Skip to main content
SpringerLink
Log in

Can rating agencies look through the cycle?

  • Original Research
  • Published:
Review of Quantitative Finance and Accounting Aims and scope Submit manuscript

Abstract

Rating agencies claim to look through the cycle when assigning corporate credit ratings, which entails that they are able to separate trend components of default risk from transitory ones. To test whether agencies possess this competence, I take market-based estimates of 1-year default probabilities of corporate bond issuers and estimate their long-run trend using the Hodrick-Prescott filter, local regression, or centered moving averages. I find that ratings help identify the current split into trend and cycle. In addition, rating stability is similar to the one of hypothetical ratings based on long-term trends. The results are robust to the use of different filter techniques. They are confirmed by a model-free analysis, which shows that ratings predict future changes in market-based default probability estimates. Since the examined trends are forward-looking in the sense that the trend filtering algorithms use future data, agency ratings exhibit important characteristics one would expect from ratings that see through the cycle.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1

Similar content being viewed by others

Notes

  1. Note that this paper deals with corporate credit ratings. Structured finance ratings, which feature prominently in the discussion of the subprime crisis, are assigned by different rating units based on a different rating approach. An assessment of rating quality should therefore be conducted separately for the two types of ratings. The methodology for sovereign ratings is more similar. Due to significant differences in the determinants of corporate and sovereign defaults, however, the conclusions of the paper should not be translated to sovereign ratings.

  2. Cf. Standard and Poor’s (2003, pp. 41–43).

  3. Cf. Campbell, Lo and MacKinlay (1997, pp. 78–80).

  4. Let me give an example for each combination: a change in the cyclical component of a firm’s default risk would be systematic if it is associated with aggregate changes in profitability over the business cycle; a change in the trend component would be systematic if it is part of a permanent, market-wide increase in leverage. Non-systematic cycles can arise if firms respond to firm-specific shocks by rebalancing their capital structure (cf. Graham and Harvey 2001; Fama and French 2002; Flannery and Rangan 2006; Leary and Roberts 2005). If a firm decides to increase its leverage permanently without the average firm doing this at the same time, there would be a non-systematic change in the firm’s trend component.

  5. One could argue that this statement is incomplete because it misses the stress scenario approach described in the literature. I discuss this aspect in Sect. 5.2.

  6. “In any case, purely cyclical factors are difficult to differentiate from coincident secular changes in industry fundamentals, such as the emergence of new competitors, changes in technology, or shifts in customer preferences” (Standard and Poor’s 2003, p. 42).

  7. If the stock market declines, for example, the EDFs will on average go up, resulting in the error terms being correlated across firms. Running the regressions without de-meaning does not change the conclusions. Note that de-meaning also purges the data of the effects of aggregate stock market fluctuations.

  8. Firm-specific clusters also take account of firm-specific serial correlation. The estimator is implemented using the procedure discussed and provided by Petersen (2009).

  9. For example, the last 36 month forecast horizon entering the analysis is from December 1998 to December 2001.

  10. Data is from S&P Compustat, for US firms only.

  11. It cannot be ruled that the influence of the rating variable is affected by endogeneity. For example, if ratings and EDFs are two different measures of the underlying credit quality, errors in measuring this credit quality could lead to endogeneity. This would not invalidate the conclusion that ratings are useful for predicting EDFs. It could even be linked with a through-the-cycle interpretation in the sense that ratings look through error cycles in EDFs. The standard econometric response—instrumental variables—does not appear feasible due to the lack of suitable variables.

  12. Pedersen (2001) also recommends λ > 100,000 for monthly data.

  13. Conclusions do not change if defaults are ignored and trends are computed over all observations of a company. I define emergence from default as an upgrade to B3 or better.

  14. Consider a Hodrick-Prescott trend computed over one sinusoidal cycle. Rather than being horizontal, it is a downward sloping line. This trend line is as smooth as a horizontal line, but the squared deviations from the sinusoid are smaller. Many short series are similar in that they contain just one peak and one trough.

  15. Standard deviations of variance ratios are high because they reach high values (>1) in some cases. The reason is that the sample selection excludes observations at the start and end of the series. Though excluded from the final analysis they are used in the computation of trends and their variability. If the variability of EDFs in the middle of the series is lower than at the start and end one can observe low EDF variability but high trend variability.

  16. I also produced predictions based on ordered logit or ordered probit regressions. Results are very similar.

  17. Even though ratings and EDFs are bounded between 1 and 21 and 0.02% and 20%, respectively, one cannot rule out that trending behavior within the bounds affects the distribution of t-statistics. Regressions using levels are available upon request; they strengthen the conclusions drawn from the analysis using differences.

  18. If the trend is computed with the Hodrick-Prescott Filter, for example, TREND t |t − is obtained by applying the Hodrick-Prescott filter (2) to observations t 0(i),…t instead of t 0(i),…,T(i).

