My post yesterday was about the experiment about social media and the job market for economists. I only noticed later that the PNAS also posted a comment on our article, by Professor Peter Rousseau, the secretary of the American Economic Association, who has a long and intimate familiarity with that job market, which the AEA has played a giant role in organizing.
Improving the job market in economics (and beyond…) by Peter L. Rousseau PNAS May 4, 2026 https://doi.org/10.1073/pnas.2609971123
Here is the part of his comment directly connected to our paper:
" the authors make a welcome and useful contribution to the market design literature with a fascinating experiment designed to substitute for and even improve upon the informal information channels lost to the economics job market in the new postpandemic normal. Given that some job candidates are less active self-promoters than others and that, conversely, excessive self-promotion can in some cases be viewed as a negative by prospective recruiters, the authors’ proposed mechanism offers serious promise for leveling the playing field, even if just modestly, for economics job candidates in terms of their visibilities, and perhaps even for expanding the number of jobs actually filled over the course of a recruiting season.
"In the experiment, an AI-based algorithm, supplemented with some human checking and reassignments, matched selected economists on social media (i.e., the “influencers”) with willing job candidates based on the closeness of their research. About 43 percent of willing candidates were selected for this treatment. The key to the experiment lies in the matches themselves, which were assigned in a manner that did not take the relative prominence or institutional ranking of an influencer directly into account. All candidate participants were invited to post a tweet about their job market papers on a social media site created for this purpose, and the influencers were asked to post neutral quote-tweets about the members of the treatment group to which they had been assigned. If executed according to design, recruiters viewing the quote-tweets receive information about the closeness of a given candidate’s research interests to those of the influencer. This may function as a partial substitute for the painstaking process of deducing such information across the hundreds of application packets that recruiters receive with only a brief period for making initial decisions. Knowing that a candidate’s research is close to that of Professor “X” is a tangible signal that could make that candidate more likely to be interviewed or receive a campus flyout or job offer from an institution seeking an entry-level economist like Professor X. The experiment indicates that individuals in the treatment group did indeed receive more campus visits and job offers than candidates assigned to the control group, and that the effect on job offers was especially strong for women. It also finds, however, that these effects were more pronounced for candidates matched to influencers with relatively higher citation counts than for those matched to influencers with relatively more followers, as these two measures of prominence in the profession are not that highly correlated.
" The question of scalability then becomes paramount. Considering the experiment’s positive findings, it is natural to assume that, if universally available, all job candidates would choose to participate and receive the treatment. The process would otherwise go on as stated with perhaps additional influencers being selected by the organizers to serve the larger pool of candidates. Two observations seem reasonable at this point: first, in such a setup, better information about matches could lead to more open positions being filled, which would be a better aggregate outcome; and second, the treatment might in practice benefit candidates from outside the very top departments the most. This is because candidates from the highest ranked departments, who are often perceived by recruiters as having a higher probability of eventually becoming a star, will typically receive more interviews, campus visits, and offers, but in the end can still only accept one offer. With an enlarged set of viable matches, this means that some candidates who may have been otherwise overlooked will find jobs. Of course, the job market may take longer to clear under this mechanism as candidates will have more options to consider before departments go to second or third rounds of offers.
"Casual observations of the job market among economics departments and their chairs do suggest that a number of recruiters are unable to fill positions they have posted. The AEA does not currently collect information on just how many, but the very existence of the “AEA Job Market Scramble,” where recruiters and unmatched candidates can post their availabilities on an online message board each March, is indicative of the challenge (3). The design of a job signaling mechanism by the AEA and its implementation in December of each year (4), where job candidates can list two departments to which they would like to express interest in an interview, is another such intervention aimed at easing the congestion.
Another interesting result is that women appear to benefit most from the treatment, while this benefit does not extend to members of other groups traditionally underrepresented in economics. The authors point to existing evidence indicating that women on average tend to be less active promoters of their own research on social media than others and suggest that the additional visibility provided by the quote-tweets could be leading to more job offers. This potential channel, of course, could also be viable for any candidate with a tendency to self-promote less. To explain a special advantage for women, one could note the possibility of forces in the 2022–2023 job market where departments seeking to improve the gender balances of their faculties became aware of candidates through the mechanism who they may have otherwise overlooked. If this is the case, the next question to ask is why does the effect not carry over for members of other underrepresented groups? The answer, though no doubt a speculative one, may lie in the preexistence of other mechanisms and informal channels for promoting such candidates, rendering the marginal effects of the authors’ particular intervention not statistically significant.
Finally, while having the potential to increase the number of matches and raise their average quality, the effects of the authors’ intervention will be subject to some randomness based on the assignment of a given candidate’s influencer. For example, when any influencer posts a quote-tweet about a candidate who has been independently and objectively determined to have close research interests, that candidate’s post tends to receive more views and likes on X than those in the control group, and the extent of this visibility correlates with the size of the influencer’s following. Yet these effects do not seem to transfer downstream to job outcomes, where candidates receiving quote-tweets from highly cited influencers are the ones tending to see more offers. In a real sense, the adage “all publicity is good publicity,” often applied to economics research, may not be always true. The assignment of influencers to candidates, even if randomized, will matter for individual outcomes even though the aggregate effects of the intervention are positive. Given the potential individual benefits compared to nontreatment, however, job candidates would likely embrace the residual uncertainty and participate in the mechanism.
"The intervention designed by Qiu et al. may hold even greater promise outside of the economics discipline. In the natural sciences, for example, recruiting for scarce academic postdoctoral positions among new PhDs at a similar career stage, which are markets typically saturated with candidates, often moves directly to a very limited allocation of campus visits based in no small part on letters and other communications from mentors, some of whom could be less than ideally matched with their students or less well known than would-be assigned influencers. These cases are ones in which an enhanced visibility of candidates, when coupled with independent information about the closeness of their work to what senior researchers and their groups might be seeking, could lead to the greater advancement of science more generally.
:
"Competing interests P.L.R. has served since 2012 as Secretary-Treasurer of the American Economic Association, a 501(c)(3) non-profit deeply committed to improving the job market for new Ph.D. economists, and for which one of the companion article’s co-authors (A. E. Roth) served as President in 2017.
#######
Peter's comment and our paper appeared online, but won't appear in print until next week in the May 12, 2026 | vol. 123 | no. 19 issue of PNAS.
Yesterday's post:
Tuesday, May 5, 2026 Social media, job market outcomes, and ethics of field experiments, by Qiu, Chen, Cohn and Roth in PNAS
Posts
No comments:
Post a Comment