This is the second in this year’s series of posts by PhD students on the job market. Reminder that submissions close tomorrow at noon.
When I was in school, I used to skip some classes; my parents probably never learned about it (at least until this blog post). Now imagine there was a system in place to tell my parents how often I skipped classes in the previous weeks. When Nina’s parents get a message pointing out that she missed school yesterday, while they learn at no cost that her behavior was not in line with their expectations, they may also realize that attendance is an important dimension of their daughter’s behavior to which they should attend moving forward. Therefore, informing parents about their child’s attendance, for instance, might also draw their attention to the importance of attending class, increasing their awareness about monitoring benefits. This is particularly relevant in face of the evidence on how poverty captures attention (Mani et al., 2013), and on how, given limited attention, individuals may fail to learn from dimensions they do not notice (Hanna et al., 2014). Therefore, while informing parents lowers monitoring costs – alleviating moral hazard and inducing better school outcomes, in line with parent’s objectives –, communication may also increase the salience of monitoring benefits. My job market paper investigates the mechanisms behind the effects of communication with parents.
Why does it matter?
Informing parents about children’s attendance and grades has been shown to significantly raise educational achievement (Bergman, 2017; Dizon-Ross, 2017; Berlinski et al., 2016; Kraft and Dougherty, 2011; Rogers and Feller, 2016). However, there is no evidence for why communication with parents works: is it mainly because information lowers monitoring costs, or is it mainly because it increases the salience of monitoring benefits? The distinction matters for two reasons. First, in developing countries, real-time information systems are often absent, making informational interventions expensive. If salience explains most of the effects of communication, similar effects could be achieved at much lower costs, as interventions to capture attention do not require such information systems. For instance, telling Nina’s parents that it is important that Nina attend classes every day does not require any real-time information system and could be implemented at a very low cost. Second, and most importantly, if salience is the key driver of the effects of communication, potentially the effects of communication could be much larger. Once the objective is capturing parents’ attention, sharing information is just one tool in a much richer toolbox; nudging also allows manipulating other features, such as time of delivery and interactivity. Moreover, while informational interventions are constrained by the frequency at which information is available, nudging can be implemented at much higher frequency. It is also worth noting that certain pieces of information may not be as effective in raising perceived returns to monitoring as nudges. Consider a parent who thinks their kid is missing more classes than s/he actually is; information may induce him or her to monitor even less.
To decompose the effects of communication into lower monitoring costs and salience of monitoring benefits, we run a field experiment with 19,300 ninth graders across 287 public schools in São Paulo, Brazil. In the experiment, over the course of 18 weeks, Math teachers have to fill-in a platform weekly with information about their students’ behavior (attendance, tardiness and assignment completion). We randomly assign parents to different messages, shared by the platform over SMS: some parents receive information that the teacher filled in (e.g., “Nina missed less than 3 classes over the last 3 weeks"), some receive an awareness message, emphasizing the importance of paying attention to that dimension (e.g., “It is important that Nina attends class every day"), while others receive no message at all (the control group). All parents in the information treatment received messages, whether the student was underperforming or not. Because we anticipate that parents’ or peer interactions may generate large spillovers, we include a pure control group, which we use as counterfactual in most of our analyses.
What did we find?
In line with previous findings, we find that weekly communication has large impacts on attendance (2.1 percentage points, or about 5 additional classes a year), Math GPA and standardized test scores (0.09 standard deviation) and promotion rates (3.2 percentage points). We find that treated parents ask their children systematically more about school, incentivize studying to a greater extent, and have higher aspirations about their children’s making it to college. Children in treated households report engaging in academic and reading activities to a greater extent. Most strikingly, comparing the estimates of information and salience, we find that salience can account for most of the effects of information: information and salience coefficients are never statistically different at the 10% significance level, as you can see below.
Based on our findings, a nudge program targeted at capturing parents’ attention could potentially have larger effects on educational achievement. Through a different nudging program that suggests weekly activities for parents to do with their children, we test whether the combination of different features can deliver higher impacts. We cross-randomize 4 sets of features to test whether SMS frequency, time of delivery, consistency of SMS delivery time, and interactivity impact students’ outcomes in line with the behavioral mechanism. Consistent with inattention, higher frequency and alternating delivery times significantly increase effect sizes. As one would expect, however, there are decreasing returns to getting parenting to the top of mind: for attendance, we find evidence that saturation kicks-in beyond 2 messages a week. Also consistent with the mechanism, nudging affects Math and Portuguese standardized test scores to the same extent, whereas the effects of information are mostly confined to the subject it targets. Strikingly, the optimal combination of features –sending parents 3 messages a week plus an interaction message, at work hours, and not alternating time of delivery – increases students’ test scores by 0.33 standard deviation, almost 4-fold the effect of the informational intervention.
Discussion and policy implications
This is the first study to investigate the mechanisms behind the effects of communication with parents. Our findings challenge previous results about the drivers of the effects of communication. In particular, different from Bergman (2017), which attributes most of the effects of an intensive communication campaign to more accurate beliefs, we show that our effects are not driven by belief updating, but rather by higher parental engagement in response to changes in the salience of monitoring benefits.
Changing poor families’ dynamics is daunting: little is known about the intra-household education production function. Low-cost interventions, such as asking parents to study with their children, are often infeasible, as in poor families many children’s educational attainment is equal to or higher than that of their parents (69% of the ninth graders in our sample). While communication interventions have showed promise, scaling them up to developing countries involve cost challenges. This study suggests that nudge programs can be both cheaper and more effective to alleviate the moral hazard problem between parents and children.
Nina Cunha is a PhD student in Stanford University, and you can read her paper here.
- job market papers 2017