Resumen de Essays on skills formation
Two main trends have characterized the economy of developed countries in the twentieth century, the evocative "race between education and technology" and the so-called "quiet revolution" in women’s labor supply.
The compelling and interconnected debates that have developed around these two expressions have greatly inspired my three "essays on skills formation". The growing literature on the dynamics of the supply of college graduates clearly influenced the first chapter of this work, where I focus on the importance of skills endowments in explaining educational attainments. The other two chapters deal instead with the role of institutions in promoting the formation and use of skills at work. In particular, in the second one, together with Clémentine Van Effenterre, we enter into the last chapter of women’s labor supply dynamics and analyze the impact of institutional constraints and flexible working schedules on mothers’ employment decisions. Finally, the works on the task-replacing technological change led me to study the factors shaping firms and workers’ incentives to invest in on-the-job training, by specifically focusing on the effects of employment protection legislation on training investments.
Before introducing the three chapters in more details, I find it opportune to briefly summarize the key elements of the "race between education and technology", the crucial steps in the "women’s quiet revolution", and the literature that has arisen around these topics. The former expression was firstly coined by Tinbergen in 1974 (Tinbergen 1974) and lately became the title of a famous book by Goldin and Katz (Goldin and Katz 2009). According to these authors skill-biased technological change constantly increased the relative demand for skilled workers over the twentieth century. As the supply of educated labor followed the pace of demand, returns to education and the college wage premium, at their highest at the beginning of the century, consequently declined. However, starting from the 1980s, the rate of growth of high-skilled workers began to slow down. Accordingly, the college wage premium rose once more until recovering the levels encountered at the beginning of the century. All researchers recognize the merits of Goldin and Katz’s book in accounting for the main labor market trends that characterized an entire century. However, several studies also highlight that this work cannot account for two major (and complementary) trends that started in the early 1990s (Acemoglu and Autor 2012, Autor, Katz, and Kearney 2006, Autor, Levy, and Murnane 2003). First, earnings dynamics have been characterized by a rapid growth at the upper and lower deciles of the wage distribution, than at the median. Secondly, a similar ushaped pattern has characterized employment, exhibiting a more rapid growth in high-skilled and low-skilled occupations, than in middle-skilled jobs. Acemoglu and Autor (2012), in their review of Goldin and Katz’s book, define these trends as wage and job polarization. To account for them, the authors propose to increase Tinbergen canonical model of demand and supply of high- versus low-skilled workers with two elements. First, they introduce a distinction between high-, middle- and low-skilled workers. Secondly, they relax the equivalence between workers’ skills and tasks, as this allows them to account for the fact that the assignment of skills to tasks can evolve over time, in particular when the set of tasks demanded in the economy is altered by technological advancements, or the dynamics of globalization. Observing that machines have replaced the routine tasks performed primarily by medium-skilled workers, a task-replacing technological change clearly becomes more suited to explain the recent trends of wage and job polarization. Yet, we are still far from understanding how to govern these phenomena (Acemoglu and Restrepo 2016).
Another strand of literature focuses on the slow-down in the supply of college graduates and highlights the challenges that raising educational attainment involves (Angrist, Lang, and Oreopoulos 2009, Dynarski and Scott-Clayton 2013, Oreopoulos and Dunn 2013, Scott-Clayton, Crosta, and Belfield 2014, Stinebrickner and Stinebrickner 2012, Turner 2004). These studies show that information asymmetries, financial constraints, motivation and preparation are all key factors to take into account when designing policies aimed at raising college completion. Finally, it is important to stress that Goldin and Katz’ work refers to the experience of the United States, and the same is true for most of the studies that analyze the topics of college enrollment and completion. However, phenomena like the rise in income inequality (Dustmann, Ludsteck, and Schönberg 2009, Piketty 2015), wage and job polarization (Goos and Manning 2007, Michaels, Natraj, and Van Reenen 2014), and the increase in college drop-out (De Paola and Scoppa 2014, Fack and Grenet 2015, Hübner 2012, Garibaldi, Giavazzi, Ichino, and Rettore 2012) are not confined to the United States and affect most European countries as well.
