


















Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Community
Ask the community for help and clear up your study doubts
Discover the best universities in your country according to Docsity users
Free resources
Download our free guides on studying techniques, anxiety management strategies, and thesis advice from Docsity tutors
A senior capstone project exploring the relationship between bond rating changes and the stock market. The author investigates whether bond ratings impact stock prices, the information content of bond rating downgrades, and whether equity markets react faster than rating agencies to changes in a company's financial condition. The project uses data from Mergent's Fixed Income Securities Database and analyzes abnormal returns following bond rating downgrades.
What you will learn
Typology: Summaries
1 / 26
This page cannot be seen from the preview
Don't miss anything!
ADVISOR • Alan Olinsky
Submitted in partial fulfillment of the requirements for graduation with honors in the Bryant University HonorsProgram
APRIL 2013
Senior Capstone Project for Michael Leonard
INTRODUCTION
There was a lot of criticism regarding bond rating agencies as a result of the financial crisis. The quality of the ratings these companies assigns to bonds was called into question and heavily criticized. People asked what the ratings were worth and if they were accurate or up- to-date. The rating agencies were blasted for assigning some of their highest ratings to the financial instruments most closely tied to the crisis, instruments which ended up being very risky and worthless. However, many of these instruments are traded in over-the-counter markets with little data available regarding the price of the product and the underlying assets and the value of them. Corporate bonds are rated by the same agencies, and these ratings are publically available and comprehensive historical records of them are kept. In addition, these bonds are issued by entities that also have stock traded in public markets. Similar to the bond ratings information, the current and historical price data for stocks is easily accessible. This provides a variable, the company’s stock price, which could be related to the rating of the company’s bonds, and could allow us to test the value of a rating. To do this, we can look at the impact of a ratings downgrade on the stock market price of the company.
Corporate bonds are rated by three agencies: Moody’s, Standard & Poor’s, and Fitch Ratings. The ratings range from the highest, AAA, to D or DDD, in default, and there are between 21 and 23 different ratings depending on the agency. A chart of the different ratings by each agency is provided immediately following this paragraph to help illustrate the ratings. The corporate bonds are reviewed by the agencies and given an initial rating at the time of creation. Thereafter, the bond ratings can be changed for two reasons, either a company
Senior Capstone Project for Michael Leonard specific event or a non-company specific event. A company specific event is any event that impacts the condition or structure of the firm. These events can include a new equity or bond offering which affects the capital structure of the firm, the retirement of debt which also affects the capital structure of the firm, or the acquisition of another company. If these or any other material company specific events take place, the rating agencies will review the ratings of the company’s bonds. If there is no company specific event that causes the rating agencies to reevaluate the ratings they have assigned to a company’s bonds, they will reevaluate the ratings after a certain period of time, possibly a year or 18 months. This is what is referred to as a non-company specific event, and this reevaluation is simply due to a certain amount of time passing since the last evaluation. When a bond is reevaluated by one of the agencies, there are several different actions the agency can take as a result of the reevaluation. The rating can be upgraded, downgraded, or put on watch, which means although it has not yet deteriorated enough to warrant a downgrade, due to the company’s condition it is at risk of being downgraded in the future. If the bond is downgraded, it can be downgraded one step or level, or multiple steps or levels. As an example of what a step or level is, using the S&P ratings in the chart on the next page, a one-step change would be from A+ to A, or BB to BB-, where a two-step change would be BBB+ to BBB-, or B+ to B-, and a three and four or more step changes follow the same pattern.
Senior Capstone Project for Michael Leonard being downgraded. A final question, which was also addressed in some previous papers, is whether equity markets overall are quicker to react than rating agencies are to a change in a company’s financial or operating condition. To summarize these questions into a statistically testable hypothesis, I will look for an abnormal return in the stock to determine if the bond rating change has an impact.
LITERATURE REVIEW
As stocks and bonds have been around for over a century, there has been much academic work on all aspects of these financial instruments and the companies that issue them. As time goes on, some of this work becomes outdated because of changes in the markets such as the advancement of technology. There are also new things to look at, or old things to look at but in a different way. Although it was not related to my topic as closely as some of the other work I will discuss later in this section, Craig MacKinley’s work entitled “Event Studies in Economics and Finance” was a very informative paper that I looked at often during my analysis. This paper presents a comprehensive discussion of event studies as they pertain to finance, specifically the stock market. It presents several different styles of event studies, and then presents an example using these different styles. This paper was very informative, written almost like an instruction manual on performing an event study. Although it is from 1997, the information it presents is just as relevant today and this was one of the more vital papers I uncovered during my literature review.
