When Risk Cuts Both Ways

Abstract

Australian women face financial hardship at higher rates than men across two decades of HILDA panel waves, but within each gender the dispositional **financial risk-taking** orientation discriminates sharply between hardship-resilient and hardship-vulnerable workers. Among Australian women aged 18–60 with comparable income, education, and household composition, those reporting greater willingness to take financial risk exhibit lower hardship rates than female non-risk-takers, and the existing Oaxaca evidence suggests that female risk-takers also exit hardship spells at faster rates following negative life events. The reverse asymmetry holds among Australian men: male risk-takers face hardship rates closer to male non-risk-takers, with the recovery advantage concentrated in temporary rather than sustained post-event reductions.

The conceptual framework draws on **gender-asymmetric loss-aversion theories** in behavioural finance, which predict that women's stated risk-tolerance is more deeply integrated with deliberate-planning behaviour than men's. Under this account, female risk-takers are not merely 'less risk-averse' — they are a specific subgroup of women who have transcended the default female-loss-aversion pattern through deliberate financial-decision integration, and the resulting behavioural integration is what produces the sustained-recovery advantage after shocks.

The empirical setting is HILDA waves 2002–2022, restricted to working-age individuals (18–60). The analytic sample is 177,163 person-years across multiple model specifications; event-conditioning produces subsamples of 94,476, 82,687, 15,954, and 7,682 for the various life-event types. The existing analysis applies the classical identification stack: OLS regression with panel fixed-effects, difference-in-differences around life-event timing, and Oaxaca-Blinder decomposition of the risk-taker-vs-non-risk-taker hardship gap. The existing Oaxaca evidence attributes the largest share of the gap to income (dominant for both genders), with employment second for women and financial capability second for men.

We propose extending the existing analysis with two ML layers to recover the recovery-dynamics interpretation: a Causal Forest CATE estimator on the gender × risk-taking interaction across income, employment, and education strata, and a Double-ML estimator partialling out the high-dimensional demographic control vector. The proposed extensions, once implemented, would identify whether women's financial-education programmes that integrate risk-tolerance training with deliberate-planning behaviour yield larger hardship-recovery returns than gender-neutral programmes.

Data & Methods