Can We Prove A Bank Guilty Of Creating Systemic Risk? A Minority Report
London School of Economics – Systemic Risk Centre
London School of Economics; Financial Conduct Authority
University of Chile
Federal Reserve Board
November 17, 2015
Since increasing a bank’s capital requirement to improve the stability of the financial system imposes costs upon the bank, a regulator should ideally be able to prove beyond a reasonable doubt that banks classified as systemically risky really do create systemic risk before subjecting them to this capital punishment. Evaluating the performance of two leading systemic risk models, we show that estimation error alone prevents the reliable identification of the most systemically risky banks. We conclude that it will be a considerable challenge to develop a riskometer that is both sound and reliable enough to provide an adequate foundation for macroprudential policy.
Can We Prove A Bank Guilty Of Creating Systemic Risk? A Minority Report – Introduction
In the film Minority Report the PreCrime unit of the DC Police uses psychics to detect people who are about to commit crimes and then locks them up before they do so. As it happens, Andrew Crockett’s original vision of macroprudential policy (which we call Targeted Macroprudential Policy or TMP) is envisioned to work in more or less the same way (Crockett, 2000). The crime in this case is to cause a financial crisis; the perpetrators are banks and other financial firms (henceforth banks) pursuing systemically risky strategies; the task of the PreCrime units of banking supervisors is to detect these banks and require them to increase the proportion of costly capital in their balance sheets so that they do not create a crisis by becoming stressed or failing. The crucial role of the psychics is played by algorithms or riskometers that empirically implement a theoretical measure of how much systemic risk each bank creates. In the movie the government shuts down PreCrime as it turns out that the psychics’ visions are not accurate enough to justify punishing people on the basis of the information they provide. In this paper we examine the limits of riskometer performance to see if riskometers can be reliable enough to provide a sound foundation for TMP. We are, to the best of our knowledge, the first to analyze the relationship between the reliability of systemic risk measurements and macroprudential policy.
In our analysis we assume that a regulator acts to minimize the sum of the expected social cost of a crisis and the social cost of bank capital. The regulator is endowed with a riskometer that it can use to get a risk reading for each bank, with the reading equal to either Guilty (beyond a reasonable doubt of creating systemic risk) or Safe. The regulator chooses between two policy options: i) the TMP option in which it requires only banks with a Guilty reading to operate with a high level of capital (the level appropriate for a systemically risky bank); and ii) a Blanket Macroprudential Policy or BMP option in which the regulator requires all banks to operate with a high level of capital. The BMP option eliminates crisis risk, but it does so inefficiently as both actually risky banks and safe banks must operate with high levels of costly capital. The TMP option conserves on costly bank capital, but at the cost of increased crisis risk: banks that do create systemic risk but cannot be shown beyond a reasonable doubt to do so can operate with a low level of capital. We define a riskometer’s reliability to be equal to the proportion of risky banks that get a Guilty reading when using that riskometer. The relative advantage of the TMP option increases with the reliability of the regulator’s riskometer, and the TMP option is always the optimal choice if the regulator has a perfectly reliable riskometer.
We investigate the practicality of TMP by first estimating a lower bound for how reliable a riskometer needs to be in order for TMP to dominate BMP and then estimating an upper bound for how reliable riskometers can be by examining riskometer reliability in a best case scenario. Based upon recent Basel Committee on Banking Supervision (2010) estimates of the cost of a crisis, the level of capital needed to eliminate crisis risk, and the cost of bank capital, we find that the TMP option dominates the BMP option if the reliability of the regulator’s riskometer exceeds 75% (in our base specification).2 Our estimates of riskometer reliability in a best case scenario fall far short of the required 75% level. Consequently, we think that it will be a considerable challenge to develop a riskometer that is reliable enough to provide a sound foundation for TMP.
See full PDF below.