Jacob Loosen

The present study investigates the predictive validity of letters of recommendation and potential racial bias in letters of recommendation. Despite the widespread use of letters of recommendation in admissions and hiring processes, there is limited research supporting their predictive validity. Based on the small body of existing literature, it was hypothesized that letters of recommendation exhibit a weak correlation with performance outcomes, and that differences across racial groups would arise in certain aspects of letters of recommendation. For predictive validity, a meta-analysis was conducted in which relevant studies were identified from databases, and their data was compiled and coded in a spreadsheet. All data was converted to provide effect sizes, then statistical analyses were conducted using a Microsoft Excel program. Across performance outcomes and forms of letters of recommendations, a small-to-moderate relationship was observed between letters and outcomes. The pool of data investigating racial differences across letters of recommendation was limited, so a mainly qualitative approach was used to analyze data. Small differences across racial groups were observed. The results of this meta-analysis indicate that letters of recommendation are poor predictors of subsequent performance, and they may contribute to racial bias. Thus, their use in hiring and admissions decisions should be given less weight.