Risk-Weighted versus Total Assets

Under the 1988 Basel I agreement, standardized risk weights were applied (to five different risk categories) to arrive at risk-weighted assets (RWA). Required capital was then set at 4 percent of risk-weighted assets for tier 1 capital (equity and retained earnings).

In 1996 an amendment to the agreement provided for weights based on market risk to be applied to the securities held in banks' trading books (as opposed to the “banking” portfolio meant to be held to maturity). It also allowed the largest banks to determine their risk weights using their internal (typically value at risk) models (Neumann and Turner 2005). The final Basel II rules issued in 2004 provided for the use of internal models for credit risk as well as market risk. The shift to internal models was especially important for European banks that were moving from commercial to uni­versal banking.^[8]

In part because of the resulting incentive to apply less conservative in­ternal models, the clear correlation between capital to risk-weighted assets and capital to total assets ratios that had existed across major European banks in 1996 became a random scatter by 2002 (Bayoumi forthcoming). By 2002 the two measures were still closely correlated for mainly commer­cial banks, but the universal banks had much lower ratios of equity to total assets and no clear correlation between the two concepts; by 2008 the posi­tive relationship had broken down even for European commercial banks.

Hoenig (2013) argues that the low risk weights on AAA collateralized debt obligations (such as mortgage-backed securities) and the zero weights on sovereign debt induced banks to leverage these assets excessively onto their balance sheets even as the risks of these asset classes escalated.

Similarly, Goldstein (2017) is acerbic in his critique of risk-weighted assets. He strongly favors capital requirements based on total rather than risk-weighted assets.

The divergence between the two capital measures was most pronounced for investment banks, in part reflecting the fact that in the United States they were regulated by the Securities and Exchange Commission (SEC) rather than the Federal Reserve (which regulates the largest banks) or the Comptroller of the Currency (which regulates other banks). US commercial banks have been subject to prompt corrective action triggered by a leverage ratio of 5 percent for high-quality capital relative to total assets, whereas investment banks (and investment bank subsidiaries of the largest bank holding companies) have not. Thus whereas the largely commercial bank Wells Fargo reported risk-weighted assets equal to 84 percent of total assets in 2008, universal bank JPMorgan Chase reported this ratio at 57 percent and investment bank Morgan Stanley reported it at 42 percent, according to SEC 10-K reports. The potential for a distorted picture from the risk- weighted ratio is highlighted by the fact that in mid-2007 the reported tier 1 capital ratio was actually higher for ill-fated Lehman Brothers (11 percent of risk-weighted assets) than for Wells Fargo (8.5 percent) (Kato, Kobayashi, and Saita 2010). In Europe universal banks with large investment bank oper­ations tended to have much lower risk weighting than standard commercial banks. Thus by 2008 the ratio of risk-weighted to total assets at Deutsche Bank was only about 1.7 percent whereas the same ratio for Italian bank Intesa Sanpaolo was about 8 percent (Bayoumi forthcoming).

In the United States, the minimum leverage ratio for a bank holding company with more than $500 million in assets is 3 percent for “strong” bank holding companies and 4 percent for all others. Bank holding com­panies with weaknesses or in the process of rapid expansion are expected to hold capital “well above” these levels (FDIC 2015). For a bank to be con­sidered well capitalized, it must pass a leverage ratio threshold of 5 percent (a buffer of 2 percent above the Basel III minimum). Below this threshold, “prompt corrective action” can involve restrictions on dividends, buybacks, and discretionary bonus payments. In 2014 US regulators increased this threshold to 6 percent for insured depository bank subsidiaries of the largest bank holding companies (institutions with $700 billion or more in assets) and set the leverage ratio at 5 percent for these institutions (including their nonbank subsidiaries) as a whole (Comptroller 2014). These “supplemen­tary” leverage ratios use the Basel III definition of exposure for the base; they thus include not only balance sheet assets but also off-balance-sheet assets, such as derivatives and repo-style transactions (Davis Polk 2014). The en­hanced supplementary leverage ratio of 6 percent for US G-SIBs thus adds a buffer of 3 percentage points to the Basel III requirement of 3 percent.

