RDP 2003-08: A Tale of Two Surveys: Household Debt and Financial Constraints in Australia 6. Analysis and Implications
July 2003
- Download the Paper 169KB
While the results so far are interesting in their own right, they also provide us with a framework with which to address some other related issues. For instance, we are now in a position to reconcile the household evidence on debt to income ratios with the aggregate evidence presented in the introduction. Also, by examining the characteristics of the households taking on debt we can draw out some implications for financial fragility.
6.1 Components of the Rise in the Aggregate Debt to Income Ratio
Between September 1993 and December 2001, the period covered by the three surveys, the aggregate debt to income ratio rose by nearly 51 percentage points and this may reflect three separate effects:
- Households that hold debt hold higher levels of debt; and/or
- The proportion of households that hold debt has increased; or
- More debt is held by higher-income households, all other things being equal.^{[16]}
The data in Table 4 suggest that all of these factors have been at work. Most importantly, a large part of the rise in the aggregate debt to income ratio appears to be explained by households now taking on higher levels of debt, on average. Secondly, we see that around 72 per cent of all households had at least one loan in 1993/94 and this increased to 76 per cent in 1998/99. And while we cannot compare these figures to the total number of loans in 2001 as we have imperfect data on credit cards and personal loans, we can use the number of housing loans as a proxy, especially as housing loans are likely to dominate the aggregate debt stock. Thus, while the data suggest that the proportion of people who hold standard mortgages has not increased significantly over the period, there has apparently been a rise in the proportion of households holding other types of housing-related loans, such as home equity loans, secured against their principal place of residence.
1993/94 | 1998/99 | 2001 | |
---|---|---|---|
Proportion of HHs with at least one outstanding loan | 72% |
76% |
N/A |
Proportion of HHs with at least one home loan | 28% | 30% | 32% |
Proportion of HHs with at least one standard mortgage^{(a)} | 26% | 29% | 29% |
Average level of housing debt (1999 $) (HH with debt) | N/A | $76,500 | $87,200 |
Average disposable income (1999 $) (HH with debt) | $45,800 | $48,000 | $51,600 |
Average housing debt to income ratio (HH with debt) | N/A | 177% | 212% |
Average housing debt to income ratio | N/A | 53% | 67% |
Average mortgage repayments to income ratio (HH with debt) | 23% | 24% | 24% |
Average mortgage repayments to income ratio | 6.1% | 6.4% | 7.3% |
Notes: (a) Each of the surveys asks slightly different questions so it is very difficult to be definitive about trends in housing-related lending. An alternative measure can be obtained from the Census. The Census classifies the person's house either as ‘fully owned’, ‘being purchased’, ‘being rented’, or ‘other’. If the respondent says it is ‘being purchased’ they are classified as having a mortgage. The Census data suggests that 28.5 per cent of households had a mortgage in 1991, 27.2 per cent in 1996 and 28.6 per cent in 2001. Alternatively, both the HES and HILDA specifically ask how many loans the household has and the purposes of those loans. In particular, the HES asks whether the housing loans are ‘to buy/build the principal dwelling’, ‘to buy or build other property’ (generally holiday homes and short-lived investment properties), and ‘loans for alterations and additions to the principal dwelling and other property’. The HILDA Survey asks for loans from financial institutions or family and friends taken out to help pay for the principal dwelling and other home loans secured against the property (e.g., home equity loans). Our broader measure of housing debt includes all these types of loans in the HES and HILDA while the standard measure shown here is calculated on a comparable basis to the Census and broadly matches those numbers. |
We can also examine the effect of distributional factors using household survey data. In general, we know that households at the lower end of the income distribution account for less than proportionate amounts of debt while high-income households account for more than their proportionate share of debt.^{[17]} Calculating the aggregate debt to income ratio based on the survey data, we find that the aggregate debt to income ratio grew by 12.3 per cent between 1998/99 and 2001. Doing the same for the average debt to income ratio, we find that the average debt to income ratio grew by 15.3 per cent between 1998/99 and 2001. As such, the aggregate measure (which gives greater weight to higher-income households) grew by less than the average measure (which gives equal weight to all households). This implies that there is now a more equal distribution of debt through the income distribution with lower-income households holding proportionately more debt in 2001 than in 1998/99. This distributional effect would serve to hold the aggregate debt to income ratio down, all other things being equal, compared with the average debt to income ratio reported in our tables.
So, while there has been a relatively small rise in the proportion of households with housing debt, in percentage terms it appears that the growth in the average size of loans has been the main contributor to the rise in the aggregate debt to income ratio. Slightly offsetting this, there has been a redistribution of debt to lower-income households and this has served to hold down growth in the aggregate debt to income ratio, relative to what it would have been had there been no change in the distribution.
6.2 Financial Fragility
To gain a better understanding of whether the increase in debt to income ratios is associated with greater financial fragility we look at the split between constrained and unconstrained households. In particular, we look at who is holding the higher levels of debt.
