From Inference to Influence: Applying Causal Game Theory to Complex Security Environments Maarten Vonk BASIC SKILLS Elke Claes PhD dissertation From Inference to Influence: Applying Causal Game Theory to Complex Security Environments Maarten Vonk Elke Claes PhD dissertation
BASIC SKILLS
© Elke Claes, 2025. All rights reserved. No part of this publication may be reproduced, stored in an automated data system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior permission of the author. Published by ROA Postbus 616 6200 MD Maastricht ISBN: 978-90-5321-643-9 Printed in the Netherlands by Ipskamp Printing
BASIC SKILLS Dissertation To obtain the degree of Doctor at Maastricht University, on the authority of the Rector Magnificus, Prof. Dr. Pamela Habibovic´, in accordance with the decision of the Board of Deans, to be defended in public on Wednesday, 10 December 2025, at 13:00 hours by ElkeClaes
Supervisors Prof. Dr. Inge de Wolf, Maastricht University Prof. Dr. Bart Golsteyn, Maastricht University Co-supervisor Dr. Suzanne de Leeuw, Maastricht University Assessment Committee Prof. Dr. Didier Fouarge, Chair, Maastricht University Prof. Dr. Monique de Haan, University of Amsterdam Prof. Dr. Dinand Webbink, Erasmus University Prof. Dr. Ines Wilms, Maastricht University
Acknowledgments A few years ago, pursuing a PhD was something I had not seriously considered. In fact, I didn’t even know what a PhD was. The idea of dedicating myself to research seemed far from where I thought I would be. Yet, here I am, not only having completed this PhD dissertation, but also having discovered a deep interest in and appreciation for research. It has been an enriching experience, made possible through the support of many people. First, my deepest appreciation goes to all who have directly contributed to my PhD journey: my supervisors, ROA colleagues, MILE colleagues, and many others. To all of you: thank you for your guidance, support, and willingness to engage with my work. I am grateful for our discussions and for your patience, expertise, and encouragement. In the same way, I want to extend my thanks to those who supported me in more indirect ways. From providing a collaborative and welcoming environment, to offering a listening ear, to helping maintain a positive atmosphere during less pleasant times, each of these contributions is sincerely valued and has made a meaningful difference throughout this journey. In closing, I hope this work, in some small way, reflects the impact you have had on my development. As I move forward, I take with me the lesson that every challenge is an opportunity to learn, and that every ending can mark the beginning of something greater. v
Executive summary This dissertation aims to contribute to the knowledge on basic skills (literacy and numeracy) taking three distinct angles, each addressed in a separate part of the dissertation. The first part of this dissertation explores the importance of basic skills for life satisfaction. The second part, consisting of two chapters, concentrates on the development of basic skills throughout primary school. The third part investigates whether environmental factors affect student performance. The first part of this dissertation (Chapter 2) concludes that ten-yearsolds’ early basic skills are strongly related to adult life satisfaction. The chapter shows, using British data, that the relationship between early math skills and life satisfaction is significantly positive and stable between ages 29 and 46, whereas the relationship between early reading skills and life satisfaction diminishes in magnitude throughout this period. These findings highlight that early basic skills can serve as signals for potential difficulties in individuals’ later life outcomes, particularly regarding individuals’ satisfaction. The second part of this dissertation addresses the development of basic skills throughout primary education and consists of two chapters. Chapter 3 demonstrates that students’ skill proficiency in the early years of primary education is highly associated with their proficiency at the end of primary education. Furthermore, it shows that a significant part of the sixth-grade achievement gap regarding parental education finds its roots in the early years of (or before) primary edvii
Executive summary ucation. This advocates policymakers to invest as early as possible in basic skills, as students’ early stock of skills seem determinant for their later stock of skills. Chapter 4 establishes that basic skills are interrelated both contemporaneously and dynamically throughout primary education, suggesting that an investment in one skill can be associated with benefits in others. The third part of this dissertation accentuates that environmental factors can affect student performance. Chapter 5 reports that temperature, especially heat, has adverse effects on reading and math test scores. The chapter highlights the need for policymakers to establish a favorable learning and testing environment for students. viii
Contents Acknowledgments v Executive summary vii 1 Introduction 1 2 Early Skills and Adult Life Satisfaction 13 2.1 Introduction......................... 15 2.2 Data, measures and descriptives . . . . . . . . . . . . . 19 2.3 Methodology ........................ 26 2.4 Results............................ 27 2.5 Conclusion ......................... 36 3 Primary School Skill Development: from First to Sixth Grade 39 3.1 Introduction......................... 41 3.2 Background......................... 45 3.3 Dataanddescriptives ................... 49 3.4 Empiricalstrategy ..................... 55 3.5 Results............................ 60 3.6 Conclusion ......................... 73 4 The Interplay of Reading, Spelling, and Math in Primary Education: A Panel VAR Analysis 77 4.1 Introduction......................... 79 4.2 Data ............................. 84 4.3 Methodology ........................ 89
Contents 4.4 Results............................ 93 4.5 Conclusion .........................102 5 Heat and Learning in a Moderate Climate: Temperature Effects on Primary School Students in the Netherlands 107 5.1 Introduction.........................109 5.2 Backgroundanddata ...................114 5.3 Empiricalstrategy .....................123 5.4 Results............................125 5.5 Conclusion .........................134 6 Conclusion 137 Impact 153 Summary 157 Samenvatting 161 Appendices 167 AppendixA ...........................167 AppendixB ...........................199 AppendixC ...........................221 AppendixD ...........................225 Bibliography 237 About the author 261 ROA Dissertation Series 263
