Empirical support for deleveraging lays groundwork for theoretical advances
As if to underline the importance of a recovery in private consumption to broader growth prospects, the IMF’s recent World Economic Outlook dedicated a chapter to the issue of household debt and its consequences for recovering economies. Based on analysis of advanced economies, the study found that recessions and housing busts that were preceded by large accumulations of household debt tended to be both more severe and protracted. These recessions tended to see large reductions in real GDP and private consumption, large rises in unemployment, and reduced economic activity that persisted for around five years.
Gross debt does matter
Gross debt matters in it’s relation to income, so it’s not just ‘net wealth’ as some commentators have argued. Net wealth is an attractive concept, as it judges our debt in relation to our assets. Evidence from the IMF research shows that it is not simply the decline in house prices and associated reduction in wealth that lead to large declines in economic activity. It is the combination of house price declines with pre-bust leverage that explains the depth of these recessions.
The role of gross household debt in amplifying slumps
The IMF’s sample consisted of 25 countries. Using nominal house price series, they identified 99 housing busts. They then classified economies prior to these busts as either high debt or low debt economies. The model specification saw real consumption regressed on lagged real consumption, housing bust dummies and an interaction of this dummy with the dummies that refer to the high debt and low debt status of the economy. The main result was quite dramatic: the decline in real household consumption is 4.3% after 5 years for the high debt group and 0.4% for the low debt group.
Importantly, it is not simply the fall in house prices, reflecting a fall in wealth, which leads to such a dramatic reduction in consumption. The report states that at most the fall in house prices explain one quarter of the reduction in consumption. Via different specifications of the model, the report strongly suggests that it is the combination of house price declines and pre-bust leverage that explain the severity of the reduction in household consumption.
This is a very important result from the IMF. Again, a lot of commentators expect that net wealth is all that matters, which implies that is only the relation between a households’ assets and liabilities that matter, not the levels. This assumption is questioned by the results of the IMF research, as it seems that gross debt really does matter.
A couple of thoughts come to mind about this result. One interpretation could be that gross debt matters because it links directly with someone’s ability to pay. Gross debt is manageable when income (or one could think of liquid wealth) is steady and secure, but it becomes unmanageable when that income is lost or is reduced significantly. Secondly, it may be that net wealth does play a role in determining the comfortable or desired level of gross debt, but again, the role of income and the ability to pay would seem to also influence this level of debt.
The IMF concludes the empirical section by citing the work of Mian & Sufi, discussed here, demonstrating the consistency of results between the two pieces of research.
Theoretical reasons why this debt can deepen contractions
One of the main theoretical areas that is explored in the literature is that of the difference between lenders and borrowers. This is the notion that lenders have a higher marginal propensity to consume, thus a shock that causes them to reduce their consumption will lead to a decline in aggregate activity. This shock may come via a sharp revision in income expectations, an increase in economic uncertainty, a tightening of credit standards, or a drastic realisation that house prices are overvalued.
Crucially, the report highlights the significant role that the zero lower bound plays in these models. It could be the case that a sufficiently large reduction in interest rates would encourage creditors to increase consumption, thus offsetting the decline in spending from the borrowers. However, where interest rates are already at zero – a classic liquidity trap scenario – it is not possible to lower the nominal rate any more.
The cubic IS model tries to capture some of these factors. Firstly, it uses a heterogeneous agent model, where households are distributed along a spectrum according to their willingness to take on debt. The shock in the model that causes the deleveraging comes from a sharp revision in house price expectations. These expectations drive the absorption of debt on the way up, but also the paying off of debt on the way down.
Finally, although not modelled explicitly, the model enables an exploration of aggregate consumption behaviour in a liquidity trap. Due to the design of the cubic consumption function, a central bank can shift it rightwards by raising inflation expectations, but this is countered by households deleveraging that pulls the curve leftwards as it flattens out. The next post on the cubic IS curve will outline this new consumption function.