Issuer Quality and Corporate Bond Returns Robin Greenwood and Sam Hanson Harvard Business School QWAFAFEW Presentation: October 2013
Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary The Credit Cycle How does the quantity/quality of credit evolve over time? Research in corporate finance and macroeconomics has emphasized time-varying financing frictions Recent research hints that time-varying returns due to shifting investor sentiment may also play a significant role: Junk bond boom of the 1980s Credit boom of the 2000s Jeremy Stein of the Federal Reserve has suggested that the Fed should actively monitor the composition of issuance This paper: Historically, what is the relationship between quantity/quality of credit and future investor returns? Junk bond boom of the 1980s (Kaplan and Stein 1993) Credit boom of the 2000s (Coval, Jurek and Stafford 2010; Ivashina and Sun 2010, Axelson, Jenkinson, Stromberg, and Weisbach 2010)
Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Quantity and Quality Existing market-timing literature uses financing quantities to forecast returns Firm-level stock returns: Loughran and Ritter (1995), Daniel and Titman (2006), Fama and French (2008) Market-wide or factor-level stock returns: Baker and Wurgler (2000), Greenwood and Hanson (2010) Why focus on the credit quality of debt issuers? Firms borrow more when expected credit returns are lower Broad changes in pricing of credit have a larger impact on the cost of debt for low quality firms (i.e., high default probability firms) “Credit Beta” Low quality issuance responds more to shifts in pricing of credit → Movements in expected credit returns trace out variation in the average quality of debt issuers
Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Overview Construct time-series measures of corporate debt issuer quality. Use quality measures to forecast corporate bond excess returns. Main finding: When issuers are of low quality, future excess corporate bond returns are low, and often significantly negative Incremental forecasting power over various controls and the total quantity of corporate debt financing What drives time variation in expected returns? Countercyclical risk premia Changes in the health of intermediary balance sheets Excessive risk-taking due to agency problems: “reaching for yield” Over-extrapolation by investors Evidence of mispricing suggests #3 or 4 may be part of the story
Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Firms of differing credit quality choose debt issuance Credit spreads: reflect expected losses and expected excess returns, both of which vary over time Shifts in expected excess returns can reflect changes in rational price of risk, mispricing, or both Firms: Issue more debt when expected returns are lower But issuance is impacted by other factors (shifts in investment opportunities or target leverage) → issuance is a noisy reflection of expected returns Identifying assumption: Expected excess returns on low quality bonds are more exposed to broad changes in the pricing of credit e.g., if E[AAA return] falls by 10 bps, E[HY return] falls by 100bps →Low quality issuance responds more to broad shifts in credit pricing
Forecasting Returns w/ Quality, Quantity, and Spreads Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Forecasting Returns w/ Quality, Quantity, and Spreads Quantity = sum of issuance of low and high credit quality firms Impacted by common shocks to factors unrelated to expected returns (shifts in investment opportunities or target leverage) Quality = difference in issuance between low & high quality firms Removes common shocks, better isolating movements in expected returns Forecasting excess returns using quantity and quality: Quality more informative than quantity if important common shocks unrelated to expected returns impact debt issuance of all firms
Measuring Issuer Quality: ISSEDF Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: ISSEDF What is the default probability of firms with high vs. low debt issuance? EDFi,t = Merton (1974) Expected Default Frequency, computed following Bharath and Shumway (2008) Easiest to think of this as the difference in the “credit rating” between high and low debt issuers
Measuring Issuer Quality: ISSEDF Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: ISSEDF ISSEDF is high when issuing firms are of poor credit quality
Measuring Issuer Quality: ISSEDF Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: ISSEDF ISSEDF is high when issuing firms are of poor credit quality ISSEDF correlated with business cycle, but removing macro variation doesn’t change basic character of series.
