How to fix the regulation of ‘too big to fail’ banks
Some banks are too big to fail, meaning they'll likely be bailed out by the government if facing bankruptcy. To avoid such banks behaving recklessly at the expense of the taxpayer, banking regulators have imposed safety nets, based on risk calculation. A trio of researchers uncovers flaws in the risk-scoring system and proposes simple improvements – but will they be heard?
Banking is a risky business. If financial institutions make rash bets on lousy investments, isn't it only fair that they sometimes crash and burn? It only appears like sound business logic. But some banks have an escape clause. They provide liquidity for so many companies, have such tight ties to other financial institutions, and buy such large fractions of government debt, that their bankruptcy would have dramatic social and economic impacts. These banks are known as SIFIs: Systemically Important Financial Institutions, or more colloquially, “too big to fail”. When they teeter on the brink of disaster, governments prefer to bail them out rather than risk seeing them take down the rest of the economy in a disastrous domino effect.
In the wake of the 2008 financial meltdown, and to avoid taxpayers being again “held hostage by bank[s] that [are] too big to fail” (in former US President Barack Obama's words), international banking regulations were toughened. In particular, the 2010 Basel III Accords imposed minimum amounts of equity capital for SIFIs, so as to increase their capacity to weather unexpected losses.
“The rationale for increasing the regulatory capital of the financial institutions that contribute most to the risk of the system is to force such banks to internalize the costs they inflict on the system and thereby create incentives for them to reduce such externalities,” explains Professor Christophe Pérignon, who co-holds the Chair in Regulation and Systemic Risk at HEC Paris.
The current systemic-risk scoring system...
Mandatory levels of buffer capital are decided by the Basel Committee on Banking Supervision (BCBS) and the Financial Stability Board. Their methodology is simple and intuitive. They assess each bank's systemic risk, as opposed to the risk posed by an institution in isolation, by aggregating information about five broad categories: size; interconnectedness; substitutability, or the presence of readily available substitutes for the services the bank provides; its cross-jurisdictional activity; and complexity. For non-eurozone banks, categories are converted into euros.
According to their scores, banks are sorted into five buckets, each with additional requirements in terms of regulatory capital, from 1% in the first to an extra 3.5% in the fifth. Compared to the 8% default capital ratio, these extra capital charges appear sizable.
Increased regulatory capital can cut lending, which is an issue of course, as it negatively affect household consumption and firm investment levels and, eventually, economic growth."
Higher regulatory capital effectively limits the amount of risky assets a bank can invest in, because one percentage point or a half actually translates into more than 10 billion euros for the largest SIFIs. Such protection comes at a cost: “There is empirical evidence showing that increased regulatory capital can cut lending, which is an issue of course, as it negatively affect household consumption and firm investment levels and, eventually, economic growth,” says Christophe Pérignon.
...and its flaws
More worryingly, he and two fellow researchers identify two major shortcomings in the current systemic-risk scoring system.
Firstly, it biases scores towards the categories which are the most volatile (without going into details of statistics, because the system aggregates variables without standardizing them, it gives more weight to categories where values show the highest dispersion across banks). To correct for the statistical bias, the BCBS uses one of two possible methods, which is to trim outliers, or extreme values, for the substitutability category. Although this method works in theory, it could encourage risk-taking among banks, because there is no incentive to reduce risk once the cap is exceeded, according to the researchers.
“An analogy would be to give speeding tickets to anyone driving faster than 80 km/h, and to increase the fine as the speed increases from 80 to 120 km/h, but to give the same fine to anybody driving faster than 120 km/h, including at 200 km/h! Guess what would be the effect of capping the fine on the speed of the fastest drivers and on the number of fatal accidents.” explains Professor Pérignon.
The other flaw is a foreign exchange effect: any depreciation of a currency with respect to the euro mechanically lowers the score of banks headquartered in this currency zone and increases the score of eurozone banks; and vice versa. “Japanese banks complain about the fact that the main driver of their regulatory capital is the yen/USD exchange rate, which is not under the operational control of banks,” says Christophe Pérignon, again pointing out that such fluctuations translate into billions of yens of equity capital.
Will improvements to the risk scoring system ever be applied?
In their forthcoming paper, the team of researchers offers possible corrections for these shortcomings.
Regarding the foreign exchange effect, they suggest using a constant reference exchange rate, which could be calculated for example as the average rate over the past three years.
Regarding the first statistical bias, the researchers suggest standardizing each category by its own volatility; a process akin to that by which the grades of essays in a national exam are smoothed so as to eliminate any unfairness caused by the varying strictness of the teachers grading the papers.
“We didn't invent a new technique, it's just Stats 101,” quips Christophe Pérignon.
The researchers empirically tested their improved methodology on data collected from more than 100 of the world's largest banks. Overall, the three systemic-risk scores – the BCBS's, one adjusted for volatility and another for volatility and foreign exchange – were strongly correlated, but 11 banks switched buckets (7 upwards, 4 downwards).
Based on the new scheme, the total extra capital requirement was higher (276 billion euros) than the current level (259 billion euros). Besides the sheer magnitude of the discrepancy in amounts, what Christophe Pérignon believes is most important is the correction in the statistical distortion: “So far, we have not been asking the right banks to contribute.”
Will the researchers' advice be heeded? Hopefully. Professor Pérignon testified before the Basel Committee in November 2017 to explain the problems and ways to fix them. But he points out that the process is also a political game, with regulators not only regulating but sometimes also acting as “lobbyists” for the institutions of their own countries: “The capping method currently used offers a lot of flexibility, in the sense that you can put the cap on any dimension and massage the outcome of your regulation to protect the banks you want.”
Based on an interview with Christophe Pérignon and on his paper “Pitfalls in Systemic-Risk Scoring” (Journal of Financial Intermediation, forthcoming) co-authored with Sylvain Benoit and Christophe Hurlin.
The methodology could be applied immediately.
Professor Pérignon is cautiously optimistic about the Basel Committee adopting the suggested improvement concerning the foreign exchange effect. He is less confident that it will drop the cap system, highlighting the fact that US lobbyists have so far been very good at getting the system to protect their banks: “They decided to lower the capital of the largest custodian banks (high substitutability score), all of them being US banks,” he notes.
The researchers put together a website, sifiwatch.org, with all the data they collected about SIFIs. Every year, the website also discloses the new list of SIFIs, several months before the official announcement by the Financial Stability Board, with so far remarkable accuracy. “The economic impact of SIFIs is huge, but the risk data for these banks are not available in a centralized way and sometimes hard to get. Our website provides transparency,” says Pérignon. Last but not least, their methodology is potentially applicable to sectors that require similar risk scoring, such as insurance or asset management.
The researchers collected regulatory data for a sample of 119 global banks from 22 countries between 2014 and 2016. They obtained data on leading European banks from the European Banking Authority and data on large US banks from the Banking Organization Systemic Risk Report. But they had to crawl the individual websites of sample banks outside the US and the EU, including sites in Chinese, to gather data indicating systemic risk (including contagion, amplification mechanisms, etc.).