The influence of credit rating changes on capital structure

(1)

The influence of credit rating changes on

capital structure

Menno Altena1 Master Thesis Finance University of Groningen

January 2014

Abstract: This paper examines the relation between credit rating changes and the capital structure of firms. Fitch credit ratings in the period 2002 till 2012 are used to test the reaction of managers on credit rating changes. The results show that firms near a rating change issue around 1.4% more debt relative to equity compared to firms not near a rating change. European firms react significantly stronger on credit ratings changes than North American firms. Firms around the investment/speculative grade issue around 3.7% less net debt relative to equity and are significant more sensitive to near rating changes than other firms.

Keywords: credit ratings, capital structure, CR-CS, net debt issuance, financing policy JEL classification: G32

1

University of Groningen, Faculty of Economics and Business, MSc Finance Student number: s1909320, Email: menno.altena@gmail.com

(2)

2 1. Introduction

The relation between credit ratings and capital structure is a relative new topic in the financial literature. Graham and Harvey (2001) were the first who explicitly mentioned the importance of credit ratings in determining capital structure. They found in a survey along CFO’s about capital structure that CFO’s see credit ratings as the second most important factor in determining the capital structure of a firm. Kisgen (2006) was the first who actually studied the effect of credit rating changes on capital structure. He argued that different credit ratings have different discrete costs and benefits for firms. This can for example be that firms with higher credit ratings can easier borrow money at lower costs because investors’ see them as of higher quality (Michelsen and Klein, 2011). Because of these different discrete costs and benefits

associated with different rating levels firms near a credit rating change issue less net debt relative to equity to avoid being downgraded or to increase the chance of an upgrade (Kisgen, 2006). The results of Kisgen (2006) confirm this relation between credit ratings and capital structure. Firms near a credit rating change issue less net relative to equity compared to firms not near a rating change. Kisgen concludes that capital structure decisions are made with credit ratings in mind. This paper builds further on Kisgen’s (2006) model and recent literature to test the relation between credit ratings and capital structure. The main research question of this paper is: what influence do credit rating changes have on a firms’ capital structure? My hypothesis is that firms near a rating changes issue less net debt relative to equity to avoid being downgraded or increase the change of an upgrade. Furthermore this paper tries to provide answers to the following research questions: Do managers react symmetrically to credit rating upgrades and downgrades? Do firm which experienced a credit rating change immediately adjust their capital structure or do they adjust their capital structure slowly over time? Do firms near the investment/speculative grade react stronger on credit rating changes than firms not near the investment/speculative grade? Do euro-country firms react different on credit rating changes than North American firms?

This paper contributes to the existing literature about credit ratings and capital structure. Previous studies all used Standard and Poor’s credit rating data. I will use Fitch credit rating data to test the relation between credit ratings and capital structure. This study is the first in which the impact of credit ratings on capital structure for European firms2 is studied and the difference between European and North American firms is investigated. Furthermore, previous studies assumed that credit ratings have a linear rating scale from 24 (AAA) to 1 (D). As explained later I

2

(3)

3

think this linear rating scale is not the best way to describe the credit ratings and therefore I include a polynomial in the model to increase the explanatory power of the model.

The results show that firms near a credit rating change issue approximately 1.4% more net debt relative to equity compared to firms not near a rating change. Firms near an upgrade issue 1.8% more net debt relative to equity and for firms near a downgrade no significant results are found. These results are not in line with most of the literature. A possible explanation is that it is important for firms to keep their current debt levels and that this overweight’s the cost

associated with those debt levels. It might be that firms are not able to react immediately on near rating changes and that they therefore slowly adjust debt levels over time (Chowdhury and Maung, 2011). However, the results of this study provide no evidence that firms slowly adjust their debt levels after a rating change. The effect of near credit rating changes is much higher for firms around the investment/speculative grade compared to other firms. Firms around the investment/speculative grade issue approximately 3.7% less net debt relative to equity compared to firms not near the investment/speculative grade. Finally, the results indicate that European firms react stronger to near credit rating changes than North American firms. European firms issue around 3.7% more net debt relative to equity when near a rating change compared to 0.3% of North American firms. This difference can partly be explained by the higher leverage ratio and smaller size of North American firms in this sample. Higher leverage ratios and smaller firm size increase the probability of default for firms. A higher probability of default makes it more difficult for firms to borrow money and this can therefore explain the lower net debt issuance of North American firms.

The remainder of this paper is organized as follows. Section 2 gives a literature review in which the existing literature about the credit rating and capital structure relation is discussed. Section 3 contains the empirical design for testing the relation between credit ratings and capital structure. Section 4 gives a description of the data. Section 5 gives the empirical results and section 6 concludes.

2. Literature

(4)

4

ratings are of major influence in managers’ capital structure decisions due to the discrete costs and benefits associated with different rating levels (Kisgen, 2006). Recent literature mentions several examples for the relation between the different discrete cost and benefits associated with different rating levels. First, Michelsen and Klein (2011) argue that investors use credit ratings as an indicator of a firms’ quality and as an indicator of the probability of default of a firm. The higher the credit rating of a firm, the lower the probability of default and the more

interesting a firm is for investors. Secondly, Partnoy (2001) argues that investors rely heavily on credit ratings because credit rating agencies have impressive reputations for investors. Investors believe therefore that credit rating agencies can better judge a firm than the investors’

themselves. Finally, in the Basel II Accord capital requirements for banks are partly determined by credit ratings. This affects the firms in which banks’ can invest, firms with lower ratings are less attractive and this raises therefore the cost of capital for firms with lower credit ratings (Altman et al., 2002). Overall, a firm’s discrete costs and benefits are associated with different credit ratings (Kisgen (2006), Michelsen and Klein (2011)).The main hypothesis of Kisgen’s (2006) study to test the CR-CS is that firms near a rating downgrade should issue less debt (or more equity) to avoid being downgraded and firms near a rating upgrade should issue less debt (or more equity) to increase the chance of an upgrade.

The methodology of all studies in table A1 is based on Kisgen’s (2006) model to test the CR-CS. Kisgen developed a regression model to test his hypothesis with dummy variables that account for if a firm is near a credit rating change and some control variables to control for the financial condition of a firm. Identical to Kisgen (2006), all but one studies use the issuance of net debt (NetDIss) (defined as the net debt minus net equity divided by the total assets of a firm i at time t) as dependent variable. Only Chowdhury and Maung (2011) use the change in leverage (leverage is defined as the ratio of total debt to total assets) of a firm as dependent variable. The control variables used to control for the financial condition of a firm are mostly leverage,

profitability and size. Chowdhury and Maung (2011) included control variables for investment opportunities and cash levels and excluded the control variables used by the other studies (leverage, profitability and size).

