Do Firms Target Credit Ratings or Leverage Levels?

Darren J. Kisgen* Boston College

This Version: July 21, 2006

Abstract

Firms reduce leverage following credit rating downgrades. A rating downgrade predicts capital

structure behavior better than changes in leverage, profitability, or bankruptcy probability.

Maintaining a particular rating level provides benefits to a firm, such as commercial paper access

or access to a broader bond investor pool, so attempting to regain a rating following a downgrade

is consistent with firm value maximization. The effect of a downgrade is larger at downgrades to

a speculative grade rating and if commercial paper access is directly affected. Rating upgrades

do not affect subsequent capital structure activity, suggesting that firms target minimum rating

levels.

JEL Classification: G32

Keywords: Capital Structure, Credit Ratings, Leverage, Financing Policy

______________________________________________________________________________ *Finance Department, Boston College, Fulton Hall, 140 Commonwealth Avenue, Chestnut Hill, MA, 02467-3961, phone: 617-552-2681, fax: 617-552-0431, e-mail: kisgen@bc.edu. This paper has benefited from comments provided by Thomas Chemmanur, Clifford Holderness, Edith Hotchkiss, Jennifer Koski, Alan Marcus, Mitchell Petersen, Jeffrey Pontiff, Edward Rice, Michael Roberts, Philip Strahan, James Weston, workshop participants at Boston College and Washington University in St. Louis, and seminar participants at the 2005 Financial Management Association meeting and the 2006 Western Finance Association meeting.

1

Credit ratings are an increasingly important component of finance theory and practice. The

Securities and Exchange Commission, in a report on the role and function of credit rating agencies

required by the Sarbanes-Oxley Act (SEC (2003)), concludes “the importance of credit ratings to

investors and other market participants had increased significantly, impacting an issuer’s access to

and cost of capital, the structure of financial transactions, and the ability of fiduciaries and others

to make particular investments”. Boot, Milbourn and Schmeits (2006) argue that credit ratings

provide an “economically meaningful role” by facilitating equilibrium in bond investment. Kisgen

(2006) finds that firms that are near a rating change issue less debt than other firms to avoid

downgrades and achieve upgrades. The Economist (2005) asserts that credit rating agencies “are

among the most powerful voices in today’s capital markets”.

Maintaining a particular credit rating level provides benefits to a firm. These benefits include the

ability to issue commercial paper, access to investors otherwise restricted from investing in the

firm’s bonds, lower firm disclosure requirements, reduced investor capital reserve requirements,

improved third-party relationships, and access to interest rate swap or Eurobond markets. Since a

firm’s leverage affects its credit rating, these benefits are relevant for determining a firm’s optimal

leverage in addition to traditional considerations (e.g., the tax benefits of debt). Moreover, unlike

traditional capital structure factors that generally imply firm value is a continuous function of

leverage, credit rating levels impart discrete benefits, implying that firm value jumps at certain

levels of leverage. The benefits of credit rating levels and the discreteness of ratings imply unique

capital structure behavior for a firm.

This paper investigates how changes in credit ratings affect a firm’s subsequent capital structure

decisions. I find that a firm whose credit rating is downgraded subsequently reduces leverage

whereas a firm that is upgraded subsequently increases leverage. This behavior is consistent with

2

targeting a credit rating or targeting a leverage level, since changes in leverage and ratings are

correlated. To identify credit rating effects specifically, I conduct tests that include lagged changes

in a firm’s leverage, profitability and probability of bankruptcy as additional explanatory variables

for capital structure decisions. In these tests I find that downgrades remain predictive for

subsequent capital structure behavior. Further, I find that a downgrade is a better predictor of

capital structure behavior than changes to leverage, profitability or probability of bankruptcy.

With these additional variables an upgrade is no longer significant, suggesting that capital structure

behavior is not affected by an upgrade once other changes to the firm are considered. These two

results are consistent with firms targeting minimum credit rating levels.1

The asymmetric behavior following downgrades and upgrades is optimal for a firm assuming that

adjusting leverage is costly and that credit rating levels provide benefits to a firm. To illustrate,

consider a firm whose leverage is optimal, given traditional factors and the benefits of higher

credit rating levels. Now consider if the firm has an increase to leverage that leads to a rating

downgrade, and for simplicity assume the firm’s optimal leverage has not changed. Since the firm

is no longer at its optimal leverage, the tradeoff theory implies the firm will adjust back to the

optimum if the benefits outweigh the adjustment costs. The benefits of moving back to the

optimum are an increasing function of the distance between the firm’s current higher leverage and

its optimal leverage. The rating downgrade however imparts specific costs to the firm that alone

may exceed adjustment costs, regardless of the size of the departure from the firm’s optimal

leverage. The downgrade itself may therefore instigate a subsequent leverage reduction

independent of other changes to the firm. On the other hand, consider if the firm is upgraded due

1 Anecdotal evidence indicates that firms target credit ratings. For example, PepsiCo surprised participants in the Stern Stewart Roundtable on Capital Structure (2001) by stating a firm policy of maintaining a single-A credit rating. Several recent papers have also provided evidence of the influence of credit ratings on capital structure (Graham and Harvey (2001), Faulkender and Petersen (2006), and Kisgen (2006)).

3

to a decrease in leverage with no other changes to the firm. The tradeoff theory again implies that

the firm will adjust back to its optimal leverage depending on the benefits relative to adjustment

costs. However since the rating upgrade imparts specific benefits to the firm, the upgrade itself

makes it less likely that the firm will bear the adjustment costs for moving back to the optimal

leverage. An upgrade therefore makes it less likely that the firm subsequently changes its

leverage, all else equal.

The results for downgraded firms are both economically significant and statistically robust. Firms

that have been downgraded issue over 2.0% less net debt as a percentage of assets relative to

equity the subsequent year than control firms, after controlling for several other potential effects.

The effects are present when both book and market measures of leverage are used, and when two

years of previous changes in leverage are included as a control. The results hold by rating, by

industry, by year, and including firm fixed effects. A downgrade predicts both a lower probability

of a subsequent debt issuance as well as a higher probability of a debt reduction. The effects are

larger at particular ratings, including ratings at the investment grade to speculative grade

distinction and such that commercial paper access is affected, providing additional evidence that

the effects are rating specific.

This paper also considers three other explanations for the behavior around credit rating changes.

First, distress costs might explain the results since credit ratings measure a firm’s probability of

bankruptcy. Second, a downgrade might affect debt issuance timing such that issuing debt is less

attractive following a downgrade or more attractive prior to the downgrade if the firm anticipates

the downgrade. Lastly, firms may be downgraded more frequently during economic downturns

when debt issuance might be less attractive in general.

4

The empirical evidence supports the target credit rating explanation compared to these three

alternate explanations. The tests control for changes in a firm’s leverage, sales, z-score and

profitability to identify credit rating effects specifically. The tests identify significantly larger

credit rating effects following downgrades compared with upgrades, consistent with targeting

minimum credit ratings and inconsistent with distress effects or timing. The results hold in tests

by particular rating, in which I find that firms downgraded to a particular rating issue less debt than

firms that have that same rating but were not downgraded. In these tests the control group should

have similar distress concerns and similar concerns regarding capital market timing. The results

are also strongest around particular ratings, whereas the alternate explanations generally imply

results should hold more evenly across all ratings. The empirical tests also control for past

leverage changes, which measure any previous market timing. Lastly, the credit rating effects are

evident in tests by individual year, with consistently strong results across years.

