Excluding missing cases and stating the number of cases excluded—in this case the 296 PhD recipients out of our sample of 2,510 whose current employment we were unable to find—is standard practice in analyses of this kind. The MLA study assigned an individual to an employment category only when we could confirm that the person whose employment we found was in fact the same person whose dissertation abstract record was in our sample. Without a positive identification, a case was categorized as missing.

I wish we knew more about the experiences of ABDs and those who withdraw from their doctoral programs before reaching ABD status. Unfortunately, the precondition for studying ABDs systematically—a compilation of all students who have entered doctoral study and reached ABD status—does not exist. The MLA study required a well-defined universe, a total population from which a random sample could be extracted. The source for the universe had to be comprehensive in its coverage, and to be useful for discovering public information about current employment it had to provide names. It also had to be based in a marker—receipt of the PhD—that distinguished the total population in a definitive way. Dissertation Abstracts records from the MLA International Bibliography may be the sole source available to us that fulfilled these requirements reasonably well. The alternative would have been to crowd-source the data, with all the problems crowd-sourced data involve.

We would have included ABDs in the study had there been a plausible way to do so. But you work with the data that you have, not the data that you wish you had. Educating Scholars (Princeton UP 2010), Ronald Ehrenberg’s, Harriet Zuckerman’s, Jeffrey Groen’s, and Sharon Brucker’s report on the Andrew W. Mellon Foundation’s Graduate Education Initiative, is the only study I know where the project’s tracking practices developed systematic information of a type that permitted identification of the ABD population and discovery of outcomes for them as well as for degree recipients.

I, too, am curious about the quality of our sample of 2,590 PhDs and have looked for ways to test it. How well or poorly, for example, would the distribution of tenure-track and non-tenure-track faculty members from the 2004 National Study of Postsecondary Faculty (NSOPF:04) compare with the employment of the parallel subset of PhDs in our sample—those employed as faculty members in United States postsecondary institutions?

Below are the results of a query on the NSOPF:04 dataset for the percentage distribution by tenure status of PhD-holding faculty members whose principal teaching fields are English or other modern languages; I further limited the query to PhDs who received their terminal degrees after 1970 so that the results would reflect employment as of fall 2003 of PhDs who graduated into the constricted academic job market that has characterized the entire post-1970 period. Following the results from NSOPF:04 is the parallel percentage distribution of PhDs in the MLA sample who are employed as faculty members in United States postsecondary institutions.

NSOPF:04
Faculty, tenured: 52.4
Faculty, tenure-track: 22.0
Faculty, non-tenure-track: 22.7
Faculty, no tenure system: 2.9

MLA Sample
Faculty, tenured: 52.5
Faculty, tenure-track: 22.0
Faculty, non-tenure-track: 25.4
Faculty, no tenure system: —

Using the NSOPF as a basis for comparison has its limitations; the comparison applies only to the 1,184 cases of PhDs that the MLA’s research identified as faculty members in United States postsecondary institutions. But if the quality of either the MLA sampling procedure or procedures for assigning employment were as terrible as the comments assert, would the results align so closely with those of NSOPF:04?