Bank Regulatory Capital Requirements

Herring argues that 75 percent of the various capital requirements for US GSIBs could be eliminated without weakening capital regulation, and offers that as evidence of the growing complexity of US capital requirements. An alternative way to view the growing complexity embodied in US capital adequacy standards comes from RegData 3.1. The database provides counts of the number of regulatory restrictions in the CFR, which include words, such as "may not", "must", "required", "shall", and "prohibited". More restrictions that are embodied in the CFR in principle means that banking organizations must spend more resources on compliance. In the CFR, Title 12 concerns banks and banking. Parts 1–199 concern the Office of the Comptroller of the Currency (OCC), parts 200–299 concern the Federal Reserve System (FRS) and parts 300–399 concern the Federal Deposit Insurance Corporation (FDIC). The parts of the CFR that concern commercial bank capital requirements in Title 12 include "Part 3 – Capital Adequacy Standards", "Part 217 – Capital Adequacy of BHCs, S&L Holding Companies, and State Member Banks" for the FRS and "Part 324 – Capital Adequacy of FDIC-Supervised Institutions" and "Part 325 – Capital Maintenance" for the FDIC. We can present measures of the rising complexity of bank capital requirements by dividing the total number of regulatory restrictions that concern bank capital by the total number of regulatory restrictions, for each agency.

Figure 1 depicts the fraction of regulatory restrictions for the FDIC, OCC and Federal Reserve that concern capital requirements since 1970. The figure shows that the fraction of restrictions that concern bank capital for the FDIC and OCC has increased greatly relative to what existed under Basel I, and now equals nearly 25 percent. For the Federal Reserve, almost all of the regulatory restrictions have come since Basel III, and by 2017, exceeded 10 percent.

Figure 1. The Fraction of Regulatory Restrictions Arising from Bank Capital Requirements, 1970–2017.

As a suggestive exercise, we estimate how capital requirements contribute to overall regulatory complexity by agency by applying Mora and Reggio's fully flexible approach to estimating average treatment effects. The idea is to estimate how many additional restrictions or word counts are on average generated by regulatory capital requirements.

Mora and Reggio show that, if the trends for the treatment and control groups are the same, the appropriate estimator for the average treatment effect for the difference-in-differences estimator under the parallel paths assumption yields the same treatment effects as the difference in double-differences estimator under the parallel growths assumption. This no longer holds if the pre-treatment dynamics for the treatment and control groups differ. They also propose a test for equal pre-treatment dynamics.

The outcome variable here is either the total number of regulatory restrictions or total word counts in the CFR, by year, for the OCC and FDIC; we exclude the Federal Reserve given that there were many years when Part 217 generated no regulatory restrictions. As a treatment variable, we use a dummy variable that equals one if the CFR part equals 3 for the OCC, or 324 or 325 for the FDIC, and equals zero otherwise. The assumption is that restrictions (or word counts) in the parts of the CFR that concern capital contribute to the total number of restrictions (or word counts), and not vice versa. As a post-treatment period, we use the period starting in 1989, when Basel I was finalized by US regulators. As an alternative, we estimate similar average treatment effects after replacing total regulatory restrictions with total word counts. Given that we do not reject the null hypothesis of equal pre-treatment dynamics, we report results for the post-treatment effects assuming parallel paths, together with the 95 percent confidence interval, from 1989–2017 in Figure 2.

Figure 2. Average Treatment Effects of Bank Capital Requirements on Regulatory Restrictions and Word Counts Since Basel I, 1989–2017.

The figure shows that the CFR parts that concern capital requirements on average generate a substantial number of restrictions relative to all other parts. For instance, Part 3, on average, generated roughly 600 (1300) more regulatory restrictions and 70,000 (150,000) extra words than other parts under Basel II after 2007 (Basel III after 2013). For comparison, the other parts of the CFR for the OCC on average generated 55 restrictions (7000 words) after 2007 and over 65 restrictions (7000 words) after 2013. Basel II and Similarly, Parts 324 and 325 for the FDIC on average generated about 300 (900) more regulatory restrictions and 40,000 (100,000) extra words than other parts under Basel II after 2007 (Basel III after 2013). For comparison, the other parts of the CFR for the FDIC on average generated about 60 restrictions (6000–7000 words) after 2007 and over 100 restrictions (11,000 words) after 2013.

Beyond the added compliance costs, regulatory complexity also can have unintended consequences including opportunities for regulatory arbitrage, which have been observed by academics long before the 2007–2009 crisis. Although the risk weights have become much more complex since the introduction of Basel I, the basic framework – setting minimum capital requirements as a fraction of RWAs with risk weights assigned to asset categories – has remained the same. At the same time, Acharya et al. argue that "risk weights are flawed measures of bank risks cross-sectionally as banks game their risk-weighted assets (cherry-pick on risky but low risk-weight assets) to meet regulatory capital requirements, which does not necessarily reduce economic leverage". Other studies find that non-risk-based measures of capital better predict bank stock returns or bank risk than risk-based measures. Moreover, Flannery observes that banks satisfied regulatory capital requirements, which rely on book values, while market valuations of capital plunged well below book values during the crisis; we examine such problems in more detail next.