Stealing Deposits: Deposit Insurance, Risk-Taking and the Removal of Market Discipline in Early 20th Century Banks

Deposit insurance spread throughout the world in the latter half of the 20^th century as a result of external and internal political pressures favoring its adoption (Demirgüç-Kunt, Kane and Laeven 2008). Despite its overwhelming political support, there is a large empirical literature suggesting that the moral-hazard costs of deposit insurance have out-weighed its liquidity-risk-reduction benefits and have contributed to the unprecedented waves of banking crises that have washed over the world during the past four decades. [1] The separation between policy recommendations and economic studies begs the question of whether empirical studies may have failed to properly control for the other contributing influences that produced both the rise of deposit insurance and banking instability.

Most studies of deposit insurance are based on cross-country comparisons or comparisons across time within countries that contrast the behavior of insured banking systems with uninsured banking systems. Despite attempts by authors to control for factors that coincide with the creation or expansion of deposit insurance through explicit controls or through instruments that explain the creation of deposit insurance, it is conceivable that some of the positive association between deposit insurance and increased bank risk may reflect exogenous increases in risk that encourage the passage of deposit insurance. (Even the few studies that examine within country variation such as Brewer (1995) and Yan et al. (2014) still could suffer possible endogeneity bias as they compare fundamentally different types of institutions.) If true, then the risk-creating effects of deposit insurance would be exaggerated.

In our recent paper, Stealing Deposits: Deposit Insurance, Risk-Taking and the Removal of Market Discipline in Early 20th Century Banks, we examine a near ideal environment from the standpoint of identification—the state deposit insurance experiments of the early 20th century in the United States. These systems installed deposit insurance for unit state-chartered commercial banks that operated in parallel to the uninsured system of national banks (i.e., unit banks that were chartered by the Comptroller of the Currency) within the same states and to uninsured state and national banks operating in bordering states. Substantially mitigating the problem of endogeneity bias due to omitted variables (a common criticism of cross-country studies), the setting thus allows for the study of insured and uninsured depository institutions operating at the same time and place as well as under the same legal system, currency, and language. We employ detailed information about the locations, economic environments, and bank-level balance sheet characteristics of insured and uninsured banks for many states and years. Specifically, we implement a difference-in-difference-in-difference model that measures the effect of deposit insurance on insured banks controlling both for the change in uninsured banks in the deposit insurance states and for the change of uninsured banks in other states. Moreover, because several of the laws were passed in the same year but implemented in different subsequent years, we are able to use placebo tests to show that the effects of deposit insurance do not merely reflect the environment that shaped its passage. When full implementation was delayed after passage, so were the observed moral-hazard consequences of insurance. Furthermore, we are able to show that our results are not the result of region-wide differential growth.

Our findings not only corroborate the prior literature on the moral hazard consequences of deposit insurance, but also show how the introduction of deposit insurance created systemic risk. We find conclusive evidence that deposit insurance caused risk to increase in the banking system by removing the market discipline that had been constraining uninsured banks’ decision-making. Deposit insurance increased insured banks’ deposits and loans, and lowered their cash to asset ratios and capital to asset ratios. At the same time that depositors applied strict market discipline on uninsured banks when evaluating whether to place their deposits, they seem to have ignored the deteriorating financial soundness of insured banks, allowing insured banks to continue to compete away deposits from uninsured banks until the collapse of the insurance systems during the 1920s.

The extent to which insured banks attracted deposits away from uninsured banks, and used those funds to expand their lending, depended on the risk opportunities available in their local economic environment. Variation across counties in the extent to which they produced commodities that appreciated during the World War I agricultural price boom explains between one-third and two-thirds of the observed effects of deposit insurance on deposit growth, loan growth and increased risk taking by insured banks. The fact that a large part of the moral hazard associated with deposit insurance is dependent on the time-varying and location-specific opportunities for risk taking has important implications for empirical analysis of the consequences of deposit insurance in other contexts. The potential costs of deposit insurance may appear low in environments that are relatively lacking in risk-taking opportunities, but those costs can appear much higher when greater risk taking opportunities present themselves.

Insured banks seemingly were betting on the permanence of agricultural price increases that had occurred during World War I, and depositors seemingly believed in the insurance systems’ ability to protect them. As banks most often used those deposits to fund new loans, the implementation of deposit insurance thus allowed an asset price bubble to quickly form (Rajan and Ramcharan 2016). When prices reversed in the early 1920s, insured banks suffered much higher failure rates at the end of World War I, and the collapse of these insurance systems resulted in losses to depositors (Alston, Grove, and Wheelock 1994). The history of deposit insurance in the United States thus has been a process of increasing systemic risk in the name of reducing liquidity risk. The deeper lesson of that history is that economic models that attempt to explain the attraction of deposit insurance are less relevant than political ones (Demirgüç-Kunt, Kane and Laeven 2008; Calomiris 2010, Calomiris and Haber 2014; Calomiris and Jaremski 2016).

The full paper is available for download here.