Crime Rates, Income and Housing Costs Emerge as Top Predictors of Personal Default in Australia

Australia is one of the most indebted countries on Earth, with household debt reaching roughly 185 percent of disposable income in 2021 and 120 percent of gross domestic product by 2023 — a level that exceeds even the United States after the 2008 financial crisis. Against that backdrop, a new study published in Heliyon has taken one of the most detailed looks yet at why individual Australians actually default on their financial obligations, and the results upend several conventional assumptions about credit risk. Analysing 4,749 borrower records spanning 2013 to 2023, researchers Sharmeen Ara Rakhi, Simone D. Scagnelli and Rashid Zaman found that the strongest single predictor of default was not age, gender or even income alone, but the rate of financial offences in the borrower’s local community.

The dataset came from Credit Fix Solutions, an Australian credit repair provider founded in 2013, and covered clients across multiple states and territories. After excluding 337 records containing late repayment information — which reflects missed minimum payments rather than actual defaults — and removing 127 incomplete records, the team retained 4,749 usable observations. Each record included the default amount, defined as outstanding principal plus accrued interest and late fees, alongside demographic details such as age, gender, ethnicity and postal code. From those postcodes, the researchers mapped clients to roughly 250 local government areas, allowing them to attach area-level socioeconomic data: employment rates, education levels, median weekly income and the ratio of financial crimes to population, sourced from the Australian Bureau of Statistics and state police websites.

On the macroeconomic side, the study linked each observation to state-level gross domestic product per capita, the consumer price inflation rate, the Reserve Bank of Australia’s cash rate and median house prices, together with local rent expenses and mortgage repayments. The dependent variable — the default amount — was heavily right-skewed, so the authors used its natural logarithm to compress outliers. To test whether their findings were artefacts of any single statistical technique, they deployed an unusually broad methodological arsenal: ordinary least squares regression as a baseline, quantile regression at the 25th, 50th and 75th percentiles, logistic regression on a binary default indicator, and a random forest machine learning model built from 500 decision trees with seven variables considered at each split.

The baseline OLS model explained about 66.9 percent of the variation in default amounts, and its results were striking. The offence rate carried a large, highly significant positive coefficient, meaning that people living in areas with more financial crime defaulted more heavily even though those crimes were independent of their own defaults. Income showed the opposite pattern: higher median weekly income strongly reduced default, a result consistent with classic findings that financially secure households can absorb shocks and resist the pull of risky borrowing. Employment, counterintuitively, was positively associated with default. The authors interpret this through a credit-access channel: employed people qualify for more credit, accumulate larger obligations, and — in line with behavioural economics — may fall prey to optimism bias, borrowing as though job security will last indefinitely. That mechanism became especially visible after 2022, when the Reserve Bank’s sharp rate increases raised the debt-service burden of employed mortgage holders.

Not everything behaved as theory predicts. Age and gender had no significant effect on default in any model. Ethnicity did matter: non-Caucasian clients showed higher default rates than Caucasian clients, a pattern the researchers attribute to systemic disparities in lending access and financial resources documented in earlier international work. Inflation raised default risk in the baseline model by squeezing disposable income, while the RBA cash rate carried a negative coefficient — higher rates were associated with fewer defaults, plausibly because tighter monetary policy discourages new borrowing and curbs risky credit expansion. House prices showed a positive but only marginally significant effect in the OLS framework, though they became strongly significant in the more flexible models.

That is where the machine learning results added genuine value. The random forest model explained 95.15 percent of the variance in default amounts — far more than the linear specification — and its variable importance rankings confirmed the hierarchy of risk factors. The offence rate ranked first by a wide margin, followed by income, education, house prices and employment. Age, gender and ethnicity ranked near the bottom. The gap between the linear and nonlinear models is itself informative: it suggests the true drivers of default interact in complex, higher-order ways that a single regression equation cannot capture, which is precisely why the authors argue for combining interpretable regression with flexible machine learning in credit risk assessment.

Perhaps the most policy-relevant finding came from the moderation analysis. The researchers tested whether rent expenditures and mortgage repayments change the strength of the relationship between other factors and default — and they do, substantially. Interaction terms between housing costs and employment, income and offence rates were all highly significant, meaning that the effect of socioeconomic conditions on default depends on how much of a household’s budget is consumed by housing. Higher rent or mortgage obligations leave fewer resources to meet other commitments and to absorb adverse shocks, so the same income level protects a renter in an expensive suburb far less than it protects someone with cheap housing. Housing costs, in other words, do not merely add to default risk directly; they amplify the damage done by every other vulnerability.

The state-level breakdown reinforced the robustness of the core findings while revealing regional texture. The offence rate significantly increased default in five of the six states examined, and income significantly reduced default in all of them, making these two variables the most consistent determinants nationwide. Employment raised default in New South Wales, Queensland and Western Australia but showed a weakly negative effect in Victoria, suggesting the employment–default link depends on local labour market conditions. Macroeconomic variables such as GDP, inflation and house prices were significant mainly in Queensland, implying their effects are largely captured at the national level. An industry-level analysis of default origins — spanning telecommunications, utilities, debt collection, banking and other sectors — told the same story: employment and offence rates increased default and income reduced it across every category.

The theoretical scaffolding behind the study draws on two frameworks rarely combined in credit research. Behavioural economics theory explains how cognitive biases such as present bias and loss aversion distort financial decisions — younger borrowers prioritising immediate consumption, confident borrowers overextending themselves. General strain theory, imported from criminology, holds that economic and social stressors push people toward deviant coping behaviour, including defaulting on obligations. Together they explain why community-level crime rates predict individual default so powerfully: crime signals economic instability and social strain that ripple through households’ finances, independent of any single person’s characteristics.

The practical implications are considerable. For lenders, the study argues that credit scoring models should incorporate socioeconomic context — local offence rates, income volatility, housing cost burdens — rather than relying on employment status alone, since a job can coexist with dangerous levels of debt. For policymakers, the findings point toward financial literacy programmes targeted at newly employed and minority borrowers, more inclusive lending rules to prevent discriminatory access, stronger social safety nets during downturns, and affordable housing initiatives to relieve the pressure that amplifies default risk across the board. The authors caution that their sample, drawn from a single credit repair provider, may not perfectly represent the broader population, and that unmeasured factors such as health status and family support could also matter. Still, in a country where mortgage defaults rose 18 percent in 2021 and more than 31 percent of individuals reported increased borrowing, knowing that neighbourhood crime, income and housing costs form the true fault lines of financial failure is a finding that lenders, regulators and households alike cannot afford to ignore.

Subject of Research: Socioeconomic and macroeconomic determinants of individual financial default in Australia

Article Title: Predicting individual default in Australia: The role of socioeconomic conditions, macroeconomic shocks, and housing costs

Article References: Rakhi, S. A., Scagnelli, S. D., & Zaman, R. (2026). Predicting individual default in Australia: The role of socioeconomic conditions, macroeconomic shocks, and housing costs. Heliyon, 12(15), Article e45575. https://doi.org/10.1016/j.heliyon.2026.e45575

Image Credits: AI Generated

DOI: 10.1016/j.heliyon.2026.e45575

Keywords: individual default, credit risk, Australia, household debt, machine learning, random forest, offence rate, housing costs, inflation, RBA cash rate, quantile regression, financial stability