Financial Preparedness and Retirement Planning

Abstract

New Zealand adults with diagnosed disabilities — physical, cognitive, sensory, or psychiatric — face systematically lower financial-preparedness scores and weaker retirement-readiness trajectories than peers without disabilities, and the gap survives standard NZ-specific controls for income, education, household composition, and KiwiSaver participation. The household-finance literature has begun to attribute the penalty to a joint operation of three channels: labour-market friction (employment penalty and lower income), non-cognitive capital deficits (lower financial confidence and self-efficacy), and welfare-policy buffering (KiwiSaver participation gaps and accessibility barriers to financial advice).

The behavioural-economics construct most directly relevant — and the one most consistently measured in the NZ disability-finance surveys — is **financial confidence**: the subjective belief in one's capacity to plan and execute long-horizon financial decisions. Financial confidence is conceptually adjacent to self-efficacy and to financial capability, but is operationalised at the household-decision level rather than the general-disposition level.

The thesis pursues three explicit objectives: to assess saving behaviours of individuals with disabilities; to analyse debt-management strategies and their impact on financial security; and to deliver practical recommendations for policymakers and financial institutions. The empirical setting is NZ disability survey microdata linked to Stats NZ population and labour-market indicators. The analytic samples are 1,813 individuals in the primary specification, with subsamples of 894 and 800 for specific disability-type and life-stage conditions. The classical identification stack applies OLS regression, ordinal probit estimation with maximum-likelihood parameter recovery, Oaxaca-Blinder decomposition of the disability-vs-non-disability gap, Oster endogeneity bounds, logit estimation for binary saving and debt-management indicators, and panel fixed-effects.

We propose extending the existing analysis with two ML layers: a Causal Forest that would recover the conditional confidence × disability-type interaction (replacing the linear-interaction coefficient with a CATE distribution), and a Random Survival Forest for retirement-readiness spell duration without the proportional-hazards constraint. The Master of Finance thesis is the NZ baseline for the disability-and-finance research line and serves as the companion to the 'Planning Against the Odds' working paper and the international HILDA-based learning-disability flagship.

Data & Methods