Summary of Empirical Results

Size of the Modigliani-Miller Offset

The tests in chapter 3 find that for US banks, only 45 percent of the M&M offset attains in practice. As banks increase the share of their equity capital, which has a considerably higher unit cost (say, 10 percent real) than their debt finance (say, 2.5 percent real), the average cost of capital to the banks will rise, because investors will reduce the unit cost of equity capital only enough to offset 45 percent rather than 100 percent of the poten­tial increase in average cost of capital.

Long-Term Costs of Banking Crises

Chapter 4 provides new estimates of the economic cost of banking crises. After taking account of the facts that on the eve of crises economies are frequently at above-potential output and productive capital not created because of the crisis would have had a finite life, it estimates the discounted total costs of a banking crisis at 64 percent of one year's GDP, almost the same as the median estimate in BCBS (2010a). Whereas some might argue that damages are far higher because of hysteresis (a permanent decline in productivity or labor force participation), it can conversely be argued that the damage estimate is overstated because several advanced economies that did not have banking crises in the Great Recession nonetheless experienced major output losses. Moreover, applying what would appear to be a more meaningful historical base period (1977-2015) significantly reduces the central expected probability of a banking crisis otherwise computed if at­tention is instead limited to a shorter period more heavily dominated by the Great Recession (1985-2007), as in BCBS (2010a).

Benefits of Higher Capital Requirements

The benefit of higher bank capital requirements is the reduction in the probability of a banking crisis multiplied by the damage from such a crisis. Chapter 4 adopts the evidence in BCBS (2010a) on the response of crisis probability to the level of capital. This probability shows substantial curva­ture, such that after capital reaches about 7 percent of total assets, there is very little additional reduction in the probability of a crisis (see table 4.2). The calculations apply the ratio of equity capital to total assets. Based on US and euro area bank averages, risk-weighted assets (which for example apply zero weight to OECD sovereign debt) are only about 55 percent of total assets. The corresponding level at which the crisis probability curve turns nearly flat is about 13 percent for capital relative to risk-weighted assets.

Economic Cost of Higher Capital Requirements

The analysis then turns to the impact on the economy on the side of higher cost of capital. Because the M&M offset is not complete, higher capital requirements raise the average cost of capital to banks and thus the rate they charge borrowers, with a likely partial spillover to rates charged for nonbank finance as well. After considering the base average cost of capital to the economy, the proportionate increase in this cost as a consequence of higher capital requirements, and the response of capital formation in a long-term aggregate production function, chapter 4 estimates that an in­crease in the capital requirement by 1 percent of total assets would reduce long-term GDP by 0.15 percentage points. This figure is substantial, because it refers to the full future time path of GDP and thus occurs annually. The impact is larger than implied by typical macroeconomic models, which refer to shorter time horizons and thus do not capture the full effect of reduction in the capital stock and may include offsetting feedback effects (especially those models that build in a monetary policy response).

The Optimal Capital Ratio

The optimal ratio of capital for banks occurs where the slope of the crisis­avoidance benefits curve equals the slope of the upward-sloping line re­lating economic cost to the capital ratio. After considering more than 2,000 combinations of low, central, and high magnitudes for the key parameters in the model, chapter 4 finds that the median result for the optimal capital ratio is 7 percent of total assets and the 75th percentile prudential result is 8 percent. The optimal capital ratio for banks is thus 7 to 8 percent of total assets. This level corresponds to 12 to 14 percent of risk-weighted assets. In comparison, Basel III sets the requirement even for the largest banks at only 9.5 percent of risk-weighted assets. By implication, the Basel III target for capital should be increased by 25 to 45 percent. Although this increase would be substantial, it is far less than the increase needed to reach the capital requirement of 20 to 30 percent of total assets called for by Admati and Hellwig (2013)—more than three times as high as the optimal range identified in chapter 4.

