Bank credit and financial crises: a historical perspective (Research of the month)

Moritz Schularick and Alan Taylor, “Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870–2008,” American Economic Review, 2012

Oscar Jorda, Moritz Schularick, and Alan Taylor, “When Credit Bites Back,” Journal of Money, Credit, and Banking, 2013

Abstract:

• In these articles, the authors construct a macro-financial database dating back to 1870 that can be used to analyze financial crises.
• The authors show that growth in bank credit predicts the probability of a financial crisis occurring and the severity of crises in general

What is the role of the financial system in economic dynamics? The economy is characterized by cycles punctuated by crises: does the analysis of financial data provide a better understanding and anticipation of these events?

The idea that credit overheating is at the root of financial crises is central to the work of economists such as Charles Kindleberger and Hyman Minsky—James Tobin already referred to these episodes as « the Achilles heel of capitalism » when commenting on the latter’s work. However, as crises are by nature rare events, we need to delve into history if we want to study them systematically. With global debt reaching 256% of GDP at the end of²⁰²⁰¹ according to the IMF – two and a half times more than 50 years ago! – it is worth examining the historical links between debt and crises.

This is the goal set by Oscar Jorda, Moritz Schularick, and Alan Taylor in constructing a macro-financial database dating back to the Industrial Revolution. This database, available^online, initially covered 14 developed countries from 1870 to 2008. In particular, the authors were the first to reconstruct long, comparative historical series of bank credit and monetary aggregates.

To analyze crises, the first step is to classify these events. To do so, the authors draw on existing methodologies. This is primarily a matter of definition: classic recessions are the most standardized phenomena, precisely classified by^national dating committees. The generally accepted definition of recession is that of the US National Bureau of Economic Research (NBER): a significant decline in economic activity across various sectors, lasting more than^a few^months. Classifying financial or banking crises is more difficult, given their more heterogeneous nature. Economists Luc Laeven and Fabian Valencia have identified two^{configurations:}

• The first is characterized by the emergence of severe tensions in the banking sector, corresponding to more than 20% of non-performing loans or banking system restructuring costs exceeding 5% of the country’s GDP.
• The second is characterized by severe tensions, countered by^massive government intervention.

These methods, applied to the panel of data collected by the authors, identify 298 recessions, including 67 financial crises, over the period.

The first contribution of the papers is to characterize these economic and financial cycles.

• First, it appears that the duration of economic cycles is lengthening: from less than three years on average before World War I to more than ten years in the period following the end of the Bretton Woods agreements in 1944. At the same time, the duration of recessions has remained stable—at around one year.
• Second, the severity of these recessions has decreased: while they were accompanied by an average loss of production of nearly 6% between the two wars, this has fallen to only 1-2% since 1945. Crises are no longer accompanied by a fall in monetary aggregates and deflationary loops, but are instead followed by periods of inflation. This is the potential result of increasing interventionism by central banks.
• Finally, financial crises differ from recessions in their intensity and duration. The loss of production is around 3% after one year, worsens in subsequent years, and remains 2% below the pre-crisis level for up to five years.

Is the financial system simply a conduit for savings to investment, or is it a real sounding board that amplifies or even generates its own shocks? The 2008 crisis seems to have settled the question. A more indebted economy is more vulnerable to confidence shocks and more sensitive to procyclical price variations affecting financial balance sheets. The papers « Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870–2008 » and « When Credit Bites Back » do not seek to validate a specific theory, but rather to identify historical patterns. They answer the following questions in turn: What is the role of credit in these financial crises? Is there a link between the intensity of credit dynamics and the severity of crises?

• The first result is a predictive relationship between real credit growth and the probability of a crisis occurring. A one standard deviation change in this growth rate over five years increases the probability of a crisis by 2.8 percentage points, given that their usual frequency is around 4%.The first paper establishes, in particular, the superior predictive power of credit compared to other variables such as currency.
• The second paper finds that the intensity of^pre-crisis credit growth⁷ predicts the intensity of real financial crises. These results appear significant when using the « local projections ^{» method8} to measure a predictive link at variable horizons and controlling for a number of macroeconomic and financial variables (see chart).

Average change in real GDP per year since the real or financial crisis; dotted lines show scenarios with credit-to-GDP ratio growth 1/2/3 points higher in the pre-crisis period. Example: a financial crisis following a period of credit-to-GDP ratio growth at a rate 2 points higher than the average pre-financial crisis rate will lead to a 6% drop in real GDP per capita after two years.

These papers have empirical limitations. The choice to focus on bank credit while excluding debt securities is becoming less meaningful as forms of debt diversify. In the second paper, the authors clearly show that the inclusion of debt securities predicts a more severe crisis ahead of the 2008 « Great Recession. » The transmission channels may also differ between credit, which involves bank balance sheets with their specific characteristics, and debt securities, which constitute debt granted by other types of investors. Furthermore, the papers do not claim to identify a causal relationship or to weigh in favor of a specific transmission mechanism, which limits their theoretical scope.

However, these papers offer, for the first time, a historical and quantitative perspective on financial crises, which sheds light on our understanding of these rare events.

Notes:

1 – See an IMF blog.

2 – https://www.macrohistory.net/database/

3 – In France, this is the French Business Cycle Dating Committee (CDCEF).

4 – See the NBER website.

5 –Luc Laeven and Fabian Valencia, “Systemic Banking Crises: A New Database, ” IMF Working Paper, 2008.

6 – In this second scenario, a financial crisis is deemed to have occurred if three of the following six measures are adopted: a freeze on deposits, nationalization of banks, restructuring of the banking system costing more than 3% of GDP, injection of liquidity exceeding 5% of GDP, significant implementation of guarantees, or asset purchases worth more than 5% of GDP.

7 – This excess is measured as the growth rate of the ratio of domestic bank credit to the non-financial private sector to GDP, relative to its historical average.

8 –Oscar Jorda, « Estimation and Inference of Impulse Responses by Local Projections, «  American Economic Review, 2005.