SOAC 2017
What if Australia’s Housing Affordability Stress Measure doesn’t actually measure Housing Affordability Stress?
Lyrian Daniel, Emma Baker, Laurence Lester
The University of Adelaide
Abstract:The long-run worsening of Australia’s housing affordability problems are especially concentrated in our cities. Underlying any policy solutions proposed by governments, is a pressing need to understand who has unaffordable housing, and how they are affected. We have previously proposed that the most widely used affordability measure underlying policy and research is only able to identify those households ‘at risk’ of housing affordability stress (HAS), rather than households who are actually experiencing HAS. By failing to identify and enumerate who is affected by housing affordability problems Australian Governments are at high risk of mis-directing any policy response.
In this paper we propose a new measure of Experienced HAS, for individual households. This measure incorporates an understanding of material deprivation, alongside the simple ratio of housing costs to income (HAS). Utilising longitudinal data from the Household Income and Labour Dynamics in Australia (HILDA) survey we test the usefulness of this measure; map relative change across an important health measure over time and provide revised estimates of the number and character of Australians affected by housing affordability problems.
Keywords:Housing; affordability; material deprivation; estimates
Introduction
Housing affordability problems have been understood to exist in Australia for decades, however since the global financial crisis, and particularly in recent years, we have seen a worsening of these issues with limited government intervention. The accuracy with which we are able to identify those in our community experiencing affordability problems not only contributes to the effectiveness of assistance but also structures our ability to rigorously examine the outcomes of such problems.
The mostly widely used measure of housing affordability problems is the 30/40 ratio rule, which classifies housing affordability stress (HAS) as occurring when 30 per cent or more of a household’s equivalised income is allocated to housing costs and where that household’s income is in the lowest two quintiles of the income distribution (i.e. lower 40 per cent).We suggest that this measure is likely to misclassify households actually experiencing affordability problems and would be more appropriately conceptualised as representative of the risk of households experiencing housing affordability stress.
This paper contributes to the complex methodological problem of accurately identifying households experiencing housing stress in Australia by providing revised estimates of the prevalence of these problems using a novel measure of Experienced HAS. A growing pool of literature casts HAS as “poverty by another name” (SERC, 2015), and our previous work has sought to develop this, conceptually linking the traditional 30/40 ratio measure of housing affordability stress with relative material deprivation (Daniel et al, 2017).The addition of a measure of relative material deprivation to our conceptual understanding of housing affordability problems seeks to quantify the impact of high housing expenditure on households’ ability to access a socially determined standard of living.
There has been a steady evolution of Australian work that has explored the prevalence of housing affordability stress and contributed to our understanding of its definition and measurement. Some research work in this area uses data collected by the Australia Bureau of Statistics (ABS).The ABS produce estimates for the proportion of Australian households in ‘rental stress’ (in addition to reporting average housing costs across all three main tenures). The ABS 30/40 ratio measure is modified to only include householdswith equivalised disposable income (using the modified OECD equivalence scale) between the 3rd and 40th percentiles (previously this was between the 10th and 40th percentiles) with the justification citing poor data fidelity at the 1st and 2nd income percentiles. In the determination of rental stress, housing costs equate to rent payments (ABS, 2015, cat no. 6553.0).Using the modified 30/40 ratio measure,in 2013-2014, 42.5 per cent (1,353,200 households) of low income private rental households were classified as in rental stress, this increased in both proportional (35.4 per cent) and absolute (1,143,400) terms since the earliest reported figures in 2007-2008 (ABS, 2015, cat no. 4130.0).
Earlier work (Yates & Milligan, 2007) utilising the ABS Survey of Income and Housing, estimated that in 2002-3 there were 862,000 lower-income household defined as being in HAS using the 30/40 ratio rule. Lower income private renters had greater likelihood of being in housing stress than lower income purchasers. Importantly, the authors note that not all households that are classified as in HAS experience problems, such as housing trade-offs, compromises across non-housing essential expenditure, lack of financial reserves, amongst others (Yates & Milligan, 2007, p. 21). To explore this, the paper applies a series of additional objective and subjective checks to define households into different degrees of housing stress: ‘coping but at risk’ (approx. 25 per cent or 200,000 households), ‘housing affordability problems’ (approx. 50 per cent, or 450,000 households) and ‘severe housing affordability problems’ (approx. 25 per cent or 200,000 households) (Yates & Milligan, 2007, p. 26). These additional objective and subjective checks include indicators of financial stress, deprivation and social inclusion, and provides a valuable precedent for enhancing the 30/40 ratio measure with additional deprivation related variables.
