradually and with a weaker degree compared to the first group. The last group is that of the “slow modernizers” or “latecomer countries” (the Mediterranean and East-European countries), characterized by less advanced or totally absent reforms dealing with marketization. This does not mean that there are no activation policies at all in these countries, but that there is still a strong reliance on the public bodies for their provision. The literature to date is focused mainly on cases with an early tradition of market-based interventions, thus concentrating on the first group of countries, which have represented benchmarking experiences for the political debate. Recall, too, that marketization is often characterized by ambiguous effects. This suggests that it would be possible to reverse the analytical perspective and put more attention to contexts characterized by less marketization. Among the group of latecomer countries, the present work will use the context of Italy to analyze the outcome of activation policies based on public-oriented services.

4. Activation policies in Italy: the case of Emilia-Romagna region

The variation in activation policies across European countries are mainly variations of scope and intensity, reflecting the different countries’ particular starting points, histories, institutional settings, and cultures. In this context, Italy has been characterized by continuous tension to introduce market service provisions. Italy represents a case of a “latecomer” country because the implementation of such reforms has proceeded very slowly and in a very fragmented way (Arcidiacono et al., 2011). The Mediterranean welfare regime of Italy for a long-time has been characterized by a rigid labor market with a familistic system of social protection (Ferrera, 1996). Italy has been historically marked by strong gender, education, and age-specific dualization, creating “outsiders” groups of the labor market, with very weak attachment. In this context, the quality of placement has always been very inefficient, facilitating the little matching of labor demand and supply and negatively affected by an overly centralized and bureaucratic system (Gualmini, 1998). From the beginning of the Nineties, the empowerment of the PES became a central issue in the political debate. The introduction of private intermediation and a change of mission for PES were considered unavoidable. Extensive reforms between the end of the Nineties and the early Two-Thousands aimed at these goals. Decentralization was supposed to improve the adjustment of labor market policy schemes to the regional demand (Zirra and Buchkremer, 2008). Regions took over the previous placement offices of the ministry, but there was hardly any coordination of the regional labor market policies (Borghi and Van Berkel, 2007). After these reforms, national programs for the improvement of the PES were no longer possible. Instead, the marketization process had to be completed by the regions[1]. Decentralized decisions on outsourcing contributed to further slowing down of the process. In recent years following, the government has tried to accelerate the implementation of marketization. In particular, in 2014 and 2015, the Italian government highlighted the need to modernize labor market institutions by implementing the “Jobs Act” as a major reform of the labor market. In September 2015, the decree concerning activation policies was adopted, which included the introduction of the “reintegration voucher”— a national scheme based on a quasi-market model in employment services—as well as the attempt to recentralize the marketization process. Nevertheless, Italian activation policies have remained only marginally touched by marketization trends. In the vast majority of the Italian regions, public bodies have been the crucial actors in setting the overall rules and procedures, and they have also been directly involved in the delivery process. The further aspect that passive labor market policies are still managed by national public bodies has contributed to this situation.

As previously pointed out, there is little sense to talk of Italian national activation policies considering that they are decentralized[2]. In consideration of this specific context, the analysis will focus on a specific region, Emilia-Romagna. Looking at this region allows for the analysis of an area that has both high economic development and public-oriented services. In the present study, this case is considered as representative of activation policies model with less reliance on market-based interventions. This corresponds to an important Italian economic area, historically based on a wide network of small and medium enterprises organized by industrial districts. In particular, this productive model has been fostered by a traditional left-wing political subculture, which has contributed to strengthening a network of trust relationships that is very important for the development of small businesses, and which is supported by a tradition of cooperative and coordinated industrial relations (Trigilia, 1986). It is not an aim of this work to explore the politics behind models of activation policies, but it is possible to say that this regional case has been the result of the precise choices of its policymakers. Regional employment services provision has generally been conceived of as a response to a perceived market failure, thus requiring publicly controlled employment centers (Rinaldi, 2002). Private providers just support, but never replace, the role of the public ones that remain at the core of activation policies management. The regulation of marketization by means of accreditation criteria, in this case, would have arrived only at the end of 2016, after the pressure induced by the Jobs Act national reform. This analysis, starting from the consideration of the more ambiguous effects of marketization, will look at the performance of this activation policy model concerning the integration into the labor market for the weakest groups.

5. Data and methods

Empirical data have been collected from administrative sources of the metropolitan area of the regional main city of Bologna, considering that this represents the most important economic area in the region. These data were released from the regional labor market agency (Agenzia regionale per il Lavoro dell'Emilia-Romagna). The main advantage of administrative data is that it is possible to include almost the complete population of interest in the sample and reconstruct the labor market history of these subjects in a particular geographical area. The main limitation is that there are a few observable personal characteristics to be controlled. However, the information available is rich enough to follow a counterfactual approach. This is necessary to determine the extent to which post-assistance outcomes are a function of referral to and participation in the activation policies. In these terms, the “counterfactual situation” is the one that would have been observed in the absence of activation policies, since it may be that job seekers who obtained a job after the activation measures would have done so anyway. Reconstructing a counterfactual situation means to exploit the possibility of observing the work situation of those subjects not participating in the activation measures, defined as “nonparticipants” or the “control group,” which are comparable to the “participants” or “treatment group” in terms of observable characteristics relevant for the selection process (Martini and Sisti, 2009).

In the present study, the reference period for the activation policies observed was from the 1st of January 2015 to the 30th of September 2015[3]. The “treatment” then corresponded to any activation measure that a first-time unemployed person participated in during the period observed. Job seekers may cumulate different measures, thereby ending up with different activation paths[4]. For simplicity reasons, the regional model has been evaluated in its totality, considering all the various measures aggregated as it was a single “program” (Sianesi, 2004).

