|
|
In the ongoing debate on the design of central bank digital currency (CBDC), several authors have expressed important concerns about financial stability risk. As CBDC would provide depositors with an account at the central bank, they may trigger a run on the banking system. In case of systemic uncertainty about banks, holding risk-free CBDC could indeed become more attractive than bank deposits (Allen et al. 2020, Bindseil and Panetta 2020, Fernández-Villaverde et al. 2020, Juks 2018, Mersch 2018). These concerns have inspired proposals for a special design of CBDC (Bindseil and Panetta 2020) and are considered in official reports of central banks about the future of money (BIS et al. 2020, ECB 2020). Nonetheless, no empirical evidence exists to date to inform such a claim. Can central bank digital currencies cause bank runs?
Flight-to-safety and the French banking crises of 1930-1931
Our research confirms empirically the intuition that the existence of safer deposits than bank deposits can trigger bank runs (Monnet et al. 2021). We examine the causal effect of bank runs on real economic activity during the French Great Depression (1930-1931). Our identification strategy exploits a key and original feature of this crisis (and of the French interwar banking system more generally) – namely, the existence of safe alternatives to banks and their heterogeneous establishment across the country – that generated an exogenous variation in the probability of bank runs.
Starting early November 1930, French depositors – households and firms – suddenly withdrew their funds from commercial banks all across the country (Baubeau et al. 2021, Monnet et al. 2021). French banks were still unregulated at that time (the first banking laws were passed in 1941). Depositors transferred the funds withdrawn from banks to savings institutions (Caisses d’épargne ordinaires, or CEOs). Contrary to banks, CEOs were regulated and they benefited from the guarantee of the government, but they were not allowed to lend nor to provide payment services.
We hypothesise that the pre-crisis density of CEO accounts at the local level increased the probability of bank runs, in a way that was independent of the health of local banks. Because CEO deposits were as liquid as bank deposits, people who already had a CEO account had incentives to withdraw their deposits from banks as soon as uncertainty emerged at the national level.
The pre-crisis density of savings institutions (measured by the number of CEO accounts per capita) was heterogeneous across departments and randomly distributed for historical reasons unrelated to local economic performances and bank characteristics. CEOs were not allowed to lend (they could only hold government securities), so other sources of borrowing were cut at the local level. Our study is conducted at the departmental level (France was divided into 90 departments), with yearly data. Banking activity is measured by the number of bank branches (the only statistics on banks available at the local level over this period). We have reconstructed a new measure of real GDP for the French departments in the interwar period, based on comprehensive tax data.
The causes and effects of bank runs
The right-hand panel in Figure 1 shows the correlation between the pre-crisis density of savings institutions, measured by the number of CEO accounts per capita in 1924, and the growth rate of bank branches between 1929 and 1932. The strong negative relationship means that more CEO-dense departments were those that witnessed a sharper decline in banking activity. Departments with fewer CEO accounts per capita experienced a smaller decline, or even an increase, in bank branches.
This dynamic was not the continuation of a pre-trend. The left-hand panel in Figure 1 shows no correlation between pre-crisis CEO density (in 1924) and the growth rate of bank branches before the crisis, between 1925 and 1928. CEOs and banks were not substitutes but complement before the crisis. When the panics started, however, bank runs spread across the country, and deposits in CEOs and banks became negatively correlated.
Figure 1 Number of accounts per capita in CEO pre-crisis (1924) vs growth rate of bank branches between 1925 and 1928: Before the crisis (left) and the crisis years of 1929–1932 (right)
Source: Monnet et al. (2021).
We then use the pre-crisis density of the CEOs to instrument the growth rate of bank branches during the crisis. Using this method, we find a strong effect of a decline in banking activity on GDP. A 1% decrease in bank branches caused a drop in income by about 1% per year between 1929 and 1932. To give an order of magnitude, a 1% annual decrease in bank branches was associated with a 2% decrease in bank credit at the bank level (using national data). A back-of-the-envelope calculation suggests that our identified causal effect of bank runs may explain one-third of the drop in real GDP in 1930-1931.
Consequences for the design of CBDC
Besides studying the real effects of bank runs, our paper shows how competition between unregulated and safer financial institutions affects financial stability. This holds important lessons for current banking regulation and the debate on digital central bank currency. Our results confirm empirically that the existence of safe deposits other than banks can play a substantial role in triggering bank runs (Bindseil and Panetta 2020, Fernández-Villaverde et al. 2020). This risk is not just a theoretical curiosity.
Moreover, our study provides two specific insights for the design of CBDC. In particular, we can provide evidence on two elements of the current discussion: ceilings on safe deposits and interest rates.
On the one hand, we show that ceilings applied to safe deposit accounts greatly matter during a crisis. In March 1931 – in the midst of the crisis – the French Parliament raised the maximum amount that depositors could deposit in their CEO accounts from 12,000 to 20,000 francs for individuals and 50,000 to 100,000 for firms. This decision was motivated by political pressure arguing that deposits with CEOs were the safest way to protect depositors during the crisis. According to several contemporary observers, the increase in the CEO deposits ceiling was a fatal blow that worsened the severity of the second wave of banking panic at the end of 1931.
On the other hand, the interest rate differential between safe institutions and banks did not matter in the pre-crisis period, but certainly contributed to exacerbating the flight-to-safety when the bank runs started. Figure 2 shows that, from 1927 through 1932, interest rates paid on CEO deposits, regulated by the government, were regularly higher than interest rates paid by commercial banks to their depositors. Depositors accepted a lower interest rate on their bank deposits in normal times because banks provided additional services (credit relationship, means of payment, investment advice, and management of securities portfolios often for free), although CEO deposits were also liquid. They rushed on safe savings accounts only during the crisis. In other words, the spread between interest rates has very different behavioral consequences during a systemic crisis, because the risk is perceived differently. Our historical study thus suggests that it would be possible for CBDC to pay higher interest rates on deposits in normal times, as long as there is a deposit ceiling that also prevents runs in uncertain times...
more at Vox