· After this week’s post by Owen Ozier on whether AI can program Stata, here is discussion of a new working paper that shows that AI can help job seekers improve their resumes and increase their chance of getting a job. “Candidates who tapped into AI to fix spelling and grammar errors on their resumes had an 8% increase in the probability of getting hired according to an MIT working paper that fielded the experiment with nearly a half-million job seekers.” The AI algorithm helps fix spelling, punctuation, and suggests better ways of phrasing sentences – allowing workers to better signal their abilities. Employers seem no less satisfied with the workers hired afterwards, suggesting it is not helping bad matches trick employers into hiring them.
· More on AI – on the Equitable growth blog, Kathryn Zickuhr discusses and links to a lot of research on how automated and algorithmic management is affecting job quality for workers in the U.S. “U.S. companies use these new technologies to “offload” management tasks and roles to their lower-level employees, gig workers, and even customers….These middle management tasks include constant decisions about how to optimize earnings, purchase and manage equipment and other costs, and even crowdsourcing ways to improve performance with other drivers in online forums. These companies also use technology to shift management functions directly to customers, whose feedback or ratings of workers may directly impact their performance evaluation with no manager or mediating actor.”
· On VoxDev, Michael Peters and Fabrizio Zilibotti argue that developing countries like India have so many small firms, not because small firms are constrained from growing, but because there is not enough creative destruction with larger firms not replacing the small firms quickly enough.
· Also on VoxDev, Jessica Leight and Eric Mvukiyehe examine whether public works have lasting impacts in Tunisia “we utilise a randomised controlled trial in Tunisia to evaluate the short- and long-term effects of the Community Works and Local Participation (CWLP) pilot, a public works programme that provided short-term paid employment to the long-term unemployed for three months… the intervention had significant and large short-term effects on both primary economic outcomes and secondary psychosocial outcomes… By the five-year follow-up, however, these effects have substantially attenuated toward zero”
· In PNAS, Athey et al. conduct a meta-analysis of 819 experiments conducted by 174 public health organizations on the cost-effectiveness of social media advertising (on Facebook and Instagram for influencing attitudes, beliefs, and take-up of COVID-19 vaccinations. “Although each experiment individually has too few subjects to reliably detect small effects, by pooling the data, we obtain more precise estimates of the overall average effect….We find that these campaigns are, on average, effective at influencing self-reported beliefs—shifting opinions close to 1% at baseline with a cost per influenced person of about $3.41. Combining this result with an estimate of the relationship between survey outcomes and vaccination rates derived from observational data yields an estimated cost per additional vaccination of about $5.68.”
· Rasheed Griffith on what policies a Caribbean think-tank should be promoting for the Caribbean (h/t Marginal Revolution) – his priorities for the region include full dollarization, public sector and talent internationalization, having a regional rather than country-by-country set of embassies and foreign policy, embracing tourism and having industrial policy designed around it, and transitioning to energy independence.
· Conference call for papers: the Impact Evaluation Network of LACEA will hold its annual meeting at Vanderbilt on April 20-21, with submission of papers due next week (Feb 10)
Join the Conversation