Information about the control group was obtained from the certifications of unemployment status registered during the same period. In fact, a general pool of job seekers was registered as unemployed at a certain point in time, but only some of them participated in the activation measures since the participation was voluntary. Many job seekers then registered without following any program, despite having the characteristics required for eligibility. The final dataset included 20,014 cases, 7,036 participants and 12,978 nonparticipants.

The analysis relied on the quasi-experimental technique of propensity score matching (PSM). This represents one of the most commonly used techniques, with regard to studies in this field (Bryson et al., 2002). Participants and nonparticipants were matched on the propensity score, which is the estimated probability of belonging to the treatment group. It is estimated on the basis of all the variables that can influence both the probability of becoming treated and the outcome in case of nonparticipation. The greatest limitation faced when choosing similar techniques results from the consideration of the nonobservable characteristics relevant to the selection process. This is the case, for example, of motivational influences that can play an important role in the choice to participate in a program, especially in cases of voluntary participation. In such cases, it is crucial to carefully select the variables to be taken into account, in order to exploit all the observable features available and to ensure that they are as close as possible to the nonobservable ones. In the present study, the propensity score was calculated on the basis of the following characteristics: age, gender, educational level, citizenship, maximum qualification achieved in previous experiences, previous labor market history, and month of registration[5]. Descriptive statistics are presented in Table AI in the appendix.

The propensity score was estimated by means of a probit regression using as independent variables the observable characteristics of the job seekers (see Table AII in the appendix). Using probit regression makes it possible to identify those sociodemographic characteristics that have more influence on the decision to participate in activation policies. The youngest and eldest unemployed persons, with a good level of education and no previous work experience, have a higher probability of becoming participants. The intermediate age group (between 30 and 45 years old), as well as being a woman and having already qualified working experiences, are not characteristics that significantly affect the probability of participating in the activation measures. Other characteristics influencing the propensity score are associated with more disadvantaged conditions in the labor market, such as being a foreign unemployed person. Regarding the month of registration, there may be evidence of some seasonality of unemployment. With respect to the beginning of the period of observation, the registration during the “winter months,” such as February and March, is significant and associated with a negative probability of participation, whereas the same probability is positive for registrations during the “summer months,” from June to September. Probably, this aspect reflects the seasonality of those fixed-term contracts ending before summer. As for the previous labor market history, there is a significant and positive relationship with respect to the probability of having worked recently before registering as unemployed. This could be due to the number of unemployment benefit recipients.

Outcome variables used to measure the effects of activation policies correspond with the employment status at 3, 6, 9, 12, 15, 18, 21, and 24 months after beginning the treatment (dummies variables: 1 = employed, 0 = unemployed). The observation period for the outcome, then, was up to September 2017, in order to take into account the long-term consequences of the participation in active labor market initiatives.

In particular, in order to assess the outcome of the model also from the equity point of view, the effect was specifically observed for subgroups defined on the basis of their sociodemographic characteristics observed. Information on the outcome variables and the labor market histories is derived from “compulsory communications” (COs – Comunicazioni Obbligatorie). These represent administrative communications that employers have to send to employment offices whenever an employment relationship is started, modified, renewed, or terminated.

For the purpose of the estimation strategy, radius matching on the propensity score was performed. Furthermore, the matching was limited only to those observations concerning the common support. Although a strict caliper of 1 percent was imposed, a lack of common support turned out to never be an issue (see the last column in Table AIII in the appendix). Overall, matching of the estimated propensity score, sufficiently balanced the distribution of all covariates between the two groups, as shown in Table AIII in the appendix. The mean standardized bias (MSB) is considerably reduced after matching, 1 percent on average, where Caliendo and Kopening (2008) considers that a reduction of the bias to 3 percent could be sufficient.

The impact was estimated as the average effects of participation on activation policies (D = 1) versus nonparticipation, (D = 0) on employment status, (Y) for participants (average treatment effect on the treated – ATT).

ATT=E(Y1−Y0|D=1)=E(Y1|D=1)−E(Y0|D=1) 

The conditional independence assumption (CIA) must hold to correctly identify the effects and avoid that the results could be biased when the outcome is partly dependent on the selection of the individuals to be treated. Selection bias should be present then if the assignment to the treatment correlates with the outcome. In the present study, information related to the previous labor market history over a long period of time should assume that this condition is satisfied. Moreover, since the standard errors of the ATT did not take into account that the propensity score was estimated, they were replaced by means of the bootstrap technique, performed with 100 re-samplings.

In order to support the interpretation of the quantitative evidence with qualitative evidence too, semi-structured interviews were conducted with five regional representative actors of the regional training and labor policies department between 2014 and 2016, both at the political and administrative level[6]. These actors were in charge of activation policies’ implementation[7].

6. Findings

In this section, the results of estimates of the effects on the probability of being employed after participating in the activation policies are reported. Table I shows the overall effect of participating in the activation measures at 3, 6, 9, 12, 15, 18, 21, and 24 months after the beginning of treatment.

Table I shows that the impact is generally very low. In particular, the effect tends to increase until 1.6 percentage points after one year from the beginning of the treatment. It decreases after this threshold, remaining one percentage point superior for the participants with respect to the nonparticipants at two years of distance from the treatment, but probably continuing to decrease. The increase in the effect from 3 to 12 months could be due to the progressive overcoming of an initial lock-in effect. These results could provide evidence that activation policies are not very effective, even though we could say they are still sufficient if we consider that a (small) positive impact remains in the long-term. This means that unemployed persons participating in the activation measures may always have a slightly higher probability of being employed compared to those who do not participate. In particular, we can also register how, after balancing for the observable characteristics, the probability of being employed is almost always reduced. This suggests the presence of a positive selection at the entrance of the program, that is, before the matching effect of participation in the active labor market initiatives was overestimated. Considering the variables that would affect the probability of being included in the program (Table AII), the positive selection could be a consequence of the number of well-educated unemployed persons and those with a short unemployment spell, characteristics generally associated with easier job-finding.