  19. The regressions are of the form \( {\text{TREND}}_{it} = \beta_{0} + \beta_{1} {\text{EDF}}_{it} + \beta_{2} {\text{Rating}}_{it} + u_{it} \).

References

  • Altman EI, Rijken HA (2004) How rating agencies achieve rating stability. J Bank Financ 28:2679–2714

    Article  Google Scholar 

  • Altman EI, Rijken HA (2006) A point-in-time perspective on through-the-cycle ratings. Financ Anal J 62:54–70

    Article  Google Scholar 

  • Amato JD, Furfine CH (2004) Are credit ratings procyclical? J Bank Financ 28:2641–2677

    Article  Google Scholar 

  • Baker HK, Mansi SA (2002) Assessing credit rating agencies by bond issuers and institutional investors. J Bus Financ Acc 29:1367–1368

    Article  Google Scholar 

  • Bangia A, Diebold FX, Kronimus A, Schagen C, Schuermann T (2002) Ratings migration and the business cycle, with application to credit portfolio stress testing. J Bank Financ 26:445–474

    Article  Google Scholar 

  • Basel Committee on Banking Supervision (2000) Credit ratings and complementary sources of credit quality information. BCBS working papers no 3, Basel

  • Bharath ST, Shumway T (2008) Forecasting default with the Merton distance to default model. Rev Financ Stud 21:1339–1369

    Article  Google Scholar 

  • Bohn JR, Arora N, Korablev I (2005) Power and level validation of the EDF credit measure in the US market. Working paper, Moodys KMV

  • Butera G, Faff R (2006) An integrated multi-model credit rating system for private firms. Rev Quant Finan Acc 27:311–340

    Article  Google Scholar 

  • Campbell JY, Lo A, MacKinlay C (1997) The econometrics of financial markets. Princeton University Press, Princeton

    Google Scholar 

  • Campbell JY, Lettau M, Malkiel BG, Xu Y (2001) Have individual stocks become more volatile? An empirical exploration of idiosyncratic risk. J Financ 56:1–43

    Article  Google Scholar 

  • Campbell JY, Hilscher J, Szilagyi J (2008) In search of distress risk. J Finance 63:2899–2939

    Article  Google Scholar 

  • Cantor R (2001) Moody’s investors service response to the consultative paper issued by the Basel Committee on Banking Supervision and its implications for the rating agency industry. J Bank Financ 25:171–186

    Article  Google Scholar 

  • Carey M, Hrycay M (2001) Parameterizing credit risk models with rating data. J Bank Financ 25:171–270

    Article  Google Scholar 

  • Cleveland WS (1979) Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 74:829–837

    Article  Google Scholar 

  • Cogley T, Nason JM (1995) Effects of the Hodrick-Prescott on trend and difference stationary time series. Implications for business cycle research. J Econ Dyn Control 19:253–278

    Article  Google Scholar 

  • Crosbie P, Bohn J (2003) Modeling default risk, Moody’s KMV technical paper

  • Daníelsson J, Embrechts P, Goodhart C, Keating C, Muennich F, Renault O, Shin HS (2001) An academic response to Basel II. Working paper, London School of Economics

  • Delianedis G, Geske R (2003) Credit risk and risk neutral default probabilities: Information about rating migrations and defaults, EFA 2003 annual conference paper no. 962

  • Egan SJ, Jones WB (2002) Statement of Egan Jones on credit rating agencies. Securities and Exchange Commission. http://www.sec.gov/news/extra/credrate/eganjones2.htm

  • Ehlgen J (1998) Distortionary effects of the optimal Hodrick-Prescott filter. Econ Lett 61:345–349

    Article  Google Scholar 

  • Fama EF, French KR (2002) Testing tradeoff and pecking order predictions about dividends and debt. Rev Financ Stud 15:1–33

    Article  Google Scholar 

  • Flannery MJ, Rangan KP (2006) Partial adjustment toward target capital structure. J Financ Econ 79:469–506

    Article  Google Scholar 

  • Fons J (2002) Understanding Moody’s corporate bond ratings and rating process. Special comment, Moody’s investors service

  • Geske R (1977) The valuation of corporate liabilities as compound options. J Financ Quant Anal 12:541–552

    Article  Google Scholar 

  • Geske R, Johnson HE (1984) The valuation of corporate liabilities as compound options: a correction. J Financ Quant Anal 19:231–232

    Article  Google Scholar 

  • Graham JR, Harvey CR (2001) The theory and practice of corporate finance: evidence from the field. J Financ Econ 60:187–243

    Article  Google Scholar 

  • Hamilton DT (2004) Default and recovery rates of corporate bond issuers. Moody’s investors service

  • Harvey AC, Jaeger A (1993) Detrending, stylized facts, and the business cycle. J Appl Econ 8:231–247