The papers I will now discuss are more similar to my work in that they look at bond rating changes and stock prices, either individually or the relationship between the two. In 1978,
Senior Capstone Project for Michael Leonard Pinches and Singleton published a study entitle “The Adjustment of Stock Prices to Bond Rating Changes.” This research, published over 35 years ago, is the most similar to my own work. In the paper, the authors examined monthly stock returns surrounding a ratings change. This paper found that, in certain months prior to either an upgrade or a downgrade, cumulative abnormal returns were present. These returns were negative for downgrades and positive for upgrades. However, this paper also found that after the ratings change took place, cumulative residuals for both upgrades and downgrades were stable and in some instances actually moved slightly opposite to the direction of the ratings change. This paper concluded that changes in a firm’s condition are recognized by the investment community before they are recognized by bond rating agencies and reflected in the ratings. It attempted to determine the length of this lag, and concluded that it could be between six months and two years depending on the situation. A main difference between this study and my work is the time scale used. This research used monthly returns and looked at longer term affects, while I used daily returns and focused on the immediate impact in the market.
My paper looks at the immediate impact of a ratings change; however, some studies look at the long-term impact. Dichev and Piotroski’s 2001 paper “The Long-Run Stock Returns following Bond Rating Changes” is one such paper. It discusses several important ideas. First, it claims that underperformance following downgrades is larger for small or low-credit quality firms. This could be due to greater information inefficiencies for smaller companies due to less analyst coverage. The paper also concludes that the market fails to fully anticipate the negative implications of downgrades on future firm performance and that the market underreacts to these downgrades. This work failed to find any abnormal returns following
Senior Capstone Project for Michael Leonard METHODOLOGY
Event Study
An event study is a method commonly used in the fields of finance and economics when one is seeking to determine the impact of an event on a particular variable. In this paper, the event being examined is a ratings downgrade for a particular company’s bond, and the variable of interest is the stock price. However, before looking at how the stock price acted on the day of, and the days surrounding the event, it is important to determine how we expected the price to act if there was no event. There are a few choices for determining the expected return, and while some studies use an average of the returns over some period of time prior to the event, in this paper I use a market model, allowing me to make a more accurate prediction of expected return.
The first step in this market model is to select the estimation window. The estimation window is some period of time prior to the event at time 𝑡 0 , over which the relationship between the movement of the market and the movement of the stock will be calculated. We also need to select the event window. The event window is the day of the event, 𝑡 0 , and the days surrounding the event during which we will look for an unexpected change in the stock price of the company. It is good practice to not include the event window as part of the estimation window, as using the data from the event window in the estimation can impact the integrity of the estimation. Once the estimation and the event window have been selected, the following market model is used to calculate beta, 𝛽:
Senior Capstone Project for Michael Leonard
(1) Market Model: 𝑟𝑠 = 𝛼 + 𝑟𝑚 𝛽 + 𝜀
where 𝑟𝑠 is the return of the stock, 𝛼 is excess return, 𝑟𝑚 is the return of the market, 𝛽 is the correlation of volatility between the market and the stock, which is what we are looking for,
and 𝜀 is a random error which is assumed to be normally distributed with a mean of zero. A value-weighted market index from the Center for Research in Security Prices (CRSP) using data from NYSE, AMEX, NASDAQ, and Arca was used as a proxy for the market return. A simple linear regression using ordinary least-squares is run on the data to calculate 𝛽. This number represents the correlation between the movement of the market and the movement of the stock. A beta of one means that for each one percent movement in the market, the stock will move one percent in the same direction, while a beta of one-half means that for each market movement of one percent, the stock will move half a percent in the same direction. This correlation value, 𝛽, will be used to calculate the expected return during the event window.
Calculating the abnormal return during the event window is the next step in this study. To calculate the abnormal return, 𝐴𝑅, I subtract the expected return, 𝐸(𝑅), from the actual return, 𝑅. This relationship is given by the equation
Senior Capstone Project for Michael Leonard where 𝑁 is the total number of events. In addition, the variance of CAR is given by the equation
𝑁 𝑖=
where 𝑁 again is the total number of events, and 𝜎𝑖 is the standard deviation of 𝐴𝑅𝑖. To aggregate the abnormal returns over all the days in the event window collectively, equation 5 and 6 are reused using the abnormal returns and the standard deviations for all the days.
Once the cumulative abnormal return, CAR, and the variance of it have been calculated, we have the numbers required to calculate a test statistic and determine if the result is significant. Our hypothesis, which we are looking to disprove by finding a statistically significant result, is 𝐶𝐴𝑅 = 0, meaning the ratings downgrade has no effect on stock price. Our alternative is 𝐶𝐴𝑅 ≠ 0 , meaning the ratings downgrade has a statistically significant effect on stock price. Our test statistic is given by the equation
Once we have this test statistic, it is used to determine if 𝐶𝐴𝑅 is statistically significant from zero and, if so, at what significance level.
Senior Capstone Project for Michael Leonard Data
The bond ratings data was all retrieved from Mergent’s Fixed Income Securities Database. This is a comprehensive database of large, publicly traded fixed income securities, with over 140,000 securities in the portfolio including corporate, supranationals, U.S. Agency, and U.S. Treasury issues. The stock data came from the Center for Research in Security Prices, CRSP. This database has stock and indices returns going back to 1925, and a huge selection of data items to choose from. It also has its own indices, including the one used earlier as a proxy for the return of the market. Finally, SAS was utilized for all of the data cleaning and analysis. The stock universe used was S&P 500 companies in the following sectors: Consumer Staples, Financials, Industrials, and Materials. Appendix A contains a list of the companies included. These sectors were selected as they tend to vary in their volatility as well as other price- related characteristics, helping the study to be well-rounded in the securities analyzed. The date range used for the study was January 1st, 2001 through September 30, 2011. This produced a total of 172 companies and 20,271 downgrades. However, many of these downgrades were duplicates, meaning the agencies downgraded several bonds for the same company on the same day. After eliminating the duplicates, the downgrades were separated by the number of steps involved. The downgrades were separated into four groups: one-step changes, two-step changes, three-step changes, and four-plus step changes. The graph below gives an overview of how many downgrades existed after removing the duplicates.