For assessing the equity dimension of the outcomes, it is necessary to observe if and how the impact may change for different groups of job seekers. Differences relative to the personal characteristics might not be fully captured simply by using these as control variables. In the following paragraphs, the effects are estimated for subgroups defined according to the observable sociodemographic characteristics. The outcome variable remains the probability of being employed at 3, 6, 9, 12, 15, 18, 21, and 24 months after the beginning of the treatment. For analyzing deeply, it is possible to discuss some intersections, particularly between ethnic origins and skill level, interacting with two very important characteristics to establish the job seekers’ labor market attachment.

6.1 Impact by age

Table II illustrates that the effect of participating in activation policies is substantially higher for the eldest job seekers. Similarly, there is an initial lock-in phase also here. After that, the impact increases, arriving at a maximum of almost six percentage points in one year, then decreasing but remaining between 3.5 and 4.5 percentage points in the long term. The impact for the other two age groups is null or negative (always particularly negative for the 30–45 age group). This picture makes it quite clear that activation policies concentrate their efforts on the eldest unemployed people.

6.2 Impact by gender

Table III shows that the impact of activation policies is stronger for women than men. In particular, it is always higher in all the post-assistance periods considered. For women, the general trend has been confirmed: the impact tends to reach its maximum after one year from the beginning of the treatment. For men, the impact is generally null and particularly negative in the short term. After the maximum level of 2.8 percentage points is reached, the impact for women decreases but remains over one percentage point in the long term. An initial lock-in effect can also be seen here for women.

6.3 Impact by the level of education achieved

Table IV presents how activation policies tend to be more effective as the educational level increases. For graduate job seekers, after the initial lock-in phase, the impact is more concentrated at the one-year peak, where it reaches 3.5 percentage points. Then, it decreases and remains around two percentage points in the long term. For job seekers with the lower levels of education, the effect is much lower. It becomes remarkable, after one year, just for job seekers with secondary educational attainment. For job seekers with compulsory education only, the initial negative effect is more pronounced.

6.4 Impact by citizenship

The effect of participating in activation policies here appears similar to that seen for the eldest people.

Table V shows that, after the initial lock-in phase, the effect reaches its maximum in one year at 3.5 percentage points, then it slightly decreases and remains at almost 3 percentage points. On the contrary, for Italians, the impact is always null and particularly negative in the very short term.

6.5 Impact by the level of qualification achieved in previous experiences

For simplicity reasons, in this study, the categories of qualified experience and executives and managers have been aggregated as single “qualified experience.” The effect of participating in activation policies for different groups, defined according to the previous level of qualification, seems to be stronger after one year for the job seekers without experience. However, the picture here does not seem to be completely unambiguous.

As Table VI shows, considering job seekers without previous experience, the impact increases until one year, registering at almost three percentage points, then it decreases in the long term, becoming null. For the group of job seekers with nonqualified experience, there is a sort of “polarization” of impact over time: low effect in the very short term and high effect in the very long term. For job seekers with previous qualified experience, the initial negative effect is more pronounced.

6.6 Impact by unemployment spell

For simplicity reasons, the categories used for unemployment spell have been synthesized into two main groups, distinguishing long-term unemployed (unemployment spell of more than one year) and short-term unemployed (unemployment spell of less than one year). From Table VII, it is possible to see that the impact is much stronger for job seekers with a longer unemployment spell.

Short-term unemployed always register a null effect and a particularly negative effect in the short term. For the long-term unemployed, the one-year peak reaches almost five percentage points, but then the impact decreases between one and two percentage points in the long term. Differently from the trend seen for the eldest people, for women and foreigners, there is a shorter initial lock-in phase here, and the impact is more concentrated in the short term (see Table VII).

6.7 Impact by intersections between skill level and ethnic origins

For adding depth to the analysis, it is possible to discuss some intersections, particularly between ethnic origins and skill level—these two interacting characteristics are very important for establishing job seekers’ labor market attachment.

The variables chosen to communicate information on the skill level are education and the level of qualification achieved in previous experiences. As indicated in Tables VIII and IX, this type of intersection allows us to obtain very interesting information regarding foreigners. Those with higher skill levels have greater effects. This applies both to foreigners who have achieved more qualified experiences in the past, but to a much greater extent, for graduates. For graduated foreigners, after the initial nine months lock-in phase, the impact increases reaching almost eight percentage points in the long term. Concerning qualification achieved in previous experiences, the impact for foreigners is always slightly superior to that for Italians. It is particularly relevant for foreigners with previous qualified experience, overcoming three percentage points after one year and in the long term.

These data also lend to the explanation of the concentration of effects observed on graduates previously (Table IV), since that impact is pushed by the outcomes on foreigners. It could be said that, for the same skill level, activation policies return better performance for foreigners.

7. Addressing the weakest targets—aims and goals of the Emilian model

For supporting the interpretation of the findings outlined in the previous paragraph, further evidence is presented here from semi-structured interviews conducted with local representative actors. For this purpose, a few short representative extracts are presented as display citations.

Stakeholders in the Emilian LMPs have always made no secret of their hostility toward market-based interventions. This is the reason why the implementation of regional accreditation rules has proceeded very slowly over time, considering that it is meant to strengthen private-oriented services. According to Representative C, this reason can also be related to the outbreak of the economic crisis:

Our system has shown that it can work very well, even without accreditation criteria. This has been true, especially for the very hard period of the economic crisis. We have obtained what we needed to support our public employment centers from specific agreements or ad-hoc procedures designed for particular projects and targets. We have found time by time the providers to involve, in consideration of our priorities at each specific moment.

(Representative C)

Activation policies specifically designed to address the consequences of the economic crisis seemed to have concentrated their efforts on the eldest job seekers. All the other subjects interviewed stressed that an increasing number of people experienced unemployment for the first time in many years as a consequence of the economic crisis. The difficult economic situation resulted in a context where employment services became overloaded by the huge number of people who remained unemployed as a consequence of the economic crisis.

