Accounting Determinants of Credit Ratings



The Role of Accounting Determinants in Predicting Long Term Credit Ratings

Gorana Roje

Research Assistant

Institute of Economics, Zagreb

Croatia

groje@eizg.hr

Abstract

This paper tests the model of accounting determinants that are anticipated to predict S&P long-term credit ratings. Specifically, it examines the relative importance of various accounting variables in simplifying the process of determining credit ratings. Further, the study explores if various accounting determinants, that prove to make impact on predicting long term credit ratings, vary across industries. The model appears to be suitable for predicting the long-term S&P credit ratings. Further still, the model is additionally simplified. The author tests the hypothesis that although the model is additionally simplified, the predictions for long term credit ratings remain almost as good as for the original model with a very small loss in model explanatory power. Specifically, the study gains to distinguish sets of variables that proxy for some common factors with influence on credit ratings. The analysis of empirical data points out the distinctive differences between industries as far as the accounting determinants of credit ratings are concerned.

Key words: credit ratings, accounting determinants, simplified model, cross- industry disparity

JEL classification: M 41

Acknowledgements

I am grateful to my numerous colleagues for worthy suggestions.

Introduction

The goal of this paper is to investigate the relation between the accounting variables that can influence financial analysts in determining companies' credit worthiness, as well as their influence in predicting long term credit ratings. When assessing long term credit ratings, Standard and Poor(S&P) uses its own proprietary system. The information S&P uses to determine rankings of companies’ financial strength is obtained from a combination of both public sources (e.g., the annual report and accounts) and private information (e.g., managerial statements). Cantor and Packer (1995) show that credit rating agencies often use both quantitative and qualitative information when formulating their rating of a company’s financial condition.

This paper, however, does not analyse the way credit rating firms formulate credit ratings, but tries to determine whether the process of assessing creditworthiness by reducing the number of accounting variables and further more by distinguishing sets of variables, as factors, can be somewhat simplified.

Overall, the paper seeks to gain a better understanding of the relationship between accounting variables and credit ratings. Specifically, the study distinguishes sets of variables that proxy for some common factors, which further on influence credit ratings. By distinguishing sets of variables, the model is additionally simplified with a small loss in explanatory power. Finally, the study shows how various credit determinants differ across industries.

In the first part of the paper a more detailed description of data is given. Research questions and appropriate research techniques are developed in the second section, while data analysis and results are provided in the third part of this paper.

1. Data sample

The sample consists of firms from Compustat database[1]. Industrial Annual file, containing annual accounting data, was used. The initial sample consists of all firms that have credit ratings in years 1998-2002. There are 10,940 such firm-year observations. The sample is reduced due to the availability of the accounting data on such firms needed for further analysis. Consequently, 58.4% of observations were lost and the remaining final sample resulted in 4,550 firm-year observations, around 900 per year.

The financial institutions are excluded because the structure of their financial statements is very different from the financial statements' structure of firms in a non banking sector. Consequently the accounting determinants that are considered to be important in a non banking sector are not the same as the accounting variables common for firms in the data sample of this paper.

Description of variables

The dependent variable in the model is an S&P long-term credit rating, assigned to each company in a particular year. The credit ratings range from AAA to D. There were 22 categories which were initially reduced to 8. However, since the highest and lowest categories included a low number of observations, they were merged with the nearest categories. Consequently, the study compiles 6 credit ratings ranging from 2 to 7, where 2 is the highest rating and 7 is the lowest rating. The list of ratings is as follows:

• 2 (AAA, AA+, AA, AA-) – indicating categories that consist of firms with very strong capacity to pay interest and repay the principal,

• 3 (A+, A, A-) – referring to firms that indicate a strong capacity to pay interest and repay the principal. However, these companies are somewhat susceptible to adverse changes in particular circumstances and economic conditions,

• 4 (BBB+, BBB, BBB-) – indicating firms with an adequate capacity to pay interest and repay the principal. Still, adverse economic conditions or changing circumstances are more likely to lead to a weakened capacity to meet payments than in a case of firms of higher ratings,

• 5 (BB+, BB, BB-) – representing a credit rating that indicates less near-term vulnerability to default than other speculative issues. However, firms with such credit rating face major ongoing uncertainties or exposure to adverse, financial or economic conditions that could lead to inadequate capacity to meet timely interest and principal payments,

• 6 (B+, B, B-) – indicating firms with greater vulnerability to default but still with current capacity to meet interest payments and principal payments. Adverse financial or economic conditions will likely impair capability or willingness to meet these payments.

