Appendix I: Relationship Between Equity and Liquidity Positions and Historical Credit Downgrades and Defaults
While the focus of the paper is on estimating equity and liquidity positions under stress, these variables can of course also be the drivers of credit rating changes and default events.
Figure A1 plots the firms’ equity and cash positions against their changes, and show whether each firm-year data point corresponds to an upgrade, a downgrade, or a default event (top charts). Firms experiencing downgrades or defaults tend to be those that have fallen into negative equity positions. In other words, credit downgrades and defaults are associated to weaker equity positions. This pattern, however, does not hold when looking at cash positions. This distinction between equity and cash positions remains when looking at the time dimension (middle charts). The figures show the share of firms downgraded in different quartiles of the equity and cash distribution of each year. While weaker equity positions (1st quartile) distinctively saw a greater share of downgrades, this pattern is less clear when looking at liquidity positions.
Controlling for equity positions, however, firms’ liquidity conditions appear as important in their association with credit rating downgrades. This is shown in Figure A1 (bottom charts), which also plots the estimated downgrade probability from a probit model with year fixed effects. Both equity and cash to asset ratios are statistically significant in the probit. With equity positions evaluated at their means, the bottom-right chart illustrates the steady decline in the estimated downgrade probability as cash to assets increase. In other words, if the goal were to estimate downgrades, both cash and equity positions would be important determinants to consider. In contrast, a probit with default as the dependent variable shows a statistically significant relationship with equity, but not with cash positions. This result may reflect the fact that insolvent firms are likely to end up in default while it is likely that illiquid but solvent firms are typically able to maintain access to financing, and continue operating.
Appendix II: Sensitivity Analysis
This appendix provides a sensitivity analysis of the main results presented in the paper. A key parameter in the stress test is the elasticity αr, which determines the ability of firms to cut costs when revenues fall. Another parameter whose value is assumed in this paper relates to the dividend payout rate. This section shows the evolution of the main numerical results when the values of these parameters are allowed to vary within their entire possible range.
Bank of England, Financial Stability Report, November 2018, Issue No. 44.
Bank of England, Financial Stability Report, July 2019, Issue No. 45.
Barua, A.P. Buckley, 2019. Rising Corporate Debt: Should We Worry?” Deloitte Insights, April 2019. Available from https://www2.deloitte.com/content/dam/insights/us/articles/5042_ibtn-april-2019/DI_IbtN-April-2019.pdf, accessed March 5, 2020.
Baudino, P.; Goetschmann, R.; Henry, J.; Taniguchi, K.; and Zhu, W., 2018, “Stress-testing banks – a comparative analysis,” Financial Stability Institute, FSI Insights on policy implementation No. 12
Board of Governors of the Federal Reserve System, Financial Stability Report, May 2019.
Board of Governors of the Federal Reserve System, Financial Stability Report, November 2019.
Board of Governors of the Federal Reserve System, Financial Stability Report, May 2020.
Financial Stability Board, “Vulnerabilities Associated with Leveraged Loans and Collateralised Loan Obligations,” December 2019.
International Monetary Fund, 2020, “Technical Note on The Risk Analysis and Stress Testing the Financial Sector,” 2020 U.S. FSAP.
Kohn, D. and Liang, N., 2019, “Understanding the effects of the US stress tests”, prepared for the Federal Reserve System Conference Stress Testing: A Discussion and Review, 9 July 2019. Updated slides prepared for the 2020 ASSA Annual Meeting.
Xie, Hong, 2019. “Increasing Share of BBB-Rated Bonds and Changing Credit Fundamentals in the Investment-Grade Corporate Bond Market,” May 2019, S&P Indexology Blog.
We are grateful to Peter Breuer, Nigel Chalk, Michaela Erbenova, Mindaugas Leika, Tom Piontek, Dulani Seneviratne, Thierry Tressel, and numerous IMF seminar participants for helpful comments. All errors are solely the authors’ responsibility.
The Basel Committee on Banking Supervision issued principles for sound stress testing in 2009, with an update in 2018 (BCBS, 2009 and 2018).