    Article  Google Scholar 

  • Hodrick RJ, Prescott EC (1997) Postwar US business cycles: an empirical investigation. J Money Credit Bank 29:1–16

    Article  Google Scholar 

  • Kaiser R, Maravall A (1999) Estimation of the business cycle: a modified Hodrick-Prescott filter. Span Econ Rev 1:175–206

    Article  Google Scholar 

  • Kealhofer S (2003) Quantifying credit risk I: default prediction. Financ Anal J 59:30–44

    Article  Google Scholar 

  • King RG, Rebelo ST (1993) Low frequency filtering and real business cycles. J Econ Dyn Control 17:207–231

    Google Scholar 

  • Kisgen D (2006) Credit ratings and capital structure. J Financ 61:1035–1072

    Article  Google Scholar 

  • Leary MT, Roberts M (2005) Do firms rebalance their capital structure? J Financ 60:2575–2619

    Article  Google Scholar 

  • Leland HE (2004) Predictions of default probabilities in structural models of debt. J Invest Manag 2:5–20

    Google Scholar 

  • Leland HE, Toft KB (1996) Optimal capital structure, endogenous bankruptcy, and the term structure of credit spreads. J Financ 51:987–1019

    Article  Google Scholar 

  • Löffler G (2004) An anatomy of rating through the cycle. J Bank Financ 28:695–720

    Article  Google Scholar 

  • Löffler G (2007) The complementary nature of ratings and market-based measures of default risk. J Fixed Income 17:38–47

    Article  Google Scholar 

  • Longstaff FA, Schwartz ES (1995) A simple approach to valuing risky debt. J Financ 50:789–821

    Article  Google Scholar 

  • Merton RC (1974) On the pricing of corporate debt: the risk structure of interest rates. J Financ 29:449–470

    Google Scholar 

  • Meyer M, Winker P (2005) Using HP filtered data for econometric analysis: some evidence from Monte Carlo simulations. Allg Stat Arch 89:303–320

    Google Scholar 

  • Nickell P, Perraudin W, Varotto S (2000) Stability of rating transitions. J Bank Financ 24:203–227

    Article  Google Scholar 

  • Pedersen TM (2001) The Hodrick-Prescott filter, the Slutzky effect, and the distortionary effect of filters. J Econ Dyn Control 25:1081–1101

    Article  Google Scholar 

  • Petersen MA (2009) Estimating standard errors in finance panel data sets: comparing approaches. Rev Financ Stud 22:435–480

    Article  Google Scholar 

  • Standard & Poor’s (2003) Corporate ratings criteria. Standard & Poor’s

  • Stohs MH, Mauer D (1996) The determinants of corporate debt maturity structure. J Bus 69:279–313

    Article  Google Scholar 

  • Sun L (2007) A re-evaluation of auditors’ opinions versus statistical models in bankruptcy prediction. Rev Quant Finan Acc 28:55–78

    Article  Google Scholar 

  • Treacy W, Carey M (1998) Credit risk rating at large US banks. Fed Reserve Bull 84:898–921

    Google Scholar 

  • Trustorff JH, Konrad PM, Leker J (2011) Credit risk prediction using support vector machines. Rev Quant Finan Acc 36:565–581

    Article  Google Scholar 

  • Warga A, Welch I (1993) Bondholder losses in leveraged buyouts. Rev Financ Stud 6:959–982

    Article  Google Scholar 

  • Yip A (2006) Business failure prediction: a case-based reasoning approach. Rev Pac Basin Financ Mark Policies 9:491–508

    Article  Google Scholar 

  • Yoshizawa Y (2003) Moody’s approach to rating synthetic CDOs. Special comment, Moody’s investors service

Download references

Acknowledgments

I am grateful to Moody’s Investors Service for providing data used in this paper. I wish to thank the anonymous referee as well as seminar participants at the annual meeting of the German Finance Association, the GRETA credit conference, the Federal Reserve Board, and the universities of Frankfurt and Mannheim for their many helpful comments. Parts of the paper were written while the author was enjoying the hospitality of the College of Management at NC State.

Author information

Authors and Affiliations

  1. Institute of Finance, University of Ulm, Helmholtzstrasse 18, 89069, Ulm, Germany

    Gunter Löffler

Authors
  1. Gunter Löffler
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Gunter Löffler.

Appendices

Appendix 1

See Table 9.

Table 9 Selected results when rating information is captured through the log of the associated idealized default probability
Full size table

Appendix 2

See Table 10.

Table 10 Sensitivity analysis of the regressions in Table 2
Full size table

Rights and permissions

Reprints and permissions

About this article

Cite this article

Löffler, G. Can rating agencies look through the cycle?. Rev Quant Finan Acc 40, 623–646 (2013). https://doi.org/10.1007/s11156-012-0289-9

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11156-012-0289-9

Keywords

JEL Classification

Search

Navigation