Concerning the other main trend, the so-called quiet revolution in women’s labor supply, Goldin (2006) shows that, in parallel to the race between education and technology and strictly connected to it, the last century has been characterized by a slow convergence in the roles of men and women, and this is clearly true for all developed countries (Blau and Kahn 2013). A narrowing has occurred between men and women in labor force participation, working hours, life-time working experience, occupations, and college majors. Women have even overtaken men in educational attainments. According to Goldin, these dynamics have occurred thanks to three fundamental changes in women’s believes – especially of married women – that became clear in the 1970s: women changed their time horizon with respect to life-time labor force participation, from intermittent and brief to long and continuous, and this clearly affected their human capital investments; women changes the way they perceived their job, from a necessity to sustain household income to an expression of their own identity; women affirmed their role in the household decision making process. All these changes were favored by several and cumulative exogenous factors: the increase in demand for clerical workers in the 1920s; the creation of scheduled part-time work in the 1940s; the diffusion of electric household technologies since the 1950s; the spreading use of the contraceptive pill at the end of the 1960s. The changes in beliefs accordingly translated in a constant decline of women’s income elasticity and the correspondent increase of the substitution elasticity in their Slutsky equation. These patterns have in turn found a reflection in a steady increase of women’s labor force participation and hours worked. Goldin concludes her seminal work by looking at the current situation – a prelude to Goldin (2014). Since 1990, female labor force participation rates and the fraction of women working full-time are no longer raising. The proportion of active women in their thirties is stacked at around 75 percent in the United States. Blau and Kahn (2005) claim that this is due to a strong reduction in women’s own wage elasticity. According to Goldin, this pattern must instead be read in conjunction with the demographic changes that led women to postpone childbearing. If this is the case, how to manage work and family duties becomes the focus of the last chapter of the revolution in female labor supply. Goldin (2014) claims that in this last chapter the gender gap in participation and wages may well be eliminated if only women’s quest for flexibility in the work environment was satisfied. In some occupations working longer hours and/or a regular presence at work might indeed be more rewarded than in others. This could be the case, in particular, in those professions where it is important to build solid relationships with co-workers, attend frequent meetings, take key decisions, and perform tasks under pressure. The continuous presence at work and the availability to work long hours should be particularly valuable in these contexts, or, in other words, the cost of a flexible working schedule might be especially high in these occupations. This argument may well explain why the gender wage gap remains largest at the top of the wage distribution, as highlighted in the last review on the topic made by Blau and Kahn (Blau and Kahn 2016). As flexibility is particularly valuable for women, measuring and reducing its costs should then be the focus of the current gender debate, according to Goldin.
At the same time, other studies such as Wiswall and Zafar (2016) insist on the importance of gender preferences in explaining labor market choices and career patterns. And the experimental literature pioneered by Niederle (Buser, Niederle, and Oosterbeek 2014, Niederle 2014) shows that gender differences in reaction to competition exist and should not be neglected in the gender debate. Finally, Blau and Kahn (2013) warn that the introduction of "family-friendly" policies such as parental leave and part-time work may well boost female labor force participation, but also encourage part-time work and employment in lower level positions. This would explain why the proportion of active women in the United States has decreased in the last twenty years relative to many OECD countries that promoted these policies, but also why women in the United States are more likely than women in other countries to have full-time jobs and to work as managers or professionals. The debate on this last chapter of women’s labor supply could never be more open.
With this work, I hope to contribute to these debates by drawing from the experience of three European countries. In the first chapter, I study the phenomenon of college drop-out in Italy. The fact that the United States, with a tough selection at entrance and very high tuition fees, have approximately the same college drop-out rate, at 20 percent, as Italy, with basically no selection at entrance and relatively low tuition fees, led me to think that students’ preparation for college might play an important role in explaining this phenomenon. For this reason, I analyze the impact of college remedial education. In the second chapter, together with Clémentine Van Effenterre, we look at female labor supply in France. Studying the last chapter of the "quiet revolution" in a country where the proportion of active women is beyond 80 percent seems particularly appropriate to me. Finally, in the last chapter, I analyze the relationship between employment protection legislation (EPL hereafter) and on-the-job training in the United Kingdom, a country where job protection rises with seniority, more than 50 percent of workers engage in employer-financed training – against an European average that is lower than 30 percent – and yet, wage and job polarization are constantly increasing.
In detail, in "Is College Remedial Education a Worthy Investment? New Evidence From a Sharp Regression Discontinuity Design" I analyze the impact of college remedial courses on students’ decisions and performance. To enhance college completion, an increasing number of higher-education institutions are introducing college remedial courses. The goal of these courses is to foster students’ readiness for college. However, assigning students to remedial education may also discourage them from continuing their studies if they interpret this as a negative signal on their chances of succeeding in college, they fear a stigma effect, or they doubt to be able to manage the increase in workload these courses involve. Moreover, assignment to remedial education is usually based on the performance in a placement test that students are required to take prior to entering college. This implies that some students may even decide not to enroll in college if placed in remediation. To assess the overall implications of this initiative, and in particular to study its impact on college enrollment, drop-out and performance, I collected a novel data set from the department of economics of an Italian university that introduced its own remedial policy in 2009. To estimate causal effects, I implement a sharp regression discontinuity design, that exploits the cut-off rule used to assign students to remediation. Results indicate that students do not get discouraged when placed in remedial courses. However, the assignment to remediation does not trigger any positive and significant effect on either persistence in college, credit accumulation, or the probability of passing the college-level exam in the remedial subject. My findings, which differ from previous ones obtained in a similar context by De Paola and Scoppa (2014), suggest that the specific structure of college remediation may play an important role in determining its success. With these conclusions, I aim to contribute to the growing literature that analyzes alternative measures designed to enhance the supply of college-educated workers.