Senior Capstone Project for Michael Leonard scatterplots of the abnormal returns for one-, two-, and three-step changes using a three-day event window.
Three-Day Event Window
Table 2 below shows the results of the test of one-step changes using a three-day event window. The CAR on all three days was very small, although slightly negative. However, a t- score of -1.96 is needed for a 95% confidence level, and the t-scores calculated are not close enough for any high level of significance, and therefore we are unable to reject the null hypothesis. There is no evidence to support abnormal stock returns on the day of and days immediately preceding and following the downgrade of a company’s bond rating for one-step downgrades.
Table 3 on the next page shows the results of the test of two-step changes using a three-day event window. The CAR on all three days was very small, although slightly negative, and on the day after the event larger than for one-step changes. However, as stated previously, a t- score of -1.96 is needed for a 95% confidence level, and the t-scores calculated are not close enough for any high level of significance, and therefore we are unable to reject the null hypothesis. There is no evidence to support abnormal stock returns on the day of or days
Day CAR t-Score -1 -0.004707 -0. 0 -0.005323 -0. 1 -0.000611 -0. Overall -0.003547 -0.
One-Step Downgrades: Three-Day Window
Table 2
Senior Capstone Project for Michael Leonard immediately preceding and following the downgrade of a company’s bond rating for two-step downgrades.
Table 4 below shows the results of the test of three-step changes using a three-day event window. Again, while the CAR is negative, and is slightly larger for all three days compared to one- and two-step changes, a t-score of -1.96 is needed for a 95% confidence level. The t- scores calculated are not close enough for any high level of significance, and therefore we are unable to reject the null hypothesis. There is no evidence to support abnormal stock returns on the day of and days immediately preceding and following the downgrade of a company’s bond rating for three-step downgrades.
Day CAR t-Score -1 -0.017782 -1. 0 -0.002620 -0. 1 -0.005967 -0. Overall -0.008790 -0.
Two-Step Downgrades: Three-Day Window
Table 3
Day CAR t-Score -1 -0.011256 -0. 0 -0.004641 -0. 1 -0.010958 -0. Overall -0.008952 -0. Table 4
Three-Step Downgrades: Three-Day Window
Senior Capstone Project for Michael Leonard
Table 7 below shows the results of the test of three-step changes using a five-day event window. Again, while the CAR is negative and is slightly larger for all 5 days compared to one- and two-step changes, a t-score of -1.96 is needed for a 95% confidence level, and the t- scores calculated are not close enough for any high level of significance, and therefore we are unable to reject the null hypothesis. There is no evidence to support abnormal stock returns on the day of and on the two days immediately preceding and following the downgrade of a company’s bond rating for three-step downgrades.
Day CAR t-Score -2 -0.024875 -0. -1 -0.011256 -0. 0 -0.004641 -0. 1 -0.010958 -0. 2 0.010181 0. Overall -0.008310 -0. Table 7
Three-Step Downgrades: Five-Day Window
Day CAR t-Score -2 -0.003650 -0. -1 -0.017782 -1. 0 -0.002620 -0. 1 -0.005967 -0. 2 0.002508 0. Overall -0.005502 -0.
Two-Step Downgrades: Five-Day Window
Table 6
Senior Capstone Project for Michael Leonard CONCLUSION
Based on my analysis and results, there is no evidence to support the claim that the downgrade of a company’s bond rating has an impact on its stock price on the day of and on the days surrounding the downgrade. Of the several individual tests I ran, none of them had a t-score high enough to confirm at any level of significance that the CAR was statistically different from zero. Although the CARs did grow slightly as the number of steps increased, none of the CARs were statistically different from zero, so we cannot conclude that a larger number of steps results in a larger CAR. I can conclude that the information content of bond rating downgrades to equity investors is very low. The firm deterioration that results in this downgrade has most likely already been recognized by investors and the stock price already reflects the deterioration. The findings also confirm the existence of a ratings change lag, the period of time after equity investors recognize firm deterioration compared to when rating agencies react to this change. Although there could be a long-term impact as some previous studies have concluded, I would argue this is not necessarily a result of the downgrade, but rather that a downgrade is foreshadowing continued firm deterioration resulting in stock price decreases.
One final question that I asked when reflecting on my research is whether different results, and statistically significant results, could be found by looking at different sectors individually. When I selected the four sectors I looked at, I consciously selected sectors that had different volatility and characteristics, so I had a well-rounded universe of stocks. However, I began to wonder if more volatile sectors, such as Financials, might have a statistically significant abnormal return which is being masked by the lack of abnormal returns in the Consumer