Demographic influences on investment: A causal discovery study in Japan

This study examines potential causal relations among demographic factors, risk preferences, and investment behavior in Japan using data from the 2022 Preference Parameter Study. While previous research has identified correlations among these variables, few studies have examined the underlying causal mechanisms. To address this gap, we apply the Fast Causal Inference algorithm, together with a modified kernel-based conditional independence test designed to accommodate mixed data types. The resulting partial ancestral graphs suggest stable upstream relations from age and gender to risk-related measures. For investment behavior, stock ownership shows recurring ancestral relations involving gender and investment-trust ownership. Investment-trust ownership appears statistically upstream of stock ownership in PAGs consistent with the observed conditional independencies, although latent confounding remains possible. In contrast, no robust set of observed direct parents was identified for investment-trust ownership, suggesting that unobserved behavioral or institutional factors may be important. We assess reliability using bootstrap stability analysis and additional robustness checks. These checks support the stability of the main investment-related findings while motivating caution for weak or ambiguously oriented edges. The study contributes to a deeper understanding of investment behavior and offer practical implications for financial institutions seeking to promote inclusive investment participation.