In general, the hostility toward market-based interventions has been accompanied by strong attention to social goals. As Representative A said, activation in Emilia-Romagna aims for social integration and social cohesion, and these are typical traits of other regional social policy areas:

By means of the public providers, we aim at ensuring a minimum level of service provision. Around these structures, we have a network of further providers, both public and nonpublic, from which we get specialized services that the public employment centers alone are not able to provide. It is fundamental for us that the provision of active labor market policies should be in line with our equity standards, as well as what happens with social care and health care services.

(Representative A)

All the other persons interviewed confirmed that much of the activation policies’ goals are focused on those job seekers who, in addition to being unemployed, suffer further social hardships, which often involve the rest of their households. These are situations in which there are social needs not solely and directly related to the unemployment condition.

For the same reasons, the accreditation regime implemented in 2016 could be considered as “heavy accreditation,” in consideration of its structure. The most important element concerning the design of the Emilian accreditation regime is its division into two separate lists of accredited operators. It is possible to identify the first area of accreditation concerning “standard interventions,” conceived for job seekers and employers, and a second area concerning “inclusive interventions,” specifically oriented toward most disadvantaged categories.

8. Concluding remarks

The findings outlined in this paper may open the debate concerning models of activation policies based on public-oriented services. The model has effectively demonstrated that its efforts are primarily oriented toward some of the weakest groups of job seekers. However, the overall level of effectiveness does not seem to be particularly high. The stronger efforts made to assist those disadvantaged groups are probably related to a loss in the overall effectiveness of the system.

The particular case analyzed would suggest that, despite the poor level of effectiveness, such models may prove to work in the integration of weakest groups during particular economic downturns.

The evidence about the eldest and long-term unemployed job seekers may be consistent with the priority given by policymakers to people who suffered most from the economic crisis by losing their jobs. Further evidence about women and foreigners could be a result of the aim to increase labor market participation and investment in groups previously not in the active population. In particular, it is interesting to consider how outcomes of activation concentrate their effects on foreigner groups characterized by high skill levels. The skill level then represents a sort of “accelerator” for foreigners; the higher it is, the greater the impact of activation policies. This aspect may also be related to the economic crisis and the need to attract and involve new skills and expertise developed in other contexts.

The analysis of the outcomes of activation frequently reveals an initial lock-in mechanism. It is likely that the job search is delayed because job seekers in the short term are attending to specific initiatives. In this way, the approach seems to be characterized more by long-term human capital development, similar to approaches described as “enablement” or “human investment” (Torfing, 1999; Barbier, 2004; Taylor-Goby, 2004; Dingelday, 2007; Bonoli, 2010; Martin, 2014). This is different from “work-first” or “workfare” approaches, in which incentive-based policies have the aim of moving unemployed people toward employment as soon as possible (King, 1995; Peck, 2001).

It seems plausible to support the idea that public-oriented models could represent a proficient alternative to proposals relying heavily on market-based interventions. Interventions are targeted at expanding employment opportunities for the weakest groups. The findings presented here suggest that the principles needed to govern such design of activation policies do not need to be invented from scratch. They can be borrowed from existing innovative governance arrangements, already developed in response to the economic uncertainties faced by past and present governments (Rodrik and Sabel, 2019). However, to call for the replication of such models, further detailed information would be needed. Italy, as well as other Southern European countries (above all Spain), are characterized by deep regional fragmentation in the field of social services, and in particular, employment policies.

This work has focused on a substantially homogeneous and economically developed region. In Italy, the reforms of activation policies have been implemented differently across the country, and this may represent a limitation for the analysis. There are significant implementation gaps due to varying administrative capacities, which are higher in the Northern and Central regions and more limited in Southern ones in comparative terms (Graziano and Winkler, 2009). The dependence on differentiated local conditions calls for high-dimensional policy space. In such contexts, it becomes necessary to develop effective multilevel governance. At the same time, it also becomes fundamental to set the conditions that prevent multilevel governance from leaving room for further territorial fragmentation. In looking forward, larger, systematic studies could, therefore, offer a more in-depth and reliable representation of the evolution of empowerment dynamics in the provision of public-oriented services.

Further research is also needed to evaluate the economic sustainability of such models. Economic sustainability refers to the use of strategies for employing existing resources optimally in order to achieve a beneficial balance between public spending and social goals over the long term. The lack of resources is a historically strong deficiency in Italian PES, which was especially evident during the years of the harsh economic crisis. During those years, more activation policies were required, in line with what occurred in other European countries (OECD, 2013, 2019).

A recent reform introducing a new minimum income scheme, The Citizens Income, has allocated two billion euros for Italian activation policies between 2019 and 2020. These funds are destined for resources and staff of public employment centers. This may represent a massive injection of resources, paving the way for research on further developments as compared to the present situation.

Table I

Impact evaluating the employment condition at a certain point in time after the beginning of the treatment

Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Before matching	0.0061	0.0155	0.0203	0.0366	0.0293	0.0283	0.0289	0.0301
	(0.0073)	(0.0072)	(0.0071)	(0.0069)	(0.0068)	(0.0067)	(0.0066)	(0.0065)
ATT after matching	−0.0154**	−0.0055	0.0001	0.0157**	0.0090	0.0088	0.0094	0.0108*
	(0.0064)	(0.0072)	(0.0065)	(0.0064)	(0.0065)	(0.0088)	(0.0077)	(0.0057)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table II

Impact evaluating the employment condition after the beginning of the treatment by age groups