Implications for Basel Targets

The optimal range of bank capital estimated in chapter 4 is the same as the median across 15 other studies reviewed in chapter 2. Those studies show a median of 13 percent as the optimal ratio of common equity tier 1 capital to risk-weighted assets (table 2.1), the midpoint of the 12-14 percent range identified in chapter 4. Even if one accepts this benchmark as the desirable target, however, two key issues of implementation arise.

Minimum or Behavioral Norm?

The first issue is whether to think of the optimal target as a minimum that banks should maintain or as an actual level that they tend to maintain in practice that exceeds a lower required minimum. At end-2014, 31 large bank holding companies in the United States representing more than 80 percent of total bank assets held common equity tier 1 capital amounting to 11.9 percent of risk-weighted assets (Federal Reserve 2015a), even though the Basel III minimum for G-SIBs is only 9.5 percent.

How to treat the behavioral capital cushion has been a source of ambi­guity in this literature. Thus, as discussed in chapter 2, a major reason why two OECD researchers (Slovik and Cournede 2011) conclude that adverse growth effects of higher capital requirements would be considerably larger than calculated by the Macroeconomic Assessment Group chaired by the Bank for International Settlements (MAG 2010) is that they assume that banks would maintain a large cushion of about 3.5 percentage points above the Basel III target.

One way to think about the cushion is that it is a voluntary counter­cyclical component that could be expected to melt away under economic conditions less favorable than business as usual. Another way to think about this issue is to infer that the behavioral cushion implies that the system would be safe enough if the target were set at the lower end of the optimal range (12 percent of risk-weighted assets) rather than the higher end (14 percent). By implication, if the Basel III minimum for global systemically important banks (G-SIBs) were raised from 9.5 to 12 percent, large banks could be expected to hold somewhat more than 12 percent in normal years, perhaps placing them at the 14 percent higher end of the optimal range in practice. It would be difficult to argue that such a consequence would be undesirable, especially to the extent that one would expect some erosion of the cushion under more unfavorable conditions. In short, the “behavioral cushion” consideration does not fundamentally change the estimates of chapter 4, but it suggests that the lower end rather than the upper end of the optimal range might reasonably be pursued in fine-tuning Basel III.

Equity or Alternative Total Loss-Absorbing Capacity?

A closely related question is whether to count nonequity total loss-ab­sorbing capacity—namely, contingent convertible and subordinated debt— as effectively meeting part of the capital requirement.

Dealing with the Risk-Weighting Problem

During the Great Recession, it became evident that measured capital ad­equacy could become dangerously misleading as a consequence of the risk-weighting process, including through the use of internal models—as vividly illustrated by the case of Lehman Brothers. With the average ratio of risk-weighted to total assets at about 55 percent for US and European banks (BCBS 2010a), it would seem desirable to give serious consideration to setting a minimum acceptable ratio of risk-weighted to total assets of, say, 45 percent. If a bank's ratio of risk-weighted assets to total assets fell below this threshold, for purposes of capital adequacy, its “adjusted” risk- weighted assets would simply be set at 45 percent of total assets. The overall effect of risk weighting would thus not be allowed to shrink risk-weighted assets to less than about four-fifths the amount that would be expected from the usual average.

In early 2016 the Basel Committee on Banking Supervision released a consultative document on revisions to Basel III rules that would reduce the variation in credit risk weights resulting from the application of internal risk models (BCBS 2016b). One approach the document mentioned was an “aggregate output floor” set at 60 to 90 percent of the risk weighting that would result from standardized approaches rather than internal models.

