Salient Themes in the Literature
Chapter 2 frames the issues and examines the state of analytical work in a critical review of the literature. It surveys four categories of studies: “heuristic seawall” studies, calibrated studies of transitional costs, calibrated long-term cost-benefit studies, and a variant of calibrated studies using dynamic stochastic general equilibrium models.[4]
The analysis of Admati and Hellwig (2013) is prominent in the first group.
It is based on the observation ofhigher historical levels ofbank capital; the argument that banking is no different from other corporate sectors, where much higher equity capital is common; a strong M&M assumption; and the argument that analysis should be based on social costs, correcting for the distortion of tax deductibility of debt interest but not equity earnings. I argue that banking is different, if only because more than half of its production activity is inherently financed by debt in the form of deposits and that it would not be optimal to apply social pricing to one sector while leaving the financing cost distortion unchanged in the rest of the economy.An important study in the first group is an IMF-based analysis by Dagher et al. (2016). It uses a seawall approach to examine what levels of capital would have been required to cover 85 percent of bank losses in past banking crises based on nonperforming loan data and alternative assumptions about the loss incidence of nonperforming loans.
Two studies stand out in the second set ofstudies, on transitional effects. Based on a large number of macroeconomic models, the Macroeconomic Assessment Group (MAG 2010) estimates that the phase-in of Basel III requirements would temporarily reduce output by a maximum of about 0.25 percentage point from baseline after three years. As noted, the IIF (2011) instead estimated an output loss more than an order of magnitude larger.
Writing in late 2014, Cecchetti (2014) pronounced that “the jury is in” and that the effects had been even milder than anticipated by the Macroeconomic Assessment Group. The exceptionally low interest rate environment in this period may cloud this diagnosis, however. Cohen (2013) provides support for the moderate or low-cost diagnosis. He finds that banks adjusted mainly by building up retained earnings, thereby avoiding the most severe adverse effects from either sharp reductions in outstanding claims or forced recourse to high-cost share issuance under adverse conditions.In the third group of studies, the preeminent calibrated long-term study remains the Long-Term Economic Impact (LEI) analysis of the Basel Committee on Banking Supervision (BCBS 2010a). Placing the median long-term cost of a banking crisis at 63 percent of one year's GDP and considering the extent of reduction in probability of a crisis at successive levels of bank capital, the study shows a wide range over which higher tangible common equity capital would be socially beneficial, with the optimal level at 13 percent of risk-weighted assets.
An important study by Miles, Yang, and Marcheggiano (2012) places the desirable range at 16 to 20 percent of risk-weighted assets. The median estimate across 16 leading studies (including those in the seawall and dynamic stochastic general equilibrium groupings) places the optimal capital ratio at 13 percent for common equity tier 1 capital relative to risk-weighted assets (see table 2.1 in chapter 2).
My estimates (in chapter 4) are also at this level: 12 to 14 percent of risk- weighted assets (7 to 8 percent of total assets). This outcome is the same as that of the BCBS (2010a), even though I allow for a considerable (45 percent) M&M offset. The production function approach of Miles, Yang, and Marcheggiano (2012), which I use, generates a substantially larger adverse impact on output from a given rise in average borrowing cost than does the compilation of macroeconomic models. Those models may inappropriately incorporate monetary policy feedbacks and may not be designed to capture the effects of permanent long-term reductions in available capital stock as opposed to short-term fluctuations in demand.
The last group of studies, the dynamic stochastic general equilibrium models, provides methodologically elegant but ultimately less credible estimates of optimal capital requirements. One study concludes that capital requirements were already too high before the Great Recession. None of the main dynamic stochastic general equilibrium models incorporates the centerpieces of this issue: the response of the probability of a banking crisis to capital levels and the economic cost of a banking crisis. Instead, they are driven by such influences as an initial phase of benefit from reducing excessive risk taking as capital rises followed by a phase of cost to output from reduced capital stock.