Age	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
15–29	Before matching	−0.0397	−0.0263	−0.0262	−0.0204	−0.0295	−0.0278	−0.0217	−0.0205
		(0.0140)	(0.0135)	(0.0132)	(0.0126)	(0.0123)	(0.0119)	(0.0117)	(0.0115)
	ATT after matching	−0.0049	0.0077	−0.0005	−0.0032	−0.0106	−0.0115	−0.0085	−0.0086
		(0.0123)	(0.0115)	(0.0134)	(0.0121)	(0.115)	(0.0115)	(0.0119)	(0.0114)
30–45	Before matching	−0.0068	−0.0014	−0.0011	0.0111	0.0055	0.0053	0.0007	−0.0004
		(0.0120)	(0.0117)	(0.0115)	(0.0112)	(0.0109)	(0.0108)	(0.0106)	(0.0105)
	ATT after matching	−0.0413***	−0.0312**	−0.0312***	−0.0172	−0.0188	−0.0170	−0.0202**	−0.0209**
		(0.0115)	(0.0123)	(0.0106)	(0.0113)	(0.0118)	(0.0106)	(0.0087)	(0.0104)
Over45	Before matching	0.0573	0.0677	0.0806	0.1089	0.1009	0.0973	0.0991	0.1028
		(0.0122)	(0.0123)	(0.0122)	(0.0120)	(0.0119)	(0.0118)	(0.0117)	(0.0117)
	ATT after matching	−0.0025	0.0093	0.0244**	0.0559***	0.0438***	0.0381***	0.0386***	0.0427***
		(0.0105)	(0.0114)	(0.0123)	(0.0126)	(0.0119)	(0.0108)	(0.0099)	(0.0109)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table III

Impact evaluating the employment condition after the beginning of the treatment by gender groups

Gender	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Women	Before matching	0.0195	0.0296	0.0369	0.0535	0.0426	0.0373	0.0378	0.0371
		(0.0100)	(0.0099)	(0.0097)	(0.0095)	(0.0093)	(0.0092)	(0.0090)	(0.0089)
	ATT after matching	−0.0113	0.0004	0.0101	0.0282***	0.0187**	0.0129	0.0142*	0.0137
		(0.0104)	(0.0091)	(0.0089)	(0.0087)	(0.0094)	(0.0088)	(0.0080)	(0.0090)
Men	Before matching	−0.0094	−0.0006	0.0010	0.0170	0.0139	0.0181	0.0186	0.0217
		(0.0108)	(0.0106)	(0.0105)	(0.0102)	(0.0100)	(0.0099)	(0.0098)	(0.0097)
	ATT after matching	−0.0234**	−0.0127	−0.0120	0.0016	−0.0002	0.0054	0.0049	0.0086
		(0.0093)	(0.0091)	(0.0092)	(0.0087)	(0.0091)	(0.0096)	(0.0097)	(0.0089)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table IV

Impact evaluating the employment condition after the beginning of the treatment by educational level groups

Education	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Compulsory	Before matching	0.0220	0.0337	0.0385	0.0473	0.0486	0.0480	0.0512	0.0455
		(0.0125)	(0.0124)	(0.0123)	(0.0120)	(0.0118)	(0.0118)	(0.0116)	(0.0115)
	ATT after matching	−0.0253***	−0.0104	−0.0054	0.0034	0.0069	0.0072	0.0109	0.0061
		(0.0093)	(0.0114)	(0.0115)	(0.0107)	(0.0116)	(0.0116)	(0.0112)	(0.0106)
Diploma	Before matching	0.0001	0.0083	0.0106	0.0313	0.0177	0.0209	0.0168	0.0229
		(0.0113)	(0.0110)	(0.0108)	(0.0105)	(0.0102)	(0.0101)	(0.0100)	(0.0099)
	ATT after matching	−0.0154	−0.0075	−0.0047	0.0168*	0.0043	0.0060	0.0016	0.0087
		(0.0109)	(0.0106)	(0.0103)	(0.0091)	(0.0108)	(0.0098)	(0.0095)	(0.0090)
University degree	Before matching	−0.0098	−0.0023	0.0063	0.0260	0.0172	0.0079	0.0139	0.0168
		(0.0154)	(0.0150)	(0.0148)	(0.0143)	(0.0140)	(0.0138)	(0.0135)	(0.0134)
	ATT after matching	−0.0009	0.0088	0.0161	0.0353**	0.0238*	0.0132	0.0206	0.0186
		(0.0136)	(0.0133)	(0.0138)	(0.0144)	(0.0138)	(0.0133)	(0.0137)	(0.0129)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table V

Impact evaluating the employment condition after the beginning of the treatment by citizenship groups

Citizenship	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Foreigner	Before matching	0.0397	0.0385	0.0427	0.0580	0.0504	0.0510	0.493	0.0500
		(0.0136)	(0.0133)	(0.0130)	(0.0124)	(0.0120)	(0.0118)	(0.0117)	(0.0115)
	ATT after matching	0.0006	0.0047	0.0147	0.0353***	0.0272**	0.0277**	0.0261**	0.0279***
		(0.0137)	(0.0133)	(0.0115)	(0.0126)	(0.0110)	(0.0110)	(0.0124)	(0.0105)
Italian	Before matching	−0.0164	−0.0020	0.0020	0.0167	0.0095	0.0076	0.0099	0.0108
		(0.0088)	(0.0087)	(0.0086)	(0.0084)	(0.0083)	(0.0082)	(0.0081)	(0.0080)
	ATT after matching	−0.0307***	−0.0163**	−0.0119	0.0033	−0.0035	−0.0058	−0.0038	−0.0023
		(0.0080)	(0.0082)	(0.0088)	(0.0076)	(0.0078)	(0.0077)	(0.0082)	(0.0077)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table VI

Impact evaluating the employment condition after the beginning of the treatment by previous experience groups

