Strong and Liquid Banks Helped Prevent a Financial Crisis This Spring

But I think as I sit here right today in the middle of May, the banking system is really a source of strength and a source of credit in the economy.  And that’s important.

— Vice Chair Richard Clarida

Between mid-February and the end of March, the financial system was hit with a staggering blow.  By several measures (see Table 1), the shock was at least as bad as that caused by Lehman’s failure in September 2008.  The reaction in financial markets to COVID-19 is not surprising.  In a few short weeks, as the inevitability of the pandemic became clear, the economic outlook went from strong to abysmal.

What is surprising is that the pandemic shock did not cause a financial crisis.  Instead, while the initial shock to the financial system made measures of financial crisis intensity blink red, those measures quickly reverted to yellow or even green.  The Systemic Risk Indicator of the Cleveland Fed and the Financial Stress Index of the St. Louis Fed, for example, both rose to high levels in mid-March but had returned to near-normal levels by mid-April.  It is widely and correctly recognized that the reversal owed importantly to the rapid and massive response of the Federal Reserve, but it also owed to the strength of the banking sector going into the crisis.

This note does three things:  It reintroduces a financial stress index that was designed to judge whether financial conditions are consistent with past instances when the Fed has engaged in emergency financial intervention.  It discusses and assesses the impact of the Federal Reserve’s actions.  And it discusses and assesses the role of the banking system in preventing a crisis.

Financial stress index based on past Fed Emergency Intervention

The financial stress index we report is designed to determine the probability that a situation warrants emergency intervention by the Federal Reserve judging by past situations where the Fed intervened.  The index uses a model introduced in Nelson and Perli (2005).  The approach is similar to those used for models of whether the economy is in recession such as Chauvet, M. and J. Piger (2008) and Chauvet (1998).[1]  In particular, Nelson and Perli identified—using judgment—past periods of financial stress and taught a statistical model to identify such periods based on a dozen standard measures of financial stress such as implied volatilities, risk spreads, and off-the-run spreads.  Carlson, Lewis, and Nelson (2012) revised Nelson and Perli (2005) by using emergency interventions by the Federal Reserve rather than judgment to define stress episodes and re-estimating the statistical model with a sample that extended through the 2007-09 financial crisis.

We now have extended and re-estimated the model used by Carlson et al. to cover three additional stress events: the current emergency actions of the Fed in response to the coronavirus crisis, the expansion of the Fed’s balance sheet in reaction to repo market volatility in late 2019, and the emergency reopening of the central bank swap lines in late 2011.  We also changed slightly the variables included to measure stress by adding the root-mean-squared-error (RSME) of a yield curve fitting model, as well as an index of bank debt spreads, and removing the series on certificate of deposit spreads for which we do not have data.

In a nutshell, following Carlson et al. (2012), we calculate the average level, volatility, and correlation of the financial variables and use those three variables to fit a logit model for Fed emergency intervention.  Details on the model and the data are provided in the appendix.  The result is shown in Exhibit 1.

Exhibit 1:  Probability that financial conditions are consistent with Fed emergency intervention

As can be seen, the model places nearly a 100 percent probability that financial conditions after the coronavirus crisis first hit were consistent with those associated with past instances of Fed emergency intervention.  The only period when the index registered such high levels of stress was after Lehman failed in 2008, although the period of the past financial crisis before the Lehman failures and the episode around the Russian default and LTCM bailout in 1998 were nearly as high.

The striking difference between the Lehman and current episode is that the index remained elevated for nearly 6 months after Lehman but fell sharply within a few weeks after the coronavirus shock.  As of June 12, the index assesses an approximately 1 percent probability that current conditions are consistent with emergency Fed intervention.

Federal Reserve Actions

No doubt, the Fed’s quick and massive response to the severe market dislocation in mid-March helped prevent a larger conflagration.  Perhaps most importantly among those actions, in just three weeks beginning on March 15, the Federal Reserve bought $1 trillion in Treasury securities.  The purchases were a response to a severe oversupply of securities as many leveraged investors sought to sell their holdings simultaneously, overwhelming the balance sheet space on the books of broker-dealers (see Duffie (2020)).

In addition, the Federal Reserve eased terms on the discount window and central bank swap lines and, between March 17 and April 9, announced nine credit facilities.  The three facilities that began operations quickly and reportedly helped ease conditions the most were the Primary Dealer Credit Facility, the Money Market Mutual Fund Liquidity Facility (MMLF) and the Commercial Paper Funding Facility (CPFF).  All three facilities are direct copies of facilities opened in the wake of the Lehman failure – the PDCF, the AMLF, and the CPFF.  The initial uptake of all three has been much lower this time around, though, as indicated by the orange lines in Exhibit 2.

