Quantifying household deleveraging following the 2007 South African financial cycle peak

Abstract

Many countries experienced a credit and housing boom over the period 2003 - 2007. This was followed by a burst in the US housing bubble, which contributed to the deep global economic and financial crisis which began in 2007. Consequently households found themselves in highly leveraged positions, and as a result, sought to restore the health of their balance sheets by following a process of deleveraging, particularly in the mortgage market. The global financial crisis raised important questions about the timing of the South African financial cycle, its characteristics, and how it impacted the real economy. Some suggest that it is the post-crisis debt overhang that has been a drag on consumption and thus economic growth. We date the South African financial cycle turning points and determine stylised facts about the South African financial cycle, as well as the behaviour of household debt over the cycles. We evaluate mortgage, vehicle and other consumer debt at the macroeconomic level, as well as their respective ratios to income. South Africa entered the current financial downward phase in May 2007, and as of December 2017 has not reached a trough or lower turning point. In general we find that South African financial cycles last 17.3 years, around 3 times longer than business cycles (5.8 years). We find that on average financial cycle contractions (10.3 years) last longer than expansions (7.0 years) showing that deleveraging is a long process. However, as information regarding the distribution of debt across the income groups gets masked at the aggregate level, we use the National Income Dynamic Study (NIDS) to determine who deleverages and in which debt categories most of the deleveraging occurs following the current financial cycle peak in May 2007. We create a panel from the NIDS data by following household representatives that were the main or joint decision maker on expenses in the household (different from other studies where the household head is usually followed). We also use multiple imputations by chained equations (MICE) with predictive mean matching to impute point values for bracket responses for debt and income variables, as well as for those who reported that they have debt, but did not provide a value. By imputing the individual debt variables for those who responded that they were debt participants, we increased the total number of observations for all debt types by 43.8% in Wave 1, 35.4% in Wave 2, 26.3% in Wave 3 and 13.4% in Wave 4. We also increased the household income response rate from 75% to 81% in Wave 1, 84% to 94% in Wave 2, 90% to 97% in Wave 3 and 95% to 98% in Wave 4. Our debt categories in the micro data include information on informal debt, which we refer to as other debt. We therefore have four categories: mortgage, vehicle, consumer and other debt. Lastly, we estimate the probability of deleveraging, while controlling for individual, household, and financial characteristics. We show that deleveraging after the recent financial cycle peak is mostly driven by married households in urban areas and those who are in the highest income quintile. We further show that employment, although a major factor in obtaining debt, is not a major driver of deleveraging. This suggests that even as the economy starts to recover, and employment opportunities are created, this will likely not translate into significant deleveraging. This implies that having a job, in itself, does not necessary provide adequate means for households to deleverage in South Africa. By assessing which characteristics assist deleveraging, policy makers can determine in which debt categories an uptake in credit will likely resume, and consequently, consumption expenditure and an economic recovery. Although growth rates in mortgage, vehicle and other consumer debt have decelerated, debt to income levels across all these debt types remain high, compared to the latest pre-financial cycle peak. This suggests that consumers are still in a more vulnerable state than in the period leading up to the May 2007 financial cycle peak. As such any major negative income shocks, either caused by external factors, domestic economic or political conditions, could prolong the financial cycle downward phase. This could halt any consumer-led recovery and the renewed uptake of credit. A further important policy conclusion is that it is mostly those with the financial resources (i.e. higher income earners) to deleverage that do so and mainly in the form of mortgage and consumer debt. This suggests that it is only a small section of the economy that is able to deleverage. Disconcertingly, this means that indebted households in the lower income quintiles are most severely impacted and are unlikely to deleverage.