This is the 12th in this year’s series of posts by PhD students on the job market.
Search frictions are prevalent in low- and middle-income countries. Many burgeoning cities in these countries have few on-line search platforms, which mostly serve for recruitment at large firms for high-skill jobs. Researchers find that the difficulty of learning about job opportunities leads to low job search effort and inaccurate beliefs for job seekers. While such reasoning may also apply to firms, we know very little about how search frictions affect firm hiring due to the lack firm data in low- and middle-income countries, and the common belief among scholars and policymakers that private firms may possess more accurate information of the labor market.
In my job market paper, we collect hiring data from 799 private formal firms with an active job vacancy (which is most of them) in Addis Ababa, Ethiopia, between May 2022 and April 2023. We sample from a wide range of sectors, including manufacturing, construction, and hospitality. Our data reveals that firms also face high search frictions and have particularly hard time finding college-educated applicants, where despite reporting a vacancy, the majority of these firms do not have any college applicants in the last 5 months. We thus design a randomized controlled trial which reduces search frictions for a random subset of firms by providing subsidized access to a new search platform – employment agency (EA).
Employment Agency: A new platform for college-educated job seekers
Responding to the high unemployment rate among college graduates and high demand for college graduates, many previously informal job brokers started to tailor their recruitment service to college graduates. These EAs are strategically located at the city center where many unemployed college graduates go and seek for jobs. Currently, most of these EAs collaborate with a few international hotels and match college graduates to white-collar jobs such as receptionists and accountants, and not yet known to most firms in other sectors. We thus collaborate with 11 such new EAs in Addis Ababa, randomly match them with 326 vacancies collected from our sample, and request each EA to recommend 1-2 additional applicants to the vacancy. Only a small fraction (8%) of our firms have ever used an EA.
Search frictions for firms in Addis Ababa
The demand for college graduates in our sample is high. 35% firms in our sample require applicants to have at least a college-level diploma, compared to only an estimated of 12% young people aged 18-23 attending any college in 2018 (Ethiopian Socioeconomic Survey). Yet, one surprising fact from our data is the low number of job applicants. For each vacancy, the survey team collects all the applicants from multiple hiring channels, including notice boards, online channels such as Telegram, and personal recommendations. Over the course of five months, the median number of total applicants is only one. When it comes to college graduates, 65% of vacancies receive zero applicant; for those requesting a college graduate at baseline, this number is still as high as 42%. These facts suggest a high level of search frictions for firms to match with a college-educated applicant.
EAs thus increase firms’ access to college-educated applicant by a large extent. 80% applicants provided by EAs have a college diploma or degree, significantly higher than that of applicants from other hiring channels (42%). If our intervention only reduces the level of search frictions, we should expect to see treated firms hire more workers provided by the EA and hire more college graduates as a result.
Not a simple story of reduced search frictions
We first examine whether our intervention helps firms fill the vacancy faster. One month after baseline (midline), treated firms are 10.1 percentage points more likely to hire at least one applicant, a 14% increase compared to that of control firms, suggesting faster hiring decision and lower cost of waiting. However, this effect is not driven by hiring from the EA. Although mechanically, treated firms interview and hire more agency applicants, the majority of new hires come from non-agency hiring channels. The findings cannot be explained by a simple decrease in search frictions.
Less optimistic perceptions of college graduates
We thus hypothesize the following: Treated firms update on the productivity of college graduates. To test this hypothesis, we directly measure firms’ perceptions in two ways: i) Five months after the baseline (endline), we ask each firm whether an average college graduate is more productive than a non-college educated worker; ii) for each applicant, we elicit firm’s perception of the productivity if the applicant were hired on the job. We find significant, negative updates on both measures of college graduates’ productivity, suggesting treated firms may observe additional signals from reading the resumes from the college graduates or interacting with college-educated applicants during interviews, but what they learn makes them less optimistic of college graduates’ productivity in general.
Switch in hiring preferences
Such a negative update manifests itself in the hiring of college graduates. Among firms requesting a college graduate at baseline, treated firms are less likely to interview or hire a college graduate; instead, they are more likely to interview or hire a non-college educated worker. Among firms that do not request a college graduate at baseline, we observe no such treatment effects. Results are consistent with the fact that firms become less optimistic of college graduates after the intervention. We also find that such treatment effects are stronger among firms with below-median percentage of employees with a college diploma or degree (college share), a proxy for past exposure to college graduates.
What signals do firms observe that may lead to such negative updates? We provide descriptive evidence by comparing characteristics of college applicants to non-college applicants applying to the same firm. College applicants do not have more relevant experience or have more outside offers, both indicators of high productivity. In addition, from the elicited firms’ perceptions of applicants, firms do not perceive of college applicants more productive than non-college applicants. The descriptive evidence suggests that if anything, firms do not observe signals of college premium from college applicants, which may explain the negative updates on their perceived productivity.
Last, we do not find quality trade-off by hiring a non-college educated worker. For the workers hired for the posted vacancy, we collect endline information on turnover, performance, absenteeism, and overtime work. We find no treatment effects on any measures. Instead, complier firms may take advantage of the existing salary ladder for non-college educated worker and pay a lower salary. Our findings thus suggest a net increase on profit for the complier firms.
Our findings suggest that firms also face high uncertainty of worker’s productivity. Many firms may use college education as a heuristic to find skilled workers, but search frictions in the labor market prevent firms from developing accurate beliefs of college graduates’ productivity, potentially rendering such a heuristic not optimal. Our collaboration with new employment agencies suggests that policymakers may leverage the existing labor market intermediaries to facilitate information exchange. In particular, properly incentivizing employment agencies with good credibility may effectively increase the interaction between different labor market participants, leading to lower information frictions and more efficient matching.
On the other hand, the fact that college graduates are considered less productive echoes with the anecdotes of decreasing quality of college education in Ethiopia. Despite the rapid growth in tertiary education, many newly-built private colleges are poorly funded and may not provide useful skill training to students. Future research can further explore the general-equilibrium implications of such a trend in tertiary education in low- and middle-income countries.
David Qihang Wu is a PhD candidate at UC Berkeley.