Two types of credit rating data are used in the studies. Kisgen (2006, 2009), Kemper and Rao (2013b) and Chowdhury and Maung (2011) use actual credit ratings from Standard and Poor’s (S&P’s)3. Kisgen developed a model in which the underlying assumption is that firms with a so

3

(5)

5

called ‘plus’ or ‘minus’ rating (for example BB+ or AA-) are near a credit rating change. Michelsen and Klein (2011) and Kemper and Rao (2013a) use credit rating outlooks4 to test if near credit rating changes influence mangers’ capital structure decisions. They argue this is a more adequate measure if a firm is close to a rating change than the Plus and Minus approach from previous studies. All studies use S&P’s credit rating data and almost every study uses North American data. Only Michelsen and Klein (2011) use worldwide firm data.

In his first study to test the CR-CS, Kisgen (2006) found that firms near a credit rating change issue approximately 1.0% less net debt relative to net equity compared to firms not near a credit rating change. Firms near a credit rating upgrade or downgrade issue approximately 0.6% and 0.5% less net debt relative to equity. Kisgen’s follow up study tested how firms adjust their capital structure after an actual credit rating change. Kisgen (2009) found that firms that were downgraded in period t-1 issued between 1.5% and 2.0% less net debt in period t compared to firms that were not downgraded in period t-1. For upgraded firms Kisgen did not find significant results. Michelsen and Klein (2011), who used the S&P credit rating outlook to measure how close a firm is to a credit rating change, found economically much stronger results than Kisgen (2006) for firms with a negative credit rating outlook. Michelsen and Klein (2011) found that firms with a negative outlook issue approximately 2.1% less debt relative to equity. For firms with either a positive or negative outlook Michelsen and Klein found also significant results, but these results seem largely be driven by the firms with a negative outlook. Chowdhury and Maung (2011) and Kemper and Rao (2013a) both found remarkable results for respectively firms that were actually downgraded at time t and for firms with a negative credit rating outlook. Firms that experienced an actual downgrade at time t (Chowdhury and Maung, 2011) or firms with a

negative rating outlook (Kemper and Rao, 2013a) issued more debt relative to equity compared to firms that were not downgraded or not had a negative rating outlook. This is not in line with the hypothesis that firms with a negative outlook issue less debt to avoid being downgraded. Kemper and Rao (2013a) argue that firms near a rating downgrade borrow at time t before the cost of borrowing increases after the actual downgrade at time t+1. Another explanation is that the acquisition of funds is very important for the firms that issue more debt when near to a rating downgrade. For these firms the acquisition of firms is more important than the costs of these funds (Kemper and Rao, 2013a). Fama and French (2002) found that firms slowly adjust their debt levels to their target debt levels. Chowdhury and Maung (2011) therefore included previous

4

(6)

6

downgrades till time t-5 in their model and they found that firms slowly adjust their leverage ratios after a downgrade and firms increase their leverage ratios in the years following an upgrade. They found that it is probably difficult for firms to immediately adjust their debt levels and that firms do this slowly over time.

Overall, most studies found significant negative results for the relation between near credit ratings changes and capital structure. The effect of negative rating changes is stronger than the effect of positive rating changes which is in line with the CR-CS. It is more important for firms to keep their current rating than to increase the chance of an upgrade. Most studies shown in table A1 only use North American data and every study in table A1 uses S&P’s credit rating data to test the CR-CS.

3. Methodology

This section gives a detailed description of the methodology I use in this study. The

(7)

7

and capital structure. They included credit rating dummy variables that accounted for actual credit rating changes in year t or previous years.

As shown in table A1 most previous studies included the same three control variables in their model which control for the financial condition of a firm. These control variables may influence capital structure decisions of a firm as well and are the three variables describe below.5

In the regressions in this study the variable is included which is a set of control variables that control for the financial condition of a firm. The control variables included in are the

following:

Leverage: defined as total debt of a firm i at time t-1 divided by total debt plus total equity of a firm i at time t-1 ( /( + )). Firms with a higher leverage ratio have higher levels of financial distress compared to firms with lower leverage ratios. This indicates that when the variable leverage has a negative (positive) coefficient, firms with a higher (lower) probability of default issue less (more) debt relative to equity.

Profitability: defined as EBITDA (Earnings Before Interest , Taxes, Depreciation and Amortization) of a firm i at time t-1 divided by the total assets of a firm i at time t-1 ( / . Profitable firms should have lower financial distress levels and should therefore have higher credit ratings. This indicates that when the variable

profitability has a positive (negative) coefficient, firms with a lower (higher) probability of default issue more (less) debt relative to equity.

Size: defined as the natural logarithm of the sales of a firm i at time t-1 (ln( )). Generally, larger firms should have a lower probability of financial distress. This indicates that when the variable size has a positive (negative) coefficient firms with a lower (higher) probability of default issue more (less) debt relative to equity (Michelsen and Klein, 2011).

Table A2 in the Appendix describes all the variables used in this study. The first regression models are equation (1) till equation (4) which test hypothesis H1à. Equations (1) and (2) test the

CR-CS with a set of control variables and equation (3) and (4) do not control for the financial condition of a firm.

H1a: firms near a credit rating change issue less debt relative to equity as a percentage of total

assets compared to firms not near a credit rating change.

5

(8)

8

(1)

(2)

(3)

(4)

(9)

9

leverage ratios are not adjusted after the period t-3. Equations (5) till (8) below include the dummy variables for actual (previous) credit ratings and test the following hypothesis:

H2a: firms will slowly adjust their capital structure after a credit rating change.

(5)

(6)

(7)

(8)

Table A3 in the Appendix shows the Fitch credit rating scale which I use in this study. The Thomson and Reuters’ Datastream credit ratings are ranked on an a scale from AAA (24 points) to D (1 point). This scale implies that all one step rating changes (for example from AAA to AA+, of CC- to CCC+) are of the same size. However a rating change from AAA to AA+ can have smaller or larger effects than a rating change from CC- to CCC+. And a rating change from for example A+ (20) to AA (22) is not twice as good as a rating change from A+ (20) to AA- (21). Especially around the investment/speculative grade rating changes can be more sensitive than rating changes further away from the investment/speculative grade. This is supported by Cantor and Packer (1997) who mentioned that credit ratings are used as a threshold to determine if institutional investors may hold the debt of a specific firm. This means that falling below the

investment/speculative grade can have great cost impacts for a firm.

(10)

10

Minus variable is found significant. In the sample used in this study there is a relative high number of firms with an investment grade compared to firms with a speculative grade. I will therefore run the regressions of equations (1) and (2) for both subsamples of firms with an investment grade and firms with a speculative grade. Kisgen (2006) and Michelsen and Klein (2011) used a different approach to test if firms with an investment grade react different on near rating changes than firms with a speculative grade. Both studies included an additional dummy variable that measures if a firm is near the investment/speculative grade. Kisgen (2006) and Michelsen and Klein (2011) found significant results for firms near the investment/speculative grade6_{. In this study I include a dummy variable IG/SG that has a value of 1 if a firm is near the}

investment/speculative grade and 0 otherwise. For the definition of the IG/SG credit rating dummy the approach of Kisgen (2006) is followed. The dummy variable IG/SG is defined in two ways, first IG/SG exists of firms with a BBB- or BB+ rating and secondly firms with a BBB, BBB-, BB+ or BB are included in IG/SG. Equation (9) and (10) include this IG/SG dummy variable and test the following hypothesis:

H3a: firms near the investment/speculative grade react stronger to near credit rating changes

than firms not near the investment/speculative grade.

