Stress tests in the wake of the 2008 financial crisis helped restore public confidence in the banking system by showing that banks had enough loss-absorbing capacity to weather an even more serious economic decline, noted Wharton finance professor Richard Herring in an interview with Knowledge at Wharton.
Since that time, stress tests have become a regular feature of regulatory oversight of the banking system to help officials assess the resilience of the financial system and guide capital planning by banks. The stress tests should be strengthened to capture the second-round effects of shocks, address a wider range of risks, and incorporate feedback effects within the banking system and between the real economy and banks, Herring said. He also called for the extension of stress tests to a wider range of institutions such as some shadow banks, the harmonization of stress tests globally, the improvement of data and market intelligence to monitor emerging risk, and robust safeguards to ensure that scenarios are not manipulated to suit political objectives. “Stress testing … may, in fact, be one of the key measures that has prevented the economic damage from the pandemic from being compounded by a banking crisis,” he added.
These concerns are the focus of a December 2019 paper titled “Objectives and Challenges for Stress Testing,” which Herring co-authored with Til Schuermann, a partner and co-head of the risk and public policy practice in the Americas at Oliver Wyman, a consulting firm. It discusses a range of challenges to improving the effectiveness of stress tests, such as “incorporating nonfinancial risks like cyber [threats], taking into account second-round effects of shocks, broadening the scope beyond just banks, and resisting a tendency to disaster myopia (or the tendency to underestimate adverse outcomes) as memories of the financial crisis recede into the past.” Herring shared his insights on tackling some of those challenges. An edited transcript of the conversation follows.
Knowledge at Wharton: Before we talk about your paper, perhaps we could spend some time on what’s going on at present. In response to the coronavirus pandemic, the Federal Reserve has cut interest rates close to zero, and it has provided enormous liquidity support to a wide range of markets. What is your assessment of the crisis that the coronavirus pandemic has sparked? What do you think of the Federal Reserve’s actions?
Richard Herring: I think the intense focus on the role of the Fed in these circumstances is fundamentally misplaced. At best, the Fed can play a supporting but significant role to ensure that the macroeconomic shock caused by the pandemic and the policy response to combat it is not compounded by a financial crisis. The first-order concern should be to support and improve the healthcare infrastructure and provide support for people who are suddenly without income. The Fed cannot stop the pandemic, but it can provide enough liquidity to keep the financial system functioning. One of the key challenges ahead for the financial system will be coping with the credit risks that inevitably materialize because much of the economy has come to a full stop. If firms don’t have revenue, they won’t be able to meet their commitments. They will be under enormous pressure to reduce costs and, sadly, that will involve terminating much of their workforce. Of course, this response will exacerbate economic conditions.
Economic conditions are made still worse by the panicky responses among consumers who are trying to stock up on everything from canned goods to toilet paper. This behavior can create shortages where they need not exist. We have seen parallel behavior in financial markets with the sudden dash to cash, that roiled markets. Luckily the Fed can remedy that problem by supplying more cash, but it is powerless to deal with shortages in the real economy. Hoarding reflects a loss of confidence in the system and a worrisome unraveling of society.
We would have been in much better shape if the responsible officials had taken appropriate precautions in response to the many simulations that have been conducted showing the potentially devastating consequences of a pandemic. For example, in 2005 the Wharton Financial Institutions Center co-sponsored a conference with the Brookings Institution and The Goldman Sachs Global Markets Institute to identify the top ten financial risks to the global economy. These risks included a global pandemic and featured a simulation showing how much more rapidly the Spanish flu of 1918 would have traveled and how much more devastating it would be because of modern international air travel. Although governments and international organizations have conducted these kinds of stress tests, there has been no corresponding obligation to take appropriate measures to protect against these adverse scenarios.
In contrast, banks have been required to raise capital to absorb losses in response to severely adverse economic scenarios. Unfortunately, the responsible officials do not appear to have subjected themselves to comparable disciplines such as stockpiling enough protective gear for healthcare workers and first responders, necessary life-support equipment and investing in tests to detect infection and vaccines to protect. These measures might have enabled leaders to put in place a more effective response to the advance of the virus without shutting down much of the economy. But that’s not where we are now. Policymakers completely underestimated the severity and speed of contagion, and we are ill-prepared to deal with a pandemic.