Exhibit 2:  Fed emergency facilities in 2008 and 2020

Thus, while the Fed responded quickly and forcefully, the conditions did not, in the end, require much emergency lending, in part because simply opening the programs helped calm financial markets.  In contrast, while Fed stood up the AMLF and CPFF quickly in 2008 as well, and the PDCF had been opened several months before, the financial turmoil persisted, and the programs lent hundreds of billions of dollars.

Role of banks

A critical difference between the two periods is that in 2007-09, banks contributed to the disarray in financial markets and the broader economy, whereas now they are part of the solution.  Leading up to the 2007-09 crisis, banks were weakly capitalized, held a relatively small quantity of liquid assets, and were exposed to losses from private mortgage-backed securities and other structured products that were at the heart of the financial rot.  Since the past financial crisis, much has been done to heal the banking sector and make it safer and more resilient.  For example, according to the data provided in the Fed’s most recent Supervision and Regulation report (available here), large banks have doubled their capital ratios and quadrupled their liquidity holdings.  Moreover, banks are rigorously tested each year for their ability to withstand a massive downgrade of the economic outlook such as the one we are experiencing today.

To compare banks’ contribution to financial market stress during the two periods, we provide a counterfactual representation of the financial stress index.  In Exhibit 3, we plot the main index along with its counterfactual that holds bank credit spreads constant throughout the 2007-09 and coronavirus episodes.[2]  Note that the main index shown is computationally equivalent to that in Exhibit 1, but to improve viewability, we do not convert the index into probabilities.[3]

Exhibit 3: Continuous financial stress index and counterfactual scenario

As can be seen, approximately 42 percent of the sharp increase in the index that followed the Lehman collapse reflected a widening of bank credit spreads.  In contrast, increases in bank credit spreads contributed only 17 percent of the jump in the index during the COVID period.

To understand the important role of banks in stemming the crisis, consider this description of the early stages of the past financial crisis from Borio and Nelson (2008):

The turmoil was triggered by a sharp and disorderly repricing of credit risk, with the US subprime mortgage market at its epicentre. Given the leverage built up in the system and the opaqueness of valuations of new structured products and of their distribution within the system, the repricing led to, and was exacerbated by, an evaporation of liquidity in many markets, including in the interbank market. As the strains spread, banks became very concerned with the liquidity and capital implications of potential large-scale involuntary reintermediation and distrusted their counterparties. The reintermediation was primarily associated with banks’ backup credit lines for securitised vehicles and with the inability to dispose of assets intended to be sold off, in line with the originate and-distribute model. In August, tensions were thus transmitted to the heart of the financial system – the interbank market, both in the United States and in a number of other mature markets. (pp. 37-8)

In March 2020, just as in August 2007, banks were again hit by massive draws on lines of credit, but this time the consequences were much different.  Between February 12, 2020 and April 1, bank loans increased a bit over $700 billion, in large part because banks were funding draws on lines of credit as large and small businesses sought to stockpile cash.  By contrast, Fed lending peaked at about $130 billion at the beginning of April.[4]  Banks and broker-dealers were the first responders on the scene in this crisis, not the arsonists.  But despite these massive draws, banks faced no material liquidity challenges, and counterparty concerns remained largely subdued.

In Exhibit 4, we plot two comparable measures of bank credit risk from JP Morgan and Barclays.  Both series measure the banking sector investment-grade corporate bond spreads over comparable Treasury yields.  The JP Morgan and Barclays indices recently peaked at 385 basis points on March 20, and 378 basis points on March 23, respectively, only two-thirds of the levels recorded in the past financial crisis and similar to levels observed during the European Banking Crisis.  Moreover, both indices fell below 200 basis points by mid-April, below levels that had generally prevailed from 2008 to 2013.

Exhibit 4:  Measures of bank credit risk

When banks are concerned about whether counterparties will repay their loans, and about their own ability to fund themselves, they stop lending to each other at term or entirely.  In that way, a liquidity shock becomes a liquidity crisis.  In March 2020, there was a profound liquidity shock, but with no counterparty or own liquidity concerns to amplify it; therefore, it quickly subsided. As noted last week in the Fed’s June 2020 Monetary Policy Report to Congress:

Funding markets proved less fragile than during the 2007–09 episode in the face of the COVID-19 outbreak and the associated financial market turmoil. The subdued reliance of large bank holding companies on short-term funding and their robust holdings of high-quality liquid assets have prevented any considerable stress in the banking sector. (p.30)

Final thoughts

We are, of course, not out of the woods yet.  While the May employment figures are encouraging, and the economy may recover, at least partially, in the second half of this year, the damage already done to the economy and possible further waves of the coronavirus could generate a long and deep recession.  Under those conditions, loan losses will mount and banks’ capital positions will be strained.  Thus far, however, as reflected in the quote from Fed Vice Chair Clarida at the top of the post, banks are a source of strength and credit for the economy.  And that’s important.