(9)

(10)

(11)

Including the IG/SG dummy variable still assumes that the 1 to 24 rating scale is the true rating scale. To capture this problem I include a third order polynomial in my original equation (1) and (2). With this third order polynomial I assume that around the investment/speculative grade rating changes are more sensitive than further away from the investment/speculative grade. Equation (11) and (12) below show the model with the polynomial included. There are no other studies which included a polynomial to give a better representation of the relation between credit ratings.

(12)

(13)

6

(11)

11

As mentioned before reactions on near credit rating changes might not be the same for all rating categories. Kisgen (2006, 2009), Kemper and Rao (2013a), Michelsen and Klein (2011) and Kemper and Rao (2013b) found different results per rating category in their studies. Therefore I will run the regressions of equations (1) and (2) for each rating category individual.

4. Data

The sample for this study exists of listed North American firms and euro country firms with a Fitch credit rating available in the Thomson and Reuters Datastream database between 2002 and 2012. Previous studies did not test the CR-CS for European firms specifically. In this study I therefore include European and North American firms to test the CR-CS for European firms and to test if there is a difference between reactions of European and North American firms. Fitch credit ratings represent the rating agency’s view of the relative vulnerability of a firm to default (Fitch Ratings, 2013). To my knowledge there are no other studies that used Fitch credit ratings to test the CR-CS. I exclude firms from euro countries that became a euro member during the period studied because their credit ratings might be influenced by the introduction of the euro7. Book values of the variables are used because Fitch emphasizes these (Fitch Ratings, 2013). Furthermore I exclude firm years in which a firm has missing data in the regular used variables in this model8. Finally, dead and alive firms are included in the sample to avoid the survivorship bias.

Summary statistics for the sample are shown in table 1 and figures 1 and 2. The final sample exists of 422 firms with 3,297 firm years. The sample is not equally distributed over all rating categories. Especially firm years with relative high credit ratings are over weighted in this sample. Therefore I will perform some tests to see if the results of this study are not mainly driven by one specific rating category. The sample contains of 269 North American firms with 2,141 firm years and 153 European firms with 1,156 firm years. Table 1 shows the mean, median and standard deviations of the book value based leverage ratios. Figure 2 shows that average debt levels increase as credit ratings decreases, but this relationship is weaker than expected. I expect that the mean of leverage ratios increase as credit ratings decrease because firms with relative high debt levels have a relative high probability of default (Fitch Ratings, 2013), and vice versa for firms with relative low debt levels. Kisgen (2006) found in his study a linear relationship between

7

Euro-countries which are included in the sample are: Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal and Spain.

8

(12)

12

leverage ratio and credit rating. In this sample the relationship is non-linear. However, on average firms with the lowest credit ratings have higher leverage ratios than firms with higher credit ratings. The four highest rating categories have an average leverage ratio of 53.1%, the four lowest rating categories have an average leverage ratio of 72.6%9. A remarkable outlier is the high average leverage ratio of 71.8% of firms with an AA+ rating which can be explained by the relative low numbers of firms with an AA+ rating (19) in the sample. Below the investment grade of BBB- the mean debt levels increase when credit ratings decrease.

Table 1

Summary statistics of leverage ratios per rating category

This table presents the mean, medians and standard deviations of the book value based leverage ratios as a percentage per rating category. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firms with missing values for regularly used variables are excluded from the sample. No. of firm years is the number of firm years for a specific rating category at the beginning of the firm year. D/D+E is the book long term debt plus book short term debt divided by the book long term debt plus book short term debt divided by the book value of shareholders equity. SD is the standard deviation of the leverage ratios. Leverage ratios greater than 1 or less than 0 are excluded from these summary statistics.

Credit rating AAA AA+ AA AA- A+ A

No. of firm years 42 19 128 307 431 451

Mean D/D+E 29.4% 71.8% 53.8% 57.2% 54.3% 45.8%

Median D/D+E 20.7% 81.4% 46.2% 58.2% 53.3% 39.4%

SD D/D+E 29.7% 30.4% 26.3% 25.2% 20.3% 19.5%

Credit rating A- BBB+ BBB BBB- BB+ BB

No. of firm years 386 412 486 215 107 120

Mean D/D+E 49.6% 45.1% 47.6% 41.5% 47.1% 51.6%

Median D/D+E 48.1% 43.1% 48.9% 40.4% 48.3% 49.6%

SD D/D+E 17.1% 16.3% 16.9% 15.0% 22.0% 17.7%

Credit rating BB- B+ B B- CCC+ or lower

No. of firm years 91 32 36 17 17

Mean D/D+E 52.6% 70.6% 66.6% 82.8% 70.6%

Median D/D+E 54.0% 79.0% 71.8% 86.1% 64.9%

SD D/D+E 19.9% 21.1% 23.7% 15.1% 21.5%

9

(13)

13

Figure 1: number of firm years per rating category. This figure shows the number of firm years per rating

category for the entire sample. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firms with missing values for regularly used variables are excluded from the sample.

Figure 2: mean leverage ratios per rating category. This figure shows the mean leverage ratios (D/D+E) as

a percentage per rating category for the entire sample. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firms with missing values for regularly used variables are excluded from the sample. D/D+E is the book long term debt plus book short term debt divided by the book long term debt plus book short term debt divided by the book value of shareholders’ equity. 42 19 128 307 431 451 386 412 486 215 107 120 91 32 36 17 _{7 9 0 0 0 0 1 0} 0 100 200 300 400 500 600

AAA AA+ AA AA- A+ A

A-BB B+ BB B BB B-BB + BB _BB- B+ B B-CC C+ CC C CC C-CC + CC _CC- C D N u mb er o f firm yera s

Number of firm years per rating category

y = 0.003x2_{- 0.034x + 0.581} 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% M e an (D/D+ E)

(14)

14

Figure 3 gives a graphical description of the debt and equity issuance of firms per rating category. Panel A of figure 3 shows the average net debt issuance per rating category and panel B and panel C of figure 3 show respectively the average debt and average equity issuance per rating category10. Panel A shows that firms with a lower credit rating issue less net debt relative to equity compared to firms with a higher credit rating. This is in line with financial distress arguments which state that firms with higher credit ratings issue more debt compared to firms with lower credit ratings because of their better credit quality (Kisgen, 2006). However, the CR-CS implies that firms near either an upgrade or a downgrade will issue less net debt relative to equity, so this can be at both sides of any broad rating category. Because of these differences between distress concerns and the CR-CS, the model should include control factors that control for the financial condition of a firm (Kisgen, 2006). The average net debt issuance per rating category in panel B indicates that firms with higher credit ratings issue more debt relative to equity compared to firms with lower credit ratings. This is in line with the CR-CS which states that firms with lower ratings try to increase their credit quality by issuing less debt. Furthermore, for firms with higher credit ratings it is easier to borrow money and therefore debt is more attractive for them compared to firms with lower credit ratings.

