Jonathan Bridges, Georgina Green and Mark Joy
Any distributional effects on credit of macroprudential policies are only one part of the distributional story. Relatively little is known about how such policies affect the income distribution in the longer term via their role in preventing crises or mitigating their severity. Our paper helps to fill that gap in the literature by looking at the impact of past recessions and crises on inequality, and the amplifying roles of credit and capital within that. This helps to shed light on the distributional implications of not intervening – in the form of an amplified recession. We find that inequality rises following recessions and that rapid credit growth prior to recessions exacerbates that effect by around 40%.
To shed light on this issue we extend findings that link measures of the financial cycle – such as credit growth – with the probability and severity of macroeconomic tail events. We use a cross-country data set spanning the five decades prior to the Covid-19 pandemic to investigate whether rapid credit growth in the lead-up to a downturn is associated with an amplification of any subsequent impact on inequality. To our knowledge, we are the first to extend those findings into distributional space.
Recessions and financial crises in our sample
Our data are annual in frequency and cover 26 advanced economies since the 1970s. Our final sample covers around 100 recessions, of which just over 20% are financial crises. We identify a recession as two consecutive quarters of negative real GDP growth (based on OECD and national statistics websites). When a recession is accompanied by a banking crisis – defined by Laeven and Valencia as the recession being within one year of a systemic banking crisis – we call it a ‘financial’ recession. When there is no banking crisis, we call these ‘normal’ recessions. Recessions are well represented across the five decades but financial recessions are mainly concentrated around the global financial crisis (GFC).
Our data source is the Standardised World Income Inequality Database. We focus on market income inequality and use the Gini coefficient as our headline measure. This captures the extent to which the Lorenz curve – which reflects the proportion of overall income assumed by different income shares ordered from lowest to highest – sags below the 45-degree line of ‘perfect equality’. If during recessions those at the bottom of the distribution bear the brunt of the shock we might expect the Lorenz curve to shift down and the gini coefficient to increase.
So what does the Gini coefficient look like in our sample? Income inequality has trended upwards over the past 50 years growing by around 20% since the 1970s (Chart 1). This trend has been the focus of a growing body of work looking at how rising inequality may have set the conditions for the GFC. But our interest is actually in the reverse of this – the effect of recessions on inequality, and not in the trend but in variation around that trend (also called cyclical variation).
Chart 1: The path of market income inequality in our sample
Source: Authors’ calculations, based on SWIID data. The red line represents the median. The blue shaded area represents the interquartile range.
To explore the relationship between recessions and inequality we use a local projections approach, where we regress lead observations (up to five years ahead) for income inequality on recession dummies. Because the dependent variable leads our explanatory variables, this helps to address endogeneity concerns ie the worry that inequality might impact the likelihood of a recession taking place.
To focus on cyclical dynamics we de-trend our dependent variable directly, subtracting the full panel average trend. Alongside that, we also control for any country and time-specific trends. This allows us to abstract from any slow-moving effects driven, for example, by different structural changes in a given country in a given decade.
We include country fixed effects to control for any bias in our estimates caused by unobserved, time-invariant variables across countries. And we also control for the domestic macroenvironment in the period before each recession, by including inflation, the size of the current account, the central bank policy rate and the output gap.
The effect of recessions on inequality
Our baseline regression reveals that income inequality rises following recessions. Recessions are associated with a significant increase in the cyclical component of income inequality three to five years out, rising to 2.7% after five years (Chart 2). When we split our sample into normal and financial recessions we find the response of the Gini to financial recessions builds to nearly 4% by year 5 and is more than 50% larger than for normal recessions (Chart 3).
Our findings are robust to a variety of alternative specifications: alternative approaches to de-trending; dropping overlapping recession episodes; dropping our macro controls; and the country-specific trend.
Chart 2: Cumulative change in de-trended Gini index (%) following recessions
Chart 3: Cumulative change in de-trended Gini index (%) following ‘financial’ and ‘normal’ recessions
Notes to Charts 3 and 4: Solid line gives the mean response of the Gini coefficient to a recession. Shaded areas represent 95% confidence intervals around the mean.
We might expect that a large amount of this rise in inequality is accounted for by a rise in unemployment. Low-income earners are most likely to lose their jobs in a recession as they’re often less skilled and more likely to be employed in cyclical industries. They are also more likely to be young with less secured job contracts. There is also an indirect link via wages, as high unemployment also weakens the bargaining power of workers, resulting in weaker wage growth which may particularly impact wages of the lowest paid.
To gauge the relative importance of the unemployment channel in driving the overall link between recessions and inequality, we control for the contemporaneous move in unemployment. This specification moves away from our baseline local projection approach, which is careful to only include explanatory variables observable in the year preceding the onset of each recession. Here we rely on reduced-form accounting rather than claiming causality.
We find that the increase in income inequality is partially accounted for by the increase in unemployment that accompanies recessions. This suggests there is a skewed impact on the income of those remaining in work, consistent with shocks loading most heavily on lower-paid workers.
The amplifying role of credit
To look at the role of credit growth as an amplifier we interact our recession dummies with credit growth. We find that a one standard deviation increase in credit growth (a 15 percentage point increase in the credit to GDP ratio in the three years prior to the crisis) is associated with around a 1 percentage point additional rise in the Gini, which is a 40% amplification by year 5. When we split our sample we find that the amplifying role of credit growth is strongest (and most statistically significant) for financial recessions (Chart 4). We find that the primary mechanism through which the rise in inequality appears to be amplified by rapid credit growth does appear to be through the unemployment channel.
Chart 4: Cumulative change in de-trended Gini index (%) following financial recessions preceded by high credit growth
Notes: Solid line gives the mean response of the Gini to a financial recession. Dashed line shows the amplified effect of a 1 standard deviation credit boom prior to the crisis. The shaded areas gives the 95% confidence interval.
Chart 5: Cumulative change in de-trended Gini index (%) following recessions preceded by low bank capital
Notes: Solid line gives the mean response of the Gini to a recession. Dashed line shows the amplified effect of 1 standard deviation lower capital prior to the recession. The shaded area gives the 95% confidence interval.
Extension: the role of bank capital
We extend our analysis to explore the role low bank capital ahead of a downturn plays in the inequality fallout that follows. Our capital data is only available for a subset of countries so we group recessions together given the more limited sample size. We include bank capital in the regression by interacting it with the recession dummy. We find that a country entering a recession with a banking sector where the aggregate tangible common equity ratio is one standard deviation (1.4 percentage points) lower, experiences around a 55% amplification of the rise in inequality that follows (Chart 5). Our preliminary results suggest that this may operate through the wage distribution of those remaining in work, rather than through the direct impact of unemployment on inequality. This is consistent with channels whereby ‘resilience gaps’ in the financial system can increase the likelihood and costs of macroeconomic tail events.
Our findings provide potential insights for a holistic assessment of the distributional implications of various macroprudential policy options. In particular, they highlight that any consideration of distributional effects needs to consider other aspects, beyond the immediate effect on credit allocation. These include: i) the distributional effects arising from crisis prevention; ii) the role of credit growth in exacerbating post-crisis inequality; and iii) the effect of greater bank capital on post-crisis inequality. All of these work in the ‘opposite direction’ to the effect on credit allocation of macroprudential measures.
Jonathan Bridges works in the Bank’s Market Intelligence and Analysis Division, Georgina Green works in the Bank’s Macro-financial Risks Division and Mark Joy works in the Bank’s Global Analysis Division.
If you want to get in touch, please email us at [email protected] or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.