RDP 2020-05: How Risky is Australian Household Debt? 1. Introduction

Household debt levels have increased considerably over the past 30 years, both in Australia and elsewhere. In Australia, and other countries with relatively high household indebtedness, this is consistently cited as a key risk to financial and macroeconomic stability. Concerns about the risks posed by household debt appear regularly in the press, and in reports from financial analysts and global institutions (such as the Bank for International Settlements and International Monetary Fund). These concerns have been heightened by the deep global contraction induced by COVID-19. As households across the world experience a significant drop in income, the extent of risk posed by the current levels of household debt will become more apparent.

These concerns are based on very clear evidence that ‘excessive’ levels of household debt give rise to serious financial stability concerns. For example, Mian and Sufi (2012) argue that almost 4 million of the 6 million jobs lost in the United States between 2007 and 2009 were because of an aggregate demand shock driven by weakening household balance sheets. Using a broader set of countries and a very long time series, Schularick and Taylor (2012) argue that accelerating growth in aggregate debt-to-GDP is the best predictor of future financial crises, especially when the starting level of indebtedness is high.^[1] Similarly, Jordà, Schularick and Taylor (2013) and Mian, Sufi and Verner (2017) conclude that recessions preceded by rapid credit growth tend to be deeper than others, even when the financial sector remains solvent. Moreover, these outcomes appear to be primarily driven by high levels of household, not business, debt (Jordà et al 2013).

Measures of aggregate household indebtedness, however, are imperfect predictors of future stress. The most commonly used measure of household indebtedness is the debt-to-income (DTI) ratio.^[2] At the onset of the 2008 global financial crisis, the DTI ratio had risen very rapidly and was at historically high levels in some countries that subsequently experienced significant financial stress (such as Ireland). But other countries that are now viewed as having had ‘too much’ household debt (e.g. Spain and the United States) did not stand out from others based on a comparison of DTI ratios, and some that had comparatively high levels of DTI ratios emerged from the crisis relatively unscathed (including Australia, the Netherlands, Norway and Switzerland).^[3] This simple observation suggests that while an accelerating, or high, household DTI ratio can be a useful signal of rising risks to financial stability, the household DTI ratio is not a perfect measure of the risk of household debt. In particular, it does not capture other factors that indicate the riskiness of household debt. These factors likely include the distribution of debt, the quality of lending, institutional structures and the resilience of the financial system. This paper seeks to look more closely at the economic and financial stability risks posed by current levels of household indebtedness in Australia.

We address this in three ways. First, we look at what factors can account for the rise in household debt over time and its relative level across countries (Section 2 of the paper). The aim of this investigation is both to shed light on developments in indebtedness and to consider whether debt levels in particular countries at particular times can be attributed to fundamentally sound considerations or not. The advantages of this approach are that it uses a very broad scope of information to place current Australian household indebtedness in an international and historical context, and that it provides results that are intuitive to understand. However, it is not always clear which factors should be considered as ‘risky’ and which should be considered as ‘safe’ explanators of indebtedness. Moreover, our empirical approach does not allow us to be confident about exogeneity, so we cannot be certain that our results are causal.

Our second way of approaching this question is to examine the risks that Australia's current pattern of household indebtedness poses to the banking system (Section 3.1). We do this by simulating a period of severe economic stress and modelling the effect this has on the capacity of indebted households to service their debt (using household-level data). Households that are found to be unable to service their debt without falling below what is commonly considered to be minimum living standards are assumed to default on their loans, thereby eroding bank capital. This analysis allows us to move beyond aggregate measures of indebtedness and so take account of how the distribution of debt across households influences systemic risk. However, it still requires assumptions to be made about the capacity (and willingness) of households to service their debt. (A caveat is that we take a partial equilibrium approach in this work by largely abstracting from stress that would likely be present in banks' business loan books.)

