Discussion of Three Papers on the Impact of Capital Regulation on Bank Behavior and Economic Performance

Discussion of Three Papers on the Impact of Capital Regulation on Bank Behavior and Economic Performance

BPI’s Chief Economist, Bill Nelson, recently discussed three papers on the economic and financial market impact of bank capital requirements.  He prefaced his remarks with some thoughts on the role of research in regulatory policy design, drawing on his experience at the Fed and also now at BPI.

…When economists reach and publish a conclusion, non-economists give it a lot of weight, perhaps too much weight. It is generally not understood that research in economics, just like in any discipline, should be thought of as a part of a dialogue, and that on any point there will probably be a dozen papers with a dozen different conclusions.

The papers under discussion today are each part of that dialogue.  All of them are about the impact capital regulations, including stress tests, are having on the economy, an issue that has important implications for the design and calibration of not only those requirements, but also of other requirements. As illustrated with the NSFR example above, especially now that the post-crisis regulatory framework is in place, regulations need to be adjusted or recalibrated with a comprehensive perspective.


 

DISCUSSION OF THREE PAPERS ON THE IMPACT OF CAPITAL REGULATION ON BANK BEHAVIOR AND ECONOMIC PERFORMANCE

Remarks at the Federal Reserve Bank of Cleveland and Office of Financial Research Financial Stability Conference

November 30, 2018

Bill Nelson, Bank Policy Institute

 

Thank you for the invitation to discuss these three papers today. For those who don’t know me, I am the chief economist at the Bank Policy Institute, a bank trade group owned by the roughly 50 largest U.S. commercial banks.  BPI is the product of a merger this past summer between The Clearing House Association, where I had also been chief economist, and the Financial Services Roundtable.

Before joining the Clearing House three years ago, I was deputy director of the Monetary Affairs Division at the Federal Reserve Board.  As many of you likely know, after the crisis, the Fed decided that there had been a major supervisory failure and redesigned how it conducted supervision.  Among other changes, it put senior representatives from the research divisions on the oversight bodies of major supervisory efforts.  I was MA’s representative.  I was also heavily involved, both in DC and in Basel, in the design of the Basel III liquidity regulations.

 

THE ROLE OF RESEARCH IN REGULATORY AND SUPERVISORY DESIGN

It would be impossible to provide a comprehensive discussion of three papers in fifteen minutes, and the conference organizers asked me to offer my perspective from my work at the Fed, at the BIS, and now at BPI.  So, while I will offer a few comments on the papers, I will first offer some thoughts on the role of research in regulatory and supervisory design, both from inside a regulatory institution and from outside.

An optimal regulatory design process has three parts. First, the regulation is crafted to meet a specific and clearly specified public policy purpose.  Second, the regulation is calibrated empirically, although often with adjustments to achieve policy objectives, to accomplish that purpose.  And third, after implementation the regulation is adjusted, if needed, in light of experience.

Research plays a critical role in all three parts.  Research informs the debate on what regulations are needed.  Research informs the calibration of the granular components of the regulation and the conclusion about the setting that best balances the social benefits and social costs of the regulation.  And research provides a disciplined and dispassionate assessment of the impact of the regulation once implemented.

Most broadly, research says to regulatory bodies and to the industry “You can’t just make things up.”  While commercial banks would like to assert that capital regulations cause scurvy, the empirical evidence just isn’t there.

In my experience, that discipline is actually much weaker on the regulators than on those of us trying to influence them from the outside, perhaps because their motives are assumed to be purer.  Examples abound, but I will give just two.

BPI takes its reputation very seriously. We recognize that analysis coming from an industry trade is automatically suspect, so we strive hard to ensure that our output is well reasoned and well supported. So in early 2017, when then Fed Chair Janet Yellen called a research note my colleague Francisco Covas had just published “deeply flawed” in Congressional testimony, we were alarmed.

Francisco’s research found that stress tests lead banks to reduce small business lending. Chair Yellen said there were multiple problems with the note, but she only listed one, namely that Francisco’s research used the average quality of banks’ loan portfolios and not the quality of new loans. The next day, we published a blog post explaining why Yellen’s point was incorrect, essentially because banks’ lending decisions are influenced by the required capital expected over the life of the loan, not just at inception.