The rest of the paper is organized as follows. In Section I, I discuss the theoretical arguments and

empirical implications for credit rating concerns in the context of tradeoff theory. In Section II, I

conduct empirical tests to identify whether firms target minimum ratings. In section III, I examine

the extent to which firms are successful in targeting credit ratings by providing specific cases of

firm capital structure activity. In Section IV, I conclude.

I. Tradeoff Theory with Discrete Credit Rating Benefits

A. The Impact of Discrete Credit Rating Benefits on Optimal Leverage

The tradeoff theory states that a value-maximizing firm will balance the benefits of debt against

the costs of debt to determine the internal optimal leverage for the firm. Traditional benefits of

5

debt include that interest payments on debt are tax deductible and that the required payments of

debt mitigate the potential for managers to misallocate free cash flow. Costs of debt include the

present value of the direct and indirect costs of bankruptcy as well as the potential risk shifting or

underinvestment incentives that can occur with high leverage.

Higher credit rating levels provide discrete benefits to the firm not traditionally considered as

benefits of lower leverage in the tradeoff theory (I henceforth refer to the tradeoff theory without

consideration of discrete credit rating benefits as TT). Credit rating levels affect whether investor

groups such as banks or pension funds are allowed to invest in a firm’s bonds and the capital

requirements for investing in a firm’s bonds for insurance companies and broker-dealers. Ratings

also signal firm quality to investors such that firms are pooled together by rating resulting in

discrete differences in a firm’s cost of capital from different rating levels.2 Additionally, bond

covenants can require a change in bond coupon rate at different rating levels, some contracts are

signed conditional on a minimum rating, and the ability to access the commercial paper, Eurobond,

or interest rate swap market requires minimum credit ratings.3 TT and credit rating effects

together imply firms will balance the traditional tradeoff benefits of higher leverage against the

traditional benefits of lower leverage and these discrete credit rating benefits of lower leverage

(henceforth referred to as “TTCR”).4

2 Regulation FD excludes credit rating agencies thereby allowing credit rating agencies to continue to receive non-public information from firms. Boot, Milbourn and Schmeits (2005) provide a model whereby credit rating agencies provide a “focal point” for investors, and Millon and Thakor (1985) propose a model for the existence of “information gathering agencies” such as credit rating agencies based on information asymmetries. 3 For example, Enron’s takeover of Dynegy was made contingent on Enron maintaining an investment grade rating (International Herald Tribune (2002)), and EDS was concerned about a lower rating because it “could make signing new customers more difficult” (Wall Street Journal, 2004). Access to commercial paper is discussed in detail in Section II.D. 4 Firms could instead balance the benefits of higher ratings versus costs of issuing equity implied by the pecking order theory. Also, the psychological impact of a credit rating downgrade (or upgrade) could influence a manager’s decision-making (and perhaps asymmetrically). These considerations are not investigated in this paper.

6

Formally, representing the benefits of debt to the firm as a function of leverage as B(L) and the

costs of debt as a function of leverage as C(L), TT argues that a financial manager maximizes firm

value, V, as a function of leverage, L, with V(L)=B(L)-C(L). For some unique LT, firm value is

maximized, implying B’(LT)=C’(LT) and B’’(LT)-C’’(LT) <0.5 TTCR implies that for any

particular firm at certain levels of leverage, Lr (r∈{AAA,AA+,…,CCC-, D}), the firm incurs a

discrete benefit, Φr, if leverage is less than or equal to Lr.6 Let Φ(L) equal the cumulative rating

benefits implied by leverage L (equal to Σ Φr ∀ r such that Lr ≥ L). Firms will then maximize

V(L)=B(L)-C(L) + Φ(L), and for some firms, the optimal leverage will be Lr* < LT. The objective

functions are:

TT: V(L)=B(L) - C(L)

TTCR: V(L)=B(L) - C(L) + Φ(L) (1)

To illustrate, consider the simple example where B(L) = L, C(L) = L2, LBBB- = 45%, ΦBBB- = .02,

and all other Φ=0. In this case, the TT optimal leverage is equal to 50%, implying a speculative

grade credit rating (since 50% > 45%). The optimal leverage under TTCR is instead at 45%, the

leverage required to obtain a BBB- rating, since V(50%)=0.25 and V(45%)=0.2675. Generally,

firm value is maximized at LBBB- if ΦBBB- > (B(LT)-C(LT)) – (B(LBBB-)-C(LBBB-)). TTCR implies a

lower optimal leverage than TT if the discrete credit rating benefit outweighs the lost traditional

tradeoff theory benefits from departing from the TT optimum.

5 While sophisticated models of tradeoff theory have been constructed (see Leland (1994) or Fischer, Heinkel and Zechner (1989)), the purpose here is simply to provide a framework sufficient to examine the implications of discrete credit rating benefits in the context of the tradeoff theory. 6 At any level of leverage, a firm may have a positive probability of a rating change, implying the cumulative expected value of rating benefits is instead continuous for certain leverage ranges. A firm’s optimal leverage is nonetheless lower given positive benefits of higher ratings with either a static or dynamic perspective. For the empirical tests of this paper, the realization of a rating downgrade alters the probability of future costs to a certain current cost, and this change may affect capital structure decisions given positive adjustment costs. For example, a firm that is near a downgrade such that it loses commercial paper access can still issue commercial paper, but on the realization of a downgrade, the firm must discontinue the use of commercial paper, and this may instigate subsequent behavior.

7

B. Testing Implications of TTCR

To test TTCR I examine changes to a firm’s leverage and/or credit rating, and consider how the

implications for subsequent capital structure decisions are different for TTCR versus TT. The

implications are tested with the following regressions:

ittitiit UpgradeDowngradeNetDIss εββα +++= −− 1,21,1 (2)

ittitititiit KLeverageUpgradeDowngradeNetDIss εφβββα ++Δ+++= −−−− 1,1,31,21,1 (3)

NetDIss is a measure of the firm’s leverage changing capital market decision at time t, equal to a

firm’s net debt issuance minus net equity issuance divided by assets (specifically, net debt issuance

is long-term debt issuance (Compustat data item no. 111) minus long-term debt reduction (no. 114)

plus changes in current debt (no. 301), and net equity issuance is sale of common and preferred

stock (no. 108) minus purchase of common and preferred stock (no. 115)). This measure identifies

direct capital market activity decisions of managers (Berger, Ofek and Yermack (1997), Leary and

Roberts (2005) and Kisgen (2006) use a similar measure). Downgrade and Upgrade are dummy

variables equal to 1 if the firm was downgraded or upgraded the previous year, respectively.

Lagged changes in ratings are used to reduce potential endogeneity. ΔLeverage is equal to the

lagged year over year change in leverage; for example, if the firm increased leverage from 50% at

time t-2 to 55% at t-1, this variable takes a value of 5%. K includes variables that measure a firm’s

change in financial condition, such as changes in profitability or z-score (these variables are

discussed in further detail in Section II.B).

8

Tests of incremental capital structure decisions of this nature implicitly assume that firms at certain

points in time are not at optimal leverage levels, and therefore specific subsequent capital market

activity is implied (see Graham (1996) and Mackie-Mason (1990) for similar approaches). I

assume that firms are at their optimal leverage and credit rating prior to the change in leverage or

rating identified in equations (2) and (3) and test the implications for NetDIss given movements

away from the optimum.7 If a firm has moved away from its optimal leverage, a manager will

undertake capital structure behavior to change leverage to the firm’s optimum if the benefits

outweigh the adjustment costs, AC, incurred. Adjustment costs can be a significant influence on

capital structure variation and policy, implying that in certain cases a firm will depart from its

optimal leverage over periods of time (Fischer, Heinkel, and Zechner (1989), Leary and Roberts

(2005), Flannery and Rangan (2005)).

