This dissertation focuses on the economics of education in developing countries with a particular focus on parental educational investment decisions in China. In Chapter 1, I explore how parents' belief affect their investments in children's education using a randomized field experiment with parents of high school students in China. I document two types of information frictions that result in systematic biases in parents' beliefs about children's ability: overconfidence in future performance and underestimating college admission requirements. I then introduce two interventions to correct parents' belief biases. In the first intervention, I use machine-learning techniques to generate predictions on children's future academic performance and distribute them to randomly selected parents. In the second intervention, I give randomly selected parents a report that lists the feasible colleges corresponding to their children's current academic performance. I find that both interventions lead to dramatic reductions in belief biases. In addition, parents report higher levels of monetary investments in children's education, which significantly improved children's academic performance. I also find significant non-linearity in the impacts of ability belief on parental educational investments around their aspirations.

In Chapter 2, I investigate the impacts of peer effects on parental educational investment decisions. Using a randomized experiment with 3379 parents of high-school students in China, I identify two channels of social influence in parents' decisions on children's educational investments: parents adjust their decisions based on other parents' behaviors because they learn from other parents' decisions ("social learning") or because their children are facing competition from peers ("competition externality"). I find that both channels have statistically significant effects on parents' investment decisions and increase their willingness to buy an educational service by over 20%. Although the average effects of the two channels are not statistically different, the main channels of peer effects are heterogeneous by parents' educational background: parents with higher education, higher income, and those who only have one child are more likely to learn from peers' decisions whereas those with lower education, lower-income, and more than one child are mainly incentivized by the competition externality.

Chapter 3 provides a theoretical explanation for the empirical findings documented in Chapter 1. I introduce the reference-dependent utility theory into the parental education investment decisions by cooperating parents' aspirations into their utility and assuming parents' utility function is discontinuous at the thresholds of achieving aspirations. The modification generates an interesting non-monotonic correlation between ability and optimal educational investments around the aspirations. When children haven't achieved their aspirations, parental educational investment is substitutive to children's ability - the lower the ability, the higher the investment. In contrast, when the aspirations are already reached, parents' investment becomes complementary to children's ability - the higher the ability, the higher the investment. This model can rationalize the remedial investments behaviors.