The issuance of net equity per rating category in panel C shows that firms with higher ratings issue more equity than firms with lower ratings. Firms with relative high credit ratings try keep their current rating and to avoid being downgraded by issuing more equity.

10

(15)

15 -4.00% -3.00% -2.00% -1.00% 0.00% 1.00% 2.00% 3.00%

AAA AA+ AA AA- A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B

N e t d e b t issuan ce ( N e tDI ss)

Panel A: Average net debt issuance

-3.00% -2.00% -1.00% 0.00% 1.00% 2.00% 3.00% 4.00% 5.00%

D e b t issuan e

Panel B: Average debt issuance

0.00% 0.50% 1.00% 1.50% 2.00% 2.50% 3.00% 3.50%

Eq

u

ity

issuan

ce

Panel C: Average equity issuance

Figure 3: average net debt issuance, debt issuance and equity issuance per rating category. This figure

shows in panel A the average net debt issuance (NetDIssi,t = ΔDi,t - ΔEi,t,)/ Ai,t)per rating category for the

(16)

16

Kisgen (2006), Michelsen and Klein (2011) and Kemper and Rao (2013a) exclude debt offerings larger than 10% from their sample. Kisgen (2006) argues that the CR-CS most directly applies to small- or medium-sized debt offerings. Small- and medium sized debt offerings have different implications than large debt offerings. A small debt offering might result in a downgrade for a firm close to a downgrade (a Minus firm) and not for a downgrade for a firm in the middle of a broad rating category. A large debt offering might always result in a downgrade, regardless of a firm is near a downgrade or not. It changes the capital structure in such a way that in almost every situation leads to a downgrade. Furthermore large debt offerings might also be a result of acquisitions, reorganizations and management changes which each in itself can have a large impact on a firm’s credit rating. Because of these arguments, the above studies excluded large debt offerings (>10%) from the sample11. In this study I will test the CR-CS without debt offerings larger than 10% and with all debt offerings included12. I test both samples to see if there is a statistically significant difference between the effects of near credit rating changes on capital structure for firms with large debt offerings and firms without large debt offerings.

In previous studies only Michelsen and Klein (2011) included European firm data in their sample (see table A1 in the Appendix). However, Michelsen and Klein (2011) did not look specifically at European firms in their study. In this study I will conduct a sensitivity analysis to test the CR-CS hypothesis on European firms. I will compare the results of European firms with the results of North American firms to see if they differ significantly from each other.

Table 2 presents the correlation matrix of the variables regularly used in the model. Credit rating changes are significant and positive related to leverage ratios and significant and negative related to EBITDA/A and ln(Sales). When correlations between variables are too high

multicollinearity may arise. Brooks (2008) argues that with multicollinearity standard errors are too high and that the regression will become very sensitive to small changes in the model. In this study some correlations are significant at the 1% level, but none of these correlations is that large that multicollinearity arises13.

11

Kisgen (2006) and Michelsen and Klein (2011) tested if the results were also robust when debt offerings larger than 5% and larger than 20% were excluded from the sample. In both studies the results were robust to those exclusions. The statistical significance was somewhat reduced, but the results were qualitatively identical.

12_{I also excluded debt offerings larger than 5% and 20% from the sample. These results give qualitative}

identical results compared to the results when debt offerings larger than 10% are excluded.

13

(17)

17 Table 2 Correlation matrix

This table presents the correlation matrix of the regularly used variables in this study. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size.

POM NetDIss D/(D+E) EBITDA/A ln(Sales)

POM 1.000***

NetDIss 0.004*** 1.000***

D/(D+E) 0.032*** 0.153*** 1.000***

EBITDA/A -0.048*** -0.210*** -0.271*** 1.000***

ln(Sales) -0.052*** -0.014*** -0.075*** 0.058*** 1.000

***, ** and * denote significance at the 1%, 5% and 10% levels, respectively

5. Results

This section presents and discuss the results of this study. Table 3 shows the results of equations (1) to (4) which test the original CR-CS developed by Kisgen (2006). To test if firms change capital structures after an actual rating change I performed the regressions of equations (5) to (8). With these regressions I also test if firms react immediately on credit rating changes or if they slowly adjust their capital structure over time. Furthermore I test if firms with different broad rating categories (for example AA of BBB) react different on near rating changes. Especially for firms around the investment/speculative grade rating changes can have great impact.

Therefore I perform the regressions of equations (1) and (2) to see if firms near the

investment/speculative grade react different on near rating changes than firms not near the investment/speculative grade. I argued before that the linear 1 to 24 rating scale might not be the best scale to describe credit ratings. I therefore included a polynomial to test if including the polynomial gives a better model. Finally I test if European firms react different on near credit rating changes compared to North American firms.

Table 3 shows the estimated coefficients of equation (1) to (4)14. The results give no evidence to reject the null hypothesis that firms near a credit rating change issue less debt relative to equity compared to firms not near a rating change. Contrary to expectations, firms near a rating change issue approximately 1.4% more debt relative to equity when controlled for the financial condition of a firm (coefficient POM, equation (1)). Firms near an upgrade issue approximately

14_{Table A4 in the Appendix shows the results for equation (1) to (4) for the entire sample. For the entire}

(18)

18

1.8% more debt relative to net equity than firms not near an upgrade. POM seems largely be driven by the Plus dummy. No significant results are found for near credit rating downgrades. Excluding control variables (equation (3) and (4)) gives no significant results which means that without controlling for financial distress firms’ capital structures are not influenced by near credit ratings. The results of table 3 are not in line with most of the existing literature. Most studies from table A1 found significant negative results for near rating changes and near rating downgrades. Positive coefficients for near upgrades were found, but none of these was significant. A possible explanation for the results of this study can be due in the relative high number of firms with an investment grade rating (AAA, AA, A and BBB) in this sample. These firms might be less concerned with an eventual credit rating change because their current rating is already high. I test this later by performing the POM and Plus/Minus tests (equations (1) and (2)) for firms with an investment grade (IG) and firms with a speculative grade (SG) separately. A second explanation for the positive coefficients of POM and Plus in table 3 is that firms might not be able to react immediately on credit rating changes and it might be possible that they slowly adjust their debt levels. This is tested in equations (5) to (9). Thirdly the costs of borrowing for firms with a relative high rating are lower compared to firms with lower ratings and it is therefore easier for firms with a relative high credit rating to borrow money.

The positive sign of the control variable leverage in table 3 indicates that firms with a higher probability of default issue more net debt relative to equity. For the profitability control variable I found remarkable results, firms with a higher profitability (and hence, a lower probability of default) issue less net debt relative to net equity. The results on the leverage and profitability control variables are not in line with financial distress theories which state the highly leveraged (profitable) firms issue less (more) debt relative to equity compared to other firms. The positive sign of the size control variable is as expected.