Measuring Issuer Quality: ISSEDF Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: ISSEDF ISSEDF is high when issuing firms are of poor credit quality ISSEDF correlated with business cycle, but removing macro variation doesn’t change basic character of series. Credit boom 2004-2007 1996-1998 1980s junk bond boom Junk bond bust 1990-1991 Telecom bust 2001-2002 Late-1960s credit boom Penn Central 1970
Measuring Issuer Quality: High Yield Share Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: High Yield Share
Measuring Issuer Quality: High Yield Share Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: High Yield Share 1962-1982: r(HYS,ISSEDF) = 0.47 1983-2008: r (HYS,ISSEDF) = 0.58
Measuring Issuer Quality: ISSEDF vs HYS Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Measuring Issuer Quality: ISSEDF vs HYS Advantages of HYS Simplicity “Natural” to use bond issuance to forecast bond returns Advantages ISSEDF Combines all sources of debt financing → not impacted by secular shifts in the bond vs. loan mix → stationary series If bonds/loans are partial substitutes, measures based on total debt issuance (loans+bonds) may be more informative about bond returns. Credit rating standards have evolved over time: agencies became more conservative in the late 1970s Based on net debt issuance as opposed to gross issuance Blume, Lim, and MacKinlay (1998) and Baghai, Servaes, and Tamayo (2010) argue that the agencies have become more conservative in assigning ratings since the late 1970s.
Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Other data Corporate bond returns by credit rating from Barclays (Lehman) and Morningstar (Ibbotson) Cumulative k-year log excess returns: Returns are in excess of Treasury bonds with comparable duration Other controls: bill yield, term spread, macro controls, etc
Issuer quality forecasts excess corporate bond returns Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Issuer quality forecasts excess corporate bond returns Figure 3, Panel A: Economic magnitudes are significant: 1-s increase in ISSEDF (0.48 deciles) → cumulative excess returns fall by 7.30 %-points over the following 2 years Same results hold with HYS (Figure 3, Panel B)
Issuer quality forecasts excess corporate bond returns Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Issuer quality forecasts excess corporate bond returns Table 2: Univariate forecasting regressions Increasing coefficients up to 3- years, levels off after Emphasize 2-year cumulative returns from here on Stronger results for HY bonds. Consistent with idea that ISSEDF reflects pricing of credit risk Results hold even with number of interest rate and macro controls →Parallel results for HYS
Quality and Quantity during Credit Booms Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Quality and Quantity during Credit Booms Measure aggregate credit growth using Compustat as DDt/Dt-1. Similar results using Flow of Funds data. A numbers of theories suggests that ISSEDF will be high (i.e., quality is low) when aggregate credit growth is high In principle, both quantity and quality may be useful for forecasting future returns Quality should outperform in a horserace if uninformative common shocks affect the issuance of both low and high quality firms r(DDAgg/DAgg,ISSEDF) = 0.45
Quantity vs. Quality Table 4: Panel A Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Quantity vs. Quality Table 4: Panel A Quality beats Quantity in a horserace Credit growth of low quality firms is most useful for forecasting returns Differential debt growth of low vs. high quality firms is a strong predictor →Similar results for HYS
What drives time-variation in expected credit returns? Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary What drives time-variation in expected credit returns? Rational consumption-based (integrated-markets) explanations: Time-varying quantity of risk Time-varying rational price of risk Frictional account: Changes in intermediary capital → changes in risk premia Agency problems: Low interest rates → “Reaching for yield” → Mispricing Investors make expectational errors: Extrapolation of recent outcomes → under/over-weight the probability of left-tail events → Mispricing
Changes in the Rational Price of Risk Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Changes in the Rational Price of Risk Counter-cyclical movements in price of risk as in representative agent consumption-based models If markets are integrated, time-varying risk premia that are reflected in credit markets should also show up in equity markets
Time-varying risk premia Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Time-varying risk premia A number of findings are consistent with these models: ISSEDF is cyclical: High debt issuers have high EDFs in expansions But ISSEDF remains a strong forecaster after controlling for macro variables Results are strongest for lower-rated bonds which are more highly exposed to macroeconomic risk Other findings cut against the integrated-markets view… ISSEDF not useful for forecasting equity returns (Table 9) ISSEDF predicts high yield excess returns after controlling for contemporaneous realizations of MKTRF or Fama-French factors
Forecasting Reliably Negative Excess Returns Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Forecasting Reliably Negative Excess Returns Consumption-based models: expected excess returns always > 0 HY underperform USTs in “bad times” → expected excess returns > 0. However, predicted excess returns are often significantly negative Figure 5, Panel B: Tests avoids joint hypothesis problem; have been used by Fama and Schwert (1977), Fama and French (1988), Kothari and Shanken (1997), Baker and Wurgler (2000). Conclusion: Overall, we argue that our results are difficult to reconcile with integrated-markets models in which the rationally-determined price of risk fluctuates in a countercyclical fashion.