The empirical specification of equations (2) and (3) examines firm behavior in the year following a

change in rating relative to other firm years. This specification could yield spurious results if some

unobserved firm factor leads a firm to, for example, regularly reduce debt and also receive

downgrades. To address this, I also include fixed effects in some tests to examine firm behavior

following a downgrade relative to its own behavior in other years. These fixed effects tests

identify activity specific to the firm itself. The empirical specification also examines behavior

only in the first year following the change in rating relative to other firm years. While a

downgrade may affect behavior beyond this year, the year after the downgrade is most likely to be

7 Since some changes in leverage and ratings may instead represent movements toward the optimum, thereby implying no specific capital market activity, coefficients identifying leverage and credit rating effects in (2) and (3) are biased toward zero. At a particular point in time, t-2, a firm might be at its target leverage (target rating) or away from its target. A change in leverage (rating) at time t-1 may indicate a move back toward the firm’s target or away from the target. A change back to the target implies no particular capital structure activity at time t. Since the firm moved back toward their target, the firm should be equally likely to issue debt or equity the subsequent period (assuming the firm does not move partially back to the target).

9

affected by the rating change. Furthermore, behavior in subsequent years may be affected by

subsequent changes in ratings and other changes to the firm.

In this context, TTCR has distinct implications compared to TT. TTCR implies firms that are

downgraded are more likely to reduce leverage the subsequent period compared to non-

downgraded firms, even when controlling for changes in leverage and firm characteristics that

measure distress. TTCR also implies that, after controlling for changes in leverage and other firm

characteristics, an upgrade will not significantly affect subsequent capital structure behavior. An

upgrade itself is beneficial to a firm, so a firm will not seek to reverse it. TTCR further implies

that downgrades at ratings levels where discrete rating costs are larger will increase the likelihood

for leverage reducing capital market decisions.

To illustrate these implications of TTCR, I continue with the simple example from I.A, with B(L)

= L, C(L) = L2, LBBB- = 45% and ΦBBB- = .02. I further assume that LBBB = 40%, ΦBBB = .005 and

AC=.01 (and all other Φ=0). In this case, the firm’s optimal leverage is still 45% under TTCR,

and the optimal rating is BBB-. Consider in this example a firm that receives exogenous shocks to

leverage with no change to the fundamentals of the firm (perhaps from exogenous changes in cash

flow) such that leverage changes each year in the following pattern: 45% to 43% to 45% to 47%.

Optimal leverage remains at 45% throughout since the fundamentals of the firm are unchanged.

However, in the first year when leverage changes from 45% to 43%, no capital structure activity is

implied since the benefits of changing leverage to 45% are not larger than the adjustment costs

(V(45%)-V(43%)<.01). The subsequent change in leverage from 43% to 45% also does not imply

capital structure behavior. The change from 45% to 47% causes a downgrade and implies the firm

will undertake leverage-reducing capital structure behavior, since V(45%)-V(47%)>.01,

specifically because of the discrete cost of the downgrade. If the firm achieves an upgrade in a

10

subsequent year by changing leverage back to 45%, no capital structure activity is implied

following the upgrade and change in leverage.

Now consider a firm that incurs exogenous shocks such that leverage changes each year from 45%

to 41% to 38% to 36%. The change in leverage from 45% to 41% is not large enough to justify

capital market activity to move back to 45%, since V(45%)-V(41%)<.01. The change in leverage

to 38% results in an upgrade and, due to the upgrade, adjustment costs are still larger than the

benefits of moving back to the optimal leverage of 45%, so no capital market activity is implied.

After the change in leverage to 36%, leverage-increasing capital market activity is implied, due to

TT considerations, and the firm will move back to its optimal leverage of 45%. Following this

change in leverage to 45%, the firm is downgraded but no capital market activity is implied.

This illustration includes two instances of upgrades and two instances of downgrades. Following

three out of four of these changes in rating, no capital structure activity is implied. Following one

of the two downgrades, leverage-reducing capital market activity is uniquely implied by TTCR

given the discrete costs of a downgrade. The larger the cost of a downgrade, the more likely that

TTCR implies subsequent leverage reducing activity following the downgrade.

This illustration is one of three cases in which a firm can have a change in rating, each more

formally modeled in the Appendix. The second case is if a rating agency changes its rating

standards. Rating agencies may become more or less stringent (for example, Blume, Lim and

Mackinlay (1998) find that rating agencies consistently became more stringent from 1978-1995).

In the previous numerical example, this implies the level of LBBB- changes. If for example LBBB-

changed to 43%, the firm would be downgraded if its leverage were 45%, and TTCR implies that

the firm would subsequently reduce leverage to 43% to regain the BBB- rating. This implication

11

is distinct from TT and consistent with the TTCR implication of the first case. The discrete cost of

the downgrade implies leverage reducing capital market activity following a downgrade. Unlike

the first case however, in certain instances more stringent rating standards may imply the firm

should actually increase debt under TTCR. This could occur if the rating agency has become so

stringent that the required leverage to maintain the previous target rating is no longer optimal.

This implication is again distinct to TTCR, and the empirical tests implicitly test which effect is

more prominent. The third case in which a firm may be downgraded is when the firm changes

fundamentally, such as from a permanent change to expected cash flows. This instance is more

complex, since this implies changes in the functional form of C(L) or B(L) as well as changes to

the levels of Lr. TT and TTCR can therefore have similar implications; however, the implications

unique to the discrete cost of the downgrade are similar to the first instance. The empirical tests

identify the TTCR specific effects by controlling for fundamental changes in the firm, including

changes to profitability and probability of bankruptcy. The implications following an upgrade in

these two additional cases follow similarly to the first case, such that an upgrade by itself does not

imply subsequent capital structure activity.

II. Empirical Tests

A. Data and Summary Statistics

The sample is constructed from all firms with a credit rating in Compustat. Approximately 78% of

outstanding debt is issued by firms with a public debt rating (Faulkender and Petersen (2006)),

suggesting this sample covers a significant portion of firms active in capital markets. The credit

rating used is Standard & Poor’s Long-Term Domestic Issuer Credit Rating (Compustat data item

no. 280). This rating is the firm’s “corporate credit rating”, which is a “current opinion on an

12

issuer’s overall capacity to pay its financial obligations” (Standard and Poor’s (2001)). Compustat

also includes subordinated ratings and commercial paper ratings for firms. Subordinated ratings

have a strict correspondence to long-term ratings, so little additional information is gained from

these ratings for tests of this paper. Commercial paper ratings are examined separately in

individual tests.

The sample period for the tests is 1987 to 2003 (1985 is the first year credit ratings are available in

Compustat and at least 2 years of lagged data is required for all tests). I exclude firm years in

which the firm has missing data in the fields required regularly for the tests in the paper.8 Some

previous papers (e.g., Fama and French (2002)) exclude financial companies and utilities (SIC

codes 4000-4999 and 6000-6999), however ratings considerations are likely to affect these firms as

well as industrial firms. I include these firms, but I also test for robustness to their exclusion from

the sample.

