The skills demanded in the labor market have changed (see, e.g., Ace- moglu & Autor 2011). Just as technological innovations have increased demand for workers capable of performing non-routine and complex tasks, demand for routine work has declined (Autor et al. 2003, Spitz-Oener 2006). Since lifelong learning can help individuals develop resilience and adapt to adverse shocks in changing labor markets, it is important to learn more about the best ways to support such learning. Skill formation starts in childhood, as does lifelong learning. By in- vesting early, and with the support of teachers and caregivers, we can equip children with a strong foundation for further development (National Research Council Institute of Medicine 2000). Societies can invest in skill formation by improving the quality of children’s environments. Indeed, the home environment, early childhood education and care (ECEC) centers, and schools can all affect children’s skill formation (Almond & Currie 2011, Blau & Currie 2006, Cunha et al. 2006, Hanushek & Rivkin 2006, Heckman & Mosso 2014). This influential role of the environment has implications for policy because it suggests a role for investments that enhance its quality. Yet gathering evidence on environmental factors that affect skill formation is no simple task, and much remains to be learned about the relevant processes. This thesis focuses on three key questions: First, what is the best way to gain a better understanding of how children acquire new skills? Econo- metric tools have been developed to evaluate the effectiveness of environ- mental factors that affect skill formation. However, these tools cannot be employed unless children are observed at equally spaced intervals. Second, which of the many possible skills should investments focus on? During early childhood, it is best to target skills that promote a child’s further development, and in the context of the crucial transition from preschool to primary school, these could be skills that help prepare children for success in school. Third, once certain skills have been nurtured during early child- hood, how can we provide effective education to sustain and build on this foundation? Although we know that teachers play a key role in education, much remains to be learned about what makes teachers effective. By ad- dressing these questions, this thesis seeks to deepen our understanding of the processes driving skill formation in childhood, yielding new and better insights into public policy design. The individual chapters are described in more detail below. Chapter 2 examines the literature on the econometrics of childhood skill formation. This literature comprises econometric tools for assess- ing the effectiveness of environmental factors and models influences on skill formation. In particular, it helps us understand what can be learned from a model of skill formation (identification), how we can best learn it (estimation), and how certain we can be about parameters we estimate (inference). Todd & Wolpin (2003), Cunha & Heckman (2008), Cunha et al. (2010), and Agostinelli & Wiswall (2016) develop tools to address several challenges. First, scales used to assess children may not have a cardinal interpretation. Second, given the difficulty of assessing children, any instrument is likely to be subject to measurement error. Third, unob- served inputs may correlate with observed inputs. Just as non-cardinality may result in biased interpretation, measurement error and the presence of unobserved inputs may cause biased estimation. I evaluate the implications of a challenge that has received little attentionin the past. When we estimate skill formation models, we typically assume that all children are observed in equally spaced intervals. The interval assumed by our model is therefore like the one observed in the data. In most longitudinal studies, however, observation intervals are not equally spaced (McKenzie 2001, Millimet & McDonough 2017). One might also argue that the substantive timing variable of child development is age, not the assessment wave. In this case, unequal spacing may occur even when survey waves are equidistant because (i) same-aged children may not be observed simultaneously, and (ii) simultaneously observed children may not be of the same age. Most longitudinal studies target “same-aged” children based on birth year. The development of children born in the same year and assessed simultaneously can differ by as much as 12 months. If, in addition, as- sessments can occur at any time during the year, then the development of children born in the same year may differ by as much as 24 months. Such differences can have an impact on cognitive skill formation (Crawford et al. 2010). Of course, we would expect these developmental differences to become smaller as children grow older (Elder & Lubotsky 2009). Still, fail- ing to account for unequally spaced intervals may lead to biased estimation, particularly for young children. Chapter 2 examines the implications of this estimation problem with the help of an often-used dataset for modeling child development. Chapter 3 draws from the early childhood literature and seeks to determine which skills should be nurtured in early childhood. We know that children start school with different skill levels, and this insight has prompted an interest in early childhood education programs that boost “school readiness” (see, e.g., Clements & Sarama 2011, Diamond & Lee 2011, Dillon et al. 2017, Rege et al. 2021). A challenge in designing these