1 Introduction 1
Chapter 1. Introduction What are basic skills? This dissertation defines basic skills as the fundamental skills that are essential both for effective functioning in everyday life and for underpinning the development of more advanced skills. The term ‘basic skills’ could be used for a varying set of skills. I focus on literacy and numeracy skills. Literacy encompasses the ability to read, write, and comprehend texts, while numeracy involves basic mathematical operations such as addition, subtraction, multiplication, and division. This set of basic skills aligns with the definition of basic skills by the Dutch Inspectorate of Education, which includes literacy, numeracy, and citizenship (see, e.g., Inspectie van het Onderwijs (2024)). I do not consider citizenship because data availability is limited. Why are basic skills important? Basic skills are part of individuals’ human capital, which several prominent economists consistently highlight as essential for economic growth. Smith (1776), for instance, laid the groundwork by emphasizing how education and skill development enhance labor productivity. Marshall (1890) later expanded on this idea and illustrated how education and the accumulation of skills are critical factors in individuals’ productive capacities and economic development. Becker (1964) formalized the concept of human capital, emphasizing that individuals and societies benefit economically from developing skills and knowledge. His theory views human capital as a critical factor for economic growth. Also the endogenous growth 2
model of Romer (1989) recognizes the importance of human capital for the economy. Romer highlighted the role of human capital in driving innovation and technological change, and consequently economic growth. He underscored that the continuous development of skills and knowledge within an economy is crucial for sustained economic progress. Besides the importance of basic skills, or human capital, for economic growth, these skills carry importance at an individual level. It is known that skills beget skills, which is part of a broader framework described by skill formation theories (e.g., the technology of skill formation from Cunha and Heckman (2007)). Early skill development facilitates future skill acquisition, meaning that the more skills you acquire early in life, the easier it is to gain more skills later on. Moreover, early skills are linked to varying economic and social outcomes, such as adult income, labor market outcomes, and health (e.g., Crawford & Cribb, 2013; Fischbach et al., 2013; Hanushek et al., 2015; Hatch et al., 2007; McIntosh & Vignoles, 2001; Ritchie & Bates, 2013; Sabates & Parsons, 2012; Von Stumm et al., 2013). The second chapter of this dissertation explores whether early basic skills also predict how satisfied individuals are with their lives in adulthood. Besides considering life satisfaction in general, the chapter examines in which life domains early basic skills predict individuals’ satisfaction in adulthood. Last, basic skills have an important role within the educational trajectory of students. In the Netherlands, for instance, which is the research context for most of this dissertation’s chapters, students are introduced to literacy and numeracy during kindergarten and begin formal lan3
Chapter 1. Introduction guage and math lessons in grade one (‘groep drie’ in Dutch). At the end of primary education, in grade six, students take a standardized test (e.g., the ‘Cito-toets’), which helps to determine their suitability for different secondary education tracks. In these tests, students’ literacy and numeracy, among others, are assessed. Therefore, their proficiency in literacy and numeracy is important for their educational trajectory. Basic skills, however, already play a crucial role in educational continuity within primary education. Proficiency levels are consistently monitored and assessed annually, with most schools even administering standardized tests twice a year. The resulting data from these assessments form a unique and rich dataset, which is utilized in most chapters of this dissertation. The current state of basic skill proficiency International studies, such as PISA, TIMMS, and PIRLS, report that children in several countries lack proficiency in their basic skills. Figure 1.1 visualizes the PISA results between 2003 and 2022 for students’ reading (left) and math (right) performances. The graphs in Figure 1.1 illustrate the results for the OECD, the EU14, and the Netherlands. As illustrated by the left graph, there is a decrease in reading performance for students in OECD countries since 2012. A similar result holds for the EU14 countries as of 2015. Before 2012, the reading proficiency of students was rather stable. As depicted in the right graph, students’ math performance has been gradually falling since 2003 in the OECD and EU14 countries. The decline has, however, remarkably accelerated since 2018. Although the COVID-19 pandemic (with school closures, 4
remote education, and limited social interaction) had an impact on students’ basic skill proficiency, the decline in their mastery of basic skills was already evident beforehand. Figure 1.1: PISA results. Note. This figure presents the PISA results for reading (left) and mathematics (right) for the Netherlands (green), the OECD (dark gray) and the EU14 countries (light gray). Source: translated from Meelissen et al. (2023), pages 21 and 52. Taking a closer look at the Netherlands, the results show an amplification of the OECD and EU14 trends. Dutch 15-years-olds demonstrate a significant decline in their reading performance since 2012, where the decline exceeds the OECD average by more than 2.5 times. Dutch students’ math performance declines continuously starting from 2003, where this drop is over 1.5 times larger than the average decline in the OECD countries. While Dutch students still score well above the OECD average for math, their reading proficiency has dropped below this OECD average since 2018. Although these PISA results demonstrate the deficiencies in the basic skill proficiency of students over the past decade, the meaning and consequences of these results for students’ daily life, also in the long run, remain abstract. Therefore, studies often report the proportion 5
