DP16059 Household Debt Overhang Did Hardly Cause a Larger Spending Fall during the Financial Crisis in the UK

Author(s): Lars E.O. Svensson
Publication Date: April 2021
Programme Areas: Monetary Economics and Fluctuations
Link to this Page: cepr.org/active/publications/discussion_papers/dp.php?dpno=16059

The "debt-overhang hypothesis" - that households cut back more on their spending in a crisis when they have higher levels of outstanding mortgage debt (Dynan, 2012) - seems to be taken for granted by macroprudential authorities in several countries in their policy decisions, as well as by the international organizations that evaluate and comment on countries' macroprudential policy. Results are presented for UK microdata that reject the debt-overhang hypothesis. The results instead support the "spending-normalization hypothesis" of Andersen, Duus, and Jensen (2016a), what can also be called the "debt-financed overspending" hypothesis - that the correlation between high pre-crisis household indebtedness and subsequent spending falls during the crisis reflects high debt-financed spending pre-crisis and a return to normal spending during the crisis. As discussed in Svensson (2019, 2020), this is consistent with the correlation reflecting debt-financed overspending through what Muellbauer (2012) calls the "housing-collateral household-demand" and Mian and Sufi (2018) the "debt-driven household demand" channel. The correlation is thus spurious and an example of omitted-variable bias. A simple model shows that consumption and debt changes are directly and strongly positively correlated, whereas consumption and debt levels are quite weakly negatively correlated. Importantly, and in contrast, examples show that there is no systematic relation between consumption cuts and levels of or changes in LTV ratios. The lack of a robust relation between consumption cuts and levels of or changes in LTV ratios implies that tests of these hypotheses should generally not be done by regressions of consumption cuts on levels of or changes in LTV ratios.