(19)

19 Table 3

Credit rating impact on capital structure decisions

This table shows the coefficients and t-statistics (in brackets) for pooled time series regressions with

NetDIss as dependent variable on credit rating dummies and some variables that control for the financial

condition of a firm. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. Plus/Minus is a dummy variable that has a value of 1 if a firm has a plus/minus rating and 0 otherwise. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Equation (1) Equation (2) Equation (3) Equation (4)

α -0.9797*** -0.9893*** -0.0277*** -0.0277*** -(5.58) -(5.62) -(6.57) -(6.57) POM 0.0142* 0.0030 (1.79) (0.57) Plus 0.0180** 0.0049 (2.00) (0.78) Minus 0.0098 0.0011 (1.05) (0.17) D/(D+E) 0.1537*** 0.1554*** (6.68) (6.73) EBITDA/A -0.7527*** -0.7514*** -(15.08) -(15.05) ln(Sales) 0.0569*** 0.0574*** (5.42) (5.46) Adjusted R² 0.1600 0.1599 -0.000315 -0.0006 N 2148 2148 2150 2150 F-value 1.98*** 1.98*** 0.32 0.32

***, **, and * denote significance at the 1%, 5% and 10% levels, respectively.

The results of table 3 might be influenced by the relative high number of firm years with an investment grade credit rating. In the sample of table 3 1,841 firm years have an investment grade and 307 firm years have a speculative grade. To test if the results of table 3 are influenced by the relative high number of firm years with an investment grade I performed the regressions of equation (1) and (2) for both the subsamples of firms with an investment grade and firms with a speculative grade. The results of these regressions are shown in table 4. Only for firms with an investment grade significant results are found. The POM and Plus dummy variables are significant

15

(20)

20

at the 5% level. This indicates that firms with an investment grade are more concerned with near rating changes than firms with a speculative grade.

Table 4

Credit rating impact on capital structure decisions – investment grade and speculative grade

NetDIss as dependent variable on credit rating dummies and some variables that control for the financial

condition of a firm. Panel A shows the results for firms with an investment grade (AAA, AA, A or BBB) and panel B shows the results for firms with an speculative grade (BB, B, CCC, CC, C or D). NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. Plus/Minus is a dummy variable that has a value of 1 if a firm has a plus/minus rating and 0 otherwise. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Panel A: investment grade Panel B: speculative grade

α -0.8811*** -0.8872*** -2.0043*** -1.9963*** -(4.65) -(4.68) -(3.55) -(3.52) POM 0.0213** 0.0020 (2.42) (0.08) Plus 0.0243** 0.0047 (2.48) (0.17) Minus 0.0170 -0.0009 (1.60) -(0.03) D/(D+E) 0.1754*** 0.1772*** 0.1452*** 0.1463*** (5.41) (5.44) (4.06) (4.03) EBITDA/A -0.5891*** -0.5881*** -1.1809*** -1.1791*** -(9.72) -(9.70) -(11.93) -(11.82) ln(Sales) 0.0492*** 0.0495*** 0.1228*** 0.1222*** (4.36) (4.38) (3.55) (3.51) Adjusted R² 0.0972 0.0969 0.3930 0.3903 N 1841 1841 307 307 F-value 1.53*** 1.53*** 3.30*** 3.25***

(21)

21

investment grade. The results of table 4 and table A4 in the appendix give not enough evidence to conclude that firms with an investment grade react different on near rating changes than firms with a speculative grade.

As argued by Chowdhury and Maung (2011), firms might not be able to react immediately on credit rating changes. This can take several periods after the actual rating change. In table 5 the results of equation (5) to (8) are presented which tested if firms react immediately on credit rating changes or that they slowly adjust capital structure over time16. Panel A shows the results of the regressions with NetDIss as dependent variable and panel B shows the regression results with ΔLEV as dependent variable. The regressions with ΔLEV as dependent variable are included to compare my results with the results of Chowdhury and Maung (2011).

The results in panel A do not reject the hypothesis that firms that experienced a rating change issue less debt relative to equity compared to other firms. No results are found that firms issue less net debt relative to equity after they actually experienced a rating change. The negative signs of previous downgrades are as expected, but not significant. The results in panel B with ΔLEV as dependent variable differ from the results with NetDIss as dependent variable. The results in panel B show that a current credit rating upgrade is associated with a current

downgrade in leverage. Firms that are upgraded reduce their leverage ratio in the same year as the upgrade with approximately 8.3%. Interesting is the significant positive coefficient of the variable Upgradet-1. This indicates that firms increase their leverage ratio (by approximately 6.0%) in the year following an upgrade. This increase in leverage ratio can be a result of the higher credit rating a firm has after an upgrade. The cost of borrowing money is lower and it is easier for a firm to borrow money. For downgraded firms I did not find significant results which mean that a downgrade has no immediate impact on a firms’ leverage ratio. The negative signs of the coefficients for the lagged variables Downgradet-1 till Downgradet-3 indicate that firms slowly reduce their leverage ratios after a downgrade, however, these results are not significant. Overall, I did not find enough evidence to reject the hypothesis that firms slowly adjust debt levels after a credit rating change.

16

Table A6 in the Appendix shows the results for equation (5) to (8) for the entire sample. Firms that are actually downgraded issue around 2.1% more net debt relative to equity (NetDIss)in the year of the downgrade. With ΔLEV as dependent variable no significant results are found. There is little evidence that

firms slightly adjust their capital structure after a rating change. Only for the variable Upgradet-1 in

equation (6) significant results at the 10% level are found which mean that firms that experienced a upgrade at time t-1 issue around 2.5% more net debt relative to equity compared to other firms. The sign

of previous downgrades dummy variables (Downgradet-1 till Downgradet-3) is negative as expected, but not

(22)

Table 5

Impact of previous credit rating changes on capital structure decisions

This table shows the coefficients and t-statistics (in brackets) for pooled time series regressions with NetDIss

(Panel A) and ΔLEV (Panel B) as dependent variables on a credit rating dummies and some variables that

control for the financial condition of a firm. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. ΔLEV is the change in leverage of a firm i and is measured as the ratio of book long term debt and book short term debt

to total assets of a firm i at time t. I-Changet, I-Upgradet and I-Downgradet are dummy variables that are 1 if

a firm’s credit rating is respectively changed, upgraded or downgraded in the current period and 0 otherwise. Change, Upgrade and Downgrade are dummy variables that are if the credit rating of a firm is respectively changed, upgraded or downgraded in a given year and 0 otherwise. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Panel A: Dependent variable: NetDIssi,t Panel B: Dependent variable: ΔLEVi,t

α -1.1739*** -1.0888*** -0.4515 -0.3343 -(4.44) -(4.12) -(1.00) -(0.74) I-Changet -0.0070 -0.0218 -(0.68) -(1.24) I-Upgradet -0.0123 -0.0825*** -(0.70) -(2.76) I-Downgradet -0.0039 0.0085 -(0.32) (0.40) Changet-1 -0.0068 -0.0013 -(0.67) -(0.07) Changet-2 0.0047 0.0070 (0.45) (0.39) Changet-3 -0.0042 -0.0108 -(0.46) -(0.69) Upgradet-1 0.0188 0.0594*** (1.09) (2.03) Upgradet-2 0.0101 0.0283 (0.63) (1.04) Upgradet-3 0.0023 -0.0114 (0.18) -(0.53) Downgradet-1 -0.0169 -0.0222 -(1.40) -(1.08) Downgradet-2 0.0028 -0.0081 (0.22) -(0.37) Downgradet-3 -0.0097 -0.0145 -(0.87) -(0.76) D/(D+E) 0.2711*** 0.2831*** 0.3720*** 0.3873*** (10.68) (11.09) (8.57) (8.89) EBITDA/A -0.7410*** -0.7487*** -0.8581*** -0.8593*** -(12.32) -(12.48) -(8.34) -(8.39) ln(Sales) 0.0649*** 0.0596*** 0.0184 0.0110 (4.15) (3.81) (0.69) (0.41) Adjusted R² 0.2682 0.2768 0.1626 0.1764 N 1106 1106 1106 1106 F-value 2.17*** 2.21*** 1.62*** 1.68***