• 7 (CCC+, CCC, CCC-, CC, C, D) – indicating firms with an identifiable current vulnerability to default. These firms are dependant upon favourable conditions to meet timely interest payments and repayment of principal. In the event of adverse conditions these firms are not likely to have the capacity to pay the interest or principal. D stands for firms with payments in default.

The frequencies for the credit rating are given in Table 1.

|Table 1. Credit rating’s frequencies |

|Ordered value |Credit rating |Total Frequency |

|1 |7 |87 |

|2 |6 |791 |

|3 |5 |1055 |

|4 |4 |1339 |

|5 |3 |1006 |

|6 |2 |272 |

When checking for the normality, some of the independent variables exhibited quite big departures from normality[2]. Therefore, all the variables were transformed into deciles in order to achieve normal distribution. This reduced the possible influence of outliers without deleting observations.

This procedure is also motivated by the fact that credit analysts do not only look at the absolute values of ratios but rather at the relative values when comparing several companies. The fact that one company has 20 times higher sales growth ratio than another could indicate that it is a start-up company or company that merged with another. However this ratio, used as an absolute measure, could have unreasonably high impact on the results and make more difficult to uncover the certain relations such as the relation between sales growth and credit worthiness. Furthermore, it is easier to compare and interpret the results when the independent variables have a common scale.

The independent variables and author's predictions concerning their influence on S&P long-term credit ratings are as follows[3]:

1. ROA= Net Income before Extraordinary Items/ Total Assets

2. ROE= Net Income before Extraordinary Items / Book Value of Equity;

3. PROFIT= Net Income before Extraordinary Items / Sales

The inclusion of the profitability ratios (ROA, ROE, PROFIT) was necessary since they indicate the financial strength of the company and as such they influence the credit rating. They should be quite highly correlated since they have the same nominator. That is why the logistic analysis is run both with profitability ratios as separate independent variables and with an aggregate profitability ratio expressed as the average of ranked profitability ratios listed above.

4. Market value of equity = Number of common shares outstanding *share price

Market value of equity generally implies the size of the company. Size should reduce the company risk. Bigger companies can be more diversified and this reduces the risk when certain business segments face adverse market conditions. Besides, bigger companies have more market power and more opportunities to secure additional financing when needed. Higher market value of equity should increase credit rating.

5. BV_TANG = Tangible Book Value / Assets

Tangible Book Value[4] represents a sort of a security for the bondholders, in case of liquidation. The intangible assets are excluded due to the uncertainty of future benefits that may be worthless in the case of bankruptcy, e.g. goodwill. Study predicts that the higher ratio reflects more equity based on tangible assets and more security and better ratings.

6. LEVERAGE= Total Liabilities / Total Assets

Leverage addresses the proportion of assets financed by liabilities. Higher ratio means higher indebtedness, so the company becomes more risky and the creditors are less secured by equity. On the other hand the higher ratio can mean that the company is doing well and is able to secure more financing as well as to use this leverage for faster growth. Nevertheless, when comparing two otherwise identical companies the analyst would assign higher credit ratings to the one with lower leverage.

7. LTDEBT= Long Term Debt / Total Assets

Long-term debt ratio represents the part of total assets financed by the long term portion of liabilities. The prediction is similar to the one for leverage.

8. CURRENT RATIO= Current Assets / Current Liabilities

Current ratio shows the short-term liquidity. In general a higher ratio indicates that the firm has a higher ability to cover the short-term liabilities with current assets that can be transformed into money much faster than long-term assets. However, this ratio is very different in different industries and the lower ratio could indicate, for example, that the company has a stable and financially strong position and can base more of its financing on cheaper short-term financing rather than on long-term financing. This study does not have a strong prediction for this ratio.

9. PEN = (Projected Benefit Obligation-Pension Plan Assets)/ Total Assets

The accounting treatment of defined benefit pension plans makes pension liability an off-balance sheet item which is not included in debt or assets. The higher liability means that there is additional liability that will have to be covered in the future. If negative, this ratio indicates that company has an additional asset and it will have to put less money in the pension plan in the future. Thus, the higher ratio should refer worse credit ratings.

10. PRSTDS= Standard deviation of earnings / Total assets

Volatility of earnings is calculated as a standard deviation of net income before extraordinary items for the last 4 years prior to assessment of credit rating divided by total assets. It should have negative impact on credit ratings since the changes in earnings indicate more risk and anticipate the danger that in case of adverse conditions it could lead to default in the debt or even bankruptcy.

11. SGROWTH = ((S3-S4)/S4+ (S2-S3)/S3+ (S1-S2)/S2)/3;

Growth in sales is calculated as the arithmetic average of growth in sales over the last 3 years prior to assigning credit ratings. The sales growth was introduced in the model to see how analysts perceive the sales growth. At this point no strong predictions concerning the influence of this variable can be stated. The higher sales growth could indicate favourable conditions of the company and industry. Still, fast growth may relate to aggressive growth and risky projects being undertaken by the management.