The main analysis of this study is conducted using the metric of tan­gible common equity capital relative to total assets, a “leverage” capital ratio. However, the comparisons frequently translate this measure into what would be the corresponding average ratio of the same capital concept to risk-weighted assets, as the Basel III requirements are stated principally against risk-weighted assets.^[9] The average relationship used in chapter 4 is that a given ratio of capital to risk-weighted assets will be 1.78 times the corresponding leverage ratio of capital to total assets (based on BCBS 2010a estimates for US and euro area banks). In principle, it would be desirable to have reliable risk weightings.

Goldstein (2017) makes a strong case that risk weighting and the use of internal models have caused risk-weighted assets to be much less reliable than total assets as the basis for determining capital needs. He argues that risk-based capital ratios were less successful in predicting bank failures in the 2007-10 crisis than simple leverage ratios. He notes that for 17 major international banks, the risk-weighted assets/total assets ratio fell from 70 percent in 1993 to 40 percent in 2011, without any evidence of an accom­panying shift toward safer banking practices. He notes the critique in the Liikanen Report (2012) that the banks with the lowest risk-weighted assets/ total assets ratios are those with the largest share of trading assets in total assets. He cites the observation by Hoenig (2013) that the very low 7 percent risk weighting on AAA collateralized debt obligations induced excessive in­vestment in these instruments before the crisis.

One approach to imposing more discipline on risk weightings would be to place a floor on the admissible ratio of risk-weighted to total assets. On average, US and European banks have had a risk-weighted assets/total assets ratio of about 55 percent (see chapter 4). Regulators could impose a minimum of, say, 45 percent of total assets as the regulatory measure of risk-weighted assets in cases where banks' internal models indicated a lower ratio. An alternative approach would be to specify that for regulatory purposes, the risk-weighted assets/total assets ratio for a particular asset category can be no lower than a specified percentage point shortfall from the standard risk weighting for that category for banks not using internal models (say, 10 percentage points).

The “Basel IV” revisions for risk weighting take the latter approach. They would include an “output floor” below which internal models would not be allowed to reduce the risk weight for the asset category in question; discussions among regulators reportedly involve a floor range of 60 to 90 percent of the standard weight. Although regulators have stated that the revisions should not “significantly raise overall capital requirements,” some estimates find that for global banks outside the United States, the conse­quence could be an increase in measured risk-weighted assets of 18 to 30 percent. European banks in particular have opposed this approach, in part because the standardized risk weight for residential mortgages (35 percent) is far above internal model weights in major European countries where de­faults have been rare.^[10] In contrast, US regulators favor relatively high output floors, partly because the Collins Amendment of the Dodd-Frank Act re­quires that risk weights be no lower than the “generally applicable risk-based capital requirements” specified by federal banking agencies (Tahyar 2010).

Basel III falls farther below the optimal capital ratio identified in this study using the leverage ratio than using the ratio of capital to risk-weighted assets. Thus the optimal requirement identified in chapter 4 is 7 to 8 percent for tangible common equity relative to total assets and 12 to 14 percent for the average ratio of capital to risk-weighted assets. In comparison, Basel III sets the minimum leverage target at 3 percent of “exposure,” including secu­rities financing transactions and off-balance-sheet assets.^[11]

A complication for risk weighting may have been created by the advent of quantitative easing, which caused massive amounts of excess bank re­serves to build up at the Federal Reserve. In October 2008 the Federal Reserve began paying interest on excess reserves, which rose from about $2 billion in 2007 to $2.5 trillion in 2014. This structural change helps explain why there was no explosion in prices despite a large increase in the money base. Instead, there was a collapse in the money multiplier from the base to broad money. In contrast, there was only a modest decline in velocity (the ratio of GDP to broad money) (Cline 2015b). It is unclear that increased bank reserves came from corresponding reductions in bank holdings of government securities and other safe assets. If they did not, then the ratio of safe to total assets rose, arguably justifying a lower leverage ratio.

In August 2016 the Bank of England recognized the incongruity posed by much higher excess reserves when it decided to exclude reserves held at the central bank from the calculation of banks' leverage ratios. Although adopted in response to recession fears following the referendum to exit the European Union (Brexit), the decision makes this exclusion permanent.^[12]