The Supervisory Capital Assessment Program (SCAP) was conducted once in 2009. This was superseded by the Comprehensive Capital Assessment Review (CCAR) which started in 2011, and the Dodd-Frank Annual Stress Test (DFAST) in 2013.
The initial sample comprising of all the firms identified as participating in the leveraged loan market as well as the corporate bond debt market included around 2,000 firms. Of those, the firms that reported all the data necessary for conducting our different stress tests were 755 firms.
Investment grade credit ratings are those above and including ‘BBB-’ by S&P and Fitch (or ‘Baa3’ by Moody’s). The ratings BBB+, BBB, BBB- or the “triple Bs” are lower medium grade. Ratings below BBB-(Baa3) sub-investment grade, also referred to as speculative grade, high-yield, or junk.
S&P Global Ratings (2019), “U.S. Corporate Debt Market: The State of Play in 2019.”
Growth in the BBB class has been due a combination of newly rated instruments, downgrades from higher ratings, and upgrades from lower ratings.
The most commonly used definition is that of the S&P Leveraged Loan Commentary and Data (LCD) which includes (i) loans rated BB+ or lower; (ii) unrated or investment grade loans with spreads > LIBOR +125 and secured by a first or second lien.
As there is no standardized definition for leveraged loans, estimates of their market size vary. The commonly cited estimate of $1.1 trillion is based on the S&P’s LCD definition and the loans included in the S&P Leveraged Loan Index (LLI). Some, including the Bank of England, have argued for a broader measure that accounts for loans not included in the LLI, which puts the estimated market size at $2.2 trillion (see the Bank of England’s November 2018 Financial Stability Report).
These dynamics do not appear to be driven by the lifecycle of the firms. Although younger firms tend to be associated with higher investment spending (in percent of assets), the age of firms does not appear to be statistically linked to their return on assets. Moreover, the higher investment of these younger firms does not seem to be linked to higher borrowing levels, given that older firms―which have been operating for longer periods of time―tend to exbibit higher leverage ratios (or lower net equity positions).
Expenditures Et in the net equity position metric relate to ‘accounting’ expenditures, which includes non-cashflow items such as depreciation. This differs from the liquidity and financing need metrics introduced subsequently, which use cashflow-based expenditures.
The computation of interest expenditures under stress allows for the quantification of potential funding risks, which are measured through the increase in interest rates (or corporate spreads) embedded in the different stress test scenarios.
Sensitivity analysis using different dividend payout rates (from 0 to 100 percent) are included in Appendix II. Results show very little sensitivity to using different dividend payout rates, as firms with an initial positive net equity position can only fall into negative equity when they have (one or more periods of) negative profits which do not entail any dividend payouts.
Other activities or usage of firms’ proceeds―such as share buybacks, mergers and acquisitions, etc.―are not envisaged in our stress tests.
At the end of 2019, these nonfinancial assets accounted for about 55 percent of U.S. nonfinancial corporates’ total assets, whereas cash and deposits accounted for only about 3½ percent of total assets.
Cash expenditures include interest payments, non-interest expenditures on cash basis, and the amount investment spending (if any) in a given year.
While circumstances could play out differently in practice—e.g., lenders may protect their investments by continuing to fund a failing firm in hopes of a turnaround or may require firms to spin-off profitable parts of their business in order to meet debt obligations—the assumed scenario is meant to illustrate the amount of funding that may be needed during a risk-off episode or a broad-based corporate sector downturn.
Factors such as business climate, regulatory framework, tax regime, adoption of new technologies, or managerial talent, can have important implications for firms’ revenues and profitability. However, none of these factors are included in the stress tests scenarios, and thus cannot be linked to the evolution of our equity and liquidity measures for the purposes of stress testing.
For all other sectors, firm revenue is positively associated with real GDP growth, albeit a few cases are not statistically significant.
This extreme case is highly unlikely in practice, as it would imply that variable costs are essentially zero, and that total costs are made of fixed costs only.