The third way we consider this question is by examining how much current patterns of debt could amplify the effect of a large economic shock on consumer spending (Section 3.2). Our approach builds on Price, Beckers and La Cava (2019), who find that Australian households curtail their spending in normal times when they have relatively high levels of debt. We extend their work by simulating the effect that debt levels may have during periods of extreme stress. We do this by applying a stress scenario to household-level data and assessing how plausible variation across households in marginal propensities to consume (MPC) out of income (MPCy) and wealth (MPCw) would affect the resulting fall in consumer spending. We calibrate these MPCs from the large international literature estimated from the global financial crisis period, given that the absence of historical data on extreme financial stress in Australia makes it difficult to accurately estimate MPCs during periods of stress for Australia. This approach enables us to broaden the existing discussion of how household debt might pose risks to the economy, while still considering the effect of its distribution. It also allows us to apply the burgeoning international literature to the Australian context and learn from periods of stress elsewhere. The results we generate are not predictions of the likely path of consumption during a period of stress, but rather a means to understanding how the level and distribution of household debt in Australia might influence the vulnerability of the economy.

The results of these three approaches suggest that the risk posed by Australian household debt is material, but the household DTI ratio is a poor measure of the extent of risk. We find that the potential decline in consumption in response to a severe downturn in the Australian economy could be large, and possibly larger than commonly assumed in past bank stress tests. This is especially true if the consumption response of households to changes in their wealth is nonlinear (that is, larger when wealth is declining than when it is rising). Moreover, the sensitivity of consumption to severe shocks appears to have increased over the past decade or two, as indebtedness has risen. However, our analysis finds that the banks would likely be resilient to such a scenario because of the significant amount of collateral backing their mortgage lending (that is, the moderate loan-to-valuation ratios on most outstanding mortgages). Furthermore, the high level of household debt in Australia compared with other countries can be accounted for by the rental stock being almost fully owned by the household sector in Australia (unlike in many other countries where the debt underpinning rental properties is owed by businesses or government) and by Australia's relatively high average income. We also find that higher levels of income and lower nominal interest rates, as well as the deregulation of the financial sector, can account for most of the rise in Australian household indebtedness over the past three decades, with only a small portion of the rise in indebtedness being either unaccounted for or related to potentially more speculative factors.

A key focus of our results is on the distribution of debt across households, and how that influences its riskiness. Consistent with previous research (particularly Bilston, Johnson and Read (2015)), we find that debt is held by those that are well placed to service it. Highly indebted households tend to be well educated, thereby reducing their risk of experiencing unemployment, and have significantly higher-than-average income. They also seem to have only modest wealth as they tend to be younger-than-average, with less time to have accumulated wealth. These latter two observations have important implications for the sensitivity of consumer spending to adverse shocks to wealth. In particular, when we assume that the spending of highly indebted households is proportionately more sensitive to shocks to wealth we actually reduce the estimated decline in aggregate consumption during stress, because the absolute size of the loss of wealth held by highly indebted households is relatively small. Rather, the risks to consumer spending from a decline in wealth are greatest when we assume that older households (including self-funded retirees or those near retirement) are more sensitive to shocks to their wealth, given that older households hold a disproportionate share of total wealth. Such a calibration is consistent with both theory and the empirical literature, but is only incidentally related to rising indebtedness.^[4] Working against this, we also show that the effect of wealth shocks on consumption is substantially increased if we assume that precautionary savings motives mean MPCs are larger for falls in wealth than for rises. Such an assumption is arguably most relevant for younger, highly indebted households who are more at risk of negative equity than others.

Our final result relates to how changes in the level and distribution of debt and wealth over time have influenced the sensitivity of aggregate consumer spending to wealth shocks. We find that a severe stress scenario would have a somewhat larger effect on consumer spending today than it would have had a decade or two earlier. This increased sensitivity of consumption to shocks arises because of a combination of increased household leverage, a shift in the composition of household assets towards more risky investments and distributional changes in consumer spending across households.

Footnotes

Above-trend credit-to-GDP ratio (the economy-wide counterpart to household debt-to-income ratio) was also found to be a strong predictor of financial crises in earlier work by Borio and Lowe (2002). A separate arm of the literature contends that high debt can have a persistent effect on the economy, by weighing on consumption growth even in normal times and, for example, contributing to permanent declines in the neutral real interest rate (Mian, Straub and Sufi 2020a).[1]

This is the household counterpart to the debt-to-GDP ratio examined by Schularick and Taylor (2012) and others.[2]

The rank correlation across countries between the level of the DTI ratio in 2007 and the subsequent fall in the output gap, consumption per capita or housing prices does not exceed 0.35 (in absolute terms), though the rank correlation is higher with the change in the DTI ratio (−0.33 for the output gap and −0.77 for the change in consumption per capita).[3]

In particular, rising debt levels among younger households has boosted the wealth of older households over the past few decades.[4]