Over the following year, Francisco’s results were confirmed in research by Acharya, Berger, and Roman (2018), and Cortes, Demyanyk, Li, Loutskina, and Strahn (2018). And Chen, Hanson, and Stein (2017) cite Francisco’s work as providing a possible explanation for the observed decline in small business lending by large banks.

Contrast that with an example of the reverse situation.  When the Net Stable Funding Ratio requirement was proposed by the U.S. banking agencies, the formal request for comment indicated that the rightly famous 2010 LEI study of the costs and benefits of Basel 3 by the Basel Committee found that the benefits of the NSFR in terms of crises averted outweighed the costs in terms of reduced lending.

BPI published a blog post pointing out that the Fed was misreading the key table in the LEI report, that at current capital levels the LEI study found that the costs of the NSFR outweigh the benefits.

When I spoke with people at the Fed about the post, you will probably not be surprised to learn that they did not immediately withdraw the proposed regulation.  Instead, the response was along the lines of “nobody takes those analyses seriously anyway.”

I disagree. I think a lot of people take economic analysis on regulation seriously. When economists reach and publish a conclusion, non-economists give it a lot of weight, perhaps too much weight. It is generally not understood that research in economics, just like in any discipline, should be thought of as a part of a dialogue, and that on any point there will probably be a dozen papers with a dozen different conclusions.

The papers under discussion today are each part of that dialogue.  All of them are about the impact capital regulations, including stress tests, are having on the economy, an issue that has important implications for the design and calibration of not only those requirements, but also of other requirements. As illustrated with the NSFR example above, especially now that the post-crisis regulatory framework is in place, regulations need to be adjusted or recalibrated with a comprehensive perspective.

 

“REPO MARKET FUNCTIONING: THE ROLE OF CAPITAL REGULATION,” ANTONIS KOTIDIS AND NEELTJE VAN HOREN

There has been an ongoing debate, going back nearly a decade now, whether, and if so, how much, leverage ratio requirements reduce financial market liquidity.  In “Repo Market Functioning: The Role of Capital Regulation,” Antonis Kotidis and Neeltje van Horen make an important, and I think dispositive contribution.  As they just discussed, when the U.K. required banks to calculate their leverage ratios on a daily-average basis, banks stopped accepting funds through repo from smaller customers. The result is sensible and robust.  I have no comments, but I do have two questions.

First, I was surprised that they found no evidence that banks also responded by lending less through reverse repos to smaller customers.  Kotidis and van Horen indicate that they expected such a finding because a bank’s balance sheet is not expanded when it conducts a reverse repo – while its cash goes down, its balance sheet increases by the security it receives. By contrast, when a bank engages in a repo, its assets go up by the cash received but the security sold remains on its books, so its balance sheet increases. OK, but I had thought a lot of repo lending and borrowing occurred in a matched repo book, and if a matched repo book declines, both repo and reverse repo decline, and leverage ratios go up.

Second, I would like to know if the customers that had been shunted aside at each month-end in 2016 were the same ones that were dropped altogether in 2017.  More broadly, rather than describe the problem in terms of volume of transactions, a perspective that can be misleading, I would like to know more about what the cross-section of creditors at different points in time looked like, specifically month end and non-month end in 2016 and month end and non-month end in 2017.  Perhaps the customers that stopped investing funds at the bank after 2017 were less dependent on the bank as a reliable and steady place to invest.

 

IN BANK BAILOUTS, BAIL-INS, OR NO REGULATORY INTERVENTION? A DYNAMIC MODEL AND EMPIRICAL TESTS OF OPTIMAL REGULATION, ALLEN BERGER, CHARLES HIMMELBERG, RALUCA ROMAN, AND SERGEY TSYPLAKOV

A critical question for research right now is whether post-crisis regulatory reforms have achieved a key objective:  eliminating (or at least nearly eliminating) the moral hazard arising because of too big to fail.  Not only is this important for determining whether reforms have been successful, but also for whether other GSIB regulations need to be adjusted.