(23)

The results of Chowdhury and Maung (2011) differ from the results of this study. Chowdhury and Maung (2011) found that upgraded firms do not change their leverage ratios in the year of the upgrade. In the years following an upgrade they slowly increase their leverage ratios. Downgraded firms increase their leverage ratio by approximately 7.0% in the year of the downgrade in the years following a downgrade they slowly reduce their leverage ratios.

Firms might react different on near credit rating changes when they have different broad rating categories (for example AA or BBB), because of the different discrete costs and benefits associated with different rating categories (Kisgen, 2006). This is tested by performing the regressions of equations (1) and (2) for each broad rating category separately. Panel A in table 6 presents the results of the POM dummy variable and panel B presents the results of the Plus and Minus dummies. For firms near a credit rating change (Panel A, POM) I only found significant results for the BBB category. Firms near a rating change with a BBB rating issue approximately 7.0% more net debt relative to equity compared to firms with a BBB rating not near a rating change. The negative coefficients of the AA, A and BB categories are as expected, however these coefficients are not significant. The results in table 6 indicate that the results for het POM variable in table 3 are mainly driven by the broad rating category BBB. Kisgen (2006) and Kemper and Rao (2013b) both found a positive coefficient for the BBB category, but their results were not significant. Both studies found significant negative results for the POM variable of B rated firms. In Kemper and Rao’s (2013b) study this was the only significant result, Kisgen (2006) also found a significant negative result for A rated firms, other rating categories were not significant.

Panel B gives a more detailed view of the impact of near credit rating changes on different rating categories because it shows the results for upgrades and downgrades separated. Firms with a BB- rating issue around 7.1% less net debt relative to equity. Firms with a BB- rating have a speculative grade and for these firms it is important to be get an investment grade as soon as possible (and to not be further downgraded towards a ‘junk’ rating category). For firms with a BBB- credit rating (just above a speculative grade) I expect also a negative coefficient for the Minus variables to avoid being downgraded to a speculative grade, but no significant results for this variable are found. The positive coefficient of the Plus dummy variable for the BBB category of 9.2% has the opposite sign as expected. It can be that these firm can issue more debt

(24)

24 Table 6

Credit rating impact on capital structure decisions per rating category

This table shows the coefficients and t-statistics (in brackets) for pooled time series regressions per broad rating category with NetDIss as dependent variable on a credit rating dummies and some variables that control for the financial condition of a firm. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t.

POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. Plus/Minus are

dummy variables that have a value of 1 if a firm has a plus/minus rating and 0 otherwise. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Credit rating AA A BBB BB B

Panel A: POM tests

α -0.3551 -0.4797** -0.6158 -1.4540** -2.3131 -(1.25) -(2.58) -(1.36) -(2.29) -(1.40) POM -0.0162 -0.0085 0.0700*** -0.0107 0.0125 -(0.63) -(1.02) (3.40) -(0.44) (0.17) D/(D+E) 0.1390** 0.1716*** 0.0853 0.2409*** 0.2037** (2.16) (5.87) (1.18) (3.16) (2.45) EBITDA/A -0.6063*** -0.3954*** -0.7801*** -0.6577*** -1.4377*** -(3.28) -(5.99) -(7.28) -(4.89) -(6.15) ln(Sales) 0.0186 0.0253** 0.0361 0.0824** 0.1428 (1.18) (2.27) (1.33) (2.09) (1.43) Adjusted R² 0.1844 0.1615 0.0880 0.1903 0.5109 N 248 771 797 230 66 F-value 1.76*** 1.73*** 1.37*** 1.76*** 3.83***

Panel B: Plus or Minus tests

α -0.0050 -0.5849*** -0.6275 -1.2129* -2.2584 -(0.07) -(3.08) -(1.39) -(1.94) -(1.28) Plus 0.0238 0.0054 0.0922*** 0.0458 0.0144 (1.09) (0.54) (4.04) (1.50) (0.19) Minus -0.0116 -0.0263** 0.0188 -0.0705** 0.0032 -(1.28) -(2.42) (0.61) -(2.26) (0.03) D/(D+E) 0.0390** 0.1863*** 0.1090 0.3276*** 0.2026** (2.04) (6.28) (1.50) (4.10) (2.39) EBITDA/A -0.0656 -0.3988*** -0.7994*** -0.5980*** -1.4432*** -(1.27) -(6.07) -(7.46) -(4.51) -(5.94) ln(Sales) -0.0011 0.0313*** 0.0364 0.0642 0.1395 -(0.26) (2.76) (1.35) (1.65) (1.31) Adjusted R² 0.0533 0.1696 0.0942 0.2286 0.4988 N 248 771 797 230 66 F-value 3.78*** 1.77*** 1.40*** 1.94*** 3.59***

(25)

25

For the Plus dummy variable only positive coefficients are found which imply that firms near an upgrade issue more debt relative to equity. However, except for BBB, none of these

coefficients is significant. The Minus variable is negative for AA, A and BB and positive for BBB and B. Only for A and BB the coefficient is significant. Overall firms near a rating downgrade issue significant less debt relative to equity in the A- and BB- category. For firms near a rating upgrade only significant results are found for firms with a BBB+ rating. The A- and BB- categories seem to be the only categories that follow a capital structure policy as is expected by the CR-CS.

Interesting to see is if firms react different on near rating upgrades or near rating downgrades. I tested this with an F-test and found that for three of the five categories in table 5 the Plus and Minus variables differ significantly. In the broad rating categories AA, BBB and B the effect of a near upgrade or near downgrade differ significantly17. Kemper and Rao (2013b) found similar positive results for firms with a Plus rating and in their sample only B- and AA+ seemed to follow a capital structure policy that is influenced by credit ratings. Kisgen (2006) findings indicate that especially firms with an AA and B rating issue less debt relative to equity. Michelsen and Klein (2011) found evidence that A- and BBB- categories follow a capital structure policy that is influenced by credit ratings. Overall, in this and the three other studies mentioned, none study found that the CR-CS holds for all rating categories.