In the United States, the use of internal models for risk weighting has already been severely constrained by the Collins Amendment to the Dodd- Frank legislation, which provides that “generally applicable” risk-based capital requirements serve as a floor for banks' regulatory capital. US regu­lators have indicated that in implementing Basel III capital rules, capital requirements calculated under the standardized approach will be used to implement the Collins Amendment (PwC 2016). Higher floors for risk weights from internal models would thus primarily affect European and perhaps other non-US banks.^[250]

A complementary way to deal with the risk-weighting problem would be to raise the Basel III leverage ratio to a more meaningful level and give it equal weight with the risk-weighted ratio in assessing capital adequacy. However, care would need to be given in setting the mapping from the ratio of capital to total assets used in this study to the Basel III leverage ratio, because the latter uses a denominator that is different from (and at least for US banks, significantly larger than) total assets. The Basel ratio is against “exposure,” which includes not only assets but also securities financing transactions, security lending and borrowing, and margin lending agree­ments (BCBS 2014a).^[251]

Large Banks versus Small Banks

The findings of this study indicate a need for higher bank capital require­ments than under existing Basel III rules. However, the analysis is pre­mised on the resulting economywide benefit from the reduced incidence of banking sector crises. Because larger banks pose the greatest systemic risks, there is a case for leaving the Basel III benchmarks unchanged for smaller banks. Smaller banks that get into trouble are much easier for regulators to place into bankruptcy, resolution, or merger, with minimal stress on the system.

Small banks constitute only a small share of banking sector assets. In the United States in 2015, for example, there were 5,309 banks (FRED 2016). Eighteen bank holding companies or banks had assets of more than $100 billion, and 3,529 banks had assets of less than $300 million (Federal Reserve 2016a). The largest banks accounted for 68 percent of total banking sector assets broadly defined;^[252] the top 50 banks accounted for 74 percent of total assets. In practice, leaving the Basel III requirements at their current level for “small” banks would affect only a modest portion of the industry by asset volume (albeit a large majority by number of institutions).

Other Key Policy Issues

Lender of Last Resort

As discussed in chapter 5, in the United States the Dodd-Frank Act of 2010 weakened the lender-of-last-resort capacity of the Federal Reserve by con­straining the use of section 13(3) to programs with wide eligibility (and re­quiring approval by the secretary of Treasury). The Fed has indicated that it would consider a program with at least five eligible institutions to meet that criterion (Scott 2016, 93). In a crisis, it would be important that the interpretation not conflate eligibility with actual inclusion, as a crisis could start with a single institution.

An important way to restore lender-of-last-resort capacity would be to change the rules such that the nonbank subsidiaries of the large bank holding companies supervised by the Federal Reserve would be given access to the discount window that is currently limited to the depository banks within those holding companies. Without such a change, a large nonbank subsidiary could cause problems for the holding company because of the difficulty of obtaining lender-of-last-resort support through the mecha­nism of the now-constrained section 13(3).

At a broader level, it is important that in the aftermath of the Great Recession the principle of forceful lender-of-last-resort support of solvent fi­nancial institutions in a panic not be lost as a consequence of political back­lash against the financial sector. Avoiding taxpayer losses to rescue banks is a proper objective, but by definition lender-of-last-resort lending should not lose money, because it is provided only to solvent banks. Support to the banks in the Troubled Asset Relief Program (TARP) yielded a profit for US taxpayers, not a loss. A more severe financial crisis and recession caused by reluctance to provide appropriate lender-of-last-resort support would cost the taxpayers much more, because of higher unemployment payments and lower general revenue in the more severe recession.

An important issue is the related question of lender-of-last-resort ac­cess for large nonbank institutions that are not subsidiaries of bank holding companies, which the tightening of section 13(3) could unduly constrain. The Dodd-Frank Act requires that the Federal Reserve provide supervision for systemically important financial institutions, which include not only the large US bank holding companies but also insurance companies AIG and Prudential and the US operations of Barclays, Credit Suisse, Deutsche Bank, and UBS (Federal Reserve 2016b). The simplest way to ensure lender-of-last- resort availability for these systemically important nonbanks (and foreign banks) would be to clarify that they would qualify for a program under sec­tion 13(3). An even more straightforward means to the same end would be to provide that, because such firms are under Federal Reserve supervision, they have access to the discount window in the face of a liquidity crisis.