programs is deciding what skills to target, as not all skills are equally important for school success (Duncan et al. 2007, Lewit & Baker 1995). An increasing number of studies have evaluated the effectiveness of early childhood education programs designed to reduce early skill dispari- ties (e.g., Attanasio et al. 2020, Conti et al. 2016, Heckman et al. 2013, Sylvia et al. 2020). These studies generally find that allocating resources to efforts to promote early skill formation can be an effective approach. However, depending on the type of skills targeted and the nature of the intervention, effects may not persist over time. For example, the effects of an investment that targets skills in early childhood that children will eventually develop irrespective of the investment may fade out. As a result, targeting such skills may not be optimal (Bailey et al. 2017). Chapter 3 studies whether executive functions, defined as the cognitive control processes necessary for concentration and thinking (Diamond & Lee 2011), are skills that programs should focus on. As they involve fun- damental skills, executive functions would appear to be a natural starting point (Diamond & Lee 2011, Howard-Jones et al. 2012). One could argue, however, that the structure provided in school enables children to develop the same level of executive functioning as they would have if they had been the beneficiaries of targeted investments in promoting such executive functions in preschool. But if executive functions are the basis for learning many other skills, children starting school with higher levels of executive functioning may be more efficient at learning other skills than their peers (i.e., skill begets skill). Understanding the processes that drive skill forma- tion in early childhood yields insights for public policy decisions about how educational resources should be used. Chapter 4 looks at how skills nurtured in early childhood can be maintained through effective education, and examines the literature regarding the education production function. This literature focuses on school inputs that are effective in promoting children’s development. While there are many kinds of school inputs (e.g., class size, number of books, number of computers), Chapter 4 focuses specifically on teachers. Teacher ef- fectiveness may vary widely, even in the same school (Aaronson et al. 2007, Araujo et al. 2016, Jackson 2018, Kraft 2019, Rivkin et al. 2005, Rockoff 2004). Moreover, effective teachers may have long-term impacts on children’s education and labor market outcomes (Chetty et al. 2014b, Opper 2019). Lastly, teachers are the largest budgetary expense in most schools (Hanushek & Rivkin 2006). While we can identify effective teachers through value-added estimates (Chetty et al. 2014a), we do not know how to replicate them (cf. Kane & Staiger 2012). The literature indicates that a teacher’s readily observable characteristics, such as education, salary, or test scores, do not consistently predict children’s academic achievement (Hanushek & Rivkin 2006). For this reason, researchers have started to focus on what goes on inside the classroom. The child development literature suggests that the quality of the child’s relationship with the teacher and classmates, as perceived by the child, is particularly important (see, e.g., Connell & Wellborn 1991, Hamre & Pianta 2001, Pianta 1997). Still, the numerous studies that suggest that teacher relationship skills, as perceived by the child, are essential for learning may be biased by a child’s (unobserved) preferences for a particular relationship. Recently, economists have started to refine our measures of investments in education to capture objective, detailed information about the quality of teacher- child interactions (Araujo et al. 2016, Kane et al. 2011). However, these classroom observations are costly and may fail to capture fundamental aspects of a child’s perceptions that ultimately drive behavior (Connell & Wellborn 1991). Moreover, it is important to evaluate teachers and what goes on inside the classroom, with the help of a variety of assessments (Kane & Staiger 2012). Chapter 4 introduces and validates a new approach for measuring teachers’ overall ability to form positive relationships in the classroom (based on the children’s perspectives). This approach has implications for public policy, as such assessments can serve as tools for identifying teachers who need support, promoting development, conduct- ing progress evaluations, and helping policymakers improve quality. A key motivation for focusing on childhood skill formation is a desire to ensure equal opportunities for all children. Section 2 explains the concept of equality of opportunity and shows that skill disparities emerge early and persist over time. Since all three chapters of this dissertation conceptualize child development in keeping with the technology of skill formation, I introduce this technology and the related literature in Section 3. All chapters address a key challenge – namely, the fact that skills are inherently unobservable. Section 4 elaborates on how, despite that fact, we can learn about a child’s skills. I begin by describing the general intuition underlying the measurement of unobserved variables. Since the measures used in each chapter have been validated in other studies, I briefly describe how such validation is typically achieved. Section 4 concludes with a description of a dedicated measurement model of the type used in each chapter. Section 5 summarizes each of the chapters.
