The role of peer effects in dynamic discrete choice models: with applications to technology adoption and educational choice
Forward looking behavior is important in many contexts. Investors pay a cost today, hoping to receive a return in the future. Students are willing to study hard, hoping this will lead to better jobs. At the same time, these individual choices can have repercussions for society as a whole. Investments in green technologies slow down climate change. Highly educated students earn higher wages and therefore provide more tax income and depend less on social security. For this reason it is important to study why individuals make these decisions. Once we understand their behavior, we can derive policy implications that can help in shaping government policies that benefit society.
The toolbox of econometrics offers many ways to investigate this. However, each specific policy question brings new challenges to empirical economists. In this dissertation, I show three topics in which different models help me to understand individual behavior and I use this knowledge to formulate policy implications that aim to improve social welfare. In each chapter I want to contribute on three levels. First, investigate the causal impact of choices on future outcomes and understand why individuals make these choices. Second, provide policy implications that could help society. Third, improve the models we use to help researchers in other contexts to investigate dynamic behavior.
I focus on two applications for which I use data from Flanders (Belgium): the adoption of solar panels and study choices of high school students. In chapter 1 and 2, we estimate structural models of green technology adoption (joint work with Frank Verboven) and study choices. In these types of models, dynamic behavior is accounted for by assuming individuals solve an optimization problem that explicitly accounts for the future impact of their choice. Connecting the theoretical model to a dataset then gives us parameters with economic meaning that help us to understand the patterns in the data, but also allow us to simulate the impact of different policies. In the third chapter, Koen Declercq and I investigate study choices by applying a different methodology. We estimate the causal effects of choices today on outcomes in the future directly using models from the treatment effect literature. To learn more about the behavior of students, we explore heterogeneity in the treatment effects and look at multiple outcomes. This in turn helps us to derive policy implications.
Chapter 1: Subsidies and time discounting in new technology adoption: evidence from solar photovoltaic systems (joint with Frank Verboven).
We study a generous program to promote the adoption of solar photovoltaic (PV) systems through subsidies on future electricity production, rather than through upfront investment subsidies. We develop a tractable dynamic model of new technology adoption, also accounting for local market heterogeneity. We identify the discount factor from demand responses to variation that shifts expected future but not current utilities. Despite the massive adoption, we find that households significantly discounted the future benefits from the new technology. This implies that an upfront investment subsidy program would have promoted the technology at a much lower budgetary cost.
Chapter 2: The effects of high school curriculum. A model of program and effort choice.
This chapter addresses the impact of study program choices in secondary education. I estimate a dynamic model of educational decisions that allows for observed and unobserved differences in initial ability. It is novel in that it adds unobserved effort as a choice variable, along with the choice of study program. This replaces traditional approaches, which assume end-of-year performance follows an exogenous law of motion. I use the model to investigate policies that aim to match students to the right program. I find that policies that encourage underperforming students to switch to less academic programs do not have a negative impact on graduation rates in higher education and they substantially reduce grade retention and drop out. Ignoring that students choose their effort level generates biases in counterfactual predictions.
Chapter 3: Tracking and specialization of high schools: heterogeneous effects of school choice (joint with Koen Declercq).
In the same context, with more recent data, we analyze the causal impact of choosing for an elite high school on study performance. Schools can offer one or more tracks, but elite schools offer only the academic track. If students underperform in this track, they can switch to a different track to avoid study delay. For students in elite schools, switching track also implies switching schools. We account for self-selection and heterogeneity in the treatment effect and derive a small and non-significant average effect. However, we find substantial heterogeneity. Students who self-select into elite schools experience the most negative effects. This result can be explained by track switching. Students with a high preference for elite schools do not want to leave this school to sort into a better suited track.