Previous experience	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
No experience	Before matching	0.0219	0.0317	0.0359	0.0429	0.0211	0.0115	0.0180	0.0222
		(0.0151)	(0.0162)	(0.0165)	(0.0165)	(0.0163)	(0.0161)	(0.0158)	(0.0155)
	ATT after matching	0.0133	0.0197	0.0231	0.0292*	0.0085	−0.0039	0.0051	0.0143
		(0.0117)	(0.0134)	(0.0157)	(0.0162)	(0.0170)	(0.0158)	(0.0158)	(0.0143)
Nonqualified experience	Before matching	0.0010	0.0127	0.0212	0.0490	0.0385	0.0552	0.0490	0.0759
		(0.0213)	(0.0209)	(0.0207)	(0.0204)	(0.0201)	(0.0198)	(0.0197)	(0.0194)
	ATT after matching	−0.0378*	−0.0267	−0.0189	0.0090	−0.0019	0.0144	0.0093	0.0368**
		(0.0199)	(0.0224)	(0.0219)	(0.0183)	(0.0209)	(0.0199)	(0.0182)	(0.0175)
Qualified experience	Before matching	0.0096	0.0174	0.0208	0.0367	0.0338	0.0314	0.0310	0.0263
		(0.0087)	(0.0085)	(0.0084)	(0.0081)	(0.0080)	(0.0079)	(0.0078)	(0.0078)
	ATT after matching	−0.0203**	−0.0096	−0.0046	0.0132*	0.0107	0.0101	0.0097	0.0054
		(0.0090)	(0.0082)	(0.0082)	(0.0076)	(0.0077)	(0.0082)	(0.0070)	(0.0071)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table VII

Impact evaluating the employment condition after the beginning of the treatment by unemployment spell groups

Unemployment spell	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Less than one year	Before matching	−0.0052	0.0056	0.0073	0.0264	0.0236	0.0257	0.0239	0.0244
		(0.0.0083)	(0.0080)	(0.0079)	(0.0077)	(0.0075)	(0.0074)	(0.0073)	(0.0073)
	ATT after matching	−0.0285***	−0.0159*	−0.0121	0.0077	0.0054	0.0081	0.0061	0.0061
		(0.0082)	(0.0082)	(0.0078)	(0.0075)	(0.0074)	(0.0069)	(0.0069)	(0.0067)
More than one year	Before matching	0.0414	0.0494	0.0759	0.0855	0.0628	0.0512	0.0555	0.0564
		(0.0106)	(0.0133)	(0.0148)	(0.0158)	(0.0161)	(0.0162)	(0.0162)	(0.0161)
	ATT after matching	0.0259**	0.0251*	0.0459***	0.0492***	0.0248	0.0106	0.0149	0.0214
		(0.0119)	(0.0138)	(0.0148)	(0.0179)	(0.0155)	(0.0180)	(0.0150)	(0.0152)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table VIII

Impact evaluating the employment condition after the beginning of the treatment by groups defined on the basis of citizenship and education

Previous experience	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Italians with compulsory education	Before matching	−0.0146	0.0039	0.0087	0.0163	0.0121	0.0117	0.0221	0.0215
		(0.0158)	(0.0158)	(0.0158)	(0.0156)	(0.0154)	(0.0154)	(0.0152)	(0.0150)
	ATT after matching	−0.0132	0.0088	0.0135	0.0182	0.0137	0.0134	0.0227	0.0200
		(0.0141)	(0.0151)	(0.0162)	(0.0150)	(0.0153)	(0.0154)	(0.0140)	(0.0129)
Foreigners with compulsory education	Before matching	0.0453	0.0427	0.0457	0.0513	0.0623	0.0602	0.0554	0.0424
		(0.0206)	(0.0202)	(0.0198)	(0.0189)	(0.0183)	(0.0181)	(0.0179)	(0.0175)
	ATT after matching	0.0069	0.0123	0.0142	0.0244	0.0350**	0.0334*	0.0246	0.0150
		(0.0181)	(0.0191)	(0.0167)	(0.0180)	(0.0173)	(0.0171)	(0.0187)	(0.0164)
Italians with diploma	Before matching	−0.0104	0.0010	0.0022	0.0196	0.0122	0.0167	0.0076	0.0116
		(0.0132)	(0.0130)	(0.0127)	(0.0124)	(0.0121)	(0.0120)	(0.0119)	(0.0118)
	ATT after matching	−0.0112	−0.0009	−0.0001	0.0168	0.0089	0.0131	0.0022	0.0069
		(0.0115)	(0.0128)	(0.0123)	(0.0111)	(0.0116)	(0.0112)	(0.0121)	(0.0116)
Foreigners with diploma	Before matching	0.0144	0.0129	0.0171	0.0433	0.0141	0.0117	0.0210	0.0321
		(0.0222)	(0.0216)	(0.0210)	(0.0199)	(0.0195)	(0.0190)	(0.0188)	(0.0185)
	ATT after matching	0.0040	0.0061	0.0184	0.0434**	0.0108	0.0082	0.0197	0.0310
		(0.0245)	(0.0228)	(0.0216)	(0.0209)	(0.0206)	(0.0181)	(0.0205)	(0.0192)
Italians with university degree	Before matching	−0.0407	−0.0278	−0.0202	−0.0009	−0.0116	−0.0276	−0.0132	−0.0143
		(0.0178)	(0.0173)	(0.0170)	(0.0166)	(0.0163)	(0.0160)	(0.0157)	(0.0156)
	ATT after matching	−0.0169	−0.0089	−0.0016	0.0146	0.0029	−0.0192	−0.0022	−0.0047
		(0.0167)	(0.0177)	(0.0142)	(0.0162)	(0.0152)	(0.0176)	(0.0158)	(0.0173)
Foreigners with university degree	Before matching	0.0780	0.0807	0.0873	0.1037	0.0962	0.1090	0.0921	0.1046
		(0.0317)	(0.0311)	(0.0303)	(0.0290)	(0.0281)	(0.0275)	(0.0271)	(0.0266)
	ATT after matching	0.0342	0.0464	0.0600**	0.0798***	0.0679**	0.0781***	0.0592**	0.0799***
		(0.0355)	(0.0331)	(0.0297)	(0.0283)	(0.0336)	(0.0334)	(0.0294)	(0.0314)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table IX

