Overview
Table 2.1 summarizes the optimal capital ratios identified in the studies surveyed here (including several dynamic stochastic general equilibrium models, described in appendix 2A, and one vector error correction study).
The ratios refer to common equity tier 1 relative to risk-weighted assets; in some cases the table entries are accordingly adjusted from the studies' main reported estimates (as indicated in the notes).[43] It is evident in the table that the study by Admati and Hellwig (2013) is an outlier, with an optimal range far higher than any other of the main estimates. As it turns out, my own estimates (presented in chapter 4) are centered at the median for the full set of estimates: an optimal ratio of 13 percent for common equity tier 1 relative to risk-weighted assets (corresponding to the 7 to 8 percent estimate in chapter 4 for the median and 75th percentile optimal ratios of common equity tier 1 to total assets).Table 2.1 Literature estimates of optimal ratio of capitala to risk- weighted assets (percent)
| Study | Method | Optimal ratio | Notes |
| Admati and Hellwig (2013) | Hist-MM | 36-53 | b |
| Hanson, Kashyap, and Stein (2010) (US) | SW | 27 | c |
| Goldstein (2017) (US) | SW | 18-32 | d |
| Dagher, Dell'Ariccia, Laeven, Ratnovski, and Tong (2016) | SW | 9-17 | e |
| Basel Committee on Banking Supervision (BCBS 2010a) | COM | 13 | f |
| Barrell, Davis, Fic, Holland, Kirby, and Liadze (2009) (UK) | COM | 9 | g |
| Miles, Yang, and Marcheggiano (2012) | CM | 16-20 | h |
| Kato, Kobayashi, and Saita (2010) | COM | 11-14 | i |
| Gambacorta (2011) (US) | VECM | 12 | j |
| Yan, Hall, and Turner (2011) (UK) | COM | 10 | k |
| Kragh-Sorensen (2012) (Norway) | COM | 16-23 | l |
| de-Ramon, Iscenko, Osborne, Straughan, and Andrews (2015) (UK) | COM | 10 | m |
| Van den Heuvel (2008) | DSGE | < 5 | n |
| Clerc et al. (2015) | DSGE | 10.5 | o |
| Mendicino, Nikolov, Suarez, and Supera (2015) (euro area) | DSGE | 10 | p |
| Minneapolis Plan (2016) (US) | COM | 23.5 | q |
| COM | 12-14 | r | |
| Median | 13 |
Hist-MM = historical, strong Modigliani & Miller assumption; SW = seawall; COM = calibrated optimization model; CM = calibrated model; VECM: vector error correction method;
DSGE = dynamic stochastic general equilibrium model
a. Common equity tier 1.
b. 20 to 30 percent of total assets.
c. 15 percent of total assets.
d. 10 to 18 percent of total assets.
e. 15 to 23 percent including intangible equity, subordinated debt, and cyclical peak buffer.
f. See BCBS (2016a) for identification of optimal 13 percent within BCBS (2010a) range.
g. 3 percentage points above 2007 actual level. Study treats risk-weighted and total assets as equivalent. Base for risk-weighted actual set at 5.7 percent (Cohen 2013).
h. Authors do not calculate optimal level but identify preferred range.
i. 6 to 8 percent of total assets.
j. Uses quarterly data for the United States.
k. UK basis.
l. Norway only.
m. Based on maximum net benefit at 4 percentage point increase in capital adequacy (p. 59), combined with 5.7 percent actual base in 2009 (Cohen 2013).
n. Study completed before the Great Recession.
o. Benefits from curbing excessive risk taking, eventually swamped by costs of reduced capital formation. No crisis costs incorporated.
p. Study finds 13.5 percent optimal but including 3.5 percent in capital other than tier 1 common equity.
q. For too big to fail G-SIBs: 38 percent, but not calculated using optimization.
r. 7 to 8 percent of total assets.
Although the studies are treated equally in arriving at the median estimate, the dynamic stochastic general equilibrium estimates are less satisfactory than most of the calibrated optimization models. (Indeed, one of these models completed shortly before the Great Recession concluded that bank capital requirements were already too high.) None of these studies contains an explicit modeling of the probability of a banking crisis or an estimation of the economic cost associated with such a crisis, although such impacts are at the heart of the policy debate about reforming the banking sector.
The median optimal capital ratio of 13 percent is substantially above the Basel III ratio of 7 percent (4.5 percent minimum common equity plus
2.5 percent capital conservation buffer); it is also well above the corresponding ratio for G-SIBs (which adds another 2.5 percent for a total of
9.5 percent). The broad implication is that the Basel III requirements fall short of the optimal levels for capital. The principal caveat is that Basel III also requires TLAC of 18 percent of risk-weighted assets for the largest banks, nearly doubling the cushion provided by common equity capital alone. If the securities included in TLAC (equity plus CoCos plus eligible subordinated debt) were all of equal quality for purposes of safety of the banking sector, one might conclude instead that Basel III had already gone more than far enough to ensure the stability of the system. However, the discussion in chapter 5 suggests that TLAC is not a fully adequate substitute for equity capital.