Policymakers completely underestimated the severity and speed of contagion, and we are ill-prepared to deal with a pandemic.
Given where we are now, what can the Fed do? It’s important to keep markets functioning. We’re going to see volatility because nobody can predict with certainty what the longer-term impacts of this epidemic will be and so the price discovery process will likely remain a source of volatility. People really don’t understand the longer-term effects of the measures we’re taking to try to stem the panic, and we don’t know how long the healthcare crisis will last. If it’s just a couple of months, we should be able to bounce back relatively quickly. If it ends up lasting much longer, however, recovery will be much more difficult.
Thankfully Congress and the Administration have taken some bold actions on the fiscal side, because the blow to income is where the most painful economic impacts are being felt. If we don’t act to help people who have suddenly lost their income the human tragedy will be appalling.
But to return to the Fed, I would give them high marks for their quick action to avert a widespread liquidity crisis. They clearly learned many lessons from 2008-2009 and acted quickly to put in place facilities to support liquidity in several important markets. They also deserve praise for rapidly activating swap lines with other major central banks in recognition of the fact this is not just a U.S. problem. The pandemic is worldwide, and an international financial crisis would cause an even larger decline in world demand and production than we now face. Although it’s difficult to achieve in the current climate, global cooperation is more important than ever. Inadequate cooperation may cause an even larger collapse and much slower recovery.
I am less enthusiastic about the Fed’s dramatic cut in interest rates. It strikes me that lowering already historically low interest rates is purely symbolic and undermines the power of a tool they may need later. It’s unrealistic to expect that investors or consumers will respond to a decline in the level of the interest rate under current conditions in which the problem is a government-directed shut down in economic activity. When we emerge from the current period of life support to begin an economic recovery a cut in interest rates might be very helpful. But if interest rates are already near zero, there’s no scope for deploying this traditional tool of monetary policy without reducing interest rates to negative levels. Having witnessed the virtual impotence of negative interest rates in other economies, the Fed seems determined to resist that course of action.
It’s important to ensure temporary aid does not become a permanent subsidy.
Knowledge at Wharton: Since you mentioned credit risk being a critical issue, what do you think should be done about that?
Herring: It depends on the nature of the problem. If it is a fundamentally sound business that we can expect to bounce back, we surely don’t want to take it through bankruptcy. That’s costly. It can destroy the value of organizational, physical and human capital.
Allocating this kind of support can be very tricky because it requires judgments about long-term viability, which are somewhat subjective. Ideally, we should offer support to any firm that is temporarily in distress. That is a rationale for helping, for example, the airline industry or hundreds of thousands of small restaurants. But it’s difficult to administer, and it’s important to ensure temporary aid does not become a permanent subsidy. In the beginning, however, it seems wise to err on the side of assuming we’re facing temporary liquidity problems until we fully understand how the economic shock is going to play out.
The Fed lacks the authority to take credit risk, but it can establish facilities underwritten by the Treasury than can do so. It resurrected several such facilities that had been deployed during the 2008-2009 crisis before passage of the CARES Act, which has greatly expanded Treasury resources for funding additional such facilities.
Knowledge at Wharton: Markets have continued to be volatile despite the Fed’s actions. Is there anything else that can be done to calm things down for investors?
Herring: We should not assume that the Fed can control expectations. In each recent case in which it has taken dramatic actions, the impact on expectations has been at best modest and transitory. The first dramatic cut in interest rates (on March 3), worked for about 14 minutes, and the bold actions taken on Sunday March 15, did not prevent markets from plunging. (The Dow fell nearly 3,000 points on March 16, the second-worst day in its history.) There is also a risk that people may take these heroic actions as a sign of panic. But more fundamentally, stabilizing expectations in these circumstances depends on limiting the spread of the virus and enabling the resumption of normal activity. This is first and foremost a healthcare crisis, which the Fed is powerless to influence.
We are in for a prolonged period of uncertainty because the market is trying to figure out what the longer-term value of various firms will be. The longer the pandemic persists, the greater the likelihood of changes in the structure of the economy. For example, we may find that many tasks can be performed more efficiently at home without wasting the time and resources of commuting. This could reduce the demand for office space and lead to different choices about where people choose to live. Although complex international supply chains and just-in-time inventory practices have made substantial contributions to reducing production costs, countries will inevitably question whether the efficiency gains justify the vulnerabilities exposed by this pandemic. It seems possible that we may experience a disintegration of the world economy that will prove costly to all countries.
