Lovisa Rambjer: Algorithms, Incentives, and Information: Essays on Labor Markets and Public Service Provision

Date
Location
Hörsal 2, Ekonomikum, Kyrkogårdsgatan 10, Uppsala
Type
Thesis defence
Thesis author
Lovisa Rambjer
External reviewer
Nikolaj Harmon
Supervisors
Peter Fredriksson, Georg Graetz
Research subject
Economics
Publication
https://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-584163

Abstract

Essay 1:  Can caseworkers improve on algorithmic assignment of unemployed jobseekers to active labor market programs (ALMPs) at the Public Employment Service (PES)?  Because ALMP assignment involves a trade-off between targeting jobseekers at highest unemployment risk and those with the largest expected treatment effects, caseworkers may add value beyond prediction alone. Using as-if random assignment of jobseekers to caseworkers, I reconstruct the algorithmic counterfactual and estimate the effects of deviations from the algorithm’s recommendations. I find that caseworker discretion weakens alignment with risk-based targeting and lowers overall employment.Essay 2 (with Arne Uhlendorff and Johan Vikström): Unemployed job seekers must comply with job-search requirements to be eligible for unemployment insurance benefits. We exploit variation in strictness across caseworkers who conduct monthly reviews of these requirements to examine the effects of stricter enforcement on job finding, search intensity, and job quality. This variation arises from central office caseworkers who are randomly assigned to review job seekers’ activity reports and whose only interaction with job seekers occurs through this process. We find that stricter enforcement increases the likelihood of reported violations and benefit sanctions, leading to persistently higher job-search intensity and shorter unemployment durations. We find no evidence of negative effects on job quality. These positive employment effects are more persistent among job seekers from low-growth industries and those with weak labor market attachment.Essay 3 (with Umrbek Allakulov, Serena Cocciolo, Binayak Das, Md. Ahasan Habib, and Anna Tompsett): Can transparency interventions improve WASH service provision? We use a randomized experiment to evaluate the impacts of a transparency intervention, a deliberative multi-stakeholder workshop initiated with a community scorecard exercise, in schools in rural Bangladesh. To measure impacts, we combine survey data, direct observations, and administrative data. The intervention leads to moderate but consistent improvements in knowledge of WASH standards and practices, and institutions for WASH service management, but does not improve school WASH service provision or change WASH facility use patterns. Drawing on rich descriptive data, we suggest several reasons why the intervention we evaluate did not improve WASH service outcomes and propose ways to improve the design of future interventions.