Journal of Undergraduate Research
Volume 5, Issue 3 - December 2003

The Predictability of Bond Ratings

Michael Novar

Historically, bond ratings have played an important role in determining a firm’s cost of capital and in assessing the potential risk of a firm from an investments perspective. Recently, bond ratings have become an important topic in the investment community due to some significant firm failures. Major bond raters, such as Moody’s, Standard & Poor and Fitch, changed their ratings only weeks before the collapse of Enron. This late warning from rating agencies begs the question: just what goes into a corporate bond rating, and how is it that the rating agencies did not foresee such an ominous disaster?

Bond ratings are used to determine how well a company can repay its debt obligations. This is measured differently by the major rating agencies. Most agencies look at financial ratios, accounting practices, the issuer’s industry, ability to generate cash flows, and management. Each agency weighs the relevant variables differently depending on their “optimal” model specification. This is often why a discrepancy in ratings over the same company’s debt occurs. For instance, Moody’s may give a company’s debt an A rating, while S&P may rate the same company’s debt as B.

Therefore, it can be concluded that bond ratings are, in part, subjective. For instance, is it possible to judge management effectiveness objectively? Enron had some of the best management on its team based on schooling, previous experience, and past performance. These qualitative variables are judged differently depending on the industry and analyst. Thus, rating companies use objective measures, past experience, and maybe even gut instinct when rating a company.

In my research project, I examine “objective” variables such as financial ratios and determine how close they predict a rating without adding subjective variables to the valuation. According to Rating Industrial Bonds by Hawkins, as many as 50 to 70 percent of bond ratings can be determined by a few variables which include: the subordination status of the issue, size of the company, degree of financial leverage, profitability of the issuer, interest coverage, and the stability of the issuer’s dividends and earnings. These key variables can be found in a firm’s financial statement. However, accounting “tricks” or management withholding pertinent information could skew the degree to which these statements are precise, hence undermining the predictability of the final rating. Therefore, in examining the ratings, it is important that these factors are not overlooked.

The initial selection criterion to create bond forecasting models requires that the firms be of a relatively high profile to insure that historical financial data and ratings are readily available. To achieve this objective, the 30 firms that make-up the Dow Jones Industrial Average are selected. For 13 of the 30 Dow Jones firms, historical ratings data is not available.1 So, 13 other high profile firms with sufficient data are selected to complete the sample. The final list of firms can be found in Table 1 (pdf format). Among those 13 additional firms, Southwest Gas Corporation and CMS Energy Corp., both energy companies, are also selected as an additional counter point to Enron because one of the objectives of this research is to determine if one could have predicted the fall of Enron.


( 1For example, since Exxon Mobil Corporation is a relatively new merger, historical bond rating data is not available during the last five years.)

Bond Ratings history of five years back (1997-2001) is collected for all firms except Enron because there was no information for 2001 due to the firm’s bankruptcy. Enron’s bond ratings are collected for five years but starting with years 1996-2000. When selecting the bond ratings, those bonds that carried the longest maturity date are used. The data is collected from Moody’s Annual Bond Record.

The next step is to find corresponding financial ratios to each year’s bond rating to determine a relationship. In the book Rating Industrial Bonds by D. Hawkins, several past studies on bond ratings are discussed. Based on these studies, the author of the book summarized the studies and concluded that the variables that determine perhaps 50% to 70% of a bond rating are the following: subordination status of issue, size of issuer, degree of financial leverage, profitability of issuer, interest coverage adequacy and stability of issuer’s dividends and earnings. With the previous studies in mind, research is conducted using a firm’s leverage, fixed charge coverage, size and profitability. Four financial ratios are found which include: long-term debt as a percentage of total capital, total assets, return on equity, and fixed charge coverage. These independent variables are found using Thompson Analytics®.

Two additional multiple regressions are then run including an analysis of all service firms and Consumer Non-cyclical firms, among the original 30 firms selected. All 30 firms are categorized by market sector. Of the thirty firms, the market sectors that existed included: basic materials, consumer cyclical, consumer non-cyclical, services, technology, healthcare, utilities, conglomerates, capital goods, and financial. The two sectors that are most frequent among the 30 firms are chosen to run the regression. They are the service and non-cyclical sectors.

Two sector models are used to determine if a particular sector would be more predictive than a regression using all companies. This would seem to be the case because firms within a sector are compared to one another helping to capture healthy correlations between financial ratios and unhealthy correlations that may cause a decrease in bond rating. Financial ratios among a particular sector will show more trends that strongly correlate within a particular sector to help gauge the relative strength of a financial ratio. This would suggest that a deviation from the norms within a sector would show up quicker and would thus, be a better indicator for a change in bond rating.