The longer the pandemic persists, the greater the likelihood of changes in the structure of the economy.
We can see evidence of this response in the nationalistic actions of many countries that are trying to limit the export of anything needed to fight the pandemic. It’s entirely understandable, but it means that you can’t rely on these sources of supply when you need them most. I suspect we will increasingly see national defense arguments advanced for protecting manufactures of inputs into things like pharmaceuticals and medical equipment. Moreover, once that sort of argument gains traction, we can expect a growing number of industries to claim that they are vital to national defense. After all, the U.S. once protected makers of ice cream under the rationale that it is vital to the morale of our troops.
Knowledge at Wharton: Some commentators have said that the coronavirus crisis is Dodd-Frank’s first real test. Do you agree with that view? (The Dodd–Frank Wall Street Reform and Consumer Protection Act was passed in 2010 as a response to the 2008 financial crisis.)
Herring: If you take the broad view that Dodd-Frank was supposed to increase the resilience of the banking sector, it seems to have served us well. We have implemented a number of policy reforms based on lessons drawn from 2008-2009 crisis – increases in the quantity and quality of capital, implementation of a leverage ratio, introduction of liquidity requirements and living wills and, of course, stress test to ensure that banks are prepared to survive a severely adverse economic scenario.
That gets us around to the topic of the paper (“Objectives and Challenges for Stress Testing”), because one of the things that we learned from the crisis was that the capital adequacy of banks shouldn’t be judged on the basis of current economic conditions alone. What we really care about is whether the banking system can continue functioning effectively even under a severe shock to the economy.
We discovered in the 2008-2009 crisis that as a result of ill-considered moves by the regulators and some very aggressive actions taken by banks, some international banks were operating with leverage above 50-to-1 (assets to equity). This assumes an incredibly optimistic view about the growth and stability of the world economy. If a bank’s assets should decline in value by a mere 2% it would be insolvent. Realistically, even a loss of 1% would be likely to bring the bank down because it would lose access to liquidity and be forced to incur losses on the forced sale of assets.
What we faced in 2008-2009 was a collapse of confidence in the plumbing of the world financial system. In recognition of the central role banks played in creating vulnerability to a financial crisis and in transmitting and amplifying the shock to the real economy, the Group of 20 (G20) — the political leaders of the 20 most economically important countries — placed reform of bank regulation at the top of the agenda. These reforms have been broad and global in reach and included, among many other measures, a requirement to institute stress tests.
The capital adequacy of banks shouldn’t be judged on the basis of current economic conditions alone. What we really care about is whether the banking system can continue functioning effectively even under a severe shock to the economy.
Obviously, stress scenarios developed for the banking system did not include a global pandemic, but this does highlight an important issue. Stress tests should include scenarios that go beyond historical experience. A fundamental limitation of most statistical analysis is that it’s based on what has happened in the past. But the future is not bounded by experience. We may find ourselves confronted by a shock that has never occurred before — an extreme tail event. Such events do not happen often, but unfortunately we’re living through one now.
The stress tests certainly didn’t anticipate the pandemic, but it did include a severe recession. Banks were required to include a substantial amount of equity in their financial structures that would enable them to remain well-capitalized, even if a very damaging recession occurred. This is happening now but not because of a contemplated, conventional macro shock. This highlights an important point: When we don’t know the future, the main source of resiliency in the banking system is its ability to absorb loss. We’re very fortunate that the U.S. banking industry seems to be quite robust entering this crisis. Unfortunately, this is less true in the European Union.
I should note that it is important not to overstate what stress tests can do. Stress tests are designed to ensure that banks do not have to reduce lending to remain well capitalized. But that doesn’t imply that banks will choose to lend. Most U.S. banks are comfortably above their required regulatory minimums, but we don’t know how they will choose to use this financial flexibility.
