Association for Public Policy Analysis & Management

In the last two decades, the subject that has received the most public, political, and academic attention in education economics is class size (Hoxby, 2000). Reducing class size to increase student achievement is a policy measure that has gained major consideration in the U.S., Europe, and Australia. Currently, many countries have enacted or are considering class size reduction with the aim of improving student achievement (Müller, 2013). Despite the remarkable amount of literature on class size, almost all studies focus on average outcomes. This focus might appear surprising, especially in light of the undeniable policy relevance of class size effects for children at different levels of achievement. Most studies attempt to deal with heterogeneous effects by using sub-sample analyses (Schanzenbach, 2006). However, evaluating the impact of smaller classes on groups that are only “more likely” to be low achieving (e.g., ethnic minorities or students eligible for free lunches) misses important dimensions of effect heterogeneity (Bitler et al., 2006).

In this paper, we use data from the Tennessee Student/Teacher Achievement Ratio experiment (also known as Project STAR) to examine the distributional effects of being assigned to a small class or a regular class with a teacher's aide, compared to a regular class. Project STAR involved the random assignment of over 11,000 K-3 pupils at roughly 80 public schools to a small class (i.e., fewer students than usual), a regular class, or a regular class with a teacher's aide. The experiment began with random assignment of teachers and pupils entering kindergarten in the mid-1980s, and their assignment to each treatment condition was intended to continue through third grade.

To estimate the effects of class size and teacher aide on the distribution of test scores, we employ the unconditional quantile regression approach recently developed by Firpo et al. (2009). This estimator provides a direct measure of how a marginal change in the level of one variable affects the distribution of achievement in the population, keeping the distribution of other characteristics equal. Our results show considerable heterogeneity in the treatment effects “small class” and “regular class with aide,” suggesting that average effects may hide crucial features about the rest of the distribution. We find that mid-achieving pupils profit the most from being assigned to a small class, whereas pupils at the bottom and top of the achievement distribution experience only minimal gains from being in a small class. We also reveal positive and significant effects of an aide in a regular class for low-achieving pupils. Interestingly, the effect of having a class aide is strongest for boys and disadvantaged children, to the point that such effect is as strong as that of small classes for the same type of pupils.

This paper shows that typical estimates of the mean gain from class size and teacher's aide provide an incomplete characterization of their real impact on achievement distribution, thus constituting a weak guide for public educational policy.