 Today I will present an impact evaluation of the public employment services in Colombia and as you would see I will try to understand in particular whether the traditional public system of labour intermediation is more or less effective than the private provision of employment services. And the motivation to conduct this research is that this partially related with in general the current state of the economy. Indeed we know that Latin America and the Caribbean has experienced a prolonged period of economic and social progress since the beginning of the 2000s. This has, of course, at least partially thanks to some external factors such as strong demand of exports from China. And this process has been, if not driven, certainly accompanied by innovative policy interventions such as conditional cash transfers which are generally credited for having contributed to the reduction of poverty and inequality in the region. However, this period of economic progress has recently been interrupted. Indeed we know first an increase in unemployment. And this figure shows how both the level and the rate of unemployment in the region have increased since 2014 and there are supposed to this increase is supposed to last at least until 2018. Then we also experience an increase in inequality, in particular the Gini index has increased in 8 out of 13 countries with available information between 2012 and 2013. And importantly the short term dynamics seem to interplay with structural economic challenges that remain unaddressed. Probably the most important one is the slow productivity growth rate that characterizes the region even during the period of economic expansion. And secondly, more from a labor market perspective, we know how many countries in the region still have a very high level of informal employment. This figure plots the relationship between the level of informal employment and gross national income per capita. And we can see how Latin American countries, which are in darker green in this figure, all tend to have a level of informal employment higher than what would have been predicted by their income level. So in this context we can ask ourselves what's next for the region, meaning which policy strategies can be adopted in order to face both these short term and structural challenges. And this presentation will be focused on the potential role in this context of actual labor market policies. As we know, these types of intervention can be divided into five different categories which broadly correspond to training, public core schemes, employment subsidies, self-employment and micro-interpretation, and labor market services, which will be the focus of this presentation. Of course the concept of actual labor market policies has been first developed in the US and then mostly applied to advanced economies. And therefore it needs to be adjusted in order to fit the context of emerging and developing countries. However, if we look at actual labor market policies specifically in Latin America and the Caribbean, we see that there is an increase in investment. So this figure shows spending in active and passive policies as a share of GDP. And we can see how spending in active policies has increased within 2000 and 2010 in almost all countries with available information. And it generally faces spending in passive policies. And this is certainly the case in Colombia. And then secondly, and within spending in active labor market policies, we see that there is a strong focus on training. So this figure shows the share of spending in active labor market policies by type of intervention. And we can see how training, which is in light green in this figure, generally accounts for the majority of spending. For instance, in Colombia, training accounts for around 85% of spending in active labor market policies, followed by labor market services, which are another 10% of spending. Training and labor market services are the main sources for spending in active labor market policies in Colombia. And interestingly, these two types of services are provided by the same institution in the country, which is the National Learning Center Service, or CENA. And Colombia is a country where there is a potentially very important role that can be played by labor market intermediation. Indeed, there is a strong level of informal employment, which is higher than the regional average, and has not decreased during the past decade despite sustained economic growth. Additionally, Colombia has a strongly segmented labor market. This figure is based on longitudinal data and shows three year transition rates across employment status. And we can see, for instance, that of those that were in an informal job in 2010, only a small minority had switched to a formal job three years later. And then finally, we know that there are very high turnover rates. So, for instance, the average job duration in Colombia is equal to six years compared to an OCD average of 10 years. So, all these indicators seem to suggest that there is a potentially important role that can be played by improving the quality of labor market matching. However, this seems to be a largely missed opportunity. In particular, only one percent of job matches in the country, of course, every year through the public employment services. And this compares badly both with respect to advanced economies. For instance, this share is equal to 10% in the European Union, but also with respect to other countries in the region. For instance, this share is equal to 4% in Brazil. The system of public employment services that is offered in Colombia is a very traditional one in which job seekers and employer can choose among different services according to their needs. So, for instance, job seekers can either opt for some simple city preparation or opt for more structured labor orientation. There are no eligibility and entry requirements. There is no connection with the unemployment benefit system, whose coverage is anyway very limited in Colombia. And all these services can be accessed either online or through the public employment services centers. We can now go to the estimation strategy of this paper. Of course, the main challenges in identifying a causal impact from participation in public employment services is that these services are generally available to everybody and free of charge. And at the same time, especially in a highly informal labor market, these services are likely to attract a specific and non-random support of the population. So, in this case, the paper will conduct the impact evaluation via preventives for matching, claiming that we can control for all variables that jointly determine the decision to participate and the outcome of interest. And in this context, I will try to overcome these identification obstacles by relying both on the institutional context and on results from previous studies. Starting with the institutional context, there has been a recent reform in 2013 that now requires private agencies to operate as public providers of labor interpretation. This means that they need to provide the same set of services free of charge to all interested participants. And at the same time, the 2013 reforms now require employers to post all vacancies in the system. And this system can be accessed both by private and public employment services. So, as we said, some help from the institutional context and some help from the econometric theory. In particular, it's true that the conditional independence assumption is generally regarded as a strong assumption. However, it's not true that we're kind of blind in front of it. In particular, previous studies have shown which variables are needed to control for in order to ensure the validity of the conditional independence assumption in the context of impact evaluation of active labor market policies. And additionally, of course, we can do some robustness tests and check in particular to what extent our estimates would be biased by the presence of unobserved heterogeneity. So, in this case, the analysis will be done by everyone seems for matching and the conditional independence assumption would be defined both with respect to the institutional context and results from previous studies. Very big on the data, it comes from the Colombian household