This is the thirteenth in our series of posts by students on the job market this year.
A standard result in industrial organization is that competition increases consumer welfare by incentivizing firms to lower prices and increase quality. Yet, education may be very different since: (1) a child’s learning depends on the match between the child and the school and (2) not all children are equally responsive to their learning in a school when they make enrollment decisions. Once these conditions are explicitly modelled, under plausible assumptions, an increase in competition can lead schools to increase their focus on wealthy, high ability children at the expense of poorer children. In my job market paper, I formalize this intuition and then combine a structural model of school choice with quasi-experimental results to show that increasing competition in a schooling market increases inequality in test-score gains by 0.1 standard deviations.
Why is this important: School competition is increasingly used as a tool to increase school quality in both rich and poor countries. In low-income countries like India and Pakistan (the focus of my paper), more than 65 percent of urban students and 28 percent of rural students are enrolled in private schools. Recent work suggests that private schools are associated with (weakly) higher learning at (strongly) lower costs, offering an alternative to publicly provided education. Yet, how such schools respond to competition given that “bad’’ schools may not be bad for everyone and “good’’ schools may not be good for everyone has not received much attention in the literature. To see how the match between a child and a school can matter for learning, consider whether a student in need of remedial math courses would benefit from a high-level calculus class or vice versa. In general, children benefit from an educational curriculum tailored to their specific level (for example, Duflo et al.). In this scenario, if greater competition moves schools away from targeting median learning levels towards targeting more or less advanced students, one sub-group will suffer. Moreover, poor students may respond less to school quality relative to rich students when they choose schools (consistent with work by Hoxby and Avery and Ajayi). For example, poor students’ parents may value learning less, or they may be less able to assess how well their child is learning if they have little education themselves.
What does the theory predict: I develop a model of private school competition that takes into account child-specific school quality and differential demand elasticities with respect to type-specific school quality. In this model, schools can cater to different types of students and poor students may be less responsive to their match to schools when they make enrollment decisions. The model predicts that schools respond to competition by targeting wealthier students at the expense of poorer students, since the returns to targeting wealthier students are higher. To see why, imagine a village with one private school. Since wealthier students are very responsive to school quality, the single private school knows they will enroll as long as they prefer it to a government school. Therefore, the school will cater as much as possible to less wealthy students without losing the wealthiest students. However, if a second school enters the market, the existing school now faces competition to retain the wealthy students and will therefore compete more intensively for them.
What do the data show: Using the Learning and Educational Attainment in Punjab Schools data (LEAPS), I first test the model’s key assumption - that poorer students are less responsive to their match when they make enrollment decisions - in rural Pakistan's competitive private schooling market. I estimate a structural model of school choice, allowing the student's perceived utility in a school to depend on her distance to the school, her predicted test score gains, school fees, and an unobserved school characteristic. I find that the coefficient on type-specific school quality in a rich student’s perceived utility function is thirteen times as large as the coefficient for poor students. This result indicates that the model’s crucial assumption is relevant to education markets in the developing world.
To test the model's predictions, I use two types of quasi-random variation in the number of private schools in the village. First, using the panel structure of the data, I exploit time series variation in the number of private schools in a village due to exit and entry. I find that an additional private school increases test score gains for rich students and reduces them for poor students within the same school. Adding an additional private school to the village increases inequality in test score gains by 0.1 standard deviations (about 25 percent of average annual learning gains).
I also examine the effects of increased private school competition using cross-sectional variation in the number of private schools in a village. Pakistani private schools rely on secondary-school educated females (who have few outside options) to serve as cheap teachers (Andrabi et al.). Therefore, villages which received a government secondary school for girls in the 1980s have more private schools today. Interestingly, the government built these schools in the 1980s in villages that had the greatest population in their sub-district. Therefore, when a village moves from being the second most populous to the most populous village in its sub-district, there is discontinuous change in the number of private schools in the village. Using this variation, I again find that an additional private school increases inequality in test score gains by rich and poor students by 0.1 standard deviations.
Therefore, match-specific quality is an important aspect of competition. The mechanisms I identify in this paper suggest that as long as schools face competitive pressures and poorer students are less responsive to their learning when they choose schools (as in secondary school applications in Ghana, or college applications in the U.S.), competition can increase educational inequality and reduce learning for poor students already enrolled in private schools.
Discussion: What drives the differential responsiveness of poor and rich students to type-specific school quality? The research to date seems to support the idea that wealthy students are more informed about their match to schools than poorer students. If this is the case, pairing policies that increase competition like vouchers with policies that increase access to information may help harness the benefits of competition for poor students. If poor students are more informed, they will make better decisions and schools will respond by catering more to them in equilibrium. Hoxby and Turner already show that customized information can have a large effect on poor students’ decision-making when they apply to college in the United States.
Finally, heterogeneity among population sub-groups matters. In the model, the effect of competition on the mean student would be close to zero since the positive effects of competition on rich students and the negative effects on poor students cancel each other out. Thus, a null or moderate effect size of competition on mean student outcomes may mask substantial and important heterogeneity. My paper suggests that researchers may want to look at the effects of competition on different subgroups who are more or less responsive to match-specific school quality.
Natalie Bau is a PhD student at Harvard University.
Join the Conversation