In Bank Bailouts, Bail-ins, or No Regulatory Intervention? A Dynamic Model and Empirical Tests of Optimal Regulation, Allen Berger, Charles Himmelberg, Raluca Roman, and Sergey Tsyplakov provide both theoretical and empirical information on the too big to fail question.

In the first part of the paper, the authors develop a model of banks’ dynamic capital structure under different regulatory regimes: bailout, bail-in and no intervention.  In the second part, the paper tests the implications of the model using regulatory data for the top 50 publicly traded bank holding companies between 2000:Q3 and 2017:Q2 by comparing the capital ratios chosen by GSIBs (treatment group) versus those of non-GSIBs (control group).

The model results have very sensible implications: bail ins are better than bail outs and both are better than a disorderly failure. The model also has the completely plausible implication that banks that go from expecting to being bailed out to no longer expecting to be bailed out would increase their capital ratios relative banks that hadn’t expected to be bailed out in the first place.

In the empirical section, they report as their “main regression result” that GSIBs boosted their capital ratios between 1 and 2.7 percentage points more than non-GSIBs between the pre- and post-crisis period, which they take to be evidence that their model is correct.

Wearing my BPI hat, I would very much like to reach the same conclusion.  But over the pre- and post-crisis period, GSIB’s capital requirements went up by more than the requirements of non-GSIBs. In terms of their point-in-time requirements, GSIBs’ capital requirements now include a GSIB surcharge, which averages 2.7 percentage points.  Moreover, GSIBs’ stress tests are harder because they include the global market shock.  As a result, I’m pretty sure that GSIBs would have increased their capital by more than non-GSIBs even if their expectations about being bailed out had not changed.

Berger et al also present evidence that GSIBs significantly increased the speed with which they adjust their capital ratios up or down toward target levels relative to non-GSIBs between the pre- and post-crisis period.  While those results also corroborate the model’s predictions, to me they seem consistent as well with GSIBs’ actual capital levels now being determined by requirements and stress tests rather than by the banks’ internal assessments of their capital needs.

 

“The Impact of Post-Stress Tests Capital on Bank Lending,” Bill Basset and Jose Berrospide

In “The Impact of Post-Stress Tests Capital on Bank Lending,” Bill Basset and Jose Berrospide address the question of whether increasing capital requirements through stress tests leads to a reduction in lending.  To do so, they define a variable that they call the “capital gap.”

The capital gap is different from the capital surplus — the amount by which a bank’s capital can be reduced before failing a capital requirement or a stress test.  When I calculated the capital surplus in the early 90s as a young economist at the Fed, it was pretty simple. There were essentially two ratios to take into account – the leverage ratio and risk-based ratio. To calculate the capital surplus now, my BPI colleague Francisco Covas looks at 18 point-in-time and stress test requirements to find the one that is closest to binding.

Basset and Berrospide’s “capital gap” is defined as the difference between the lowest CET1 ratio projected by the bank under the severely adverse scenario and the lowest CET1 ratio projected by the Fed under the severely adverse scenario.

Bassett and Berrospide see the capital gap as a measure of the extra capital that a bank must hold because of the stress tests; that is, the difference between the amount the Fed requires the banks to hold to pass the test and the amount the bank would hold if it did not have to pass the stress test.

In a nutshell, Bassett and Berrospide’s central hypothesis is that the capital gap measures how much extra capital the stress tests make banks hold and so can be used to evaluate whether tougher stress tests lead to less lending.  Because the gap does not enter with a significant negative sign in their loan growth regressions, they conclude that tougher stress tests do not reduce lending.

For a couple of reasons, however, I’m not sure their central hypothesis is correct.