We can, however, be confident that this time the banking system will not make the problem worse, which is a nontrivial accomplishment. During the crisis in 2008-2009 many banks were on the brink of insolvency (and, frankly, some had gone over the brink) and so they were forced to retrench. Very few were in shape to support an economic recovery thus contributing to the length and depth of the recession. Of course, banks will not lend to firms with no revenue, but parts of the CARES Act may provide guarantees that will enable banks to participate in current life support measures and aid the eventual recovery.
Knowledge at Wharton: You write in your paper that stress testing was an effective crisis-fighting tool that banks used during the great financial crisis 10 years ago. Do you think it’s still the right tool today, or do we need other tools to deal with the current crisis?
Herring: My co-author Til Schuermann distinguished “peace time” from “war time” stress tests. Peace time stress tests are conducted during normal economic conditions when the attempt is to anticipate shocks that might occur. War time stress tests are conducted during a crisis. This is the circumstance in which system-wide stress testing began. It was an attempt to restore confidence in banks and regulators at a time when the public had lost confidence in both. The regulatory ratios were shown to be entirely misleading. The regulators had failed to keep banks safe and had no coherent plan for resolving insolvent institutions. They needed a way to reassure the public that they now understood the dimensions of the problem and how to fix it. This was the origin of the first stress test.
In wartime, it’s relatively easy to design a stress test because the relevant stress is all too obvious. In peacetime, however, the challenge is much more difficult because no one can know what stress is relevant. It’s certainly possible, indeed likely, that reasonable people may disagree about what should be in a stress scenario and how seriously adverse it should be. And since the banking industry knows the more adverse the stress in a scenario, the more capital they will need, they will be very critical about severely adverse stress scenarios.
Some politicians have proposed subjecting stress scenarios to a period of public comment, which I think is a terrible idea. The objective should not be to arrive at some sort of consensus forecast, but rather to probe whether banks will be resilient, even if something quite unexpected and exceptional happens. These kinds of pressures to make stress tests more predictable indicate some of the challenges regulators face in maintaining the integrity of stress tests during peace time.
One obstacle to designing more flexible stress tests, however, is that they are very expensive. Collectively, banks have spent billions in developing the infrastructure to be able to conduct these tests. The Fed has made huge expenditures as well. Given these substantial sunk costs, there’s an understandable reluctance to change the models much. Fundamentally, you’re trying to see what’s likely to happen to bank income statements and balance sheets, given a certain kind of scenario. The scenarios focus on unemployment, interest rate and inflation shocks, which are certainly important. But they are not the only things we need to worry about.
The next crisis may have nothing to do with any of that. It may be a pandemic, a climate shock, a cyber-attack or a natural disaster of some sort. And, because of the huge investments in the current stress-testing infrastructure, there’s resistance to trying to evaluate other shocks that don’t neatly fit the models we have in place. Banks also have an innate reluctance to introduce new kinds of stress scenarios also because they suspect one consequence may be a requirement to accumulate more capital.
This is a serious problem in the longer term, but perhaps the fact that we are now experiencing a severe, unanticipated shock will dampen some of the pressures for reducing the severity of the adverse scenarios. Nonetheless, it’s a simple fact of human behavior that the longer the stretch of good outcomes we’ve experienced, the less our concern about bad outcomes. We’ve just experienced an abrupt end to one of the longest economic expansions in our history. This undoubtedly contributed to the record highs in equity markets we enjoyed earlier this year. Suddenly, however, we find ourselves in a very different world. We tend to forget how suddenly these shifts can happen.
It’s a simple fact of human behavior that the longer the stretch of good outcomes we’ve experienced, the less our concern about bad outcomes.
Knowledge at Wharton: Thinking back about when you first started working on the paper, what were some of the main questions you were trying to answer? What were some of the key findings from your research?
Herring: This paper was part of a large project to produce a Handbook of Financial Stress Testing that is being published by Cambridge University Press. Our assignment was to think about the rationale for stress testing, how it has evolved, and to identify some of the key challenges looking ahead.
We began by distinguishing different objectives for conducting stress tests. The idea of conducting stress tests is not particularly novel. We’ve seen them deployed in engineering and in medicine, for example. And, well before the 2008-2009 financial crisis, many financial institutions were stress testing trading positions and portfolios to anticipate how much they might lose under a variety of circumstances.