Contrary to the other estimations conducted in this paper using gross revenues, this was done by estimating directly net non-interest revenues (that is, net revenues excluding interest expenditure) as a function of the macroeconomic variables included in our stress test scenarios.
The stress tests use reported data as of end-2019 as a starting value, and changes in revenue and expenditure-to-asset ratios are computed sequentially in subsequent years using the estimated elasticities linking these to the macroeconomic variables defined by the stress test scenarios.
See Technical Note on Risk Analysis and Stress Testing the Financial Sector.
The alternative scenario presented here corresponds to the “Sensitivity Scenario 3” in the 2020 U.S. FSAP. In terms of severity, all other sensitivity scenarios used in the FSAP lie in between the baseline and the alternative scenario used in this paper. These two scenarios can thus be seen as the lower and upper bounds of the FSAP stress test scenarios. Results are available upon request.
In the baseline scenario, output losses during the lockdown period (in 2020Q2) are estimated to be less than 15 percent relative to a ‘normal’ period.
Firms that fall into negative equity in a given year are not allowed to continue—with the potential of turning their net equity position back into positive territory, as the macroeconomic environment improves in the stress test scenario. Essentially, once a firm falls into negative equity, it is treated as if it drops out of the sample.
Depending on the relative size―in terms of nominal outstanding debt or equity value―of these firms, the relative pattern of potential debt and equity losses across the different years of the stress test horizon differs slightly from the annual pattern observed in the number of firms with negative equity. For instance, potential debt losses appear to be lower in 2024 relative to 2023 in the baseline scenario (Figure 11), even though there is a slightly larger number of companies that fall into negative equity in 2024 (Figure 10). This is because those companies have lower aggregate debt levels than those that fell into negative equity in 2023.
Note that the assessment of funding risk is explicitly embedded in the solvency metrics presented in the previous section, as changes in net equity positions include the impact from rising corporate spreads―due to e.g. funding liquidity stress―on firms’ interest costs and their debt servicing capacity. In particular, stress tests assume that funding is available but at increased cost depending on the firm’s credit rating and the severity of the stress test scenario.
This essentially provides a lower and upper bound for the plausible range of capex spending by firms. The case where firms set their capex levels equal to depreciation and amortization (equivalent to assuming zero new net investment) would intrinsically fall within this range.
The combined outstanding debt of these firms with negative equity that might not have access to new financing accounts for about $130–300 billion (depending on the scenario), equivalent to 2–5 percent of total outstanding debt in the sample.
High-frequency data on loan amortization are not readily available, and thus the liquidity needs computed in this box are smaller than the broader annual liquidity need measures presented earlier.
The actual losses for creditors from a defaulted asset would depend not only on the nominal amount of the defaulted debt instrument, but also on the recovery rate (or loss given default). Estimation of recovery rates under stress is an integral part of the stress test of financial institutions, but is beyond the scope of this paper.
Common equity Tier 1 (CET1) capital for the 34 banks analyzed in the 2020 U.S. FSAP (accounting for 98 percent of total banking system assets) amounted to $1.2 trillion in 2020Q1, and their Tier 1 capital amounted to $1.4 trillion.
More detailed analyses on the potential impact on banks, mutual funds, and insurance companies were conducted within the stress test setting of the 2020 U.S. FSAP. See Technical Note on Risk Analysis and Stress Testing the Financial Sector.
Expected losses would be smaller than these ‘exposures at risk’, crucially depending on their corresponding recovery rates.
Officially known as the Financial Accounts of the United States.
The nonfinancial private sector includes other sub-sectors, such as the noncorporate business sector as well as households and nonprofit organizations.
See Technical Note on Risk Analysis and Stress Testing the Financial Sector for more details.
Historical estimates of the elasticity αr at the sectoral level suggest that for most industries αr is in the range from 0.2 to 0.7 (with several of them having an elasticity near 0.7). Thus, in weighted average terms, the potential losses using the central case value of αr = 0.5 are slightly higher than when using historical elasticities.
When net after-tax profits are negative, this parameter is always set to zero in our stress tests. Firms are only allowed to pay a fraction (which could be zero) of their after-tax profits when the latter are positive.