General Seminar - Anna Zhu

Place-Based Heterogeneity in Welfare-to-Work Reform Success: Insights from Combining Regression Discontinuity and Machine Learning Models (joint work with Barbara Broadway and Tim Robinson)

Welfare-to-work reforms are legislated nationally but experienced locally. We ask how the strength of a local labour market shaped single mothers' capacity to respond to one of Australia's largest reforms aimed at lifting single parents' employment. We focus on single mothers because they make up the large majority of recipients targeted by the reform. Using population administrative records, we estimate a separate regression discontinuity for each of Australia's 87 local labour markets. On average, the reform reduced single mothers' first-year earnings by about $820, or 15 per cent of pre-reform mean earnings. That national average conceals substantial variation across markets: earnings fell in 60 of the 87, and the size of the loss differed sharply among them. What distinguishes areas with smaller earnings losses is not general labour market strength but the availability of a particular kind of work: areas with more short part-time work saw smaller losses. Short part-time hours matter because this is the segment of the labour market single mothers in Australia rely on most heavily, allowing paid work to be fitted around care responsibilities. This area-level relationship holds after conditioning on pre-determined area characteristics selected through a double-selection LASSO procedure.