Typical of social justice research which ignores economics and incentives

Ann Althouse highlights a research paper which I blogged about last week but which is scheduled for Monday, Counterintuitively, the representation of women and minorities and bias were uncorrelated.

The core issue is that the researchers found apparent disparate impact in terms of how professors respond to unsolicited expressions of program interest by unknown potential applicants. They vary the names of the applicants to reflect different racial groups and then look at the different response rates. See the above post for details.

Revisiting the issue based on the discussion in Althouse's comment section, I would add the following.

Typical of social justice research which ignores economics and incentives. The base line to which the researches should have been comparing is relative return on invested time on the part of professors. My suspicion is that the professors of all races and gender are investing their time heuristically where they perceive it will yield the most benefit to them and their institutions (better quality candidates, candidates more likely to join, candidates more likely to produce results, etc.).

For example: If a professor receives 100 requests a month from individuals with Chinese names, including 80 of whom are from overseas and without a green card, then the likely return on invested time is very low in terms of landing a high quality degree candidate. If the yield of African-American enrolled students to those expressing interest is 1 in 100 but that of white candidates is 1 in 30, then you will likely see a much higher invested effort in responding to white candidates over African-American candidates, not because of racial animus but because of Return On Invested Effort (ROIE).

The researchers started with no discernible focus on understanding the professor’s objectives. The researchers then made the predicate assumption that all candidates are equally worthwhile to invest time in regardless of race or other distinguishing attribute. If this unstated predicate assumption is true, then A) the researchers have produced an interesting finding. If it is not true then B) they are shedding light on the likely differentials between classes of candidates and even possibly some light on the source of the differences.

In all fields of endeavor there is disparate performance based on all sorts of attributes (education attainment, IQ, ethnicity/culture, region, quality of educational institutions attended, parental class, familial structure, religion, etc.), one would expect there to be disparate quality (and therefore disparate anticipated ROIE) based on any attribute the researchers chose, including race. Since all professors, by gender and race, show the same disparate response to candidates, that is evidence that indeed practitioners in the field experience a consistent variation in quality of candidates. Consequently, I would interpret the research outcome as evidence for B over A.

With no evidentiary basis, the researchers made a critical predicate assumption that is likely untrue (all candidates are equally qualified and equally worth investing time and effort in). Having made the assumption (likely without recognizing that they were making it), they are then interpreting their results under scenario A. Consequently their conclusion is that there is shameful discrimination occurring by race. However, if B is true, as is likely, then the interpretation is much more benign. Professors are discriminating based on experiential heuristics.

In a perfect world this is also regrettable. However, we do not exist in an environment of unlimited time and unlimited money. Assumptions are made at the drop of a hat (or email) with no prior knowledge other than a sense of experiential ROIE by vestigial attribute (such as implied race by name type). It is understandable from an efficiency point of view. If the universities wish to ensure that everyone receives an equal hearing regardless of race and on the basis of all individual circumstances, they can redesign the meeting solicitation process. Take the load off the professors, route everything centrally, apply a documented screening mechanism that has a verifiable empirical relationship to the anticipated outcome. Then watch tuition go up another 15%.

The researchers found what they were looking for based on their unstated, unexamined, and likely incorrect assumptions. The rest of us in the real world can shake our heads and watch the glacial steps researchers make towards engaging with reality where disparate impact is everywhere, most heuristics are relatively efficient, most of it is not malevolent even though it might be individually unfair; where time and money are limited and goals are high; and where there are real consequences when goals are not achieved.

If the researchers wish to correct their study, they should first look at whether there is an empirically disparate ROIE based on racial attributes (or any other attributes in which they might be interested) and then look at whether the professor’s response rates match that disparate empirical ROIE. If it does, then the professors are rationally responding to incentives (optimizing their time). If not, then there is a genuine conundrum worth investigating. Without doing that, all the researchers are doing is revealing their own prejudices and biases (and lack of research sophistication).

It seems to me there are three potential outcomes from further research.

1) Professors are using their time efficiently towards their own goals and those of the university but in a manner that is unfair to individual candidates.
2) The professors are using inaccurate heuristics that are resulting not only in individual unfairness but also reduced achievement of goals.

If the outcome is 2, then the university has three choices.

1) Accept the status quo as unfair but less costly than the alternative.
2) Universities can take on the screening process in order to give all individuals an equal opportunity for consideration but doing so will likely cost a lot more.
3) Universities can attempt to coach/train/educate professors into better heuristics.