Brent Jensen and Dr. Eric Eide, Economics Department

Education is a growing area of study in the area of economics and it is often difficult to find objective answers to the questions and problems that educators and policy makers face each day. For my research, I used economic theory to try and determine if there is a causal relationship between sleep and academic achievement for students between the ages 5 and 18 and to see if that relationship is significant.

Working with Dr. Eide, I gathered data from the Panel Study of Income Dynamics (PSID) which has gathered data from families for decades and looked at a subset of the data that is called the Child Development Supplement (CDS) which focuses on adolescents. This is a rich data set and allowed me to evaluate different types of data that measure the amount of sleep to see if there are significant differences in determining the effects of sleep on academic achievement. First, the data contains self reported sleep data where the individuals would report how much the thought they slept on average. This is the sleep data that is typically used when a sleep variable is a variable of interest. In addition to the self reported sleep variable, the CDS contained time diary sleep data where the individuals would record their activities for a 24 hour period. Using time diary data theoretically provides a more accurate measure of the time spent participating in any given activity as long as the activity is a common activity that many individuals participate in.

There were different techniques I used in determining the causal relationship. First I used ordinary least squares regressions which measure the effects of sleep on academic achievement at the average. While this provides some insight, of more interest is how sleep affects students who are at the lower and upper parts of the spectrum of how well they perform academically. In order to do this, a technique called quantile regressions was used which allowed me to look at the effects of an extra hour of sleep for adolescents who were at the 10th and 90th percentiles for the amount sleep received.

The CDS allowed for quality control variables to account for other factors that could potentially affect academic achievement. For the variables of interest I used the sleep variables described earlier and for the measure of academic achievement, the PSID issued standardized tests which measured reading and math ability. Having standardized test scores was a significant advantage in this research. The data, for the most part, seemed reasonable with the exception of some of the time diary data. The amount of sleep reported in the time diary data provoked the question of how it was measured. For example, I subtracted 25 minutes of sleep from the time diary sleep data since it is suggested that it takes about 25 minutes for an individual to fall asleep. Minor adjustments made the data seem more reasonable. The following table shows the difference between the two sets of sleep data where “hrsleepout” is the self reported sleep variable and “avgsleepout” is the time diary sleep data with hours on the left axis and the bars representing the number of observations reporting a specific amount of sleep.

Without getting into the details of the methodology, some conclusions were reached based on the data. The results of this research showed that sleep has different effects on reading and math. More sleep affected reading scores more than math scores when the students were performing poorly but sleep was more beneficial for those students who were performing well. The results did show that as a student performs better, the causal effect of sleep on improving academic scores weakens. In my study, I included a variable to account for the fact that getting too much sleep could adversely affect test scores and the results show that sleep is beneficial up to a certain level then it can cause a decrease in test performance. The results were mixed when the time diary data was used and part of that can be attributed to the unexplained differences in the data that was collected with the time-diary data. Although these results were not clear, it helped me understand the importance of understanding the data that is being used especially when trying to reach meaningful conclusions about causation.

The conclusions can be influential for both parents and policy makers. This research suggests that allowing kids to get an optimal amount of sleep would help them perform their best academically. This could affects how parents raise their children and the amount of sleep they encourage their kids to receive. For education policy it could have an effect on the start time for schools which, if move later, could allow students to receive more sleep and help them perform better.

This research experience was invaluable to me and has helped me develop my way of thinking and interpreting information in a way where data can be organized, analyzed, and interpreted. This has helped me in my current career as an actuary where I am constantly evaluating data and applying similar research techniques in order to reach conclusions for clients.