More recent work by Rowley and colleagues (Rowley & Ong, 2012; Rowley et al., 2015) measured HAS using data from the Household Income and Labour Dynamics in Australian (HILDA) survey (waves 1-10). They found that in 2010, 15per cent of private rental households were in housing stress compared with just 8 per cent of owner purchasers – both proportions remained relatively stable of the previous 10 years: private rental households in housing stress up slightly from 13 per cent in 2001 and owner purchasers up from 7 per cent. In addition, they found that younger households consistently remained more likely to be in housing stress over the 10-year period. This work also reflected on how HAS could be understood, suggesting that HAS shouldbe thought of more correctly as an indicator of household risk (Rowley & Ong, 2015, p. 48). Previously suggested by Hulse et al. (2010), Rowley & Ong (2015) note that a classification of housing stress does not always equate to a position of financial stress (Rowley & Ong, 2015, p. 48).In response to this and in a similar vein to the work by Yates and Milligan (2007), the authorspropose several methods to refine the 30/40 ratio measure. They include: distinguishing between medium and high levels of housing stress by recognising cash flow problems (representing medium-level stress) and financial deprivation (representing high-level stress); taking the duration of time spent in housing stress into account; identifying whether housing stress was entered into by choice (e.g. accessing improved housing or location quality) or constraint (e.g. change of financial position through life events such as divorce); and disregarding those with high household wealth (Rowley & Ong, 2015, p. 71-77). While undoubtedly these additional measures enhance the reliability of the 30/40 ratio measure, they also underline the complexity of the definition and measurement of housing affordability problems.
Finally, while the ‘residual income’ method, the main alternative approach to measuring housing affordability problems, presents an effective way of capturinga household’s ability to meeting other non-housing essential expenditure (Stone, 2006), there remain several, non-trivial,practical limitations in its application.The residual method subtracts housing costs from household disposable income and benchmarks the remaining amount against accepted,subjective, poverty indicators (i.e. budget standards) to establish if households can be categorised as being in housing stress (Baker et al., 2013). The two key limitations in this approach centre on the reliance of budgets standards and are: 1) budget standards are based on a set of ‘family-types’ and may not necessarily be appropriate for application to an entire dataset (e.g. HILDA); and 2)until very recently (i.e. Saunders & Bedford, 2017), Australian budget standards had not been revised in two decades.Conceptually, a measure of relative material deprivation offers a useful counterpart to the residual income method by 1) providing a contemporary reflection of socially perceived living essentials or ‘prevailing community standards’, 2) not being limited to particular family or household types, and 3) relatively straightforward formation of standards and collection of data.
The research in this paper seeks to address the complex issues of the measurement of housing stress by proposing a new, more reflective measure of ‘Experienced HAS’ and providingrevised estimations of the proportion of Australian households affected. In doing so, we address the following three research questions:
RQ 1.Using a novel measure of the experience of housing affordability stress, Experienced HAS, how many households in Australia have unaffordable housing?
RQ 2. What are the characteristics of those in Experienced HAS?
RQ 3. How does the mental health of those in Experienced HAS change overtime compared to people without Experienced HAS?
The experience of housing affordability problems is understood to have outcomes across wellbeing, quality-of-life and health. In particular, a significant body of work links the experience of HAS to declining mental health (e.g. Bentley et al., 2012; Reeves et al., 2016). By examining the relative mental health of affected and unaffected cohorts, and controlling for confounding factors, we are able to demonstrate that financial hardship due to high housing costsaffectsmental health, and that the affected cohort is adequately identified by our proposed measure of Experienced HAS.