Chapter 1. Introduction of the students that are so-called ‘insufficiently literate’, indicating the proportion of students whose performance is insufficient to function independently in society. According to Meelissen et al. (2023), a third of the Dutch students are insufficiently literate in reading and a quarter of them is insufficiently literate in math. They note that this proportion exceeds the OECD average for reading and is comparable to the OECD average for math. How do basic skills develop? Given the importance of basic skills for economies and individuals, and that current trends in skill proficiency levels are cause for concern, it is essential to understand how these skills develop. Kautz et al. (2014) present a conceptual scheme, as illustrated in Figure 1.2, to understand the skill development process. The framework illustrates that skills originate before birth and depicts, as shown from top to bottom in the figure, the process of their development from early life through to adulthood. Skills develop dynamically by investments at different life stages. Skills at birth are determined by inherited traits and prenatal investments. In further life stages, the skill development process is determined by a complex interplay of the stock of skills acquired early on and investments. The process that an increase in the stock of skills in one period relates to the stock of skills in a subsequent period, is also referred to as the ‘self-productivity’ of skills (e.g., Cunha & Heckman, 2007; Kautz et al., 2014). Skill development is affected 6
Figure 1.2: The skill development process. Note. This figure shows the conceptual framework for the skill development process. Source: Kautz et al. (2014), page 32. by investments from parents, the environment and schooling but the returns to these investments are also determined by the earlier stock of skills. The larger the stock of skills in one period, the more individuals benefit from investments, which is also referred to as ‘dynamic complementarity’ (e.g., Cunha & Heckman, 2007; Kautz et al., 2014). As suggested from this framework, early skills form an important foundation for the development of future skills. The third and fourth chapter of this dissertation focus on the development of skills throughout the primary school period. In the third chapter, I consider students’ skill levels in the beginning of primary 7
Chapter 1. Introduction education and inspect the extent to which these levels relate to their levels at the end of primary education. To this end, I take into account the socioeconomic background of the students and their primary school. Furthermore, I examine the extent to which achievement gaps observed at the end of primary school originate in the early years of primary education. The analysis touches upon the question whether skill differences among categories of individuals within the educational system increase, decrease, or stabilize. While the framework of Kautz et al. (2014) does not implicitly visualize the connection between varying skills, skills do not stand alone. Cunha and Heckman (2007) notice the ‘cross-productivity’ of skills, which describes that the stock of one skill relates to the stock of another skill in the next period. The fourth chapter, which also focuses on skill development during primary education, provides insights into the interplay among various basic skills in the primary school years. I explore this interplay by simulating a hypothetical shock to a single skill within a student’s skill set and examining the resulting patterns across the entire set of skills. This thought experiment illustrates typical patterns of co-movement among skills in a descriptive and conceptual manner. It may help policymakers understand the interplay between skills and examine how different educational interventions relate to patterns of skill development. 8
Do environmental factors affect student performance? Understanding the skill development process also provides opportunities to enhance skill development. Policymakers are deliberating on how to remedy the insufficient mastery of basic skills by a large proportion of individuals. Following on the framework of Kautz et al. (2014), they could intervene in several life stages and their interventions could target varying factors important for skill formation, such as prenatal investments or parenting. As suggested by Kautz et al. (2014), one of the factors that policymakers could target is the environment. This dissertation addresses whether environmental factors, more specifically temperature, affect student performance. The temperature on Earth has increased over the past years, also in Europe. The rate of global warming in Europe is twice as fast as the global average (KNMI, 2021), while research on the effects of temperature on skill development in Europe is lacking. The fifth chapter of this dissertation closes this gap in the literature. It provides policymakers with information on whether temperature matters for test scores of students, and thus whether improving students’ thermal environment could enhance their performance. 9
Chapter 1. Introduction Research questions In sum, the main research questions for this dissertation are as follows: Part I: Why are basic skills important? 1. Do early basic skills predict adult life satisfaction, and how does this association evolve across the adult life span? (Chapter 2) 2. In which life domains do early basic skills predict individuals’ satisfaction in adulthood? (Chapter 2) Part II: How do basic skills develop? 3. To what extent does early skill proficiency relate to later skill proficiency? (Chapter 3) 4. Do significant disparities in achievement exist among students during the early years and at the end of primary education? (Chapter 3) 5. Do varying basic skills relate to each other during primary education? (Chapter 4) Part III: Do environmental factors affect student performance? 6. How does temperature affect students’ basic skill test scores? (Chapter 5) 10
Outline This thesis comprises the following related, yet independent, chapters. Chapter 2uses data from the 1970 British Cohort study. It investigates the relationship between ten-years-olds’ skills and their life satisfaction measured in adulthood. The chapter enhances the understanding of this relationship by inspecting how it evolves over adulthood, and in which specific life domains the relationship is mostly present. Furthermore, the chapter explores how income and daily functioning mediate these relationships. Chapter 3 uses Dutch cohort data to analyze how skills develop from first to sixth grade. It focuses on the relationship between first-grade skill scores and their sixth-grade counterpart. The chapter examines achievement gaps regarding parental education, sex, and migration background, and inspects whether and the extent to which the roots of achievement differences at the end of primary school lay in the beginning of primary education. Chapter 4 uses Dutch panel data with students’ reading, spelling, and math proficiencies throughout primary education. It employs a panel vector autoregression approach to examine whether skills interact with each other. The chapter shows a simulation of the responses of a student’s skill set to a hypothetical shock in one skill. The resulting patterns provide descriptive insights into how skills vary and co-move. Chapter 5 exploits exogenous variation in temperature exposure of 11