Tables I-II and Figure 1 display summary statistics for the sample. Table I indicates the

downgrade and upgrade activity at each rating. With few exceptions, each rating category has

significant firm years in which firms are both upgraded and downgraded to the particular rating.

922 downgrades are to ratings that are investment grade, and 805 downgrades are to ratings that

are speculative grade. 677 upgrades are to ratings that are investment grade and 467 upgrades are

to ratings that are speculative grade.

Table II shows the percentage of firm years that have debt and equity issuance and reduction

activity, with firm years separated by the previous year’s change in rating. An issuance or

reduction is defined as a net amount greater than 5% of beginning of period assets, as in 8 These are Compustat data item no’s, 6, 12, 13, 25, 35, 108, 111, 114, 115, and 199.

13

Hovakimian, Opler, and Titman (2001). The table indicates that downgraded firms are more likely

to reduce debt or issue equity and are less likely to issue debt or reduce equity than other firms.

For example, downgraded firms are 86% more likely to reduce debt than non-downgraded firms.

Furthermore, firms that are downgraded to a speculative grade rating from an investment grade

rating are more likely to reduce debt or issue equity and less likely to issue debt or reduce equity

than other downgraded firms. For example, firms downgraded to speculative are 50% more likely

to reduce debt than other downgraded firms, and are nearly three times more likely to reduce debt

than other firms in general.

The capital market activity following upgrades is less consistent. For example, upgraded firms are

more likely to issue debt, however firms that are upgraded to investment grade are less likely to

issue debt than other upgraded firms. Upgraded firms are also more likely to issue equity and

more likely to reduce equity. To some extent however, the pattern for upgraded firms indicates

that upgraded firms are more likely to conduct leverage increasing capital market activity.

Figure 1 shows the percentage of firm years in which a firm undertakes leverage increasing or

decreasing capital market activity, with firm years grouped based on their previous year changes in

leverage and credit rating. Leverage increasing or decreasing activity is defined as positive or

negative values of NetDIss, respectively. In Panel A of Figure 1, firms that were downgraded the

previous year are compared to firms with no rating change that had a similar change in leverage

during the same year. Panel A indicates that firms that are downgraded are consistently more

likely to undertake leverage decreasing behavior the subsequent year than firms that are not

downgraded that had a similar change in leverage. This is consistent with the implications of

TTCR described in Section I.B. This preliminary evidence also suggests that downgrades lead to

leverage reductions independent of any previous capital market timing. Panel B of Figure 1 shows

14

the same statistics but considering upgraded firms compared to firms with no rating change that

have a similar leverage change during the same year. Also consistent with TTCR, upgraded firms

are not consistently more likely to increase leverage following an upgrade. In both Panel A and B,

minimal relationship exists between leverage changes and the likelihood of leverage increasing

and decreasing behavior the subsequent year. These general findings are formally tested in the

following section with regressions of equations (2) and (3).

B. Empirical Results

Results of regressions of equations (2) and (3) are shown in Table III. Separate tests of equations

(2) and (3) are conducted using changes in book and market levels of leverage,9 with assets in

NetDIss defined corresponding to the measure used for the change in leverage explanatory

variable. Leverage is defined as in Fama and French (2002) as: (liabilities plus preferred stock

minus deferred taxes and investment tax credit)/(book assets or market value of the firm).

Control variables include lagged levels of a cardinalized value of the firm’s credit rating level

(AAA=1, AA+=2, AA=3, etc.), sales, profitability, market-to-book ratio and changes in a firm’s

size, profitability, market-to-book ratio and z-score.10 Levels of a firm’s credit rating, size,

profitability and market-to-book ratio control for the possibility that upgraded and downgraded

firms are systematically different (e.g., upgraded firms might generally be higher quality or have

higher growth potential). Changes in a firm’s size, profitability and z-score (in addition to changes

9 Rating agencies indicated they focus more on book leverage levels; Standard and Poor’s (2001) argues, ”a company’s ability to service its debt is not affected directly by such factors…” that influence the stock market value of equity. The inclusion of market levels is similar to Welch (2004) who examines to what extent changes in market values of equity explain changes in leverage; however, equation (2) is different from Welch (2004) in that past changes in leverage are the explanatory variables for future net issuing activity. 10 Sales is defined as ln (Sales (Compustat data item no. 12)), profitability is EBTIDA (no. 13) divided by assets (no. 6), market-to-book is (market equity (no.25 times no.199) plus total liabilities (no.181) plus preferred stock (no.10 or, if unavailable, no. 56) minus deferred taxes and investment tax credit (no.35) minus convertible debt (no.79))/(book assets (no.6)), and z-score is (3.3 times pretax income (no.170) plus 1.4 times retained earnings (no. 36) plus 1.2 times working capital (no.4 minus no.5))/book assets (no.6).

15

in leverage) control for changes in financial distress and other characteristics (Graham (1996)) to

identify distinct credit rating effects. Changes in the market-to-book ratio controls for potential

market timing (Baker and Wurgler (2002)).

Equation (2) does not include control variables, and the results of this test shown in column 1 and

2 of Table III indicate a large impact of changes in rating on leverage decisions. For the book

measure, a firm that has been downgraded issues over 4.0% less net debt relative to net equity than

other firms, and a firm that has been upgraded issues approximately 1.0% more debt relative to

equity than other firms. The results are significant at 1% and 5%, respectively. The significant

downgrade and upgrade results without control variables are consistent with both credit rating

effects and distress or target leverage effects. However, the null that β1 = -β2 is rejected at 1% for

the tests in columns 1 and 2 in favor of β1 > -β2, consistent with TTCR and inconsistent with

ratings only proxying for distress concerns.

Columns 3 and 4 show results using the full set of changes and levels of firm characteristics as

control variables. In these regressions, the coefficient on Downgrade remains economically and

statistically significant. Firms that have been downgraded issue over 2.0% less net debt relative to

equity in both cases, and the coefficient is statistically significant at 1% in both cases. Consistent

with TTCR, downgraded firms undertake capital structure behavior to attempt to regain their

previous rating after controlling for changes in leverage and other firm characteristics. After

controlling for these changes in the firm, the upgrade coefficient is no longer statistically

significantly different from zero for the book measures of leverage, and is only significant at 10%

for the market measures of leverage. This result is also consistent with TTCR. Firms do not

attempt to reverse an upgrade, after controlling for changes in leverage and the firm’s financial

16

condition, because the upgrade, when considered by itself, brings discrete benefits to the firm.

Conditional on other changes to the firm, a manager does not have an incentive to reverse an

upgrade, and this is confirmed by these results. This lack of symmetry is inconsistent with the

alternate explanation that firms react to changes in credit ratings because they proxy for changes in

financial distress costs. If firms are reacting to changes in distress costs as represented by changes

in ratings, firms that are upgraded should issue more debt relative to equity than other firms, as

they have lower distress costs. The null that the coefficient on the upgrade dummy variable is

equal to negative the coefficient on the downgrade dummy variable is again rejected at 1%.

I examine the predictive capability of the downgrade variable relative to changes in leverage,

profitability and z-score. In separate regressions of NetDIss on each of the four individual

variables, the coefficient on the explanatory variable has the sign predicted by theory and is

significant at 1% (tables not reported). Regressions with only the downgrade variable however

have the highest R-squared, twice that of the next best regression. Regressions with the

downgrade variable also have the lowest root mean squared error. Lastly, in regressions with all

four explanatory variables, only the downgrade variable remains statistically significant at 1%.

