Conclusion
The proposition that TBTF incentive distortions cause excessive risk taking by large banks has dominated the TLAC policy discussion. Somehow the discussion has discarded the other side of this issue: the proposition that “charter value” poses an incentive for large banks to avoid gambling despite the TBTF incentive.
Contrary to the popular image, the evidence presented in this chapter does not provide resounding proof that TBTF and excessive risk taking by large banks dominated the US financial crisis in the Great Recession. On the disciplinary role of subordinated debt, one study (Nguyen 2013) finds that this role is present in general but absent for large TBTF banks, with the plausible interpretation that investors do not penalize risk taking when they anticipate bailout. Yet the design of TLAC assumes just the opposite: that subordinate debt will indeed discipline risk taking in TBTF banks.
Surprisingly little attention has gone to the measurement of economies of scale at the largest banks, although the issue is central to the social value of imposing penalties designed to offset TBTF distortions. The Davies-Tracey (2014) study finds no economies of scale at social prices, but it suffers from serious methodological questions (the use of a parametric function where nonparametric methods have long been seen as necessary). Moreover, when the authors of another recent study that does find large economies of scale (Wheelock and Willison 2012) rerun the tests using borrowing costs applicable to smaller banks rather than larger banks, they find little change in the scale economies identified, the opposite finding from Davies-Tracey.
The interaction between risk-taking capacity and risk-taking incentive seems especially problematic in arriving at the proper policy inference. Larger banks have greater capacity to diversify, so in principle they are likely to have a comparative advantage in undertaking activities that involve greater risk.
If some risky activities have social value, as seems highly likely, greater observed risk taking will not be unambiguous evidence of incentive distortions generating external diseconomies to the system.On the practicality of TLAC, too little attention has been given to identifying proper holders who will have long-run incentives rather than behave in a herding exodus at the first sign CoCo conversions might be triggered, causing panic dynamics to be amplified.
TLAC has been popular at the political level because it implies a promise that taxpayers will not pay for bailouts in the future. It seems to have become seen as a cheap form of quasi-equity that can play much of the role of outright equity without imposing nearly as much cost on the banks as the real thing. Both of these attractions seem illusory. As long as central banks follow Bagehot's dictum (as they basically did in the Great Recession), taxpayers do not pay for losses, because only solvent banks are supported (the US Treasury made a profit on the TARP support for banks). Subordinated debt cannot be mobilized without default-like circumstances, so it cannot prevent banking panic associated with defaults in the same way that complete avoidance of default by virtue of greater depth of equity capital can.
TLAC thus broadly seems to be at best an inadequate substitute for equity capital. Basel III sets the total target, including equity, at 18 percent of risk-weighted assets. If the appropriate true equity optimum is 13 percent, it could be argued that overproviding TLAC compensates for its imperfect ability to replace equity. Boosting common equity tier 1 capital from the Basel III requirement of 9.5 percent of risk-weighted assets (for G-SIBs) to the equivalent of an optimal level of 13 percent by adding 8.5 percent of risk-weighted assets in the form of CoCos and subordinate debt would imply that the proper regulatory exchange rate between nonequity and equity TLAC is about 50 cents on the dollar ((13 - 9.5)/8.5). Even this favorable an exchange rate, however, would seem to be on the optimistic side.