Chapter 1. Introduction students over time. It examines the effect of temperature on student performance. For this purpose, it uses Dutch panel data between 2013 and 2023, combined with grid-level outdoor weather data. The chapter estimates the effect of test-day temperature on students’ test scores, and explores sleep disruption, proxied by the temperature in the night before the test, as potential mechanism. Chapter 6concludes. 12
2 Early Skills and Adult Life Satisfaction This chapter is joint work with Bart H.H. Golsteyn and Tim Huijts. This research uses data from the 1970 British Cohort Study (BCS70), conducted by the Centre for Longitudinal Studies (CLS), University College London. The data were supplied by the UK Data Service. University College London, UCL Social Research Institute, Centre for Longitudinal Studies. (2024). 1970 British Cohort Study. 11th Release. UK Data Service. SN: 200001, DOI: http://doi.org/10.5255/UKDA-Series200001. We thank participants of the 2024 AQMAPPS conference in Amsterdam and seminar participants at Maastricht University. We also thank Jacqueline Charpentier, Brian Korthout, Suzanne de Leeuw, and Inge de Wolf for their valuable feedback. 13
Chapter 2. Early Skills and Adult Life Satisfaction Abstract One of the most prominent indicators of a country’s education quality is students’ performance in early language and math skills. Evidence shows that these skills are important as they predict outcomes like educational attainment and wages. This paper examines whether early skills also predict generic measures of life quality and whether these patterns persist over time. Using data from the 1970 British Cohort Study, we investigate the relationship between early skills and life satisfaction across adulthood (ages 29-46). We find a strong, positive association between early skills and life satisfaction. The association between math skills and life satisfaction remains stable throughout adulthood, while that of reading skills diminishes. Early skills are particularly linked to satisfaction in the domains of health, financial management, and emotional stability in adulthood. 14
2.1 Introduction Early reading and math skills serve as foundations for the set of skills that people acquire in life. These early skills are linked to socioeconomic status in adulthood, higher educational attainment, higher incomes, better labor market outcomes, and better health (e.g., Crawford & Cribb, 2013; Fischbach et al., 2013; Hanushek et al., 2015; Hatch et al., 2007; McIntosh & Vignoles, 2001; Ritchie & Bates, 2013; Sabates & Parsons, 2012; Von Stumm et al., 2013). Yet, whether the value of these early skills extends beyond these outcome measures and how this value evolves during life has remained underexplored in the literature. This chapter investigates the relationship between early skills and individuals’ life satisfaction, which is a key indicator of overall life quality. We explore how ten-year-olds’ reading and math skills relate to life satisfaction at different stages of adulthood. These relationships may differ for reading and math skills, as reading presumably corresponds to social interactions, communication, and accessing information, while math is needed for logical reasoning and problem-solving. This also holds for variations throughout the life course. For instance, early reading skills might be more relevant to life satisfaction in early adulthood, when social integration and information gathering are crucial, whereas early math skills may relate more strongly to satisfaction in later years. To improve our understanding of these relationships, we analyze the relationships separately for reading and math skills, analyze satisfaction in specific life domains (e.g., health, career, relationships), and investigate factors that mediate these relationships. 15
Chapter 2. Early Skills and Adult Life Satisfaction We use longitudinal data from the 1970 British Cohort Study, which follows all English, Scottish, and Welsh people born in the same week in April 1970. Information is collected on a wide variety of topics, including education, health, and social development. The longitudinal aspect of the dataset provides invaluable input for studies that aim to analyze aspects at different life stages. We use data from ages 10 to 46, and perform separate analyses for early reading and math proficiency to inspect differences between these skills. Importantly, we are able to control for a measure of general intelligence, which is arguably an important potential confounder of our predictions. Our main results indicate that the relationship between early reading and math skills and adult life satisfaction is significantly positive. For reading skills, the size of the relationship decreases over the life course. For math skills, this relationship is remarkably stable across adulthood. An increase of 1 SD in early math skills is associated with an increase of around 0.07-0.10 SD in adult life satisfaction. This is a strong relationship. It is roughly equal to the size of the relationship between take-home income and life satisfaction at age 46. Income and daily functioning partially mediate these relationships. When inspecting domain-specific satisfaction, we find that early reading and math skills predict general health satisfaction, financial management, and emotional satisfaction in adulthood. These relationships remain significant during adulthood. A large body of literature explores the relationship between early skills and objective outcomes in adulthood, such as labor market performance and health (e.g., Crawford & Cribb, 2013; Fischbach et al., 2013; Hanushek et al., 2015; Hatch et al., 2007; McIntosh & Vignoles, 2001; 16