The downgrade variable is a better predictor of capital structure behavior than changes to leverage,

profitability or z-score.

I also conduct logit regressions in which the dependent variables are binary measures indicating

debt issuance versus no issuance and debt reduction versus no debt reduction to determine the

relative predictability of the explanatory variables (tables not reported). An issuance and reduction

is defined as greater than 5% of total assets. In these tests, a downgrade predicts both a lower

probability of debt issuance as well as a higher probability of debt reduction. Further, the

17

downgrade variable is again a greater predictor of capital structure choice than any of the other

three variables, using measures of observed fit or log-likelihood statistics.

Interpreting the behavior following a downgrade may be confounded by potential endogeneity of

the downgrade. In particular, firms may time capital structure behavior in anticipation of a change

in rating, or make other changes to the firm that precipitate a downgrade and leverage reduction.

Measuring the rating change the year prior to the capital structure decision however reduces the

potential for endogeneity. Further, including lagged changes in leverage captures capital structure

timing prior to the change in rating. Also, logit regressions indicate that a downgrade predicts both

a lower probability of debt issuance as well as a higher probability of debt reduction. While firms

may be reluctant to issue debt following a downgrade because of a higher cost of debt, a higher

cost of debt does not imply that a firm would be more likely to reduce debt (since the rate for

existing debt is already determined). Any alternate story must explain why downgrades lead to

both a higher probability of a debt reduction and a lower probability of a debt issuance, controlling

for previous changes in leverage. The empirical evidence is most consistent with firms reducing

leverage to regain a minimum target rating level.

These results are related to the results of Kisgen (2006), as both papers examine firm capital

structure behavior given benefits of credit rating levels. Kisgen finds that firms near a rating

downgrade or upgrade issue less debt relative to equity than other firms. This paper differs from

Kisgen in two important ways. First, this paper examines implied behavior following changes in

credit ratings, whereas Kisgen tests behavior before a change in rating. Second, Kisgen excludes

large offerings of debt or equity in the empirical tests, since large offerings may lead to a rating

change whether the firm is near a change in rating or not. The tests of this paper on the other hand

apply to all offerings, and thus have broader implications for firm leverage, specifically implying

18

lower levels of leverage for firms. Further, the downgrade variable is easily integrated into other

tests of capital structure behavior, since no sample restriction is made and the variable is

straightforward to construct. The results of both papers however are complementary, since it is

consistent for firms to try to avoid downgrades and achieve upgrades while also trying to reverse

downgrades (but not reverse upgrades).

To evaluate the robustness of the coefficients and standard errors, I report results from two

additional tests.11 First, I calculate standard errors clustered by firm as suggested by Rogers (1993)

and Arellano (1987). Second, I conduct a firm fixed effects (within-group) test with clustered

standard errors (see Wooldridge (2002), equation 10.59). With the firm fixed effects tests, I

exclude firm level control variables since the firm effect together with changes in those variables

fully subsumes these firm level effects (I also exclude firm years with only one year of data as

these will add no additional information in this framework). The results from these tests, shown in

columns 5-8, indicate the results are robust to these alternate econometric designs. The credit

rating results remain strong after including dummy variables for each firm, and the standard errors

are largely unchanged when a clustering approach is used. The robustness of the results to the

inclusion of fixed firm effects indicates that the behavior following downgrades is not due only to

variation across firms. The results are also robust to tests with random effects, inclusion of year

and industry dummy variables, exclusion of SIC codes 4000-4999 and 6000-6999, and with t-

statistics calculated using the approach suggested by Fama and MacBeth (1973) (tables not

reported). I conclude from these statistical robustness tests that the OLS results are not

compromised by the panel data specification, and I therefore report OLS coefficients and White’s

standard errors for the remainder of the paper.

11 See Petersen (2005) for an excellent review and evaluation of current panel data econometric approaches.

19

The results shown in Table III indicate that firms are also targeting leverage levels to some extent.

Firms with leverage increases (decreases) issue less (more) net debt relative to net equity the

subsequent year. The sizes of the coefficients are small however. The coefficient in Column 3 on

ΔLeverage indicates that a firm whose leverage increased by 10% the previous year will issue

0.3% less debt relative to equity as a percentage of assets the subsequent year (compared to the

2.0% coefficient on the Downgrade variable). These tests likely underestimate target leverage

behavior as discussed in Section I. Even with consideration for the conservative nature of this test

however, the magnitudes of the coefficients on the change in leverage variables are significantly

smaller than those on the Downgrade variable, indicating a firm’s change in rating is a greater

influence on capital structure decisions than a firm’s change in leverage.

The somewhat weak target leverage results so far might be consistent with firms adjusting toward

target leverages over periods longer than one year, as argued in Leary and Roberts (2004) and

Flannery and Rangan (2004). While this argument could equally be applied for credit rating

concerns (if adjustment costs are high, firms might not immediately try to move back to their

target rating), I conduct an additional test including changes in leverage from the two previous

years, but still only including the credit rating change from the previous year. Table IV shows

results that support the idea that firms undertake capital structure behavior to move back to target

leverages for periods longer than one year. Changes in leverage two years prior are predictive for

capital market activity today, with and without control variables. The sum of the coefficients

without control variables indicates that firms reverse over 12% of a leverage change within the 2

years following the change in leverage. Again, the conservative nature of these tests suggests this

figure is a lower bound. The implications for credit ratings are unchanged in these regressions.

Firms that are downgraded continue to issue less debt than other firms, even with these extensive

20

control variables, and the magnitude of the effect is largely unchanged. The upgrade dummy

variable is no longer statistically significant in any test suggesting that this full set of controls

account for the extent to which credit ratings proxy for distress costs. The significance of the

downgrade variable with these control variables indicates that firms are concerned with

maintaining minimum rating levels.

The coefficients on the additional explanatory variables shown in Table IV provide additional

support for the tradeoff theory. The coefficient on EBITDA is positive and significant at 1%

throughout the tests. The coefficient on the firm’s rating level is negative and significant at 1%,

consistent with financial distress concerns to the extent credit ratings are a proxy for probability of

bankruptcy. The coefficients on ΔSales are consistently positive and significant, consistent with

the tradeoff theory to the extent firms with higher sales have lower distress costs (and consistent

with Graham (1996) who uses changes in debt only as a dependent variable). The coefficient on

the level of sales is negative however, indicating firms that are larger in general issue more equity

than debt. The coefficients on market-to-book are positive or negative depending on the

dependent variable, results likely explained by the relationship between the dependent variable

measures and market-to-book. All else equal, firms with higher market-to-book ratios have lower

levels of market NetDIss, given the larger value of market assets in the denominator.12

12 For example, consider three firms with net debt minus net equity issuance of 40, 60 and 80 million, all with book assets of $1 billion. Assume these three firms have market to book ratios of 1, 2 and 4, respectively, and for simplicity assume that they are 100% equity financed at the beginning of the period. These firms have a positive relationship of book NetDIss to market-to-book and a negative relationship of market NetDIss to market-to-book.

21

D. Results by Rating

In this section I examine the impact of downgrades at particular rating levels on subsequent

leverage decisions. TTCR implies greater capital structure effects at ratings levels in which the

cost of a lower rating is larger. Testing effects by rating provides an additional test of TTCR

distinct from alternate explanations for the credit rating results, since several alternate explanations

imply uniform reactions to changes in credit rating. I test individual rating effects in two ways. I

conduct tests at each rating level comparing firms downgraded to each rating to a control group of

firms with the same rating that were not downgraded. Since the control group has the same rating

as the downgraded firm, they should have similar financial distress concerns and costs of debt.