Disclaimer: The views expressed do not necessarily reflect those of the Bank Policy Institute’s member banks, and are not intended to be, and should not be construed as, legal advice of any kind.

Appendix

We follow the financial stress events and model used by Carlson, Lewis, and Nelson (2012) and developed in Nelson and Perli (2005).   Financial stress episodes are defined events in which U.S. policymakers (mostly Federal Reserve) intervened in response to deteriorating financial market conditions.  As in those previous papers, we fit those occurrences using a logistic regression using various market measures of risk, uncertainty, and liquidity.  As explained below, we adjust the variables used in the model slightly and identify additional instances where the Fed intervened.

Methodology

The building blocks of the model are thirteen financial series, taken at a daily frequency, that in turn capture market measures of risk spreads, investor uncertainty, and liquidity premia.  These series are listed in table A.  The variables are standardized, converted to a 5-day moving average, and consolidated into three sub-indices:

• Levels: the average levels of the series
• Volatility: the average volatility of the series, calculated as the sum of squared changes over an 8-week moving average
• Co-movement: the co-movement of the daily changes of the series, calculated as the percent of total variation that can be explained by a single common factor over a 26-week moving average

These sub-indices comprise the explanatory variables in the logistic regression.

The dependent variable consists of a binary variable that is set equal to one on the four weeks before and after the announcement of government (primarily Fed) interventions designed specifically to protect financial market functioning.  A complete list of identified stress events is provided in Table B.

Changes from Carlson, Lewis, and Nelson (2012)

We updated the analysis slightly relative to Carlson et al.  First, we added additional episodes of government intervention.  In addition to the interventions in response to the COVID-19 crisis we added the reopening of the central bank swap lines on November 30, 2011, in response to pressers stemming from the European banking crisis, and the expansion of Fed asset purchases and repo lending announced on October 3, 2019, in response to volatility in repo markets.

Second, we added two measures of financial stress:  The root-mean-squared-error (RSME) of the Treasury yield curve fitting error, and, more importantly, the average corporate bond spread to comparable Treasury yields for investment-grade U.S. banks.  RMSE has become a standard measure of financial stress.  Bank risk spreads improve the fit of the model and help us to assess the role of banks during the COVID-19 financial shock.  We dropped the series on certificate of deposit spreads for which we do not have data.

Table A: Underlying data series

• Liquidity
  • On-the-run liquidity premium for the 2-year Treasury
  • On-the-run liquidity premium for the 10-year Treasury
  • Root-mean-squared-error of the Treasury yield curve fitting error
  • Federal funds target – yield on the two-year Treasury
• Risk Spreads
  • Yield spread between AA-rated corporate bonds (ICE BofA AA US Corporate Index) and 10-year Treasury securities
  • Yield spread between BBB-rated corporate bonds (ICE BofA BBB US Corporate Index) and 10-year Treasury securities
  • Yield spread between high-yield corporate bonds (ICE BofA US High Yield Index) and 10-year Treasury securities
  • Spread between the 3-month LIBOR and the 3-month Treasury rate
  • Corporate bond spread to comparable Treasury yields for investment-grade U.S. banks (Bloomberg-Barclays investment-grade bank index)
  • (12-month ahead earnings/S&P 500 earnings) – yield on 10-year Treasury (a measure of the equity premium for stocks)
• Investor Uncertainty
  • Implied volatility on 3-month swaptions
  • Implied volatility on 10-year swaptions
  • S&P100 implied volatility (VXO)

Table B: Policy Intervention Events

References

Borio, C. and Nelson, W. (2008): “Monetary Operations and Financial Turmoil,” BIS Quarterly Review, March 2008.

Carlson, M., Lewis, K., and Nelson, W. (2012): “Using Policy Intervention to Identify Financial Stress,” International Journal of Finance & Economics, 19(1), 59–72.

Nelson, W., and R. Perli (2005): “Selected indicators of financial stability,” Irving Fisher Committee’s Bulletin on Central Bank Statistics, 23, 92–105.

[1] Chauvet, M. and J. Piger (2008), “A Comparison of the Real-Time Performance of Business Cycle Dating Methods,” Journal of Business and Economic Statistics, 26, 42-49, and Chauvet (1998), “An Econometric Characterization of Business Cycle Dynamics with Factor Structure and Regime Switching,” International Economic Review, 39(4), 969-96.

[2] The two bank-credit spread measures are the 3-month LIBOR – 3-month Treasury yield spread, and the Bloomberg-Barclays investment-grade bank index.  In the counterfactual version, we hold each series constant from the end of June 2007 to the end of December 2010, and from the end of January 2020 to the present.

[3] As explained in Carlson et al. (2012), the three summary indicators of financial stress are first mapped into an index that they call the “weighted-sum” before being translated into probability of Fed emergency intervention.  Exhibit 3 uses the weighted-sum representation.

[4] Federal Reserve statistical releases H.4.1. and H.8.