CANCELED CInSt Research Seminar "The Heterogeneous Effects of Schoolmate Characteristics on Long-Term Labor Market Outcomes. Evidence from Denmark"

CANCELED
We are sorry to inform that this seminar is canceled due to unforeseen circumstances. We profusely apologise for the inconveniences we might have caused.

The Heterogeneous Effects of Schoolmate Characteristics on Long-Term Labor Market Outcomes. Evidence from Denmark
Giorgio Brunello (University of Padova (Italy), Institute of Labour Economics (Germany), CESifo Group (Germany), and Research Centre for Education and the Labour Market (ROA) of the Maastricht University School of Business and Economics (the Netherlands))

Date and time: May 22, 2018; 6 PM 
Place:  20 Myasnitskaya str.; room 120
Language: English 

Free entrance.
If you don't have an HSE ID, please send your name and surname to lovakov@hse.ru before 13:00, May 22.

Abstract
Much of the research on the effects of social interactions at school has focused on short-term outcomes (test scores, student behavior). Less is known about whether these interactions affect key long term outcomes, including lifetime earnings. Distinguishing between short and long-term effects is important for the evaluation of education admission policies such as tracking or school vouchers that affect peer composition in schools and classes. As remarked by Carrell et al, 2016, concerns on the effects of these policies are more serious when social interactions at school affect adult (long-term) outcomes. Peer characteristics have been measured using either the percentage of girls or indicators of average parental education and earnings in the class/grade/school attended by an individual. Some contributions find that a higher share of girls affects the learning outcomes of both girls and boys (Lavy and Schlosser, 2011 and Black, Devereux and Salvanes, 2013) and the choice of college major of boys (Anelli and Peri, 2017). There is also evidence that having a higher percentage of classmates with a college educated mother decreases the likelihood of dropping out of high school and increases the likelihood of attending college (Bifulco et al, 2011), and that the average paternal income of 2 schoolmates matters for the education and labor market outcomes of boys (Black, Devereux and Salvanes, 2013). Peer interactions at school can affect individual outcomes by influencing the development of skills, the attitudes toward education and individual aspirations, or by providing information on alternative opportunities that affect individual choices and actions. A higher share of girls in the class or school can improve the learning environment by reducing disruption (Lazear, 2000). Individual behavior may also change. For instance, pupils with more female schoolmates may change their attitude toward risk (see Booth and Nolen, 2012) and competitiveness (see Gneezy, Niederle and Rustichini, 2003). Peers with a better parental background may have higher ability, act as positive role models in education and facilitate access to economic and social networks. Since the interactions with differently gendered individuals and the exposure to a privileged background starts in the family well before school, a question that the existing literature has not addressed so far is whether and how social interactions in the family and at school interplay in determining both short-term school outcomes and long-term labor market outcomes. Using the framework introduced by Cunha and Heckman, 2007, exposure to a privileged background at home may foster / hamper the effects of later exposure to privileged peers in schools, depending on whether these interactions are substitutes or complements in the production of individual human capital and personality traits.
This paper contributes to the literature in two ways. First, we focus exclusively on long term outcomes, including education attainment at age 31, and earnings and employment between age 31 and age 40. Although the most comprehensive measure of adult earnings is lifetime earnings, the existing literature has investigated the effects of peer composition at school either on earnings in the twenties (between age 24 and 28 in Carrell et al, 2016, and at 3 age 24 in Anelli and Peri, 2017) or on earnings in the late twenties and early thirties (at age 27 to 32 in Bifulco et al, 2014) or finally on earnings in a three year window, starting with age 31 and ending with age 48 (Black, Devereux and Salvanes, 2013). However, current earnings can deviate significantly from lifetime earnings if age-earnings profiles are not parallel – the so called “life-cycle” bias (see Haider and Solon, 2006). Following Brenner, 2010, Bjorklund and Jantti, 2012, and Bhuller et al, 2017 among others, we minimize this bias by using individual average real earnings between age 31 and age 40. Second, we investigate whether the ffects of schoolmate characteristics on educational attainment, long term earnings and employment vary with family characteristics.
Our data are drawn from Danish registers and contain information on parents, siblings and schoolmates. We measure schoolmate characteristics when individuals are aged 15 (normally attending the 9 th grade) with both the average and the standard deviation of schoolmates’ parental education, and family characteristics with parental education. Our sample consists of siblings enrolled in the same school at different points of time. The identification of peer effects exploits plausibly random variation in peer composition within schools and between cohorts, controlling for family by school fixed effects and for school - specific trends. When we do not allow peer characteristics to have heterogeneous effects, we find no statistically significant relationship between average parental education and individual outcomes, and a statistically significant and positive relationship between the diversity in the background of schoolmates – measured by the standard deviation of parental background – and long term earnings. When we allow peer characteristics to have heterogeneous effects, we find that a higher average parental education of schoolmates affects positively pupils from disadvantaged families, by increasing their attained education and their long-term labor market earnings, and negatively pupils from privileged 4 families, suggesting that peer and family characteristics are substitutes in the production of human capital. A candidate reason why the privileged do not benefit from “better” peers is the presence of rank concerns (see Murphy and Weinhardt, 2016): assignment to a school with high average parental background reduces the probability that privileged pupils are top ranked, with negative incentives effects on their school performance and ultimately their labor market earnings. We also find that higher diversity, measured by the standard deviation of parental education, improves the labor market earnings of privileged pupils, but has no statistically significant effects on disadvantaged pupils. Therefore, a favorable parental background acts as substitute with respect to average schoolmate background and as complement with respect to its diversity. In line with the existing literature, the size of estimated effects is rather small. We estimate that a one standard deviation increase in average parental background raises the long-term earnings of disadvantaged pupils by 1 to 2 percent, which corresponds to the estimated returns to one fourth to one eighth of a year of additional education (see Bhuller et al, 2017; Brunello et al, 2017). Our results suggest that vouchers and bussing policies, which provide incentives for pupils from disadvantaged families (e.g. immigrants) to enrol in better schools, where schoolmate parental education is typically higher, payoff in terms of higher long-term earnings for the disadvantaged. Following this enrolment, both the average parental education of incumbents and its dispersion are likely to increase, with positive long-term effects for the privileged as well.