Methods
Data
This study uses data from the HILDA survey (Summerfield et al, 2016). HILDA is a large panel survey of Australian households and individuals, based upon a nation-wide probability sample and focused on housing, income, employment, health and wellbeing. Conducted annually since 2001, information is collected from household members aged 15 years and over using face-to-face interviews and self-completion questionnaires, sampling methods are detailed in Wilkins (2016). Data from waves 14 and 15 were included in the following analyses. The analytical sample was confined to people over the age of 15 years old and identified as ‘heads-of-household’ following the reasoning that these household members are likely to be the most exposed to the impact of HAS and/or material deprivation. Based upon the HILDA classification, household heads are defined in this analysis as the individuals with the highest personal income. If two people have identical incomes, then the older of the two is classified as the household head. In the interpretation and discussion of results one household head is taken to represent one household.
Approach
To address the research questions, three stages of analysis were untaken: an estimation of the prevalence of Experienced HAS in wave 14 (i.e. 2014), a description of key socio-demographic characteristics of those in Experienced HAS at wave 14; and a comparison of the changein mental health from wave 14 to wave 15 of those in Experienced HAS with those unaffected,by select socio-demographic characteristics.
Definition of key measures
Experienced HAS: Experienced HAS occurs when household heads are classified as in both HAS and as deprived in wave 14. The use of these two variables in conjunction is based on previous work (Daniel et al., 2017) that established relative material deprivation as an essential component in understanding housing affordability issues and the effects of these problems. This measure will be referred to as ‘EHAS’ from this section onwards.
Housing affordability: HAS is defined using the 30/40 ratio measure: unaffordable housing occurs when 30 per cent or more of the household’s equivalised income is allocated to housing costs and where that household’s income is in the lowest two quintiles of the income distribution (i.e. lower 40 per cent).
Material deprivation: In Wave 14 of the HILDA survey, a Material Deprivation module was introduced to the Household Questionnaire. The Material Deprivation Module consists of three tiers of inquiry: the first asks respondents to indicate whether 26 items are ‘essential’ (see Wilkins (2016) for full explanation and item list). Those items receiving over 50 per cent positive agreement are deemed ‘essentials of life’ and by the nature of their derivation represent a socially normalised living standard. The second tier of questions ask respondents whether they have these items or not. If the respondents do not have the items, the third tier of questions asks whether it is because they cannot be afforded. People who are without these items because they cannot be afforded (as opposed to choice) are considered deprived. In the following analysis, a dichotomous variable for material deprivation is used, defined using a ‘cut-off’ of two or more items. That is, a person is considered deprived if they lack two or more ‘essentials of life’. Previous work has used similar dichotomous definitions of material deprivation (e.g. Wilkins, 2016; Stephens & Leishman, 2017).
Mental health: Mental health is assessed using the Mental Component Summary (MCS) score of the Short Form 36 measure (SF-36). Self-completed, the SF-36 measure of health status (Coons et al, 2012) has been validated for use in the Australian population (Butterworth & Crosier, 2004) and to detect within-person change over time (Hemingway et al, 1997). A higher score on this 0-100 scale reflects better mental health and wellbeing.In the analysis, the mean is the average change in the MCS from wave 14 to wave 15 (not the change in mean).
Results & discussion
Estimates
Overall, we estimate that almost 3.5 per cent of our sample of household heads were classified as having EHAS in 2014. This translates to a weighted population estimate of 332,000 Australian households classified as having EHAS (based on an estimated resident population of 21,489,000, ABS 2015, cat. no. 3101.0)
The population affected by EHAS varied across Australian states. Table 1 shows that household heads in Tasmania and Queensland were more likely than household heads in other jurisdictions to be classified as having affordability problems as defined by our measure. Household heads in South Australia and Western Australia were less likely. Interestingly, the state distribution of prevalence of EHAS does not appear to reflect the spatial patterning of prevalence of affordability problems when measured using simple ratio approaches.
