PhD defence: Studies in human capital development in childhood

PhD defence

Ninja Ritter Klejnstrup


This thesis consist of an introductory chapter and four self-contained chapters.

In Chapter 1, the introduction, I provide a detailed background for each of the chapters, outlining the theoretical background for the questions that they address, as well as a summary of the empirical literature that they contribute to, and the contributions of the chapters themselves.

Chapter 2, A Comparison of Model-Based and Design-Based Impact Evaluations of Interventions in Developing Countries, which is co-authored with Henrik Hansen and Ole Winckler Andersen, represents an input into a debate about methods for learning about what works in development aid and policy. At the core of this debate is whether non-experimental estimators of causal effects of social programmes can overcome selection bias. We survey four impact studies, all using data from development interventions, that directly compare experimental and non-experimental impact estimates. Our review illustrates that when the programme participation process is well understood, and correctly modelled, the non-experimental estimators are able to overcome selection bias to the same degree as randomised controlled trials. Based on this we suggest operational guiding principles that are likely to promote credible causal inference in situations where randomised controlled trials are not feasible or desirable.

Chapter 3, Formal Education, Malaria Preventive Behaviour and Children’s Malarial Status in Tanzania, is co-authored with Joel Silas Lincoln. In this chapter, we explore formal education as a causal determinant of women’s malaria preventive behaviour, as well as children’s risk of malaria infection. Understanding determinants of children’s risk of malaria is important in the context of Tanzania, since malaria, despite substantial reductions in prevalence since the early 2000s, remains a leading cause of morbidity and mortality. For identification of the effect of women’s schooling, we rely on exogenous variation in educational attainment generated by educational reforms during the 1970s. We use data from four rounds of either Demographic and Health Surveys or Malaria Indicator Surveys, which allows us to explore variation in the relationship over time. Our results vary across both the outcomes considered and survey years. Results for the early survey years, but not the later years, indicate that women’s years of schooling increased their probability of using malaria prophylaxis during pregnancy, and their children’s probability of sleeping under insecticide treated bed-nets. At the same time, we find that the magnitude of the effect is greater for women’s use of malaria prophylaxis during pregnancy than for children’s probability of sleeping under bed nets. We argue that this difference in magnitude is likely due to differences in the mechanism linking these outcomes to education, with the latter being mediated by income to a higher degree than the former. Results for women’s own use of insecticide treated bed-nets and for children’s malaria status are inconclusive, due to low statistical power and discrepancies between graphical evidence and regression-based estimates.

Chapter 4, Early Life Malaria Exposure and Academic Performance, is coauthored with Julie Buhl-Wiggers, Sam Jones, and John Rand. In this chapter, we explore whether malaria exposure in early childhood affects educational achievement in later childhood, also in the context of Tanzania. For causal identification, we rely on differences across districts, in the pace of decline in malaria prevalence occurring over the last 15 years. We control for time-invariant district level, age, birth cohort and survey year effects, as well as district level trends and individual and household specific factors. In addition, we use sibling variation in birth-year exposure to strengthen our identification. To implement this strategy, we combine repeated cross-sectional data on children’s numeracy and literacy collected by the Uwezo project, which are representative at both the national and district levels, and pixel level estimates of malaria prevalence produced by the Malaria Atlas Project. We find that birth year malaria prevalence negatively predicts English literacy achievement, and that the effect of a ten percentage point reduction in malaria prevalence is comparable in magnitude to some of the most effective school-based interventions targeting learning in developing country contexts. We find no effect of birth year malaria prevalence on children’s Kiswahili literacy nor on numeracy, but argue that this is likely to be attributable to strong ceiling effects in these test scores. We conclude, that malaria is an important factor in explaining geographical variation in English literacy in Tanzania and that it represents a significant public health challenge to education in the country, and most likely also in other malaria-endemic areas.

In Chapter 5, Teacher Absence and Student Achievement in Danish Public Schools, I explore the frequency and distribution of teacher absences across Danish public schools, as well as their effects on ninth grade students’ performance on written Danish and math exams. The analysis is based on register data provided by Statistics Denmark, as well as data on students’ test scores and teacher-classroom links made available by the Danish Ministry of Education. I analyse the effect of teacher absences within a value-added framework, controlling for past test scores as well as school fixed effects and a range of student and peer demographic and socio-economic characteristics. I show that Danish teachers have high levels of absence from work, and that teachers often take many spells of sick leave each year. Further, I show that schools serving students of lower socio-economic status more often have persistently high rates of teacher sickness absence, than do other schools. I argue that this has consequences for the equality of opportunities in the Danish public school system, because ninth grade students do perform worse on exams, when assigned to teachers with high levels of sickness absence. Thus, I find a significant effect of teachers’ sickness absence (but not maternity absence) on students’ exam performance. Much of the effect, however, is accounted for by unobservable teacher characteristics. This has policy implications, as it indicates that policies to reduce teacher absences may not eliminate the negative effect of teachers with high levels of sickness absence, if they do not also address the at-work performance of these teachers.


Bo Jellesmark Thorsen, Professor, Department of Food and Resource Economics (IFRO), University of Copenhagen

Assessment Committee

Chair: Arne Henningsen, Associate Professor, Department of Food and Resource Economics (IFRO), University of Copenhagen
Rebecca Thornton, Associate Professor, Department of Economics, University of Illinois
Maria Humlum, Associate Professor,  Department of Economics and Business, University of Aarhus

If you are interested in a full copy of the thesis, you can contact the PhD student or the supervisor.