What was new about the kind of stress testing that emerged in response to the crisis was the attempt to look at all major banks, applying the same stresses to each institution at the same time. This permitted the authorities and the public to gain a perspective on the resilience of the banking system as a whole. The first stress test applied to 19 bank holding companies that represented about two-thirds of bank lending and to determine whether they were prepared to deal with a substantial worsening of economic conditions and still have sufficient capital to meet their regulatory capital requirements.
The results were not encouraging. Ten of the 19 institutions failed the stress test, a result the U.S. authorities disclosed to the public. This had the benefit of assuring the public that the stress test was rigorous and conducted with integrity. The obvious downside risk of identifying the banks that failed the test was limited by the availability of standby public funding, which assured anxious creditors that the failing banks would be recapitalized.
When the European Union tried to conduct comparable tests, it lacked the fiscal resources to recapitalize banks that failed the stress test. It seems likely that officials in Europe tried to finesse this problem by reducing the rigor of the stress tests so that nearly all banks passed. While this temporarily bolstered confidence, the widespread bank failures that occurred soon thereafter undermined confidence and raised troubling questions about the competence of the authorities. If the stress test isn’t credible, there’s no point in doing it.
When we don’t know the future, the main source of resiliency in the banking system is its ability to absorb loss.
The success of the wartime stress test in the U.S. encouraged Congress to include stress tests in the Dodd-Frank reform package as a way of ensuring that the banking system would remain healthy and resilient. It was a huge advance over traditional regulatory tools used to monitor the safety and soundness of banks. Before the adoption of stress tests regulators relied primarily on ratios based on accounting data, which at best provided information about what had happened in the past, but they conveyed almost no useful information about the future. Yet what we need to know is whether the banking system will be resilient even if bad things happen in the future. The introduction of stress testing reframed the way the safety and soundness questions were posed. It was no longer, “Is your bank currently well capitalized?” It became, “Is your bank sufficiently well-capitalized to deal with the kinds of stresses that might happen in the future?”
Significance of Macro Stress Testing
Stress testing serves two different objectives. One is focused on evaluating the safety and soundness of individual institutions. This corresponds to the traditional microprudential objective of bank examiners but is a much more transparent approach to prudential supervision. The other objective is termed macroprudential. It’s an attempt to evaluate the resilience of the banking system and ensure it can serve the real economy in a severely adverse downturn in activity.
These objectives are similar but not identical. Some measures taken to strengthen the safety and soundness of individual institutions may inadvertently undermine macroeconomic stability. For example, many prior efforts to reform capital regulation have aimed to make capital requirements more risk sensitive. To the extent these reforms succeed, banks will be obliged to increase their ability to absorb loss when the economy enters a recession. As a practical matter this often means they will need to reduce lending in order to comply with regulatory capital requirements. Although this may strengthen the safety and soundness of individual banks, the reduction in lending will reduce consumption and investment and exacerbate the recession. Very risk-sensitive capital requirements tend to amplify economic cycles.
Stress tests in the U.S. have been designed to reduce this pro-cyclicality of capital requirements to better support macroeconomic objectives. This has been accomplished by automatically increasing the severity of the severely adverse scenario as the unemployment rate falls below 10%.
Over the past decade, as the unemployment rate has fallen to record low levels, the stress tests have been getting tougher. This is the way macroprudential policy is supposed to work. Capital requirements increase in an economic expansion reducing the tendency of banks to over-lend and decrease in a downturn to reduce the tendency to under-lend.
Capital requirements increase in an economic expansion reducing the tendency of banks to over-lend and decrease in a downturn to reduce the tendency to under-lend.
Ironically, a more powerful countercyclical impact is due to the requirement that banks prefund all their expected distributions of capital over the next nine quarters. That has a strong counter-cyclical effect because understandably banks want increased payouts to their shareholders when times are good. This requirement means that they will need to accumulate more capital to fund the intended payout in addition to the amount they need to accumulate to cope with the more severe stress scenario.
Regrettably, in a recent reform, the Fed has relaxed the requirement to prefund planned distributions to shareholders. Banks need now prefund only four quarters of average dividends, which will reduce the countercyclical effectiveness of the stress testing regime.
Knowledge at Wharton: What are the implications of your findings for regulators and bankers?