Table 1. EHAS by household head, number and proportion
State/Territory / EHAS / Total populationNSW / 109 / 2,860
% / 3.81 / 100
VIC / 85 / 2,424
% / 3.51 / 100
QLD / 82 / 2,041
% / 4.02 / 100
SA / 23 / 890
% / 2.58 / 100
WA / 19 / 867
% / 2.19 / 100
TAS / 14 / 313
% / 4.47 / 100
NT / 0* / 69*
% / 0* / 100*
ACT / 3* / 198*
% / 1.52* / 100*
Total / 335 / 9,662
% / 3.47 / 100
*note: low reliability for NT & ACT results due to low/no observations
Descriptive
Not only are there state based variations in the prevalence of EHAS, but the characteristics of those who are affected are distinct. Table 2 summarises the major population differences between the total sample, and the cohort classified as having EHAS. Female household heads are shown to be much more vulnerable to having EHAS than male household heads. The patterning of age is also interesting, it shows that younger household heads are over-represented in the EHAS cohort, and correspondingly, older (65+) headed households are under-represented. The results describing tenure are particularly stark. While only around 30 per cent of all households are private renters, they represent more than three quarters of the total population with EHAS. Public renters are also over-represented. In contrast, while two thirds of the total population are home owners/mortgage holders this tenure type represents just 12 per cent of the population with EHAS.
EHAS does not appear to be a specifically city based problem. Table 2 shows that residents of cities are slightly under-represented in the EHAS population, and households in inner regional areas are correspondingly slightly over represented. Unsurprisingly, people who receive some form of government assistance or support are much more likely than those who don’t to have EHAS. Relatedly, household heads who are unemployed, not in the labour force, or have a limiting illness are also over represented.
Finally, even though the measure used is equivalised to account for household size and composition, household status appears to substantially influence the likelihood of having EHAS. Household heads who were never married or partnered are over-represented among those with EHAS, and household heads who are married or partnered are correspondingly under-represented.
Comparative
The final part of the analysis examines lagged (1 year) changes in mental health for household heads who are exposed and unexposed to EHAS. These results are summarised in Figure 1 below and show notable differences in mean mental health across a number of key population characteristics. Overall, we see a generalised negative change in mental health for household heads who are exposed to EHAS, compared to those who are not exposed. Male and Female household heads for example who are not exposed to EHAS both have a similar small increase in mean mental health across between annual survey waves. In comparison, males and female household heads who were exposed to EHAS both had a similar, corresponding decrease in mean mental health in the period. Across each of the population characteristics examined there appears to have been a pattern of decreased mental health among those who are exposed to EHAS. One interesting exception is homeowners/purchasers experiencing EHAS. This group had a noticeable increase in mean mental health over the period, which may reflect the acceptance of high housing costs, deprivation trade-offs, and resultant ontological security for homeownership or purchase.
Table 2. EHAS by household head, proportion by select characteristics
Variable / EHAS (%) / Total population (%)Gender (% female) / 56.7 / 42.0
Age
16-24 / 20.9 / 8.5
25-44 / 36.4 / 34.8
45-64 / 34.0 / 34.0
65+ / 8.7 / 22.7
Tenure
Owner/mortgage / 11.9 / 65.7
Renter - public / 11.9 / 4.2
Renter - private / 76.1 / 30.2
Location
Major cities / 58.2 / 61.6
Inner Regional / 29.6 / 24.9
Outer regional / 11.6 / 11.7
Remote & Very Remote / 0.6 / 1.8
Government support
Not receiving government support / 22.7 / 67.6
Receiving government support / 77.3 / 32.4
Children
No children in household / 83.0 / 86.4
Children in household / 17.0 / 13.6
Labour market status
Employed / 35.8 / 67.4
Unemployed/NLF / 64.2 / 32.6
Limiting illness
No limiting illness / 62.4 / 79.7
Has limiting illness / 37.6 / 20.3
Marital status
Married/Defacto / 24.9 / 57.7
Separate/Divorce/Widow / 32.2 / 21.7
Never Marry / 42.9 / 20.5
Figure 1. Comparison of mean change in mental health (2014-2015), by EHAS and no EHAS, by selected population characteristics