Project

Back to overview

Can Artificial Intelligence Improve Economic Policy Decisions? Optimal Assignment Rules for Labour Market Programmes

English title Can Artificial Intelligence Improve Economic Policy Decisions? Optimal Assignment Rules for Labour Market Programmes
Applicant Strittmatter Anthony
Number 190422
Funding scheme Spark
Research institution Forschungsstelle für empirische Wirtschaftsforschung Hochschule St. Gallen
Institution of higher education University of St.Gallen - SG
Main discipline Economics
Start/End 01.02.2020 - 31.01.2021
Approved amount 99'845.00
Show all

Keywords (6)

Public Policy; Active Labour Market Policies; Microeconometrics; Machine Learning; Programme Evaluation; Causal Inference

Lay Summary (German)

Lead
Dieses Projekt untersucht wie man künstliche Intelligenz verwenden kann um die Verteilung von öffentlichen (und privaten) Mitteln effizienter zu gestalten.
Lay summary

Regionalen Arbeitsvermittlungszentren (RAVs) stehen limitierte finanzielle Mittel zur Vefügung um mit arbeitsmarktlichen Massnahmen (z.B. Bewerbungstraining) die Wiedereingliederungsfähigkeit von Arbeitslosen in den ersten Arbeitsmarkt zu verbessern. Das Ziel der arbeitsmarktlichen Massnahmen is überlicherweise die Dauer der Arbeitslosigkeit zu verkürzen und die Stabilität der Beschäftigung zu verbessern. In der Arbeitsmarktliteratur werden heterogene Effekte für arbeitsmarktlichen Massnahmen dokumentiert, d.h. nicht alle Arbeitslose profitieren im gleichen Umfang von arbeitsmarktlichen Massnahmen. Ziel dieses Projektes ist es künstliche Intelligenz zu verwenden um eine effiziente Verteilung von Arbeitslosen zu den arbeitsmarktlichen Massnahmen zu finden, so dass die limitierten finanziellen Mittel der RAVs optimal verwendet werden.

 

Direct link to Lay Summary Last update: 25.01.2020

Responsible applicant and co-applicants

Abstract

Public employment services allocate limited funds to unemployed persons with the objective of reducing their unemployment duration and improving employment stability in the future, e.g., by assigning unemployed persons to job search assistance or training programmes (which are called active labour market policies, ALMPs). A large body of literature provides evidence that ALMPs have heterogeneous effects by characteristics of the unemployed and regional economic conditions. Optimal (statistical) assignment rules have the potential to improve the effectiveness of ALMPs. Recently, machine learning algorithms have been proposed to estimate optimal assignment rules. They have the potential to incorporate more covariates and outperform the out-of-sample prediction power of conventional estimation approaches. The proposed research project simulates the implementation of optimal assignment rules for ALMPs using high-quality administrative data that include 5 million unemployed persons. I consider optimal assignment rules for multiple ALMPs and several welfare functions. I test algorithmic restrictions that have the purpose to prevent discrimination. Since the effectiveness of ALMPs could vary between boom and recession periods, I consider dynamic optimal assignment rules that can adapt to economic conditions.
-