12. DUMMY VARIABLES

Year dummy variables were included (1998-2001) in order to show if the overall economic conditions influence the credit ratings. Dummy variables for industries were also included to explain whether certain industries are more or less affected by some other factors that are not accounting related[5].

2. Research Questions and Techniques

In an effort of constructing a model for predicting credit ratings by reducing the number of accounting determinants used by S&P, the following research questions are addressed:

1. How well would this model predict long-term credit ratings?

2. How various accounting variables weight in relative importance of their impact on credit rating?

3. Can the model be further simplified by distinguishing sets of variables that proxy for some common factors that make influence on credit ratings?

4. How do various credit determinants differ across industries?

Referring to the first research question, this paper tests the following hypothesis:

A simple accounting model built in this paper, by reducing the number of accounting determinants used by S&P, is to yield good predictions for long terms credit ratings.

Referring to the third research question, this paper tests the following hypothesis:

The model further simplified by extracting factors, will yield almost as good predictions for long term credit ratings as the original model addressed in the first research question does and will not result in significant loss of the explanatory power.

Since the dependent variable of the model consists of several ordered categories (from 2 to 7) this study uses logistic model. The OLS was not appropriate because this model does not contain categorical variable. The dependent variable, Credit Rating has six ordered values. Also, the distance between two values is not equal to the distance between other two categories, and that does not satisfy the assumption for the OLS regression to be applied. The distribution of the dependent variable is not continuous but discrete.

Main model to be used is ordered logistic regression. This model addresses the first two research questions. The third research question will be addressed by conducting factor analysis. ANOVA analysis is a technique used in providing the results for the last research question.

3. Data Analysis and Results

Logistic procedure

In addressing the first research question and first hypothesis, this study uses Logistic procedure. The Probit procedure gives similar results, e.g. signs and significance of estimated coefficients[6].The overall model is significant and all variables are significant at 5% level. The R-square from this logistic model is 0.57[7]. This implies that the simple model explains 57 % of variation in credit ratings. When running the logistic model with profitability ratios separately, the leverage and ROE are not significant. This could result from multicollinearity since the correlation coefficients for profitability ratios range from 0.76 to 0.87. Multicollinearity check was performed by running OLS and the variance inflation factors (from here on VIF) were all below 10. However, the VIF related to profitability ratios are quite high, ranging from 4.94 to 8.76[8].

The coefficients and significance levels are provided in Table 2. As the reader can notice, the coefficient on profitability is negative and this means that the higher profitability is associated with better credit rating, as expected. The size[9] is also predictably negative and the magnitude of the coefficient indicates that it is a very important variable. The negative coefficient on the ratio Tangible Book Value to Total Assets[10] indicates that the creditors evaluate a certain company as less risky if the Book Value is greater. However reader must keep in mind that the intangibles are not included here for the intangibles are not highly valued as security.

Looking at table 2 in more detail, the positive sign on Leverage and Long-Term Debt indicate that the more debt as a percentage of Total Assets the company is carrying the lower credit ratings it receives. It turns out that greater Current ratio, as positive coefficient, is associated with lower rating. This is to some extend surprising because the higher Current ratio means that the company has more short-tem liquidity and this should limit the default and bankruptcy risk. However, as it is stated, there are other factors that could cause companies with higher credit ratings to have lower current ratio. In addition, volatility of Earnings[11] suggests that the higher volatility also has the predicted effect of lowering the credit ratings by increasing the perceived risk.

The sales growth turned out to decrease credit ratings but the significance level is lower than for other variables. Also the magnitude of the coefficient is very small. In general the sales growth variable seems to be less important than other variable. This result can be explained by the fact that due to recent sales growth managers are eager to undertake certain risky investments just because they assume the sales growth to continue increasing. However, the analysts are aware of the risk that the sales growth can decline and that the over-invested company can face big problems in the future.

The dummy variables for the four years from table 2 indicate that the credit ratings were better in 1998 and 1999 in comparison to 2002. It indicates that probably better general economic conditions in those years resulted in generally better ratings assigned by credit analysts. The credit ratings for 2001 were the worst, probably because this marked the start of the recession and the outlook for the future was quite bleak.

|Table 2. Analysis of maximum likelihood estimates |

|Parameter |DF[12] |Estimate |Standard Error |Wald Chi-Square |Pr > Chi Sq |

|Intercept 7 |1 |-49.879 |0.2409 |4.285.622 | ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download