Next, in "How Does Maternal Labor Supply Respond to Changes in Children’s School Schedule?", joint with Clémentine Van Effenterre, we exploit a reform of the primary school schedule, that was implemented in France in 2013, to contribute to the debate on women’s labor supply, along two specific dimensions. First, this intervention, that restructured and extended the total time children can spend in school, gives us the opportunity to understand to what extent the elimination of institutional constraints can further boost women’s labor supply in the context of a developed country, characterized by high female labor force participation rates. Secondly, this reform allows us to investigate whether having a flexible schedule is especially costly for some women, as suggested by Goldin (2014). To analyze these issues, we compare employment decisions of mothers whose youngest child is in primary school with those of mothers whose youngest child is slightly older, in a difference-in-difference framework. With respect to the first dimension of response – the potential increase in labor supply driven by the implicit wage subsidy offered by the reform – we provide evidence that mothers reallocate their working hours over the week but do not increase the the total number of hours worked per week. Concerning the second dimension of response, we show that women do take into account that flexibility is costly when making their employment decisions. On the one hand, we see that women facing a higher cost of flexibility – i.e. those working in occupations where it is important to build solid relationships with co-workers, attend frequent meetings, take key decisions, and perform tasks under pressure – were already working longer hours before the reform. On the other hand, we observe that only women facing a low cost of flexibility – that is those working in professions where team work is less relevant, the worker is not responsible for important decisions, and work pressure is low - are able to immediately react to the reform, by restructuring their working schedule in accordance to the new timetable of their children. By combining the new insights of Goldin’s theory, to the evidence provided by Blau and Kahn (2016), and to the literature on childcare (Bauernschuster and Schlotter 2015, Baker, Gruber, and Milligan 2005, Berlinski and Galiani 2007, Cascio 2009, Fitzpatrick 2010, Gelbach 2002, Havnes and Mogstad 2011, these results hopefully enrich the gender debate, with a special regard to the European context.
Finally, in "Does Employment Protection Legislation Affect Training Investments? Evidence from the United Kingdom", I test the hypothesis that in flexible labor markets, firms and workers may have less incentives to invest in on-the-job training. Analyzing this topic is particularly important in the conceptual framework of the race between education and technology, as on-the-job training can be seen as a key tool to increase labor productivity and allow both firms and workers to cope better with rapid task-replacing technological changes. Moreover, the empirical evidence on the impact of EPL on training investments is basically non-existent, despite the fact that in the current debate over the effects of flexible EPL, its negative consequence for training are often cited by its opponents. In addition, many countries are considering the possibility to introduce the so-called unique contract to overcome a dual labor market structure. The main feature of this contract is that employment protection should raise with tenure and it is unclear how this could affect the level and timing of training investments. The United Kingdom offers an interesting setting to analyze this topic as it is one of the few countries where workers receive high doses of training, and all workers are hired under the so called "unified" contract, the closest version to the single contract that has been put in place so far. With this contract, firing costs rise with seniority, after an initial probationary period, with no protection. The length of this initial phase has been repeatedly modified and the last modification was introduced in 2012, in the immediate aftermath of the financial crisis. This intervention shortened the probationary period from two to one year of tenure. This allows me to study two issues. First, I can analyze the effect of loosening EPL during a recession, characterized by negative wage growth. Secondly, conditional on this effect, I can study how training levels evolve from the probationary phase to the following part of the contract and whether shortening the probationary period affects training investment. To do so, I compare, in a difference-in-difference framework, workers who have between one and two years of tenure, with those having more than two years of seniority, who are not affected by the 2012 reform. My results show that the firing hazard of treated workers does not increase after the reform.
This may happen because, in a period in which real wages are falling, firms might be less inclined to fire their workers. Secondly, conditional on this result, my findings suggest that training investments for treated employees do not decrease either, while training increases for workers with less than one year of tenure. These findings are encouraging with respect to the introduction of the unique contract, as it is currently discussed in many European countries. Moreover, they should be particularly insightful for the debate on skills-biased technological progress and the specific role that labor market institutions may have in helping workers coping with it.
To conclude, as the debates that have inspired this work evolve, so does my research agenda, but the focus remains on skills acquisition and utilization. In particular, my next works will focus on the impact of guidance program towards university choice on high-school graduates’ decisions; on the effect of cultural norms in influencing gender preferences towards education and work; and on the impact of women’s quest for flexibility on firms’ organization. Maybe, in the twenty first century, women will have a decisive role in the race between education and technology.
