I’m working through some ideas about the operation of academic labor markets, that I hope to test if I ever get around to it. Below is a VERY rough sketch of the argument. I’d be happy to see what our readers think of this. Some of this may be half-baked, so read it in that spirit.
A theory of efficient academic labor markets would predict that the “best” scholars are at the “best” institutions. So, Harvard is better than the institutions below it because, when job candidates receive offers, the best ones (measured by future scholarly impact) choose Harvard because of its superior resources and reputation. The labor market, in this understanding, is a sorting mechanism, and an efficient market effectively ranks all candidates hierarchically and places them at the institution at which they belong. This model is efficient because the richer resources of the elite institutions are given to the best scholars, who are able to make the best use of them. One implication of this model is that affirmative action is inefficient, since it would allocate resources to those who would make less effective use of them than alternative candidates.
In the model above, quality is exogenous to academic institutions. An alternative model suggests that, to a significant degree, quality is endogenously produced. This model suggests that the sorting processes of the academy are characterized by substantially incomplete information. The information that hiring committees have available to them varies in predictive quality, from the dissertation (could be more a function of the good ideas of the candidate’s advisor than the candidate), to the job interview (could be more a measure of slickness of presentation than underlying intelligence or productivity) to recommendations (which may simply operate as a proxy for the candidate’s PhD granting institution, since their substance is often remarkably similar). Furthermore, academic hiring committees are committees–because of the need to put together majorities, committees may converge on non-offensiveness, not predicted productivity. So, while the job market for freshly minted PhDs is probably a less than perfect predictor of future academic productivity.
The initial sorting of PhDs to institutions would be self-correcting if mistakes at this first stage were corrected in the “secondary” market for job candidates. But—and this is the key to the argument—there are reasons to believe that this will not occur. First, initial allocations are sticky, because PhDs are less mobile in the secondary market than they were in the initial job market. Once they have settled in to an institution, and their spouses have made a life wherever they initially landed, there may be limits on reallocation.
Second, and most important, quality (as measured by scholarly productivity) is to a significant degree endogenously produced by initial allocation to institutions. Rick Hess has published an interesting article in PS that shows that it is almost impossible to “write your way out” of low-ranked institutions. The reasons for this are obvious to anyone who has spent time in a variety of different university settings. In lower ranked institutions, a great deal of scholarly time is taken up by grading, and opportunities to teach courses that closely track with research interests are limited. Resources for research are more limited the further down the pecking order one goes (funding and time off of teaching in particular). As one goes down the pecking order, the average quality of research assistance one can obtain from graduate and undergraduate students declines. Institutions further up the pecking order have a constant flow of visiting speakers that help stimulate ideas for research and expand scholarly networks. Funding sources are also more willing to support research at higher-ranked institutions. For all these reasons, PhDs at higher ranked institutions will find it considerably easier to produce a stream of high-quality work than their counterparts at lower-ranked institutions.
The consequence of this is that initial allocations of individuals to institutions will tend to be highly sticky—high potential PhDs who end up at the “wrong” institutions in the original sort will have a hard time producing the scholarship that shows that they “deserve” to be at the higher ranked institutions (that is, that shows that at the higher ranked institutions they would produce more quality scholarship than incumbents). At the same time, those who have the good luck to end up at the “right” institutions will produce significantly more quality work than they would if they had been sorted into the institutions that matched their inherent potential.
As a consequence of this, those at the higher-ranked institutions will, as a consequence, appear as if they really do have more intrinsic “merit” than those at lower-ranked institutions. And on average, given that the original sort is not random, they will. The point, however, is that the endogenous production of academic quality will make it look like the original sort was more efficient than it actually was.
While I don’t have strong preferences on affirmative action either way, I do think this analysis has some consequences for the merits of the policy. There are good reasons to believe that,where, say, law school admissions are concerned, affirmative action may have undesirable effects, since it may cause significant numbers of uncompetitive individuals in effected groups to enter law school, who then either flunk out or manage to pass but not get pass the bar exam (this is the argument Richard Sander has recently made, and defended—the articles and responses can be found here. I’m not sure if he’s right, but based on his evidence, it’s more than plausible. But if the argument above is correct, then affirmative action in the initial hiring market for PhDs (at least in the social sciences, which I know best) may not suffer from the perverse consequences that may effect professional school admissions. While the average potential of PhDs in effected groups may be lower, presence in higher-quality institutions will, in fact, tend to substantially increase their productivity. While effected scholars may not produce as much as their competitors at higher-quality institutions, and thus be denied tenure, the endogenously produced productivity from being at higher-quality institutions will put them in a position to get positions at higher-ranked institutions than they would have originally, if the sort was not influenced by identity characteristics. While this may have consequences for the efficiency of the academy overall, affirmative action in this area would have desirable distributive effects–depending, that is, on the appropriate targeting of the effected group. This suggests that the right argument is not whether affirmative action, as a whole, is desirable, but–normative questions aside–where will it tend to have the least undesirable side effects, and the most positive impact on those for whom it is intended.
UPDATE: Jane Galt found the argument above interesting enough to paste into her blog. Thanks! She suggests that my argument implies that the consequences for whites displaced by affirmative action are even more devastating than we had heretofore recognized. I’m pretty sure I don’t buy it. This works if we assume that affirmative action has a very large but intense effect on a small number of displaced candidates—so it’s a 1-1 transfer from X number of blacks to X number of whites. I don’t think that’s how it works. To just stick to race for a second (and AA for women is probably a much bigger deal, given the unfortunately low supply of racial minorities who are even in the pool), affirmative action works by increasing the probability that minorities will be competitive for a particular job. Conversely, it reduces the probability that all the whites will be competitive. [this is on the assumption that affirmative action doesn’t increase the supply of jobs, which might be true, but it might also be true that universities transfer resources from other activities in order to fund additional slots targeted to minorities] But this cost is spread throughout the whole population of whites, so the marginal effect on any particular white candidate is rather small–the cost, in other words, is a diffuse one born by the whole class, not concentrated only on one candidate. Of course, even a relatively marginal impact will seem very large to the guy who lost out by .001%, but the important point is–that guy is exceedingly unlikely to lose out by .001% at all the other top schools he applies to. So given that this game is played out over dozens of different selection processes at different schools, it is probably not right as an analytical point to suggest that there’s one guy (or a number of guys) out there who would have been at Harvard, but ended up at [place state university with two directions in the name here].