The second way I test the effects of particular ratings is to incorporate dummy variables for a

specific change in rating directly into the regressions of equation (3). In these tests, I examine the

incremental effect of specific rating levels that have higher costs associated with them, including

the investment grade to speculative grade change, the change from B to CCC, and changes that

directly affect commercial paper access.

D.1. Individual Rating Tests

Table V shows the results by individual rating. Regressions of equation (3) are conducted on

separate samples of firms downgraded to each particular rating level with a control group of firms

that already have that credit rating. Although control variables are less important in this context

since the firms are matched by rating, I continue to control for levels of firm characteristics, and I

conduct tests with and without changes in leverage, size, z-score, market-to-book, and

profitability.

22

The results shown in Panel A of Table V indicate that firms receiving downgrades to particular

ratings attempt to move back to their previous rating. The coefficients in the market value tests

with all control variables (specification (3)), for example, are significant at 5% in 8 out of 18

ratings categories. The results are economically significant by rating as well, with firms that have

been downgraded issuing over 1% less net debt than equity versus other firms in the same rating

category for 10 out of 18 ratings. The results are stronger around certain ratings, while at other

ratings the effect of a downgrade is not significant. The null hypothesis that the effects are equal

at all ratings is rejected at 1% for the market measure tests. This result implies the effects are

greater at certain rating levels, consistent with the implications of TTCR. For the book measure

tests however the F-test does not reject the null that the effects are equal across returns, however

several F-tests comparing individual coefficients indicate significantly stronger results at certain

ratings levels versus others.

These tests by rating might still reflect some distress effects if firms that are downgraded are

downgraded again the next year and consequently have greater distress concerns than their control

groups. To account for this, I also conduct the tests by rating with control firms based on the

firm’s rating at time t, at the end of the year when capital structure activity is taking place. The

capital structure activity of firms that are downgraded is thereby compared against firms that have

the same rating at the end of that year. The downgrade dummy variable is still lagged such that the

regressions remain predictive. Results using this approach are very similar to those in Panel A of

Table V, with limited exceptions (tables not reported).

The ratings levels that are most significant largely correspond to ratings in which discrete benefits

of those ratings levels are largest. Firms target ratings around the change from investment grade to

speculative grade (BBB- to BB+). A speculative grade rating prohibits some investor groups from

23

investing in a firm’s bonds (e.g., banks and pension funds) and increases capital charges for other

investors. Ratings triggers are also most prevalent around the change in rating from investment

grade to speculative grade, and disclosure requirements increase at this change in rating.

The changes at higher ratings levels correspond to ratings that are significant for commercial

paper, asset-backed securities, and interest rate swap market access. The size of the commercial

paper market was $1.4 trillion as of 2004, and industrial firms as well as financial firms find

commercial paper to be a valuable source of short-term capital. Standard and Poor’s (2001) notes

a “strong link” between a firm’s long-term rating and its commercial paper rating: an A long-term

rating (or higher) is generally necessary to obtain an A1 commercial paper rating, a minimum BBB

rating is required to receive an A2 rating, firms with a BBB rating with a negative outlook or a

BBB- rating are rated A3, and a firm with a junk bond rating would receive a commercial paper

rating of B or below. Previous literature further indicates the importance of the commercial paper

market and the impact of commercial paper rating changes on stock prices and commercial paper

outstanding (Hahn (1993), Crabbe and Post (1994), Nayar and Nozef (1994)). Several regulations

are also tied to these higher ratings: money market funds must limit holdings of short-term bonds

to short-term ratings that correspond to an A long-term rating or better, lending is permitted

against mortgage-backed securities and foreign bonds rated AA or better, and pension funds are

allowed to invest in asset-backed securities rated A or better (Cantor and Packer (1997)).

California state regulations also prohibit California-incorporated insurance companies from

investing in bonds rated below single-A (SEC (2003)).

Results are also strong around the B to CCC distinction. A minimum B rating has regulatory

advantages and can be significant for third party relationships (for example, a condition for GE to

provide financing for a large jet order by US Airways was that US Airways credit rating not fall

24

below B- (Financial Times (2004))). Liquidity is also a significant concern around these ratings

(Patel, Evans and Burnett (1998)). This result is also consistent with anecdotal evidence of

clustering at B rating levels (Financial Times (2005)) and large increases in spreads from a change

in rating from B- to CCC+.13

A downgrade can be to a rating more than one category lower (e.g., from BBB+ to BBB- instead

of BBB+ to BBB). The significant results at the BB rating, for example, may therefore be due to

firms downgraded from an investment grade rating trying to regain that rating. Johnson (2003)

finds that firms receiving a downgrade from BBB- will “travel” more categories than other ratings

levels, implying firms that are downgraded from this rating can often be downgraded to BB or

lower. I conduct additional tests that separate the downgrade dummy variable into two dummy

variables, one that indicates the firm was downgraded one category (e.g., BBB+ to BBB) and one

that indicates the firm was downgraded more than one category (e.g., BBB+ to BBB-). Panel B of

Table V shows results of these tests at credit rating levels that were consistently significant in the

Panel A tests. The results for firms downgraded to investment grade ratings (AA-, A, BBB and

BBB-) do not indicate a stronger effect for a multiple rating change relative to a single rating

change. These results suggest that firms target ratings levels one category above the rating to

which they were downgraded. The BBB results indicate a desire to move back to the previous

rating whether it was one category above (BBB+) or two or more above (A- or higher). The effect

of being downgraded to BB however is significantly stronger for firms moving two or more

categories down to BB; that is, they went from investment grade (BBB- or above) to speculative 13 The size of the coefficients in the CCC category suggests that the overall results in Table III may be driven by CCC firms alone. Excluding CCC firms however from the regressions of equations (2) and (3) does not materially affect the results of Table III. The CCC rating levels have a small number of firm years, as shown in Table I, and thus do not significantly impact the overall results. The incremental effect of the B/CCC distinction is directly tested in the following section.

25

grade. This result underscores the significance of maintaining an investment grade rating for

firms. This result and the significance of the downgrade to BB+ in Panel A and B indicate that

certain firms target a minimum investment grade rating. For all of the CCC results, the results are

strongest for firms experiencing a multiple rating change, suggesting that certain firms also target a

minimum rating of B-.

D.2. Incremental Impact of Certain Rating Levels

Table VI shows tests of equation (3) with the inclusion of additional dummy variables that identify

downgrades at specific rating levels (investment grade to speculative and B to CCC) and

downgrades of commercial paper ratings (A1 to A2 and A2 to A3). The coefficients on these

dummy variables are incremental to the overall downgrade effect since the general downgrade

variable includes the specific rating downgrades and over 90% of firms that have their commercial

paper rating downgraded also receive long-term rating downgrades.

