Weathering the Shock

Abstract

Women face systematically higher rates of financial hardship than men following negative life events (divorce, widowhood, severe illness, job loss) and the gap survives pre-event income, education, and household-composition controls. Australian women, in particular, enter and exit hardship spells at higher rates than men with comparable pre-event resources, suggesting that the gap is not merely a consequence of disparate exposure to events but a consequence of differential recovery from events that both genders experience.

This paper frames **financial capability** as shock armour: a practical behavioural resource that distinguishes deliberate from reactive financial behaviour and that, when present, reduces the depth or duration of hardship after a life event. The construct is distinct from financial literacy (information stock) and from liquid wealth (income-smoothing capacity); it captures the non-cognitive, skill-level resources that govern whether deliberate financial action survives high-cognitive-load conditions such as a sudden separation or illness.

The empirical setting is a national household panel linked to the ANZ Financial Wellbeing Survey financial-capability scale, across the post-2016 waves that fielded the capability measure, with event-specific subsamples for divorce and widowhood. The existing analysis applies an already-rich identification stack: OLS with panel fixed-effects, instrumental-variable estimation, Oaxaca-Blinder decomposition, Oster endogeneity bounds, Cox proportional-hazards modelling for hardship-spell duration, Imai-Keele mediation analysis for the income-mediator channel, and a latent-class analysis of capability profiles. The existing classical evidence already identifies financial capability as a measurable buffer against women's life-event-induced hardship.

We propose extending the existing analysis with two further ML layers: a Causal Forest CATE estimator that would recover the conditional capability effect by life-event type, gender, and pre-event hardship history, and a Random Survival Forest that would estimate hardship-spell duration without the proportional-hazards constraint. The proposed extensions would distinguish whether divorce and job-loss events, rather than widowhood, are the contexts in which the capability buffer concentrates, and supply event-type-specific intervention recommendations for women's financial-education programmes.

Data & Methods