Panel Paper: Effects, Timing and Heterogeneity of the Provision of Information in Education: An Experimental Evaluation in Colombia

Saturday, November 10, 2018
Harding - Mezz Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Felipe Barrera-Osorio, Kathryn E. Gonzalez, Francisco Lagos and David Deming, Harvard University

In the human capital model, individuals invest in education if the present value of benefits exceeds costs (Becker, 1962). As such, information about the quality of education and student performance is important, since the benefits of education depend on the actual acquisition of skills in the classroom. Nonetheless, evidence suggests that families and students have limited information or are misinformed about school quality, their own academic performance, and the returns of education (Nguyen 2008, Jensen 2010, Loyalka et al. 2013). This lack of information may lead to suboptimal educational investment by households (Houtenville and Conway 2007, Avvisate et al 2014, Bergman 2015, Berlinski et al 2016, Dizon-Ross 2017). Providing performance information to parents may cause them to update their beliefs, which could lead to changes in investment of time and resources in education, and ultimately to increases in student achievement.

In this paper, we report the results of a randomized experiment examining the effects of providing information to families on their children’s reading and math achievement in a mid-size city in Colombia. Parents of children in grades 3 through 5 were randomly assigned to receive standardized information about the performance of their children during household visits. We also presented these families with a menu of options that they might consider in light of the information, such as asking their children about school every day, reading with their children more often, and by interacting with their children’s teachers.

We followed these families across multiple years. We show that the provision of information led to improvements in the academic performance of children. However, these effects are short-lived. Our results show an initial pattern of small negative effects (not statistically significant), then positive and significant short run effects (0.12 sd), and finally zero effects in the final follow-up waves. This pattern is consistent with effects that follow a pattern of action and backsliding observed in other interventions (e.g., Gallagher, 2013; Allcott & Rogers, 2014). This is consistent with models in which the parents cannot permanently modify their behavior (or change a "stock" or permanent variable); parents can temporarily modify their behavior in the short run but actions quickly return to a "business as usual" mode.

We find larger impacts for students with low-baseline scores (of the order of 0.26 sd, at the peak of effects). This is consistent with these students and families having less accurate information about performance, or an increase in parent-student information frictions in these households (Bergman, 2016). Yet we also observe the same pattern of action and backsliding of effects in this population.

We do not find any effect of parent internal investment. One explanation for the lack of effects on these mechanism is that all measures that proxy these variables were quite high at baseline, according to self-reported answers. We find, however, increases in the number of parents-teacher meetings and an update in the beliefs of the parents. We present some evidence suggesting that this last effect is the main channel of impact.