Ritchie & Bates, 2013; Sabates & Parsons, 2012; Von Stumm et al., 2013). Building on Sen’s capabilities approach, Heckman and Corbin (2016) highlight the role of early skills in shaping not only objective outcomes but also in enabling adults to flourish. Building on this, we focus on life satisfaction as an outcome variable, an important proxy for overall life quality. Life satisfaction, which has gained attention across disciplines such as economics, sociology, and psychology (e.g., Diener, 2000; Diener et al., 1985; Diener et al., 1999), represents a weighted average of satisfaction across life domains, varying in importance across individuals (Erdogan et al., 2012; Heller et al., 2004; Pavot & Diener, 2008). As a subjective measure, it reflects personal evaluations of life quality. Related to our work, a variety of papers analyze the relationship between early cognitive skills and well-being. Dodgeon et al. (2020) show that cognitive performance at age 7, measured by a latent variable constructed from four tests (including reading and arithmetic) relates to physical well-being and the quality of life at age 50 (proxied by control, autonomy, self-realization, and pleasure). Similarly, Von Stumm et al. (2013) show that eleven-year-olds’ cognitive ability, measured by a latent variable of verbal reasoning, arithmetic, and English tests, relates to adult physiological distress at ages 46-51. Layard et al. (2014) find a positive relationship between childhood intellectual performance at ages 5, 10, and 16 and life satisfaction at age 34. Fle`che et al. (2021) examine whether the relationship between childhood cognition and adult life satisfaction remains stable over time, showing that intellectual performance at ages 5, 10, and 16 is positively related to life satisfaction at ages 26, 30, 34, and 42. They use the British Ability Scales score at age 10, which aims at measuring general intellectual functioning. Unlike these studies, which analyze the predictions of cognitive 17
Chapter 2. Early Skills and Adult Life Satisfaction tests, we focus on early reading and math skills, controlled for intelligence. The paper closest to this chapter is Frijters et al. (2014), which analyzes the relationship between life satisfaction and various childhood skills, including reading and math test scores. Using the National Child Development Study (NCDS) and the British Cohort Study (BCS70), they examine the link between early skills and average life satisfaction in adulthood (ages 33-50 for the NCDS and ages 29-38 for the BCS70). They demonstrate a positive relationship between early math skills and average adult life satisfaction, and a negative relationship for reading. They explain that while there is a positive raw correlation between early reading skills and life satisfaction, this turns negative when controlling for other cognitive and non-cognitive measures. Unlike Frijters et al. (2014), who focus on an average measure of life satisfaction, we are specifically interested in the trajectory of this association across life. Life satisfaction likely varies over the years due to changes in life circumstances (e.g., career, relationships, health, finances), evolving personal values, and shifting perceptions and expectations. Thus, it is valuable to analyze how the association between early skills and adult life satisfaction develops over life. In sum, this chapter contributes to the literature by investigating how the relationship between early skills and life satisfaction evolves across adulthood. In addition, we contribute by analyzing which domains primarily drive this longitudinal relationship. This chapter continues as follows: Section 2.2 discusses the data, measures and descriptives. Section 2.3 explains the empirical strategy. Section 2.4 provides the results. Section 2.5 concludes. 18
2.2 Data, measures and descriptives 2.2.1 Data This chapter uses longitudinal data from the 1970 British Cohort Study (BCS70), which tracks around 18,000 individuals born in England, Scotland, and Wales during one week in April 1970. Since birth, there have been nine study sweeps, starting in 1975 at age five and ending most recently in 2021 at age 51. The data cover topics such as education, health, and social development and, essential for our study, include information on early language and math skills and life satisfaction measures across all sweeps. The longitudinal nature of the data allows us to examine long-term associations between early skills and life satisfaction. Additionally, all individuals in the sample are of the same age, eliminating age-related variability, and share similar historical and cultural contexts, moving through life stages simultaneously. 2.2.2 Measures Skill variables at age 10 We assess reading and math skills by two achievement tests taken in the sweep in 1980: the Shortened Edinburgh Reading Test and the Friendly Maths Test.1 All cohort members were ten years old at the time of these tests. We standardize the achievement test scores to a mean of zero and standard deviation of one to increase comparability among tests. 1Table A.3 contains a short description of these tests. 19
Chapter 2. Early Skills and Adult Life Satisfaction Outcome variables at ages 29, 34, 42, and 46 In various sweeps throughout their lives, individuals were asked questions about their life satisfaction. We use this life satisfaction measure as our main outcome variable. The question was asked how satisfied the respondents are with how their life has turned out so far on a scale from zero to ten, ranging from completely dissatisfied to completely satisfied. This is a common way to measure life satisfaction in this type of research (e.g., Fle`che et al., 2021; Frijters et al., 2014; Layard et al., 2014). Furthermore, we select five domains that the literature has identified as relevant to life satisfaction and assess individuals’ satisfaction in these domains: self-perceived general health, financial management, happiness in relationship, job satisfaction, and emotional satisfaction.2 First, self-perceived general health is measured by a 5-point Likert scale, assessing satisfaction with physical and mental health. Second, financial satisfaction is measured by asking individuals how well they manage financially, also on a 5-point Likert scale. Third, individuals were asked about their happiness regarding their relationship on a 7-point scale. Fourth, job satisfaction is assessed by asking how satisfied individuals are with their current job, measured on a 5-point Likert scale. And fifth, emotional satisfaction is measured by the ab2We base these five life domains on Felce and Perry’s (1995) model of quality of life, which includes: 1) Physical well-being, 2) Material well-being, 3) Social wellbeing, 4) Development and activity, and 5) Emotional well-being. In selecting our variables, we made a deliberate effort to address all five domains. These variables have been identified as predictors of life satisfaction in existing literature (e.g., Brief et al., 1993; Clark et al., 2017; Diener et al., 2000; Haller & Hadler, 2006; Hart, 1999; Margolis & Myrskyla¨, 2013; Mroczek & Spiro III, 2005; Palmore & Luikart, 1972; Seghieri et al., 2006; Tait et al., 1989). 20