Table 7 presents the regression results of capital structure decisions around the

investment/speculative grade. The discrete costs of firms below this grade are likely to be higher than for firms above this grade (Kemper and Rao, 2013b). The results of my study confirm the hypothesis that firms near the investment/speculative grade issue less debt relative to equity. The significant results at the 1% level in panel B indicate that firm with a BBB, BBB-, BB+ or BB rating issue approximately 3.6% less debt relative to equity compared to other firms. In panel A no significant results are found, however the negative sign is as expected. Both Kisgen (2006) and Michelsen and Klein (2011) found the same results for firms around the investment/speculative grade. These results imply that credit ratings changes around the investment/speculative grade have a larger impact on managers’ capital structure decisions than rating changes for firms not around the investment/speculative grade.

17

(26)

26 Table 7

Credit rating impact on capital structure decisions around investment/speculative grade

NetDIss as dependent variable on a credit rating dummies and some variables that control for the

financial condition of a firm. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. IG/SG in panel A is a dummy variable that is 1 if a firm has a rating near the investment grade (BBB- or BB+) and 0 otherwise. IG/SG in panel B is a dummy variable that is 1 if a firm has a rating near the investment grade (BBB- or BB+) and 0 otherwise.

D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The sample exists of North American and euro-country

firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Panel A: BBB- and BB+ Panel B: BBB, BBB-, BB+ and BB

Equation (9) Equation (10) Equation (11) Equation (9) Equation (10) Equation (11) α -0.9815*** -0.9808*** -0.9889*** -0.9469 -0.9484 -0.9550*** -(5.58) -(5.58) -(5.62) -(5.40) -(5.40) -(5.43) IG -0.0017 -0.0125 -0.0106 -0.0372*** -0.0352*** -0.0344*** -(0.12) -(0.84) -(0.70) -(3.42) -(2.94) -(2.86) POM 0.0168** 0.0036 (1.98) (0.41) Plus18 0.0196** 0.0061 (2.11) (0.62) Minus 0.0126 0.0011 (1.24) (0.11) D/(D+E) 0.1532*** 0.1537*** 0.1552*** 0.1490*** 0.1494*** 0.1505*** (6.66) (6.68) (6.72) (6.49) (6.50) (6.52) EBITDA/A -0.7531*** -0.7558*** -0.7542*** -0.7692*** -0.7683*** -0.7671*** -(15.04) -(15.10) -(15.05) -(15.38) -(15.34) -(15.30) ln(Sales) 0.0576*** 0.0570*** 0.0574*** 0.0564*** 0.0563*** 0.0566*** (5.48) (5.43) (5.46) (5.39) (5.37) (5.39) Adjusted R² 0.1585 0.1599 0.1597 0.1641 0.1637 0.1634 N 2148 2148 2148 2148 2148 2148 F-value 1.97*** 1.98*** 1.97*** 2.01*** 2.01*** 2.00*** ***, **, and * denote significance at the 1%, 5% and 10% levels, respectively.

Table 8 shows the results of equation (12) and (13) in which a third order polynomial is included to account for that credit ratings do not have a linear relationship. With this polynomial I assume that credit rating changes around the investment/speculative grade are more sensitive to capital structure decisions that rating changes further away on each side of the

18

(27)

27

investment/speculative grade. Including this polynomial gives similar results compared to equation (1) and (2) in table 3 and table A4.

Table 8

Credit rating impact on capital structure decisions with polynomial included

This table shows the coefficients and t-statistics (in brackets) for pooled time series regressions with NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise.

Plus/Minus is a dummy variable that has a value of 1 if a firm has a plus/minus rating

and 0 otherwise. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Equation (12) Equation (13) α -0.5535 -0.5799 -(1.41) -(1.47) POM 0.0145* (1.85) Plus19 0.0176** (1.98) Minus 0.0108 (1.16) D/(D+E) 0.1685*** 0.1699*** (7.21) (7.24) EBITDA/A -0.7911*** -0.7899*** -(15.86) -(15.83) ln(Sales) 0.0553*** 0.0559*** (5.31) (5.34) CR 0.0408* 0.0394* (1.84) (1.77) CR2 0.0222** 0.0215** (2.08) (2.00) CR3 0.0182** 0.0176* (2.04) (1.96) Adjusted R² 0.1778 0.1776 N 2148 2148 F-value 2.11*** 2.10***

19

(28)

28

Firms near a rating changes issue approximately 1.5% more net debt relative to equity compared to firms not near a rating change. In the results of the original model in table 3 firms issue approximately 1.4 more net debt relative to equity. Results of the Plus and Minus variables are also similar to the results of equation (2). The coefficients of the polynomial (CR, CR2,CR3) are all significant which indicate that the polynomial which represent the credit rating structure has a significant positive impact on a firms’ net debt issuance. The higher value of the adjusted R squared and the higher F-value indicate that the model with the polynomial included is a better way to test the CR-CS than the original model of equations (1) and (2). However the differences between the results in table 3 and table 8 are very small.

(29)

29 Table 9

Credit rating impact on capital structure decisions – Europe and North America

NetDIss as dependent variable on a credit rating dummies and some variables that control for the

financial condition of a firm. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. Plus/Minus is a dummy variable that has a value of 1 if a firm has a plus/minus rating and 0 otherwise. D/D+E is a lagged control variable for leverage,

EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size. The

sample exists of North American and euro-country firms with a long term Fitch credit rating between 2002 and 2012. Firm years with large debt offerings (>10%) are excluded from the sample.

Panel A: Europe Panel B: North America

α -1.4632*** -1.4578*** -0.7194*** -0.7268*** -(3.44) -(3.43) -(4.57) -(4.59) POM 0.0373** 0.0036 (1.96) (0.51) Plus20 0.0436** 0.0053 (2.05) (0.65) Minus 0.0307 0.0016 (1.43) (0.18) D/(D+E) 0.0808 0.0845 0.1901*** 0.1907*** (1.47) (1.53) (9.16) (9.16) EBITDA/A -0.5850*** -0.5753*** -0.7736*** -0.7734*** -(3.36) -(3.29) -(19.28) -(19.27) ln(Sales) 0.0850*** 0.0845*** 0.0416*** 0.0420*** (3.34) (3.32) (4.42) (4.44) Adjusted R² 0.0475 0.0466 0.3070 0.3065 N 729 729 1419 1419 F-value 1.24** 1.23** 3.34*** 3.32***

6. Conclusion

In this study I tested what influence near credit rating changes have on firm’s capital structure decisions. My hypothesis was that firms near a credit rating change issue less net debt relative to equity to avoid being downgraded or to increase the chance of an upgrade. It is possible that firms are not able or not willing to react immediately on credit rating changes. Therefore I tested if firms react immediately on actual credit rating changes or if firms slowly adjust their debt levels after a rating change. Furthermore I tested if firms near the

20

(30)

30

investment/speculative grade react stronger on near rating changes than firms not near the investment/speculative grade. Finally I tested if European firms react different on near rating changes than North American firms. No other studies tested the CR-CS explicitly for European firms.

