Elyas Saif

PhD candidate at Hanken School of Economics and Helsinki GSE.

My primary research interest is in applying AI and machine learning to economic questions. Currently, my research focuses on household finance, utilizing AI and econometric methods to study household financial distress, financial shocks, and interhousehold monetary transfers. My university teaching experience involves teaching predictive machine learning in economics, the use of data science tools to automate recurring empirical tasks and conduct empirical analysis, as well as the use of Agentic AI to streamline data analysis workflows.

I am on the academic Job market 2026/2027.

How Predictable is Household Financial Distress?

I examine the ex-ante predictability of consumer financial distress (CFD) using nationally representative Finnish household-level data for the period 2008–2017. Unlike studies that focus on consumer default, this predictive study uses machine learning methods (ML) to analyse the predictability of CFD from the perspective of the consumer rather than the lenders. This study provides insights into the overall accuracy of ML in predicting CFD, revealing significant variability in predictability between households with and without prior experience of distress. Specifically, I find that predictability is nearly perfect for households with prior distress but significantly lower for those without. This demonstrates that predictions of CFD can be used to accurately target support measures to households most likely to experience persistent consumer financial distress. The findings of this study hence contribute to the literature on consumer financial protection — which focuses on causally identifying measures that mitigate suboptimal borrowing decisions and consumer financial distress — by considering a predictive dimension that complements this causal focus. With more accurate predictions of CFD, consumer protection measures, such as advice on how to cope with CFD, can be targeted to households that are at greatest risk.

Read full paper (PDF)