There are three bond forecasting models. They include: an overall company model, service sector model, and a consumer non-cyclical sector model. The model’s independent variables determine perhaps 53% to 73% of a bond rating. The remaining 30-50% of a bond rating is most likely composed of qualitative factors. The equations for the models are:

Overall Company Model:
Y = 5.00899 + 0.05315 X1 - 0.00000 X2 - 0.02538 X3 - 0.08090 X4

Service Sector Model:
Y = -0.71192 + 0.14356 X1 + 0.00000 X2 - 0.05311 X3 + 0.06913 X4

Consumer Non-Cyclical Model:
Y = 4.61031 + 0.03105 X1 + 0.00001 X2 - 0.04620 X3 + 0.02256 X4

Where:
X1 is the leverage ratio
X2 is the size (Total Assets)
X3 is the return on equity
X4 is the fixed charge coverage

FORECASTING BOND RATINGS

Six companies have been selected to forecast their ratings. Of the six, three of the companies are from the non-cyclical sector and three are picked from the service sector. The companies from the non-cyclical sector are Anheuser-Busch Company, Dean Foods Company, and Tyson Foods. The service sector companies included: Clear Channel Communications, New York Times Company, and Sears Roebuck & Company. All of the companies are forecasted using the overall company model and the sector model that correlates to each company. Enron is also forecasted using the overall company model only.

To determine which model is most accurate the Mean Absolute Deviation is calculated. This is the average of the forecasting errors based on the absolute difference between the actual and the forecast ratings. Using the mean absolute deviations between the models, it is evident that the most accurate model is the overall company model. The overall company model has the lowest MAD of 1.

When comparing the distribution of bond ratings regressed against the financial ratios of each model, it becomes clear that the models tend to capture the relationships between financial ratios and bond ratings that are approximately between Aa1 to A3. This is due in large part because most of the ratings used to come up with the models fall within this range. This would suggest that this ratings range would be more precisely forecasted because the model, specifically the overall company model, is developed with a high degree of financial ratios that correlates to this range. Other ratings outside this range are less accurately forecast because the models do not have enough relationships between financial ratios and lower ratings to conclude more precisely a rating outside the range between Aa1 to A3.

It appears evident that bond ratings may be forecast to a certain extent by using selected financial ratios. Forecasted ratings are close to the actual ratings and the degree to which they vary may be captured in examining qualitative aspects of a firm. The forecasted ratings can become more precise if the forecasting models include a larger and more equally distributed sample of bond ratings. Furthermore, creating specific models to each company sector is more appropriate than an overall model because each sector has different industry averages associated with their particular financial ratios. Thus, a model is able to compare financial ratios carefully because the ratios are in a more narrow range taking into consideration the sector effect. Financial ratios need to be compared to industry averages to determine if they are good or bad for a particular industry. The sector models are able to capture a more accurate degree of financial ratios that are outliers and will cause a ratings change.

Another consideration is used when using the models to forecast future bond ratings. The observed time period is a 5 year period. During this 5 year period the economy was going from a boom to a recession. The stage of the economy, whether it be recessionary or boom, is critical when forecasting. Forecasts are to be done with data used to create the forecasting model that matches the current stage in the economy. This creates a more precise environment in which to forecast.

In addition, a model is beneficial to a firm interested in determining the risk associated with their firm. This is an inexpensive way to find out if it is worth going to a ratings agency. Most companies want the highest rating possible. By using the forecasting model a company will determine an approximation of their rating. If the forecasted rating is low than they should take measures to correct it before they get a ratings agency to rate them because once they are rated at a low level, their cost of debt may increase. In addition, it is harder to find outside investors in their company if the rating is below investment grade. The model will save money, by saving costs in evaluating debt and by giving the company additional time to increase its rating which in turn will decrease its cost of debt.

Originally, the research conducted on bond ratings was trying to determine if a model could have predicted the downfall of Enron. This is not the case because Enron had falsely created their financial statements. The model does not tell us that Enron created false financial statements but rather what bond rating is associated with the inaccurate financial statements.

New rules of reporting accurate financial statements need to be put into place to ensure investors that what they see is what they get. This might mean new GAAP accounting standards or possible penalties such as jail time and fines. Until investors are certain that financial statements are accurate then no one can be sure of their investments. This uncertainty makes for an inefficient market because false information is accounted for in bond ratings and stock prices. As investors remain unsure of corporate financial statements, investor confidence may decline.

REFERENCES

  1. Hawkins, David F., Barbara A. Brown, Walter J. Campbell. Rating Industrial Bonds. 1983.

  2. History of Standard & Poor’s. Standard & Poor’s. June 20,2002 http://www.standardpoor.com/AboutUs/History.html.

  3. Introduction to Moody’s. Moody’s Investors Service. June 15, 2002 http://www.moodys.com/moodys/cust/staticcontent/2000200000265777.asp?section=about&topic=intro.

  4. Jewels, Jeff, Miles Livingston. A Comparison of Bond Ratings. Boston: Blackwell, 1999.

  5. Mergent FIS, Inc. Mergent Bond Record. New York: Mergent FIS, December 1996-2001.

  6. Moody’s Rating Approach. Moody’s Investors Service. June 15, 2002 http://www.moodys.com/moodys/cust/staticcontent/2000200000265776.asp?section=about&topic=rapproach.

  7. Standard & Poor's 2002 Corporate Ratings Criteria. “Ratings Methodology”. Standard & Poor’s. June 20, 2002 http://www.standardpoor.com/ResourceCenter/RatingsCriteria/CorporateFinance/2002CorporateRatingsCriteria.html.

--top--

Back to the Journal of Undergraduate Research