Hard Skills, Soft Skills: The Relative Roles of Cognitive and Non-cognitive Skills in Intergenerational Social Mobility

In this article, we adopt a two-step strategy to assess the relative roles of cognitive and noncognitive skills in mediating the relationship between family SES and children’s academic achievement using data from the Early Childhood Longitudinal Study-Kindergarten Cohort. First, we decompose the total effects of family background on children’s achievement into: (1) direct effects of family background, and (2) indirect effects via cognitive and non-cognitive skills. We estimate this model using skills alternatively measured at four points in time between kindergarten and the 5th grade. Second, we use growth curve and fixed effect models to study the changing relationship between family background and skill formation over time. Overall, we find that cognitive skills are stronger mediators of family SES than non-cognitive skills. This is both because non-cognitive skills are less predictive of later achievement and because they are less affected by family SES. However, the mediating role of non-cognitive skills grows over time because the effect of family SES on non-cognitive skills significantly increases over a child’s life-course. Our findings raise important questions regarding the role of non-cognitive skills in intergenerational social mobility. Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 3 INTRODUCTION It is well understood that family background exerts strong influences on children’s educational outcomes, with children from higher-SES families academically outperforming those from lower-SES families (Becker 1993; Blau and Duncan 1967; Bourdieu 1977; Duncan and Brooks-Gunn 1997; Hauser, Tsai, and Sewell 1983; McLanahan and Sandefur 1994; Sewell, Haller, and Portes 1969). However, the question of how family SES actually affects children’s educational outcomes is still subject to debate. Broadly speaking, three categories of causal mechanisms have been considered: direct effects of hard resources, indirect effects through cognitive skills or “hard” skills measured by test scores, and indirect effects through noncognitive skills or personality traits which include a range of “soft” skills such as motivation, social skills and work habits. The first two causal mechanisms—the direct effect of family resources on achievement (Becker 1993; Mincer 1974; Kaushal, Magnuson, and Waldfogel 2011) and the indirect effect via cognitive skills (Griliches and Mason 1972; Jencks et al. 1979; Hauser, Tsai, and Sewell 1983; Sewell, Haller, and Portes 1969)—are well established. The role of non-cognitive skills in status attainment also has a long tradition in stratification research dating back to the Wisconsin socio-psychological model of status attainment (Hauser, Tsai and Sewell 1983; Sewell, Haller and Portes 1969). In recent years, however, their role has gained renewed interest. A growing body of research suggests that non-cognitive skills may be as important as cognitive skills in predicting a variety of outcomes, ranging from educational attainment (Duncan and Magnuson 2011; Lleras 2008; Rosenbaum 2001) to income and labor market performance (Cunha and Heckman 2009; Hall and Farkas 2011; Jackson 2006) to incarceration and teenage childbearing (Heckman, Stixrud and Urzua 2006). Moreover, sociological theories have long speculated that the family is instrumental in shaping critical non-cognitive skills that are important for future success, such as motivation, aspirations and self-discipline (Boudon 1974; Bourdieu 1977; Heckman 2006, 2011; Lareau 2003; Kao and Tienda 1995; Kim 2011; Goyette and Xie 1999). Yet to date, important questions remain about the role of non-cognitive skills in status attainment. While it is commonly accepted that non-cognitive skills, like cognitive skills, mediate the intergenerational transmission of family advantages or disadvantages, it is far from clear, a priori, that non-cognitive skills mediate the effects of family SES in the same manner as cognitive skills. Our current understanding is limited in at least two important ways. First, we do Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 4 not know the relative importance of cognitive versus non-cognitive skills as mediators of family origin. For skills to mediate the effects of family SES on achievement outcomes, two conditions must hold: (1) they must affect achievement outcomes, and (2) family SES must affect them. To date, neither of these two relationships has been well established for non-cognitive skills. Second, most prior studies have taken a static conceptualization of skill formation by using single-point-in-time measures of skills (Duncan et al. 2007; Lleras 2008; Hauser, Tsai, and Sewell 1983; Jencks et al. 1979; Mood, Jonsson and Bihagen forthcoming; Sewell, Haller and Portes 1969). Growing evidence, however, suggests that the developmental trajectories of cognitive skills differ from those of non-cognitive skills. We do not know how these potential differences in the evolution of cognitive versus non-cognitive skills over the life-course might shape the ways in which they mediate family SES effects in different ways. To address these gaps in our understanding, we employ a two-step strategy. First, we decompose the impact of family background on children’s academic achievement into: (1) direct effects of family background and (2) indirect effects via cognitive and non-cognitive skills. We estimate these models alternatively using skills measured at kindergarten to 5 grade. This approach allows us to evaluate the relative importance of cognitive versus non-cognitive skills as mediators of family SES effects and to determine whether the mediating roles of these two types of skills change over time. Second, we use a variety of statistical techniques—including growth curve and fixed effect models—to explicitly determine the dynamic effects of family SES on trajectories of cognitive as well as non-cognitive skills. Doing so allows us to examine a potential reason why the mediating roles of cognitive and non-cognitive skills might change over time. Overall, we find that cognitive skills are stronger mediators of family SES than noncognitive skills. This is both because non-cognitive skills are less predictive of later achievement and because they are less affected by family SES. The later findings are particularly relevant because they suggest that the role of the family in shaping personality traits is weaker than previously theorized. When we take a developmental perspective and examine these relationships as they change over a child’s life course, we find that the mediating role of noncognitive skills increases over time, although they remain weaker mediators than cognitive skills. Moreover, the increase in the role that non-cognitive skills play over time can be attributed mainly to the increasing influence of family SES on non-cognitive skills. In other words, children’s non-cognitive skills are more sensitive to family environment at later ages. Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 5 THEORETICAL AND METHODOLOGICAL ISSUES In Figure 1, we present a stylized conceptual model of status attainment. We revive the classical Wisconsin socio-psychological model as a starting point for understanding the causal pathways through which family SES influences children’s achievement outcomes (Hauser, Tsai, and Sewell 1983; Sewell, Haller and Portes 1969). In its original formulation, the Wisconsin model conjectures that the influence of family background on children’s status attainment is entirely mediated through children’s social-psychological traits defined as educational and occupational aspirations. We deviate from this position to include both a direct effect of family SES on children’s achievement outcomes and an indirect effect via children’s cognitive skills. In the following sections, we briefly review the literature pertaining to each of the three causal mechanisms linking family SES and children’s achievement (i.e., direct effects, indirect effect via cognitive skills and indirect effects via non-cognitive skills) and highlight the gaps in our knowledge. Figure 1: Stylized Model of Status Attainment Note. NC = Non-cognitive skills, SES = Mothers’ education and permanent family income Direct Effects of Family SES Family SES may exert direct effects on children’s educational outcomes because a higher family SES means more potential resources that could be expended to promote children’s education. The resource explanation has been popular in economics. According to this Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 6 explanation, parents have an intrinsic “altruistic interest” in their children’s socioeconomic wellbeing and purposely invest in their children (Becker 1993; Mincer 1974). Parental investment in children may take a variety of forms, such as purchased goods and services (e.g., quality child care, school supplies and books, recreation and entertainment activities, and private lessons) and time spent on children (e.g., help with homework and attendance at school events) (Kaushal, Magnuson, and Waldfogel 2011). Note that these examples of parental investment are all about resources, in the sense that they are subject to firm budget constraints, and parents could divert their uses towards other uses (including other children) if withholding them from a particular child. Indirect Effects through Cognitive Skills Another main causal pathway through which family SES affects children’s educational outcomes is via cognitive abilities, also called “hard skills.” We do not yet know the true extent to which cognitive abilities are determined by nature (i.e. genetics) versus nurture (i.e. environment), but it is safe to say that the two forces interact in certain ways that shape one’s cognitive abilities (e.g., Nisbett 2009; Plomin, Defries, and Loehlin 1977). Studies show that the most important period for cognitive skill development is early to middle childhood (Cunha and Heckman 2008; Guo and Harris 2000; Keane and Wolpin 1997). During this period, high-SES families enhance children’s math and verbal development by offering greater material resources (Klebanov et al. 1998; Korenman, Miller, and Sjaastad 1995) and more stimulating home environments (Bronte-Tinkewet al. 2008; Parcel and Dufur 2001) than low-SES families. These “hard” skills, in turn, go on to positively influence a variety of achievement outcomes, including children’s educational attainment (Farkas and Vicknair 1996; Griliches and Mason, 1972; Warren, Hauser and Sheridan 2002). Indirect Effects through Non-cognitive Skills Non-cognitive skills, or personality traits, have recently resurfaced as another important causal mechanism through which family SES background affects children’s academic achievement. These skills are multidimensional in nature and encompass a broad class of individual attitudes, behaviors and habits that are correlated with but distinct from cognitive traits (for a detailed review see Borghans et al. 2008). By non-cognitive skills, researchers have meant a variety of characteristics distinct from purely cognitive abilities, ranging from leadership Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 7 and perseverance (Jencks 1979) to self-esteem and locus of control (Heckman, Stixrud and Urzua 2006), aspirations and expectations (Sewell, Haller and Portes 1969), and school-related attitudes and behaviors (Rosenbaum 2001; Lleras 2008; Clasessen, Duncan and Engel 2008; Duncan and Magnuson 2011). For non-cognitive skills to mediate family SES effects on children’s achievement, they must (1) influence achievement outcomes and (2) be influenced by family SES. While consensus exists that traits, such as self-control, social skills, and attention-related capacities, positively predict children’s academic outcomes, there is intense disagreement among policy-makers and academics as to their importance relative to cognitive abilities. The “Neurons to Neighborhoods” report of the National Research Council and Institute on Medicine argues that early intervention programs should devote equal attention to enhancing children’s socio-emotional development and their cognitive development (Shonkoff and Philips 2000). On the other hand, reports from the National Research Council’s Committee on the Prevention of Reading Difficulties in Young Children (Snow, Burns, & Griffin 