Impact evaluating the employment condition after the beginning of the treatment by groups defined on the basis of citizenship and previous experience

Previous experience	Matching	After 3 months	After 6 months	After 9 months	After 12 months	After 15 months	After 18 months	After 21 months	After 24 months
Italians with no experience	Before matching	0.0199	0.0331	0.0346	0.0349	0.0110	0.0075	0.0146	0.0247
		(0.0176)	(0.0187)	(0.0190)	(0.0190)	(0.0188)	(0.0185)	(0.0182)	(0.0179)
	ATT after matching	0.0165	0.0251	0.0282	0.0243	−0.0013	−0.0063	0.0025	0.0162
		(0.0176)	(0.0176)	(0.0197)	(0.0200)	(0.0223)	(0.0189)	(0.0176)	(0.0186)
Foreigners with no experience	Before matching	0.0389	0.0376	0.0446	0.0615	0.0481	0.0213	0.0251	0.0101
		(0.0292)	(0.0322)	(0.0334)	(0.0333)	(0.0329)	(0.0325)	(0.0317)	(0.0310)
	ATT after matching	0.0305	0.0372	0.0329	0.0442	0.0403	0.0066	0.0124	0.0043
		(0.0303)	(0.0344)	(0.0375)	(0.0328)	(0.0355)	(0.0288)	(0.0347)	(0.0299)
Italians with nonqualified experience	Before matching	−0.0512	−0.0268	−0.0241	0.0057	0.0110	0.0281	0.0192	0.0363
		(0.0312)	(0.0310)	(0.0308)	(0.0306)	(0.0304)	(0.0301)	(0.0300)	(0.0297)
	ATT after matching	−0.0471	−0.0258	−0.0300	0.0047	0.0071	0.0230	0.0149	0.0266
		(0.0287)	(0.0316)	(0.0310)	(0.0326)	(0.0313)	(0.0292)	(0.0309)	(0.0316)
Foreigners with nonqualified experience	Before matching	0.0274	0.0253	0.0360	0.0606	0.0315	0.0468	0.0431	0.0828
		(0.0296)	(0.0289)	(0.0282)	(0.0276)	(0.0269)	(0.0262)	(0.0259)	(0.0256)
	ATT after matching	0.0139	0.0096	0.0181	0.0411	0.0091	0.0264	0.0203	0.0556**
		(0.0318)	(0.0312)	(0.0340)	(0.0300)	(0.0272)	(0.0283)	(0.0281)	(0.0250)
Italians with qualified experience	Before matching	−0.0094	0.0013	0.0050	0.0197	0.0148	0.0095	0.0104	0.0060
		(0.0104)	(0.0101)	(0.0100)	(0.0098)	(0.0096)	(0.0095)	(0.0095)	(0.0094)
	ATT after matching	−0.0197**	−0.0070	−0.0025	0.0127	0.0081	0.0036	0.0048	0.0005
		(0.0100)	(0.0096)	(0.0107)	(0.0096)	(0.0105)	(0.0093)	(0.0102)	(0.0086)
Foreigners with qualified experience	Before matching	0.0245	0.0263	0.0304	0.0470	0.0488	0.0527	0.0518	0.0444
		(0.0166)	(0.0160)	(0.0156)	(0.0149)	(0.0143)	(0.0142)	(0.0141)	(0.0139)
	ATT after matching	0.0100	0.0104	0.0164	0.0363***	0.0345**	0.0370***	0.0377***	0.0297***
		(0.0165)	(0.0164)	(0.0149)	(0.0141)	(0.0154)	(0.0133)	(0.0150)	(0.0127)

Note(s): Bootstrap standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table AI

Descriptive statistics of the sample (N = 20,014)

Variables	No.	Percent
Age
15–29	5,348	26.72
30–45	7,539	37.67
Over45	7,127	35.61
Gender
Female	9,314	53.46
Education
Compulsory	7,097	35.46
Diploma	8,419	42.07
University degree	4,498	22.47
Citizenship
Foreigner	5,256	26.26
Maximum qualification achieved in the previous experiences
No experience	3,896	19.47
Nonqualified experience	2,340	11.69
Qualified experience	10,888	54.40
Executives and managers	2,890	14.44
Previous labor market history
Employed last 3 months	14,638	73.14
Employed 4th–6th month before	13,973	69.82
Employed 7th–9th month before	13,592	67.91
Employed 10th–12th month before	13,501	67.46
Employed 13th–15th month before	13,451	67.21
Employed 16th–18th month before	13,201	65.96
Employed 19th–21st month before	12,948	64.69
Employed 22nd–24th month before	12,630	63.11
Month of registration
January	2,249	11.24
February	2,243	11.21
March	2,412	12.05
April	2,319	11.59
May	2,271	11.35
June	1,994	9.96
July	2,272	11.35
August	1,673	8.36
September	2,581	12.90

Table AII

Estimation of the propensity score

Variables	Treatment
Age
15–29 (ref.)	–
30–45	−0.0228
	(0.0243)
Over–45	0.0775***
	(0.0254)
Gender
Female	0.0145
	(0.0188)
Education
Compulsory (ref.)	–
Diploma	0.0840***
	(0.0217)
University degree	0.0760***
	(0.0271)
Citizenship
Foreigner	0.372***
	(0.0213)
Maximum qualification achieved in the previous experiences
Nonqualified experience (ref.)	–
No experience	0.194***
	(0.0458)
Qualified experience	0.00135
	(0.0302)
Executives and managers	−0.0391
	(0.0387)
Previous labor market history
Employed last 3 months	0.0982***
	(0.0367)
Employed 4th–6th month before	0.127***
	(0.0482)
Employed 7th–9th month before	0.0216
	(0.0512)
Employed 10th–12th month before	−0.0180
	(0.0499)
Employed 13th–15th month before	−0.0911*
	(0.0486)
Employed 16th–18th month before	0.0125
	(0.0516)
Employed 19th–21st month before	0.0843
	(0.0559)
Employed 22nd–24th month before	−0.0575
	(0.0437)
Month of registration
January (ref.)	–
February	−0.0808**
	(0.0392)
March	−0.0955**
	(0.0386)
April	0.0323
	(0.0386)
May	0.0281
	(0.0389)
June	0.118***
	(0.0400)
July	0.178***
	(0.0385)
August	0.0966**
	(0.0420)
September	0.122***
	(0.0376)
Constant	−0.769***
	(0.0548)
Observations	20,014