sence of malaise, which is captured through a 9-item survey on negative emotions and physical responses. For clarity, we invert the latter scale, where zero indicates malaise and nine indicates no signs of malaise, and refer to this as emotional satisfaction. We argue that these five measures are meaningful indicators for individual’s satisfaction in multiple life domains.3 All measures are standardized to a mean of zero and a standard deviation of one. Control variables We also include two control variables. The most important potential confounder is intelligence,4 for which we use a proxy measured at age 10. Intelligence quotients (IQ) are widely used to assess intelligence but require expensive and time-consuming tests. In the BCS70, British researchers developed the British Ability Scales (BAS), which provide a meaningful profile of general cognitive abilities. Our analysis includes the BAS Matrices Scale, a non-verbal reasoning test similar to the Raven Progressive Matrices, which proxies intelligence as a component of Spearman’s g for general intelligence. Cohort members completed this test at age 10, alongside assessments of their reading and math skills. We also control for sex in the regressions. 3Although we make every effort to provide an elaborate overview of individuals’ satisfaction in life, the measures that we select have their limitations. The job satisfaction measure, for instance, is only reported by individuals that have a job. This measure might, therefore, not capture the fact that certain individuals are not satisfied because they do not have a job, while early reading and math skills might relate to the possibility of getting a job. We examine the extent to which our selected domain satisfaction is reflected in the measure on life satisfaction in Figure A.5 in the appendix. 4The literature describes the relationship between intelligence and factors such as academic success in young adulthood and outcomes later in life (e.g., Borghans et al., 2016; Gale et al., 2009; Schneider et al., 2014). 21
Chapter 2. Early Skills and Adult Life Satisfaction Mediators We explore two broad channels which may mediate the relationship between early skills and adult life satisfaction at age 46: income and daily functioning.5 In the first channel, we consider whether the development of skills during childhood might lead to better educational and employment opportunities, resulting in higher income levels. These higher income levels, consequently, can provide more resources to individuals (e.g., improved living conditions, healthcare, leisure activities etc.) which, in turn, might contribute to higher overall life satisfaction. The relationship between early skills and adult income (e.g., Charette & Meng, 1998; Crawford & Cribb, 2013; McIntosh & Vignoles, 2001; Vignoles et al., 2011) and the relationship between income and life satisfaction (e.g., Cheung & Lucas, 2015; Diener & Biswas-Diener, 2002; Kahneman & Deaton, 2010; B. Stevenson & Wolfers, 2013) are well-documented in the literature. In this chapter, we consider the individual’s total take-home income (after tax and deductions) at age 46. In the second channel, we consider that the development of skills during childhood may affect how individuals function in daily life, which might, in turn, improve their satisfaction in life. For this purpose, we use five measures of the 36-Item Short Form Health Survey (SF-36): role limitations due to physical health, role-limitations due to emotional problems, social functioning, energy/fatigue, and pain. Role limitations refer to the constraints or restrictions that hinder an individual’s ability to carry out their usual roles and responsibilities (e.g., whether the individual had to cut down time spent on work, whether 5Note that we only examine mediation at age 46 due to data availability. 22
the individual accomplished less than they would like to, etc.). We select these measures of daily functioning as they are mentioned as relevant factors for life satisfaction in the literature (e.g., Dixon et al., 2001; Khodabakhsh, 2022; McNamee & Mendolia, 2014; State & Kern, 2017; Wahl et al., 2009). 2.2.3 Descriptives Table 2.1 presents descriptive statistics for all variables. We have approximately 11,600 observations for our skill measures at age 10. The average reading score is 40, and the average math score is 44, though these are not directly comparable due to differing test scales. Life satisfaction is measured multiple times across the life course on a 0–10 scale. At age 29, we have around 11,000 observations, with an average of 7.29, and at age 46, we observe approximately 8,500 observations, averaging 7.35. Life satisfaction is highest at age 34. While raw measures are shown in the table, we use standardized measures throughout the chapter. The descriptives show clear signs of attrition. We address this issue in the robustness section. 23
Chapter 2. Early Skills and Adult Life Satisfaction Table2.1: Descriptive statistics. Variable Obs. Mean Std. Dev. Min Max Early skills (Age 10) Reading 11,638 40.234 12.678 0 65 Math 11,629 43.953 12.322 0 72 Outcome variables Life satisfaction Age 29 11,104 7.290 1.848 0 10 Age 34 9,594 7.405 1.799 0 10 Age 42 9,696 7.356 1.999 0 10 Age 46 8,487 7.351 1.894 0 10 Self-perceived health Age 29 11,211 3.148 .715 1 4 Age 34 9,632 4.041 .888 1 5 Age 42 9,799 3.599 1.074 1 5 Age 46 8,576 3.441 1.097 1 5 Financial management Age 29 11,216 3.953 .965 1 5 Age 34 9,637 4.050 .925 1 5 Age 42 9,796 4.055 .979 1 5 Age 46 8,571 4.055 .932 1 5 Happiness in relationship Age 29 7,405 2.864 2.264 1 7 Age 42 7,378 5.729 1.704 1 7 Age 46 6,433 5.713 1.624 1 7 Job satisfaction Age 29 9,135 3.976 1.012 1 5 Age 34 8,003 4.096 .977 1 5 Age 42 8,304 4.077 .949 1 5 (Continued on next page) 24
(Continued from previous page) Variable Obs. Mean Std. Dev. Min Max Age 46 7,470 4.081 .961 1 5 Emotional satisfaction Age 29 11,109 20.468 3.486 0 24 Age 34 9,596 7.323 1.999 0 9 Age 42 8,578 7.137 1.999 0 9 Age 46 7,851 7.233 2.122 0 9 Mediators (Age 46) Income 7,909 22,696.676 73,696.261 0 4,000,000 Daily functioning Role-limitations - physical health 7,928 84.738 31.247 0 100 Role-limitations - emotional problems 8,001 83.639 32.061 0 100 Social Functioning 8,016 85.344 24.108 0 100 Energy/fatigue 8,012 57.148 22.013 0 100 Painscore 8,014 78.031 24.056 0 100 Controls (Age 46) Intelligence (BAS Matrices) 11,496 15.343 5.399 0 28 Sex 18,031 .499 .500 0 1 Note. This table presents the descriptive statistics of our sample. All continuous variables are standardized to have a mean of zero and a standard deviation of one in the remainder of the chapter. 25
