Abstract
The amount of merit-based scholarship support for graduate students in the United States has increased dramatically. Given this increased investment, does increasing the size of scholarships awarded to the most academically able admitted students substantially increase their probability of enrollment? We found no support for a positive answer to the question. Our study is unique in the literature in two important ways. First, the design randomly assigned admitted students to size of scholarship. Second, the size of scholarship varied from substantial ($10,000) to substantially larger ($18,500). The applicants in the study were the most academically able applicants to master's programs in a graduate school of education at an elite university. It may be that the difference between no scholarship support and some scholarship support (e.g., $10,000) might have a positive effect on yield. Natural variation data from the same site supported this conclusion. It may be that dramatically larger amounts of scholarship support would have a positive effect on yield as well (e.g., the full cost of tuition).
Keywords:
Notes
In interpreting the results of this investigation, it may be important to note that a substantial recession had gripped the U.S. economy starting in December 2007 and was still very much in effect when students applied and enrolled.
2009 applicants are those applying for admission during the 2008–2009 academic year who would normally matriculate in summer or fall of 2009, depending on the program to which they applied.
None of the interactions were statistically significant for the full interaction model: age at application, p =.42; male, p =.98; United States citizen, p =.38; Asian, p =.59; Black, p =.51; Other, p =.83; Local region, p =.61; Top college, p =.76; Go to graduate school, p =.19; Undergraduate Grade Point Average, p =.66; GRE verbal, p =.16; GRE quantitative, p =.52; GRE analytical writing, p =.65.
We also fit logistic regression models using nonexperimental data for all admitted students (3,204 observations) from the same graduate school of education for the period of 2007–2010 to examine the relation between amount of scholarship support and yield and academic performance measures. These nonexperimental data allowed an investigation of a full range of scholarship support from none to $18,500. It also allowed investigation of the full range of students in terms of academic promise, but that range is still restricted by the high standards for admission, even for those in Tier 4 and 5. Although the full set of control variables were in the model, it is still true that scholarship amount and academic promise of admitted applicant were positively correlated. Because yield is negatively related to applicant academic promise, that bias should push against finding an effect of scholarship amount on yield. To the contrary, the amount of scholarship was significantly positively related to yield and on-time completion but not significantly related to grade point average and final completion.