Frictional Account: Intermediary capital Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Frictional Account: Intermediary capital Fluctuations in intermediary balance sheets affect risk premia Predict that issuer quality will be poor (i.e., ISSEDF will be high) when intermediary balance sheets are strong and risk bearing capacity is high Look at several types of intermediaries: Insurers: Largest holders of corporate bonds Broker-dealers: Provide liquidity in corporate bond market Banks: Provide a close substitute for bond financing Measures of balance sheet strength: Equity/Assets, Asset Growth, Bank Credit Losses
Frictional Account: Intermediary capital Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Frictional Account: Intermediary capital We run two types of regressions: Regression 1: What is the relationship between ISSEDF and proxies for intermediary balance sheet strength Zt? Frictional models predict: b > 0 Regression 2: Do proxies for intermediary capital diminish the forecasting power of ISSEDF? Frictional models predict: b2 < 0; magnitude of b1 should decline once we control for Zt
Insurer balance sheets Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Insurer balance sheets Equity Capital, or Asset growth Some evidence of a link between balance sheets and ISSEDF But controlling for intermediary balance sheet variables does not have meaningful impact on forecasting power of ISSEDF Similar conclusions for other intermediary variables (Results here→) Frictional stories also inconsistent with negative expected returns Negative results don’t imply that intermediary capital isn’t ever part of the story. Might matter at certain points, but quality may integrate different factors over time.
Agency-based Explanation: “Reaching for Yield” Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Agency-based Explanation: “Reaching for Yield” Delegated institutional investors have incentives to reach for yield when interest rates are low or have fallen (Rajan 2005) 2004-2007 credit market boom Klarman (1991): 1980s junk bond boom Possible stories: Intermediaries with fixed liabilities have incentives to engage in risk shifting when nominal rates fall Costly for pensions to reduce return targets → reach for yield Fund managers compensated on basis of absolute nominal returns Stories may admit the possibility of negative expected returns Our analysis: Investigate impact of yields and changes in yields on ISSEDF But recall our baseline results already control for interest rates
Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary “Reaching for Yield” Table 11: Impact of yields and changes in yields on ISSEDF 1-yr changes Levels 2-yr changes Results are consistent with “reaching for yield” hypothesis, but probably far from definitive. → Similar results for HYS
Investor-beliefs based explanation Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Investor-beliefs based explanation Time-variation in expected returns may be due to mistaken investor beliefs about true creditworthiness of borrowers Natural story: over-extrapolation Wide variety of evidence on investor extrapolation Investors use a “representativeness” heuristic Intermediaries use backwards looking risk management systems (e.g., Value-at-Risk) → built-in tendency towards over-extrapolation
Extrapolative Beliefs Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Extrapolative Beliefs Potential account : Economy switches between good times in which few firms default, and bad times in which a higher fraction of firms default Investors think economy either evolves via a more persistent process or less persistent process than truth (Barberis, Shleifer, Vishny 1998) What happens? A string of low-default realizations → investors become over- optimistic that good times will last → neglect down-side risks If the high default state arrives → expectations are revised If bad state persists → investors over-estimate default probabilities Generates short-term return continuation, longer-term reversals Add a corporate sector that levers up when debt is “cheap” Growing optimism → borrower quality erodes Spreads under-react to erosion in borrower quality in booms → both quality and credit spreads forecast returns
Extrapolative Beliefs Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Extrapolative Beliefs Consistent with negative expected excess returns ✓ ISSEDF should be high following a string of low realized defaults or high returns on credit assets ✓ 1-yr changes Levels 2-yr changes → Similar results for HYS
Conclusions Summary: Interpretation: Future work: Introduction Empirical Strategy Issuer Quality Forecasting Results Interpretation Summary Conclusions Summary: Issuer quality is low → future corporate bond excess returns are low Evidence of mispricing: forecast significantly negative excess returns 2004-2007 credit boom is not without precedent – part of a recurring historical pattern, dating to at least the 1940s Interpretation: Difficult to fully explain by appealing to rationally time-varying risk aversion or other rational drivers of counter-cyclical risk premia Partially consistent with frictional and agency-based stories Some evidence that over-extrapolation plays a role Future work: Micro empirical work on excessive risk-taking? Or mistaken beliefs? Understand the real consequences of credit market booms Quality of sovereign debt issuers