Note(s): Standard errors in parentheses. ***p < 0.01; **p < 0.05; *p < 0.1

Table AIII

Indicators of covariates balancing, before and after matching

Sample	No. treated before	No. Nontreated before	Probit psR2 before	Probit psR2 after	Pr > χ2 after	Mean standardized bias before	Mean standardized bias after	No. Lost to CS after
Total	7,036	12,978	0.020	0.000	1.000	4.4	0.5	1
15–29 years old group	1,884	3,464	0.014	0.000	1.000	6.7	0.6	1
30–45 years old group	2,586	4,953	0.029	0.000	1.000	6.8	0.7	2
Over45 years old group	2,566	4,561	0.039	0.001	0.999	7.4	1.4	9
Female group	3,825	6,875	0.024	0.000	1.000	5.1	1.2	0
Male group	3,211	6,103	0.019	0.000	1.000	5.0	1.0	3
Compulsory education group	2,435	4,662	0.028	0.000	1.000	5.5	0.7	1
Diploma group	3,022	5,397	0.016	0.000	1.000	4.5	0.6	3
University degree group	1,579	2,919	0.026	0.001	1.000	6.8	1.1	1
Foreigner group	2,380	2,872	0.017	0.001	1.000	7.7	1.0	4
Italian group	4,651	10,106	0.009	0.000	1.000	4.9	0.5	1
No experience group	1,437	2,459	0.011	0.000	1.000	4.6	1.1	0
Nonqualified experience group	864	1,476	0.033	0.001	1.000	7.0	1.2	6
Qualified experience group	4,735	9,043	0.024	0.000	1.000	5.5	1.0
Short-term unemployment group	5,136	9,494	0.027	0.000	1.000	5.3	0.8	0
Long-term unemployment group	1,454	2,700	0.024	0.000	1.000	8.5	0.8	3
Italians with compulsory education group	1,397	3,384	0.010	0.000	1.000	3.8	0.9	0
Foreigners with compulsory education group	1,038	1,278	0.020	0.000	1.000	9.7	1.0	1
Italians with diploma group	2,114	4,324	0.008	0.000	1.000	3.7	0.7	0
Foreigners with diploma group	908	1,073	0.027	0.001	1.000	7.6	1.2	1
Graduated Italians group	1,141	2,398	0.017	0.001	1.000	6.4	1.0	1
Graduated foreigners group	438	521	0.062	0.003	1.000	15.2	1.3	7
Italians with no experience group	1,064	1,926	0.013	0.002	0.982	5.8	2.0	0
Foreigners with no experience group	373	533	0.007	0.001	1.000	4.1	1.4	0
Italians with nonqualified experience group	364	862	0.032	0.001	1.000	6.7	1.2	2
Foreigners with nonqualified experience group	500	614	0.029	0.002	1.000	10.4	2.0	1
Italians with qualified experience group	3,224	7,318	0.008	0.000	1.000	4.3	0.4	3
Foreigners with qualified experience group	1,511	1,725	0.031	0.001	1.000	9.0	1.4	0

Notes

1. At the regional level, marketization should have been regulated by means of accreditation criteria. Such regulation should define the selection process and the level of control of the public bodies over private agents.

2. Some regions attempted to introduce some experimental market-based schemes. In particular, one of the most consolidated experiences is that of Lombardy (Sabatinelli and Villa, 2015).

3. In this way, the last period was observed preceding the changes introduced by the national decree concerning activation policies, which came into effect in October 2015.

4. The activation policies considered in the present study refer mainly to employment assistance and training. The former corresponds to interventions, such as counseling, orientation, placement measures, and also monitoring of job seekers’ efforts. The latter concerns interventions aimed at skill contents (or its consolidation/improvement/reorientation). Training measures may differ in priority according to a different trade-off: general education versus specific skills, soft versus hard skills, and the type of learning activities, whether in the classroom, on the job, or alternating educational courses and work experience (Crepon and Van der Berg, 2016).

5. Each of the characteristics considered can distinguish the following categories:

1. age: 15-29; 30-45 and over-45

2. gender: 1 female, 0 male

3. level of education achieved: compulsory, diploma, university degree

4. citizenship: 1 foreigner, 0 Italian

5. maximum level of qualification achieved in previous experiences: no experience, non-qualified experience, qualified experience, executives and managers

6. month of registration

7. Previous labour market history is calculated on the basis of eight dummies, corresponding to the employment condition (1=employed; 0=unemployed) in each of the three-months intervals observed retrospectively from the date of registration as unemployed: last three months; between the fourth and the sixth month before; between the seventh and the ninth month before; between the 10^th and the 12^th month before; between the 13^th and the 15^th month before; between the 16^th and the 18^th month before; between the 19^th and the 21^st month before; between the 22^nd and the 24^th month before.

   Information on the labor market histories, in particular, is crucial because it should also capture unobservable variables (Caliendo et al., 2017).

6. This followup becomes important to strengthen the analysis and to counteract the weaknesses and limitations that may be associated with the PSM method (Bryson et al., 2002).

7. The persons interviewed in this study have been anonymized and listed as “Representative A,” “Representative B,” “Representative C,” “Representative D,” and “Representative E.”