Results shown in Table VI indicate that firms whose credit rating is downgraded to speculative

from investment grade issue incrementally less debt relative to equity than other downgraded

firms. For the continuous dependent variable NetDIss (columns 1 and 2), a downgrade to

speculative implies approximately 1% less debt issuance relative to equity as a percentage of assets

compared to other downgraded firms, although the coefficient is only significant at 10% for one of

the two specifications. In logit tests with binary dependent variables indicating debt issuance

(column 5) or debt reductions (column 7), firms that are downgraded to speculative are

incrementally less likely to issue debt and more likely to reduce debt, and the result for reducing

debt is significant at 1%. Average marginal probabilities computed from the test of column 7

indicate that a downgrade increases the probability of a debt reduction in a given year by 5.4%,

26

and a downgrade to speculative increases the probability an additional 4.6%. For any given year, a

firm has a probability of reducing debt of approximately 14.6%, so these increases are

economically meaningful. An odds ratio derived from the test of column 5 implies that a

downgrade, after controlling for other factors, decreases the likelihood of a debt issuance by 34%.

Table VI also indicates a strong incremental impact of a downgrade to CCC from B. In tests with

the NetDIss dependent variable, a downgrade from B to CCC implies greater subsequent leverage

reduction versus other downgraded firms, and both results are significant at 1%. A downgrade to

CCC however implies that both an issuance of debt and reduction of debt are less likely. Taken

together these results imply that firms that are downgraded to CCC reduce leverage, however the

firms are unable to retire debt directly and instead must reduce debt through equity issuance or

reduced issuance of debt.

Table VI also indicates a strong incremental impact of a change in commercial paper rating from

A2 to A3. The two main tiers of ratings in the commercial paper market are A1 and A2, and

Crabbe and Post (1994) note that 97% of commercial paper carried this rating in 1991. A

downgrade therefore to A3 significantly restricts a firm’s access to the commercial paper market.

This result corresponds to the results of Table V that indicate effects at the BBB and BBB- long-

term ratings, since these ratings levels correspond to changes in commercial paper rating from A2

to A3. The change from A1 to A2 is not incrementally significant, and in some cases is significant

in the opposite direction predicted (although when considered with the downgrade variable, the

coefficients together imply no significant capital structure activity for this change in rating).

Money market funds, who make up a significant portion of commercial paper investment, invest

nearly exclusively in A1 rated paper, and A1 rated commercial paper has more favorable firm

liquidity requirements than lower rated paper and lower spreads (Hahn (1993)). Despite this, the

27

A1 to A2 change does not lead to leverage reductions. These results are consistent with Nayar and

Rozeff (1994), who find significant negative stock price reactions to downgrades in commercial

paper ratings for firms. They also find that a downgrade to A3 has a greater negative stock price

reaction than a downgrade to A2, and that stock prices react positively to a firm receiving an initial

commercial paper rating.

E. Results by Year

The results indicating that firms target a minimum credit rating could be driven by changes in the

business cycle, if more firms are downgraded during economic downturns and firms issue less debt

during these same periods. Table VII shows results of equation (3) for market value measures by

year. The table indicates the downgrade results are significant across individual years. The

coefficient on downgrade is significant at 5% in 11 out of 17 years, and the coefficient is negative

in every year. The magnitudes across years are also significant, with downgraded firms issuing

over 1.5% more net equity than net debt in 14 out of 17 years. These results confirm that the

results of this paper are independent of business cycles.

The incentives to regain a target rating might still be strongest in years in which credit spreads are

largest. To examine this relationship, I calculate the average spread between BBB and AAA

bonds for the three months prior to each year (I also calculate the average for 12 months prior to

the issuance year, with little change in results). The simple correlation between this spread and the

coefficients reported in Table VII is -0.59. The coefficient on an interaction term between the

spread and downgrade variable is also negative and statistically significant when included in

equation (3) of the paper (tables not reported). These results suggest target credit rating behavior

is largest when credit spreads are higher.

28

III. Do Firms Succeed in Targeting a Rating?

Not all firms have the ability to achieve a desired rating, given a firm’s financial condition or

limitations on access to capital. Many downgraded firms however are successful at achieving

subsequent upgrades. I examine firms that have at least one downgrade during the sample period

and that have 5 years of data following one of the downgrades. This criterion identifies 266 firms

in the sample. Of these firms, 149 (56.0%) received an upgrade in their rating within the 5

subsequent years following the downgrade. Over half of downgraded firms are able to achieve

upgrades within a reasonable period following the downgrade.

Additionally, firms that actively target a minimum credit rating are more likely to achieve

upgrades following a downgrade. 6.3% of downgraded firms are upgraded the subsequent year if

NetDIss is negative that year, whereas 3.6% of downgraded firms are upgraded the subsequent

year if NetDIss is positive. This data indicates that the dependent variable measures behavior that

leads to regaining a target rating.

Table VIII shows examples of rating downgrades and leverage changes for 15 different firms in

the sample of this study. These situations are illustrative only, however examining the credit

rating and leverage changes of actual firms is useful to see if real examples are consistent with the

general results of the paper. These firms were identified by screening on firms that were

downgraded and that had negative values of net debt minus net equity issuances the subsequent

year (that is, they appear to be targeting a minimum credit rating). From that group of firms,

specific cases were selected with the additional criterion that the firms were successful in obtaining

29

a subsequent upgrade.14 These examples illustrate that the empirical tests of this paper identify

firms undertaking activity consistent with targeting a minimum credit rating. For example,

Panenergy Corp. was downgraded from BBB- to BB+ in 1992, NetDIss was -14.0% in 1993, and

they were subsequently upgraded to BBB- in 1993. The empirical results of the paper taken

together with these illustrations indicate that firms make long-term leverage level decisions with

consideration for minimum credit rating levels.

IV. Conclusions

A firm’s capital structure decision is affected more by whether the firm’s credit rating was

downgraded the previous year than by whether the firm’s leverage changed the previous year or

previous two years. Firms are more likely to reduce debt and less likely to issue debt following a

downgrade. The empirical evidence suggests this behavior is independent of distress concerns,

timing activity and yearly business cycle effects and is consistent with a long-term capital structure

policy of targeting a minimum credit rating.

Firms target minimum credit ratings at which regulations affect investment in a firm’s bonds and

at which commercial paper access is affected. Firms target the investment grade rating level, a

minimum B- rating, and a minimum A2 commercial paper rating. Regulations based on ratings

determine whether certain investor groups (e.g., banks and pension funds) can invest in the bonds,

the capital charges that investors (e.g., insurance companies and broker-dealers) incur from

holding the bonds, and listing and disclosure requirements for the bonds. Several of these

regulations relate particularly to the investment grade distinction and the B-/CCC+ distinction.

14 This screening identified significantly more than the firms shown here. The firms and corresponding years shown are ones that were reasonably stable over many years (e.g., no major acquisitions), and where the different issues of the paper were best illustrated.

30

Certain higher long-term bond ratings also correspond to commercial paper ratings that directly

affect a firm’s ability to issue commercial paper.

The effect of discrete credit rating level benefits on capital structure behavior is complementary to

the tradeoff theory of capital structure. A downgrade is predictive for issuance behavior after

controlling for other tradeoff theory factors, and firms exhibit capital structure behavior consistent

with traditional tradeoff theory arguments after controlling for downgrades. Firms whose leverage

has increased (decreased) the previous year or the year before are more likely to undertake

leverage reducing (increasing) capital market activity the following year, and several other tradeoff

theory factors also remain predictive for issuance behavior.

Appendix A firm can depart from its optimal leverage for the reasons described in the following three cases.

In each case, the subsequent capital structure implications are described for TT and TTCR given

the objective functions in equation (1).