Herring: The first point is that stress test should be strengthened, not relaxed. The second point is that we need to make them more versatile, and we need to think about the kinds of stresses we should worry about that don’t readily lend themselves to conventional macro-modeling. And we should consider several kinds of stresses that may emerge in the financial system outside the banking sector.
Stress tests should be extended beyond banks to parts of the rapidly growing Shadow Banking sector. At present only very large banks are scrutinized carefully for their ability to withstand shocks. As more and more activity takes place outside the banking system, we ought to have a better understanding of how these entities will react in a severely adverse scenario. We don’t have a mechanism for doing that. It requires significant coordination between bank and nonbank regulators, and some sort of oversight of important players that are not regulated by any national authority.
The FSOC (the Financial Stability Oversight Council created by Dodd-Frank) in principle should shoulder this responsibility, but it has no inclination to do so. FSOC has moved from an initial position of monitoring all financial institutions that received support from the Fed during the 2008-2009 crisis to a policy of more or less saying, “Unless we can make a convincing case that you’re a systemic threat, we’re not going to exercise any prudential oversight.” I think that’s a serious problem.
We ought to be thinking about harmonizing stress tests globally.
We also ought to be thinking about harmonizing stress tests globally, because the financial system is global and if things go badly wrong in Europe or Asia, then we’re going to have problems here as well. We ought to understand those interconnections and have a sense of the measures we could take to mitigate them, should it be necessary.
We should think about other kinds of stresses. To some extent, we are seeing the benefit of earlier attention to nonfinancial shocks. Almost all the major financial institutions are working remotely. The banks have adopted measures that facilitate a transition to working remotely at least in part because regulators have required them to have robust continuation of service and recovery plans. They have undergone examinations to show they could keep operating even if some shock put the headquarters out of commission. This isn’t a new concern. For a long time, the largest banks have understood they need backup facilities to keep operations running if something should close down operations at headquarters.
The acid test came on 9/11. The Bank of New York, which is the largest custody bank and central to the payment system, believed it was adequately backed up. But its backup facility was on the same power grid and the same transportation network as the headquarters. This redundancy proved useless against the kind of shock that occurred because people couldn’t get to work at the headquarters or the backup facility and lost power at both locations. The bank was essentially offline for two to three weeks.
Since that time, banks have been very thoughtful about how they can relocate work to facilities that have a different transportation grid, a different power grid, and access to a different pool of labor. They’ve also invested heavily in enabling employees to work from home.
We are in much better shape to deal with the operational consequences of the pandemic – not because banks anticipated it, but because of what we learned from the 9/11 tragedy: to ensure operational resilience, you have to make sure that you have diversified the whole variety of inputs you need to keep the business going.
If there’s a downside, it is that stress testing is extremely expensive.
Knowledge at Wharton: Is there a downside to stress testing?
Herring: If there’s a downside, it is that stress testing is extremely expensive in part because of the detail with which it is overseen. There are heavy demands for documentation and rigorous requirements for ensuring the quality of data and validating models. Much of this is necessary. These costs make the system much less agile than it should be.
Fortunately, many of these expenses are in the nature of start-up costs. The continuing costs are significantly lower. And improvements in in software should reduce them over time — admittedly, at the cost of substantial investments in software.
It’s important to identify the additional data we need to collect to improve the relevance and effectiveness of stress tests. A lot of the problems we faced with the [2008] subprime crisis was that the authorities had no true understanding of the scope and size of the market. That, I think, remains a difficulty with respect to other innovative markets that might cause difficulties in the future.
This is the function many hoped the Office of Financial Research (an independent body within the Treasury department) would serve, but it has not received the financial support and political backing it needs to perform this function as well as it should. It’s important to have some entity that is scanning the horizon and trying to figure out where the next shock may come from, identifying the information we need to be able to evaluate the potential dimensions of the shock and the likely consequences if the shock actually occurred.
Knowledge at Wharton: What are some of the questions for future research that you’d like to tackle about these issues?
Herring: One problem, which is technical but important, is how to take account of second-round effects. Our current models focus mainly on the initial hit to banks’ income statements and balance sheets. But the actions banks take will affect what their customers and counterparties do. And, of course, these actions will, in turn set off additional actions by banks. These general equilibrium effects can be significant. They are difficult to model but important to understand because sometimes they can be as serious as the initial shock.