1998) and the National Association for the Education of Young Children (NAEYC) argue for the primacy of “hard” skills such as literacy and math competence for success in schools. The academic literature is also mixed. Some studies demonstrate that traits such as selfcontrol explain much more of the variation in children’s academic performance than cognitive abilities (Duckworth and Seligman 2005; Wolfe and Johnson 1995). Still others argue that noncognitive skills have far more persistent effects on children’s later academic achievement than cognitive skills (Chetty et al. 2010; Heckman et al. 2007). In contrast, other studies show that “hard” skills trump “soft” skills as determinants of academic success. Based on analyses of data from six longitudinal studies, Duncan et al. (2007) find that early math and reading skills are stronger predictors of later achievement than non-cognitive traits, such as attention-related capacities, social strengths and behavioral deficits. Lleras (2008) reports similar results from a study analyzing data from the National Educational Longitudinal Study, as do Claessens, Duncan and Engel (2009) based on data from the Early Childhood Longitudinal Study. 1 In referring to such a variety of personality and psychological traits, non-cognitive skills in the social science literature do not conform to the conventional notion of “skills,” a term that normally applies only to abilities or facilities acquired through learning or training. 2 http://www.naeyc.org/about/positions/pdf/psmath.pdf Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 8 The second condition that must be fulfilled for non-cognitive skills to mediate family SES effects is that family SES must influence them. Longstanding sociological theories postulate that families play an instrumental role in shaping children’s non-cognitive skills. Sociologists have long argued that attainment of higher education is a cultural expectation specifically valued by the middle class but not shared by the lower class (Boudon 1974; Bourdieu 1977; Brand and Xie 2010; Jencks et al. 1972; Goyette and Xie 1999; Smith and Powell 1990). For example, in a study of middle and working class families, Lareau (2003) documents disparities in the types of skills that are promoted, consciously or unconsciously, by parents. In contrast to poor and working class parents, middle-class parents are more likely to “cultivate” children’s cognitive and non-cognitive traits along specific dimensions that foster academic achievement. However, the relationship between family SES and non-cognitive factors may be weaker than what some scholars would expect. For example, comparison of monozygotic and dizygotic twins typically shows that only 25% of total variation in non-cognitive traits is due to shared family traits (i.e. both genetic and environmental) (Floderus-Myrhed, Pedersen, and Rasmuson 1980; Rowe and Plomin 1981). In contrast, approximately 40% of total variation in adult IQ, a cognitive trait, is due to shared family traits. Using the National Longitudinal Study of Youth (NLSY), Duncan and colleagues (2005) find that family SES is a surprisingly weak predictor of social-psychological indicators such as shyness, self-esteem and depression. A recent study by Mood, Jonsson and Bihagen (forthcoming) attempts to directly address the question of how family SES effects on children’s educational outcomes are mediated by cognitive versus non-cognitive skills, using Swedish registry data on approximately 180,000 fathers and sons. Their measures of cognitive and non-cognitive skills (i.e. social maturity, emotional capacity and leadership skills) are assessed when sons are age 18. The study finds that the intergenerational correlation of educational attainment is mainly mediated by children’s cognitive skills and only weakly mediated by non-cognitive traits. Unfortunately, the study does not inform us why the mediating effects for non-cognitive skills are weak. Is it because noncognitive skills are poor predictors of later educational outcomes, or is it because father’s education is a poor predictor of non-cognitive skills? Another factor to consider is the potentially evolving role of cognitive and non-cognitive skills over one’s life course. This point is critical because mounting evidence demonstrates that there are important differences between the developmental trajectories of cognitive and those of Cognitive and Non-cognitive Skills in Intergenerational Social Mobility 9 non-cognitive skills (see Borghans et al. 2008 for a detailed review). Specifically, cognitive skills undergo the greatest amount of change in early childhood and stabilize by adolescence. Noncognitive skills, on the other hand, continue to undergo changes throughout childhood and into young adulthood. We call this conjecture the “sensitive period hypothesis,” or the proposition that there are sensitive periods in the life course when skills are more or less malleable to environmental influences (Borghans et al. 2008; Cunha and Heckman 2008; Guo and Harris 2000). Emerging evidence supports this hypothesis. For example, Cunha and Heckman (2008) find that parental investments, such as the number of books at home and frequency of trips to museums, have strong effects on cognitive skills at earlier ages and have strong effects on noncognitive skill formation at later ages. Static conceptualizations of skills cannot capture important dynamics in skill formation, nor can they capture the changing roles of cognitive and non-cognitive skills as mediators of family SES over the life course. For example, if the effects of family SES on non-cognitive skills grow with time, then a static measure of non-cognitive skills would lead to an underestimation or overestimation of the mediating role of non-cognitive skills, depending on the time at which noncognitive skills are measured. Therefore, a developmental perspective that explicitly considers changes in skill formation over time allows us to identify potential sensitive periods in the relationship between family origin and skill development and to address questions regarding whether the mediating role of cognitive and non-cognitive skills changes over the life course.