We are able to divide the sample into cash-constrained and unconstrained households on the basis of their actual responses to the questions about financial fragility in 1998/99 and in 2001. However, as the questions are unavailable for 1993/94 we need to adopt a different procedure to examine the characteristics of constrained and unconstrained households in 1993/94. One way of doing this would be to generate the predicted probability of being constrained for each household, based on the 1998/99 model, and then choosing some arbitrary cut-off to divide them into the two classes (e.g., less than 50 per cent predicted probability of being constrained means they are unconstrained, greater than 50 per cent predicted probability means they are constrained). However, this can be misleading when the distribution of cash flow problems is skewed towards one end (i.e., in the model many households are predicted to be a 30–40 per cent chance of being constrained). Instead, we ‘weight’ each household by their predicted probability of being constrained. If we then sum across these weighted estimates, we get an estimate for the average constrained household.^{[18]}
From Table 5 we can see that our measure of the proportion of households that hold debt has been rising for both constrained and unconstrained households. This is likely to reflect the fact that access to debt, for example, through access to home equity loans, has improved for most households following the deregulation of the Australian financial sector. And while both groups have been taking on higher debt levels, unconstrained households appear to have taken on proportionately more debt than constrained households, on average. Conversely, growth in average disposable income has been more pronounced amongst unconstrained households over the period.
1993/94 | 1998/99 | 2001 | ||
---|---|---|---|---|
Proportion of HHs with at least one home loan | Constrained Unconstrained |
25% 29% |
30% 30% |
29% 34% |
Proportion of HHs with at least one standard mortgage |
Constrained Unconstrained |
24% 27% |
30% 29% |
28% 31% |
Average level of housing debt (1999 $) (HH with debt) |
Constrained Unconstrained |
N/A N/A |
$70,600 $78,200 |
$82,100 $91,100 |
Average disposable income (1999 $) (HH with debt) |
Constrained Unconstrained |
$37,200 $48,000 |
$38,800 $50,600 |
$40,100 $56,000 |
Average debt to income ratio (HH with debt) |
Constrained Unconstrained |
N/A N/A |
203% 170% |
249% 197% |
Average debt to income ratio | Constrained Unconstrained |
N/A N/A |
61% 51% |
69% 67% |
Average mortgage repayments to income ratio (HH with debt) |
Constrained Unconstrained |
27% 22% |
25% 24% |
27% 23% |
Average mortgage repayments to income ratio |
Constrained Unconstrained |
6.5% 6.0% |
6.6% 6.3% |
7.3% 7.4% |
Overall, comparing the changes in the debt to income ratios of constrained and unconstrained households between 1998/99 and 2001, the debt to income ratio of constrained households has risen from 61 per cent to 69 per cent while the debt to income ratio of unconstrained households has risen from 51 per cent to 67 per cent – a considerably larger increase. Thus, it appears that the rise in the aggregate debt to income ratio is mainly the result of unconstrained households voluntarily taking on higher levels of debt.
And while we noted earlier that the housing debt to income ratio has increased for both groups, the mortgage debt-service burden is often seen as a better measure of households' capacity to service debt. The average mortgage debt burden has actually increased more for unconstrained households (from 6.0 per cent to 7.4 per cent) than for constrained households (from 6.5 per cent to 7.3 per cent). Among those constrained households that actually have a mortgage, the average mortgage debt burden has remained roughly constant at around 25–27 per cent between 1993/94 and 2001. We also know that the interest burden for these households fell from around 16 per cent in 1993/94 to 14 per cent of disposable income in 1998/99. This is partly explained by falling lending rates over the period. So while there is tentative evidence that constrained households are taking on more interest-sensitive debt, their capacity to service this debt (as measured by the debt burden) has not worsened, on average. As such, their financial fragility is likely to have remained relatively unchanged.
There is stronger evidence that unconstrained households are taking on more debt and in increasing amounts. This is reflected in the fact that the mortgage repayments (as a share of disposable income) of unconstrained households appear to have increased more than for constrained households. Moreover, there are now more unconstrained households than there were in the past. This suggests that relatively strong economic growth and falling nominal interest rates more than offset the increase in household indebtedness over this period. So, in summary, most of the rise in household debt at the aggregate level appears to be explained not only by more unconstrained households taking on debt, but also by unconstrained households taking on increasing amounts of debt.
Footnotes
The aggregate debt to income ratio is the weighted average of individual debt to income ratios where the weights are the shares of each household in total income. So if higher-income households (with greater weights) incur proportionately more debt, the aggregate debt to income ratio can rise even at constant household debt to income ratios. For example, suppose there are only two households, A and B. Household A earns $50,000 and Household B earns $100,000. Initially, suppose Household A has $10,000 in debt (and Household B has no debt). The average household debt to income ratio will be 10 per cent while the aggregate debt to income ratio will be 6.7 per cent. Suppose instead that Household B takes out a loan at the same debt to income ratio (i.e., a $20,000 loan). If Household A were now to repay its loan, the average household debt to income ratio would still be 10 per cent but the aggregate debt to income ratio would have risen to 13 per cent. [16]
Very high income households, those representing around the top 15 per cent of the aggregate income stock (which corresponds, approximately, to the top 5 per cent of households in the income distribution), account for less than their proportionate share of debt – as might be expected of very wealthy households. [17]
For example, suppose there were only two households (A and B) in the economy. Household A earns $100 a week while Household B earns $200 a week. Suppose Household A has a 70 per cent chance of being constrained while Household B has only a 30 per cent chance of being constrained. Then applying these weights to both households ($100 × 70/(30+70) for Household A and $200 × 30/(30+70) for Household B) and summing across the weighted estimates ($70 + $60) gives us the income of the average constrained household ($130). This can be compared with the estimate of $100 (Household A) that we would get if we used a cutoff between 0.3 and 0.7 probability. [18]