Chapter 2. Early Skills and Adult Life Satisfaction 2.3 Methodology To investigate whether and to what extent early reading and math skills relate to adult life satisfaction, we use Ordinary Least Squares regressions. We consider the following model: OUTCOMEi =β0 +β1SKILLi +β2IQi +β3SEXi +ϵi (2.1) whereOUTCOMEi represents the outcome variable for individual i. In our main analysis, the outcome variable is life satisfaction, measured at different phases in life, which we treat as a continuous variable. This is commonly done in life satisfaction research and justified by Ferreri-Carbonell and Frijters (2004), who conclude that linear and ordinal models yield essentially the same results for life satisfaction. Our variable of interest is SKILLi. This variable refers to the standardized skill level of the individual at age 10. We consider reading and math skills separately. Furthermore, we control for the individual’s general intelligence (IQi) andsex (SEXi) in most specifications. ϵi is the error term. To explore income and daily functioning as mediators, we add these variables in further specifications of the model. 26
2.4 Results 2.4.1 The relationship between early skills and adult life satisfaction We first investigate whether ten-years-olds’ reading and math skills relate to their assessment of how life has turned out so far in adulthood. Table 2.2 shows the regression estimates of the relationship between early reading and math skills at age 10, and life satisfaction of individuals at ages 29, 34, 42, and 46 without controls and when controlling for intelligence and sex. Table 2.2 demonstrates that reading skills significantly relate to adult life satisfaction. Beyond age 42, the strength of this relationship and its statistical significance decline noticeably. At age 46, a 1 SD increase in reading skills at age 10 is associated with a 0.04 SD increase in adult life satisfaction. The significant positive relationship between early math skills and adult life satisfaction remains relatively stable across different ages, with values ranging from approximately 0.07 to 0.10 SD. The magnitude of this relationship is substantial when comparing it to other relations, such as the relationship between standardized take-home income and life satisfaction at age 46, which is 0.09 (without controls). Controlling for intelligence and sex reduces the estimates for reading skills but not for math. In Table A.10 of the appendix, we extend the analysis by controlling for the social class of the cohort members’ father’s occupation (or mother’s occupation, if paternal information is missing) to proxy the cohort members’ economic situation. The father’s occupation proxies 27
Chapter 2. Early Skills and Adult Life Satisfaction Table2.2: Regression estimates of the baseline relationship between adult life satisfaction and early skills. Life satisfaction Without Controls With Controls Age 29 Age 34 Age 42 Age 46 Age 29 Age 34 Age 42 Age 46 Panel A: Reading Skill .096*** .089*** .078*** .043*** .069*** .064*** .060*** .028* (.012) (.013) (.013) (.014) (.014) (.016) (.016) (.017) Constant -.007 .002 -.005 -.008 -.061*** -.042** -.038** -.065*** (.011) (.012) (.012) (.013) (.016) (.017) (.017) (.018) Controls NO NO NO NO YES YES YES YES Obs. 8,264 7,177 7,175 6,281 8,164 7,097 7,086 6,203 R-squared .009 .007 .006 .002 .012 .011 .007 .005 Panel B: Math Skill .098*** .090*** .094*** .071*** .090*** .080*** .099*** .087*** (.012) (.013) (.012) (.013) (.015) (.016) (.016) (.017) Constant -.006 .004 -.006 -.013 -.071*** -.052*** -.049*** -.077*** (.011) (.012) (.012) (.013) (.016) (.017) (.017) (.019) Controls NO NO NO NO YES YES YES YES Obs. 8,265 7,179 7,173 6,277 8,162 7,094 7,080 6,197 R-squared .009 .008 .008 .005 .014 .012 .010 .008 Note. This table represents the results for a regression of life satisfaction at ages 29, 34, 42, and 46 (in SD) on early reading and math skills (in SD). The four leftmost columns present the results without controls. The four rightmost columns present the results, controlled for intelligence and sex. Robust standard errors are in parentheses. ***p< .01, **p< .05, *p< .1. economic, social, and cultural capital that may shape the individual’s skills and life satisfaction (through e.g., financial resources, social networks, values, and beliefs). Results show that part of the relationship between early reading skills and life satisfaction is driven by these forms of capital, with the link between early reading skills and life satisfaction becoming insignificant only at age 46. For math, 28
the results remain unchanged when including additional controls. We then also include additional childhood variables related to communication and behavior at age 10, such as the Rutter behavioral score, being sullen/sulky, and three interpersonal skills (communication, problem-solving, and teamwork). The results show that reading skills no longer significantly relate to adult life satisfaction, suggesting that part of the relationship between reading and life satisfaction is driven by interpersonal and behavior skills. For math, the results remain unaffected by these controls, albeit with a slight reduction in the magnitude of the estimated relationship. 2.4.2 Income and daily functioning as potential mediators In the appendix, Table A.13 explores income and daily functioning as potential mediators in the relationship between early skills and adult life satisfaction at age 46 (controlled for sex and intelligence). For reading, the appendix shows that income and daily functioning together (almost) fully mediate the baseline relationship between early reading skills and adult life satisfaction. For math, it shows that income and daily functioning together mediate around 64 percent of the baseline relationship between early math skills and adult life satisfaction, where the mediating role of daily functioning is stronger than the mediating role of income.6 Table A.13 in the appendix also shows that all daily functioning variables (at least) partly mediate the relationship when including them 6The coefficient for math in the last column of Table 2.2 is .087. When adding income and daily functioning in our model, this coefficient drops to .031. This means that the estimate has decreased by 64 percent. 29
