The ROC and role of the 30% rule

Abstract

In the research and policy community there is ongoing debate about the measurement of housing affordability. This debate is important because what we measure affects the perceived prevalence of housing affordability problems, who should be targeted for intervention, and ultimately it reflects how well government agencies have addressed the affordability problem. This paper examines individual self-assessments of housing affordability stress, and evaluates the degree to which the two most commonly applied ratio measures of housing stress actually reflect that experience. The two ratio measures examined are the 30%Rule (defined as households spending 30% or more of income on housing costs), and the 30/40Rule (households in the lowest 40 per cent of the income distribution who spending 30% or more of income on housing costs). We note in this paper that in addition to defining and calculating housing stress, an additional and under-acknowledged means to identify those in housing stress is to ask them: to borrow from Allport (1941), if you want to know if people are in housing stress ask them. We use survey data from low income individuals and ask if housing costs meant they could not pay for bills, meals or utilities. We classify this as self-identified housing stress. In this paper we compare the characteristics of populations who are classified by each housing stress approach (self-assessed, the 30%Rule and the 30/40Rule). We then undertake Receiver Operating Characteristic (ROC) curve analysis to evaluate the degree to which a ratio measure predicts self-identified housing stress and find that these ratio rules are not particularly accurate.