First, frequently, banks are not bound by the stress test but rather by their point in time requirements.  For example, by our calculations, in 2017, JP Morgan Chase was most tightly bound by their CET1 requirement under standardized risk weights.  Their common equity exceeded that requirement by $20.6 billion.  By contrast, JPM exceeded its CET1 stress test requirement under the severely adverse scenario by $37.3 billion, and that wasn’t even its most binding stress test requirement.  Its most binding stress test requirement was the supplementary leverage ratio, which it exceeded by $29.1 billion, still almost half again more than it exceeded its point-in-time CET1 requirement.  In so far as JPM in 2017 was being forced by regulation and supervision to hold more capital than it would have otherwise, Bassett and Berrospide’s capital gap had nothing to do with it.

Second, when I asked two bank officials responsible for stress testing and capital allocation at two giant financial institutes considered the best run in the business, they both told me that the models the banks use to conduct their stress test projections are not indicative of how banks view their own capital needs.  One noted that banks have been pressured in the past to project capital needs that are about as stringent as those projected by the Fed, and so they reverse engineer their models to deliver outcomes that resemble the Fed outcome.  The second described for me the stress test he has developed for his bank, which involved thousands of projections under a wide range of different stressful outcomes – from no deal Brexit to severe recession to runaway inflation.

Consequently, I’m not actually sure what Bassett and Berrospide’s capital gap is measuring, and I wasn’t surprised to find that it generally didn’t help explain subsequent loan growth. I encourage them to determine the binding capital requirement or stress test component for each bank/year, and the corresponding capital surplus, to evaluate the impact of stress tests on loan growth.

Thank you.

 

REFERENCES


Acharya, Viral V., Allen Berger and Raluca Roman (2018) “Lending Implications of U.S.  Bank  Stress  Tests: Costs  or  Benefits?” Journal  of  Financial Intermediation, Volume 34, April 2018, pp. 58-90.

 

Basel Committee on Banking Supervision (2010), “An assessment of long-term impact of stronger capital and liquidity requirements,” Bank for International Settlements, August 2010.

 

Berger, Allen N., Charles P. Himmelberg, Raluca A. Roman, Sergey Tsyplakov, “Bank Bailouts, Bail-ins, or No Regulatory Intervention? A Dynamic Model and Empirical Tests of Optimal Regulation,” manuscript, 7 March 2018.

 

Berrospide, Jose, William F. Basset (2018), “The Impact of Post-Stress Tests Capital on Bank Lending,” manuscript, November 2018.

 

Chen, Brian S., Samuel G. Hanson, Jeremy C. Stein (2017), “The Decline of Big-Bank Lending to Small Business: Dynamic Impacts on Local Credit and Labor Markets,” NBER Working Paper No. 23843, September 2017.

 

Cortés, Kristle, Yuliya Demyanyk, Lei Li, Elena Loutskina, Philip E. Strahan (2018), “Stress Tests and Small Business Lending,” NBER Working Paper No. 24365, March 2018.

 

Covas, Francisco (2017), “The Capital Allocation Inherent in the Federal Reserve’s Capital Stress Test,” TCH Research Note, January 2017. https://www.theclearinghouse.org/~/media/TCH/Documents/TCH%20WEEKLY/2017/20170130_WP_Implicit_Risk_Weights_in_CCAR.pdf

 

Covas, Francisco (2017), “Capital Requirements in Supervisory Stress Tests and their Adverse Impact on Small Business Lending,” Staff Working Paper 2017-2, The Clearing House, August 2017. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3071917

 

Covas, Francisco and Bill Nelson (2017), “TCH Responds to Fed’s Criticism of its CCAR Implicit Capital Requirements Research Note,” Blog post, The Clearing House, 16 February 2017. https://bpi.com/tch-responds-to-feds-criticism-of-its-ccar-implicit-capital-requirements-research-note/

 

Covas, Francisco and Bill Nelson (2017), “Foundational Basel Committee Study Estimates the Costs of NSFR Exceed its Benefits by Nearly $1 Trillion,” blog post, The Clearing House, 27 March 2017, https://bpi.com/foundational-basel-committee-study-estimates-the-costs-of-nsfr-exceed-its-benefits-by-nearly-1-trillion.

 

Kotidis, Antonis and Neeltje van Horen (2018), “Repo Market Functioning: The Role of Capital Regulation,” CEPR Discussion Paper No. DPI3090, 6 August 2018.