Can better data resolve deeply-felt issues? My entry today might be viewed as a little experiment on this question. But first, some background.
Last summer, I published The Racial Paradox of the Corporate Law Firm, a piece that explored the fate of minorities at big, elite law firms. I pointed out that while minorities are very well-represented among the starting associates at these firms, blacks and (to a somewhat lesser extent) Hispanics have far higher attrition rates than whites, leaving few black and Hispanic senior associates around when partnership decisions are made. I laid out various possible theories to explain these high attrition rates, including two different ways discrimination might affect the outcomes. And then I laid out a variety of new and fairly detailed data on the characteristics and experiences of new associates. Though I found much of the evidence consistent with more than one theory, I concluded that it was very likely that the aggressive use of racial preferences by law firms contributed to the problem. There was no doubt that firms were granting enormous preferences – in one large dataset of new associates at big firms, the median black GPA was at the 18th percentile while the median white was at the 75th percentile. Wouldn't a big disparity in grades produce a disparity in work quality? It was hard to explain many of the patterns otherwise:
--Black and Hispanic associates had experiences that closely matched those of white men on the “external indicia” of their jobs -- things like salary, hours, and representation on firm committees. But they tended to get fewer work assignments, were given less responsibility on those assignments, and were far less satisfied with the training and mentoring they received, compared with white male associates. Importantly, the problems a group had in these areas closely matched the size of their credential gap with white male associates. Blacks reported the most serious problems, and received the largest hiring preferences, then Hispanics, and then Asians. White women, who appeared to have exactly the same credentials, on average, as white men, reported virtually no differences from white men in mentoring, work assignments, or levels of responsibility -- and indeed reported very high levels of satisfaction.
--Smaller law firms appeared to be much less likely to use large racial preferences in hiring associates. I could not find any statistically significant difference in the grades of black and white associates, for example, though probably there's some disparity. Correspondingly, blacks at small firms generally are far more satisfied on issues like mentoring and work responsibility than their counterparts at big firms, and by most indicators, there's not a significant black-white gap in small-firm experiences.
Certainly, if big-firm racial preferences in hiring produce differences in performance that correlate with race, they will also produce invidious negative stereotypes in the minds of big-firm partners that could further hold back strong minority associates. But, I argue, if firms use radically different GPA thresholds for different racial groups in hiring, we ought to expect both differences in performance and negative stereotyping.
From the time I first presented this paper to a symposium at the University of North Carolina, it has evoked two types of responses. The first reaction -- always very vocally expressed -- is that my entire argument unjustifiably assumes that grades have something meaningful to do with the on-the-job performance of lawyers. The second reaction -- always expressed quietly, usually one-on-one and often from lawyers who have sat on the hiring or promotion committees of big firms -- is that my article gets it exactly right, but that any firm which publicly said so would expose itself to a lawsuit.
Public discussion of the article tends, therefore, to be a bit skewed. In November, after the New York Times ran a careful, thoughtful piece about The Racial Paradox, it ran a fluffy follow-up piece in the Week in Review, in which education reporter Jonathan Glater asked various law firm partners how important grades were for success at their firm. A typical response was 'Gee, I never would have made partner if they'd looked at my grades.'
This is disingenuous in two ways. I certainly don't contend that firms look at law school transcripts when they decide which associates to promote to partner. My point is that, other things equal, associates with high GPAs will tend to perform better, and thus will tend to build stronger reputations at the firm and succeed more often. And if this isn't true, then why would big firms set such high GPA standards in their entry-level hiring, and why would they pay such high salaries to lure the highest-GPA students?
The Racial Paradox did present some data showing that higher-GPA students tended to be more successful at big firms (e.g., Table 13), but I realized after the article was published that my presentation of the data was somewhat too abstract to really drive the point home. With further analysis, I've developed a clearer way to show the underlying patterns. The data below comes from the University of Michigan Law School's famous alumni surveys. UMLS has for many years sent out questionnaires to its alumni five and fifteen years after graduation. It links this data to actual transcript information, so that we know how each respondent performed in law school. By looking at this data, only for whites, we can accurately see the effect of grades on outcomes without the usual confounding effects of race and law school.
White Michigan Grads at Private Firms with 100+ lawyers (per 1000 graduates)
|Grade Decile||Joining big firms after graduation||At big firms, 5 years out||At big firms 15 years out (gen'ly partners)||Partnership rate (per 1000 grads)|
|Number of cases||1731||1537||755|
For this analysis, I grouped the white alumni from UMLS into ten equal groups based on their actual law school GPA. The table shows the rate at which these graduates (at rates per 1000 grads) worked at firms with 100 or more lawyers. Just after graduation, for example, white UMLS grads with grades in the 3rd decile had a 12.6% chance of working at a big firm (126 per 1000), while grads in the 10th (top) decile had a 22.1% chance -- a reflection, no doubt, of the easier time high-GPA grads would have getting jobs in big firms. The next two columns show how many grads were still at big firms 5 and 15 years after graduation. (I haven't tracked individual career histories for this analysis, so it is not identical to an attrition study, but it is very similar to one.) There is some erosion in numbers by the 5th year, and much more erosion between the 5th and 15th year (by which time partnership decisions have been made). And the rates of erosion are very highly correlated with GPA. Thus, for every 126 grads in the 3rd decile who are big firms in Year 1, there are 32 by year 15 (one-fourth as many). In the top grade decile, the Year 15 numbers are still about three-quarters as large as the Year 1 numbers. Indeed, in this population, there is nearly as much sorting by GPA after the initial hiring process as there is during the initial hiring process.
In other words, this data shows clearly that GPA matters a lot to one's success and longevity in the world of big firms. And, since this data is just for whites, we know that discrimination is not playing a role in these outcomes (unless, as seems unlikely, firms do pay attention to transcripts after hiring).
Moreover, if we think about this data carefully, we'll see that low GPAs will probably have even more harmful effects when they arise from racial preferences. In the table above, I can't distinguish between different sorts of big firms. But it is likely that the high-GPA whites are at tougher, more competitive firms (on average) than are the low-GPA whites. Their success rates are swimming upstream against more competitive environments and lower partnership rates. Naturally, low-GPA whites have both a better chance of being hired, and of succeeding, at a firm which is big but not super-elite. If we could control for firm quality, the GPA disparities in outcomes in the chart would probably be much greater.
For racial minorities, this caveat does not apply. The most elite firms are the ones most aggressive in using racial preferences, so huge grade disparities divide races within the same firm. So just how, exactly, can one argue that "grades don't matter" in the face of this data?