The results in this study indicate that firms near a credit rating change issue approximately 1.4% more net debt relative to equity compared to firms not near a rating change. No significant result for firms near a downgrade is found and firms near an upgrade issue approximately 1.8% more net debt relative to equity. These results do not reject the hypothesis that firms near a rating change issue less net debt relative to equity to avoid begin downgraded or to increase the change of an upgrade. Regressions for each broad rating category individual indicate that the results are mainly driven by firms with a BBB rating. There are several explanations for the positive results of net debt issuance. It can be that firms near a rating change issue debt now, before the cost of borrowing increases after an actual rating change (Kemper and Rao, 2013a). Secondly, the acquisition of funds can be very important for a firm and it might be that it is more important to get those funds than the rising costs of getting these funds. Thirdly, Chowdhury and Maung (2011) found that firms adjust their debt levels slowly after a credit rating change. To test the result of Chowdhury and Maung (2011) I included dummy variables that accounted for credit rating changes in previous years. I found no evidence for the hypothesis that firms near a rating change slowly adjust their debt levels over time.

The CR-CS does not seem to hold across all rating classes. Only firms in the A- and BB- category issue less net debt relative to equity when they are near to a rating downgrade. For other rating categories no support is found that the CR-CS holds. Kemper and Rao (2013b) and Kisgen (2006) also found that the CR-CS does not hold for all rating categories.

As found in other studies (Kisgen (2006) and Michelsen and Klein (2011)), firms close to the investment/speculative grade issue significant less net debt (around 3.7%) relative to equity compared to firms not close to the investment/speculative grade. This confirms the hypothesis that firms close to the investment/speculative grade are more concerned with near rating changes than firms not near the investment/speculative grade.

(31)

31

distress arguments. North American firms have much higher debt ratios and are smaller in size than European firms. Both these variables indicate that North American firms have higher probabilities of default and therefore issue less net debt relative to equity.

Overall the results indicate that firms near a rating change issue more debt relative to equity compared to other firms. These results seems largely be driven by firms near a rating upgrade. This paper contributes to the limited literature about credit ratings and capital structure. This study is the first in which Fitch credit ratings are used to test the CR-CS. Furthermore in the existing literature there are no studies which tested the impact of credit ratings on capital structure for European firms specifically. Thereby this study tests if European firms and north American firms react significantly different on near rating changes.

(32)

32 References

Altman, E. I, Bharath, S. T. and Saunders, A., 2002, Credit rating and the BIS capital adequacy reform agenda, Journal of Banking and Finance, Vol. 26, Issue 5, pp. 909-921

Brooks, C., 2008, Introductory econometrics for finance (Cambridge University Press, New York) Cantor, R. and Packer, F., 1997, Differences of opinion and selection bias in the credit rating industry, Journal of Banking and Finance, Vol. 21, Issue 10, pp. 1397-1417

Chowdhury, R.H., Maung, M., 2011, Credit rating changes and leverage adjustments: concurrent or continual, Working paper, University of Dubai

Fama, E. F. and French, K. R., 2002, Testing tradeoff and pecking order predictions about dividends and debt, The Review of Financial Studies, Vol. 15,Issue 1, pp. 1-33

Fitch Ratings, 2013, Fitch Ratings: definitions of ratings and other forms of opinion

Graham, J. R. and Harvey, C. R., 2001, The theory and practice of corporate finance: evidence from the field, Journal of Financial Economics, Vol. 60, Issue 2, pp: 187-243

Kemper, K. J. and Rao, R. P., 2013a, Credit watch and capital structure, Applied Economics Letters, Vol. 20, No. 13, pp. 1202-1205

Kemper, K. J. and Rao, R. P., 2013b, Do credit ratings really affect capital structure?, Vol. 48, Issue 4, pp. 573-595

Kisgen, D. J., 2006, Credit ratings and capital structure, Journal of Finance, Vol. 61,Issue 3, pp. 1035-1072

Kisgen, D. J., 2009, Do firms target credit ratings or leverage levels?, Journal of Financial and Quantitative Analysis, Vol. 44,Issue 6, pp. 1323-1344

Marquardt, D.W., 1970, Generalized inverses, ridge regression, biased linear estimation and nonlinear estimation, Technometrics, Vol. 12, Issue 2, pp. 591-612

Michelsen, M. and Klein, C., 2011, The relevance of external credit ratings in the capital structure decision-making process, Available at SSRN: http://ssrn.com/abstract=1960699 or

http://dx.doi.org/10.2139/ssrn.1960699

(33)

33 Appendix

Table A1

Overview existing literature about credit rating – capital structure theories

This table shows the existing literature about the relationship between credit rating and capital structure. NetDIss is the net debt issuance of a firm i at time t and is measured as the net debt minus net equity divided by the total assets of a firm i at time t. ΔLEV is the change in leverage for a firm between time t and time t-1. POM is a dummy variable that is 1 if a firm has a plus or a minus rating and 0 otherwise. Plus/Minus is a dummy variable that has a value of 1 if a firm has a plus/minus rating and 0 otherwise. Change, Upgrade and Downgrade are dummy variables that are if the credit rating of a firm is respectively changed, upgraded or downgraded in a given year and 0 otherwise. PosNegOutlook is a dummy variable that is 1 if a firm has a positive or negative rating outlook and 0 otherwise. PosOutlook/NegOutlook are dummy variables that have a value of 1 if a firm has a plus/minus rating outlook and 0 otherwise. D/D+E is a lagged control variable for leverage, EBITDA/A is a lagged control variable for profitability and ln(Sales) is a lagged control variable for size.

Paper Research question Data Methodology Main findings

Kisgen (2006)

Credit Rating – Capital Structure hypothesis (CR-CS):

Firms near a rating change will issue less debt relative to equity to avoid a downgrade or increase the chance of an upgrade.

All firms listed in Compustat from 1986 till 2001 with a credit rating at the beginning of a particular year (only North American firms). Standard & Poor’s long-term domestic issuer credit rating. Debt offerings greater than 10% of total assets are excluded.

12,336 firm years.

Pooled time series cross section regressions with NetDIss as dependent variable. Independent variables: Credit rating dummy’s:

POM, Plus, Minus, Credit Score, Credit Score High, Credit Score Low

Control variables:

D/D+E, EBITDA/A, ln(Sales)

Firms near a rating change issue approximately 1.0% less net debt relative to net equity compared to firms not near a rating change.

Firms near a rating upgrade issue approximately 0.6% less net debt relative to net equity.

Firms near a rating downgrade issue approximately 0.5% less net debt relative to net equity.

Firms near a rating change around the

investment/speculative grade issue approximately 1.6% less net debt relative to net equity.

Kisgen (2009)

How do firms change their capital structures following an actual credit rating change?

Kisgen argues that downgraded firms will lower their leverage and that upgraded firms will respond little after an upgrade.

All firms listed in Compustat from 1987 till 2003 with a two year consecutive credit rating (only North American firms). Standard & Poor’s long-term domestic issuer credit rating. 11,372 firm years.

Pooled time series cross section regressions with NetDIss as dependent variable. And a simultaneous equations model. Independent variables:

Credit rating dummy’s:

Downgradet-1, Upgradet-1 Control variables:

D/D+E, EBITDA/A, ln(Sales),

market-Firms that have been downgraded issue approximately 1.5%-2.0% less net debt relative to net equity as a percentage of total assets compared to firms not downgraded.