Chapter 2. Early Skills and Adult Life Satisfaction separately in the regression. The table suggests that, of these variables, energy/fatigue is most strongly related to life satisfaction. This is followed by social functioning, i.e. how individuals engage in social activities (e.g., visiting friends). When including all daily functioning variables together in the regression, pain does not significantly mediate the relationship anymore, whereas the other functioning measures do.7 2.4.3 The relationship between early skills and distinct adult life domains We now examine how early skills relate to adult satisfaction in multiple life domains: 1) self-perceived general health, 2) financial management, 3) happiness in relationships, 4) job satisfaction, and 5) emotional satisfaction.8 We apply the Bonferroni correction to account for multiple hypothesis testing. The estimates of the relationships between early skills and domain-specific satisfaction are shown in Figure 2.1, with the first panel illustrating estimates for reading and the second panel those for math. First, we find that early reading and math skills positively relate to adult self-perceived general health. A one standard deviation increase 7This is consistent with a study by Wahl et al. (2009), who find that fatigue mediates the relationship between chronic pain and general life satisfaction. 8In Table A.5 of the appendix, we assess how well our selected domain satisfactions reflect the general life satisfaction measure by examining the adjusted Rsquared of regressions of life satisfaction on domain-specific satisfaction. We analyze each domain separately and then all domains together. The appendix shows that domain-specific satisfaction increasingly contributes to overall life satisfaction over the years, with the five domains explaining 16.3% of its variation at age 29, rising to 33.3% by age 46. This trend holds for all domains except job satisfaction. Malaise explains most variation in overall life satisfaction at all ages. 30
Figure 2.1: The development of the relationship between adult domainspecific life satisfaction and early skills. 0.12*** 0.11*** 0.17*** 0.17*** Reading score (in SD) .05 .1 .15 .2 Reading 0.12*** 0.10*** 0.16*** 0.17*** Math score (in SD) .05 .1 .15 .2 Math General health 0.12*** 0.12*** 0.12*** 0.12*** Reading score (in SD) .08 .1 .12 .14 .16 .18 Reading 0.14*** 0.13*** 0.14*** 0.15*** Math score (in SD) .08 .1 .12 .14 .16 .18 Math Financial management -0.05** 0.04 -0.02 Reading score (in SD) -.1 -.05 0 .05 .1 Reading -0.01 0.02 -0.01 Math score (in SD) -.1 -.05 0 .05 .1 Math Happiness in relationship -0.03 -0.01 -0.04 -0.01 Reading score (in SD) -.1 -.05 0 .05 .1 Reading -0.01 -0.01 -0.00 0.04 Math score (in SD) -.1 -.05 0 .05 .1 Math Job satisfaction 0.08*** 0.07*** 0.05** 0.04* Reading score (in SD) 0 .05 .1 .15 Reading 0.11*** 0.12*** 0.11*** 0.11*** Math score (in SD) 0 .05 .1 .15 Math Emotional satisfaction Age 29 Age 34 Age 42 Age 46 Note. This figure represents the estimates of the regressions of the adult satisfaction measures (in SD) on early skills (in SD). The dots represent the estimates for satisfaction at age 29, whereas squares, diamonds, and triangles represent the estimates at ages 34, 42, and 46, respectively. The bands illustrate the 95% confidence intervals. Unreported controls are intelligence and sex. The stars complementing the estimates show the significance level of the estimates. These significance levels apply a Bonferroni correction for five hypotheses. ***p< .01, **p< .05, *p< .1. in reading or math skills at age 10 is associated with a 0.17 SD improvement in self-rated health at age 46. We explore potential explanations for this relationship in the appendix, starting with income as a mediator. Early skills may lead to higher educational attainment, better job opportunities, and ultimately higher income, providing greater access to quality healthcare, nutritious food, or safer living environments, which are all critical determinants of good health. Table A.14 in the appendix shows that income mediates around 2% and 5% of the relationship between early reading and math skills and adult general health, respectively. Next, we consider daily functioning as a media31
Chapter 2. Early Skills and Adult Life Satisfaction tor. These skills may enhance an individual’s ability to perform everyday tasks efficiently, contributing to better general health satisfaction. Daily functioning mediates around 24% and 44% of the relationship between general health satisfaction and early reading and math skills, respectively (cfr. Table A.15). Together, income and daily functioning mediate around 26% and 46% of the relationship between early skills and adult self-perceived health (cfr. Table A.16). This suggests that a significant portion of the relationship is explained by these mediators, although some part remains unexplained.9 Second, we find that early skills also positively relate to financial management at all ages in adulthood. A one standard deviation increase in early reading skills is associated with a 0.12 SD increase in satisfaction with financial management, while for early math skills, the increase is around 0.15 SD. According to Table A.16, income and daily functioning together explain approximately 16% of the relationship for reading and 27% for math. Thus, regardless of income or functioning, individuals with a higher level of early skills tend to report better financial positions. One explanation for this might be that reading and math skills enhance understanding of financial concepts (e.g., budgeting, saving, 9Early skills may relate to healthy behavior by enabling individuals to better understand health information, follow medical instructions, and navigate healthcare systems. Sabates and Parsons (2012) link low literacy and numeracy to higher risks of adverse behaviors like smoking, mediated by health knowledge. Arnold et al. (2001) find a positive link between women’s reading levels and awareness of the health effects of smoking. Unlike smoking, alcohol consumption is positively correlated with education and basic skills (De Coulon et al., 2010; Huerta & Borgonovi, 2010). Table A.17 in the appendix shows that moderate drinking and eating breakfast relate positively to life satisfaction, while smoking has a negative relationship. Healthy behavior partly mediates the link between early skills and health satisfaction, with math skills relating to health satisfaction more strongly through daily functioning than reading skills. 32
www.proefschriften.netRkJQdWJsaXNoZXIy MjY0ODMw