my deep burning hatred of letters of recommendation
Econjeff mentions my long standing critique of letters of recommendation (LoRs). Here, I describe my personal experience with them and then I restate the massive empirical research showing that LoRs are worthless.
Personal experience: In graduate school, I had enormous difficulty extracting three letters from faculty. For example, during my first year, when I was unfunded, I asked an instructor, who was very well known in sociology, for a letter. He flat out refused and told me that he didn’t think I’d succeed in this profession. In the middle of graduate school, I applied for an external fellowship and was informed by the institution that my third letter was missing. Repeatedly, I was told, “I will do it.” Never happened. Even on the job market, I had to go with only two letters. A third professor (different than the first two cases) simply refused to do it. Luckily, a sympathetic professor in another program wrote my third letter so I could be employed. Then, oddly, that recalcitrant member submitted a letter after I had gotten my job.
At that point, I had assumed that I was some sort of defective graduate student. Maybe I was just making people upset so they refused to write letters. When I was on the job, I realized that lots and lots of faculty never submit letters. During job searches at Indiana, I saw lots of files with missing letters, perhaps a third were missing at least one letter. Some were missing all letters. It was clear to me that l was not alone. Lots of faculty simply failed to complete their task of evaluating students due to incompetence, malice, or cowardice.
Research: As I grew older, I slowly realized that there are researchers in psychology, education and management dedicated to studying employment practices. Surely, if we demanded all these letters and we tolerated all these poor LoR practices, then surely there must be research showing the system works.
Wrong. With a few exceptions, LoRs are poor instruments for measuring future performance. Details are here, but here’s the summary: As early as 1962, researchers realized LoRs don’t predict performance. Then, in 1993, Aamondt, Bryan and Whitcomb show that LoRs work – but only if they are written in specific ways. The more recent literature refines this – medical school letters don’t predict performance unless the writer mentions very specific things; letter writers aren’t even reliable – their evaluations are all over the place; and even in educational settings, letters seem to have a very small correlation with a *few* outcomes. Also, recent research suggests that LoRs seem to biased against women in that writers are less likely to use “standout language” for women.
The summary from one researcher in the field: “Put another way, if letters were a new psychological test they would not come close to meeting minimum professional criteria (i.e., Standards) for use in decision making (AERA, APA, & NCME, 1999).”
The bottom line is this: Letters are unreliable (they vary too much in their measurements). They draw attention to the wrong things (people judge the status of the letter writer). They rarely focus on the few items that do predict performance (like explicit comparison). They have low correlations with performance and they used codes that bias against women.