survey between 2013 and 2016. So, after the reform of labor intermediation, this is the main survey in the country being around 150,000 pounds per year. It has a comprehensive coverage of all social and demographic aspects as it comes from the merging of two previous surveys. And then from this sample, we obtain our final sample, the first looking only at individuals that have used either the public or the private employment agencies, as we discussed before. And then a focused analysis only on individuals with previous job experience. And this is because previous studies have shown how accounting for this previous career experience is a key variable to control for in order to ensure the validity of the conditional independence assumption. And then finally, the question on the job search method that is used is asked only to individuals that are employed. This means that the analysis cannot investigate the impact of the policy on the probability of finding a job, but only on the point of the job that is found. This, of course, limits a bit the scope of the analysis, but it also has some important advantages. In particular, it limits the risk of contamination by which, for instance, individuals use different job search methods at the same time. And it also reduces the risk of the support for treatment by which, for instance, individuals might be registered with the public employment services, but then, in practice, do not use these services. And if we look at the descriptive statistics, we can see how selection into treatment follows some dynamics that we could expect given the institutional context and for which we can control. In particular, for instance, we have said how, in Colombia, the same institution is in charge of providing both vocational training and labor market services. And in the descriptive statistics, this results into a higher share of individuals using the public employment services, coming from a vocational training track equal to 51%, compared to, for instance, a 27% among those individuals that use private employment agencies. If we instead turn to household characteristics, we see that there are no notable differences between the two group of countries pointing towards a possible level of comparability and homogeneity between these two groups. So, very briefly, a more technical look at the methodology. There are three main steps to be followed in the estimation of preventive score matching. The first one is, of course, the estimation of the preventive score. And in this case, the first charge concerns the functional work form to be used. However, the previous studies have shown that it's generally equivalent to use other acrobitorologic function. And the second step, instead, includes the decision of the variable to be included in the estimation of preventive score. In this case, I include a different set of variable, one after the other, and then jointly, and then look at traditional indicators of goodness of it. The second step, instead, concerns the charge of the matching up for the rhythm. In fact, we should make sure that the different algorithms that we can use, for instance, neighborhood matching, caliper, kernel, reduce the bias between the treatment and the control group. And we can also obtain a graphical representation of the balance of the preventive score between the treatment and the control group. So, for instance, this figure shows both the density distribution and the box plots of the preventive score for the control and the treatment group before and after matching. And we can see how, in both cases, matching substantially reduces the bias between the control and the treatment group. And then, finally, the last step concerns checking data of common support. First, we need to make sure that the number of observations that is dropped is small. And in this case, it's never about one to two percent of the sample. And then we should check that the probability of participation is higher among the three than the control group. And, for instance, this is the classic figure which shows the distribution of the preventive score in the two groups. It has the traditional well-behaved form with only two observations of support here in red. We can now look at the results. I divided them into the effect of participation in standard measures of employment quality and self-recorded measures of employment quality. Starting from the format, we see that there is a small positive effect of participation on the probability of having a formal job. However, this effect is statistically significant only for women. And the effect on wages is instead different by-scales group. In particular, we see how participation in the public employment services tends to increase the wages of the low-skilled but when it has a negative, it will do not statistically significant effect on the wages of the high-skilled. And there is also a negative effect from participation in the public employment services on working hours. And this effect is relatively robust and consistent across the different groups. However, it seems that this effect, this last effect plays through higher compliance with the labor legislation. In particular, we see how participating in the public employment services reduces the probability of working an excessive number of hours, which is in this case defined as more than 48 hours per week. If we turn instead to the self-recorded measures of employment quality, we find that the public employment services performs relatively better than the private employment agencies in all these measures. So for instance, participation in the public employment services increases the probability of being consistent with the employment contract. It also increases the probability of reporting work-life balance to be satisfied with the job and consequently it reduces the probability of being willing to change the job. And with this, we can turn to the conclusions. I think that the paper makes two main contributions to the literature. The first one is that this is the first impact evaluation of the public employment services in Colombia and one of the first studies specifically targeted to emerging and developing countries. The public employment service is an important institution which most countries have. However, our understanding of the functioning and the benefits of this institution is mostly based so far on a purely intuitive economic reasoning while there is a general lack of impact evaluations of the effect of participation in the public employment services. And then secondly, it's one of the first studies that looks at the relative effectiveness of the public versus private system of services provision. And this is a question that many governments, especially in developing countries with fiscal constraints, are asking themselves. The results reveal how this two different system of public or private provision can complement each other. For instance, if we think of the results on wages, we've seen how the public employment services seem to be relatively effective in placing low-skilled individuals in better paid jobs. For instance, by guaranteeing higher compliance with minimum wage legislation. However, it's not equally successful in placing highly qualified individuals in good and good quality jobs. And at the same time, we've also seen how the public employment services is more successful than the private employment agencies on all sort of reported measures of job quality. Which leads us to some key policy conclusions. The first one is that labor intermediation can play an important role in emerging and developing countries both in terms of raising productivity but also improving job satisfaction. And the second one is that the externalization of labor market services to private providers can represent a valid strategy in the sense that these two systems of provision seem to complement each other and they can also target different groups of the workforce. However, it's important to guarantee the presence of well-defined externalization contracts in order to avoid, for instance, that private employment agencies can just care about placing individuals in the first available job without adequately taking into consideration the quality of the job that is found. Thank you very much. Thank you.