Case 1: The firm departs from its optimal leverage to a different level of leverage, LS, but all

other firm characteristics are unchanged. This move could occur because the firm did an offering

of debt alone or equity alone. Previous literature finds firms will depart from their target leverage

for periods of time and then revert to it (Marsh (1982)), and that firms in a particular year issue

debt only or equity only more often than they issue both (Kisgen (2006) and Faulkender and

Petersen (2006)). The firm could also have had an exogenous cash flow shock with no change to

future expected cash flows. In this case, the firm’s optimal level of leverage remains at LT or Lr*

under TT or TTCR, respectively.

31

TT implication: Firms will undertake capital structure behavior to move back to the target if

(B(LT)-C(LT)) – (B(LS)-C(LS))>AC. Since the TT optimal leverage is not measurable, empirical

literature has adopted several approaches to measure departure from the target leverage.15 One

way to test this implication not previously explored is to examine firms over time and test if

increases (decreases) in leverage are followed by capital structure behavior to reduce (increase)

leverage the following year. Testing the implication in this manner will bias results toward zero as

some changes in leverage are not departures from LT, however this implication provides for an

intuitive comparison of TT and TTCR and mitigates certain issues with previous tests.16

TTCR implication: Firms will undertake capital structure behavior to move back to the target if

(B(Lr*)-C(Lr*) + Φ (Lr*)) – (B(LS)-C(LS) + Φ (LS))>AC. If the change in leverage caused a one

rating downgrade, a firm will reduce leverage to move back to the target if (B(Lr*)-C(Lr*) + Φr*) –

(B(LS)-C(LS))>AC. Similar to TT, moving back to the optimal leverage will have tradeoff theory

benefits, since Lr* is on the upward sloping part of the function V(L). This inequality has several

implications that are different from TT however. Conditional on the same change in leverage, this

inequality implies a credit rating downgrade increases the probability that a firm will reduce

leverage the subsequent year. Similarly, for even a small change in leverage, a firm that receives a

downgrade might still reduce leverage to regain the higher rating if the benefits of that higher

15 One approach to estimating the target is to take the average leverage over the sample period or historically (Taggert (1977), Marsh (1982), Shyam-Sunder and Myers (1999)). An alternate approach is to regress leverage cross-sectionally on factors thought to predict leverage and use predicted values from these regressions as target leverages (Hovakimian, Opler and Titman (2001), Fama and French (2002)). 16 Using the sample period to estimate target leverage is biased towards finding a result in favor of the tradeoff theory (Shyam-Sunder and Myers (1999)), and using historical averages unrealistically assumes that leverage targets have not changed over time. If cross sectional regressions are used to estimate the target and they do not include every factor that predicts leverage, there will be a missing variable in those cross sections. Titman and Wessels (1988) note the limitations of cross sectional leverage regressions: “there may be no unique representation of the attributes we wish to measure” and since the representations chosen can be imperfect, they introduce, “an errors-in-variable problem”.

32

rating are significant. The size of Φr is also important. The more significant the cost associated

with the downgrade, the more likely the firm will attempt to reverse the downgrade.

If LS < Lr* and the firm received a one rating upgrade, TTCR implies a firm will increase leverage

to move back to Lr* if ((B(Lr*)-C(Lr*)) – (B(LS)-C(LS) + Φr*+1)>AC (with r*+1 defined as the rating

level one category higher than r*), or equivalently (B(Lr*)-C(Lr*)) – (B(LS)-C(LS))>AC+ Φr+1. The

left hand side of the inequality is positive, since B’(L)-C’(L)>0 ∀L< LT. The inequality implies a

firm will increase leverage following a leverage reduction and upgrade if the tradeoff benefits

outweigh the adjustment costs and discrete cost of a downgrade. In this case, the upgrade itself

does not increase the likelihood of a reversal in leverage. Conditional on the same change in

leverage, an upgrade makes it less likely that the firm will undertake capital structure behavior to

reverse the change in leverage. The downgrade and upgrade results together imply capital

structure behavior will be more significant following a downgrade than an upgrade, an additional

implication distinct from TT.17

Case 2: Credit rating agencies initiate a policy change such that there are new levels of Lr, which I

denote as LLr. This could occur because rating agencies have become more or less stringent, or

rating agencies could have changed their treatment of particular securities. Blume, Lim and

Mackinlay (1998) and Doherty and Phillips (2002) study the change in rating standards of credit

rating agencies over time. For the period 1978-1995, Blume, Lim and Mackinlay (1998) find that

rating agencies consistently became more stringent, such that a firm with the same levels of

accounting variables and equity risk measures had lower ratings over time. This is true at both

17 The change in leverage might also not cause a change in rating, in which case the Φ terms cancel out. The implications in this instance appear similar to TT, however there is one subtle difference. If a firm moved to LS > Lr* without a downgrade, the firm should not undertake any capital structure behavior. If a firm incurred a reduction in leverage without an upgrade, the implications are similar to TT.

33

investment grade and speculative grade levels. Doherty and Phillips (2002) draw similar

conclusions, focusing on the insurance industry specifically.

TT implication: No implication, since nothing fundamental changed in the firm.

TTCR implication: If the change in policy is more stringent such that LLr < Lr ∀ r and if the firm’s

optimal leverage was Lr*, the firm receives a downgrade and the firm’s new optimal leverage could

be at LL r*, the different leverage required to keep the same rating. Assuming a one rating

downgrade, a firm will reduce leverage the next year to LL r* if (B(LL r*)-C(LL r*) + Φr*) – (B(L r*)-

C(L r*))>AC. This implies a downgrade itself can imply a reduction in leverage the subsequent

period, even if the firm’s leverage or financial condition didn’t change the previous period.18 The

size of Φr is also significant. The greater the discrete cost of the downgrade, the more likely the

firm will reduce leverage to attempt to reverse the downgrade.

If the change in policy is more lenient such that LLr > Lr ∀ r, the firm is able to maintain the same

rating with a higher level of leverage. If the firm also received an upgrade and if LLr* is less than

LT, the firm would increase leverage to LLr* if (B(LLr*)-C(LLr*)) – (B(Lr*)-C(Lr*)+ Φr*+1) >AC, or

equivalently, if (B(LLr*)-C(LLr*)) – (B(Lr*)-C(Lr*)) >AC+ Φr*+1. This can occur if the tradeoff

benefits of increasing leverage outweigh the costs of a downgrade and adjustment costs. If the

tradeoff benefits of increasing leverage are small relative to adjustment costs, the firm would not

undertake leverage increasing capital market activity. The upgrade does not make it more likely

that the firm will increase leverage, since the upgrade provides a discrete benefit to the firm. 18 It is also possible the policy change and downgrade would cause a firm to increase leverage. The firm might increase leverage if the leverage optimum for the firm is no longer at the same minimum rating level r, but instead changed to LT or LLr*-1 (with r*-1 defined as the rating level one category lower than r*). A firm would increase leverage in this instance if (B(LT)-C(LT)) – (B(L r*)-C(L r*))>AC or (B(LL r*-1)-C(LL r*-1)) – (B(L r*)-C(L r*))>AC, respectively. These cases represent instances where the rating agency policy becomes so stringent it is no longer optimal for the firm to target that same minimum rating.

34

Conditional on the size of the tradeoff benefits from increasing leverage, a rating upgrade makes it

less likely the firm undertakes capital structure behavior to move to the optimal. The larger Φr*+1

is, the less likely the firm will increase leverage, conditional on the tradeoff benefits.19 The TTCR

implications of Case 2 in general are similar to Case 1 for capital structure behavior, with

downgrades implying subsequent reductions in leverage and upgrades by themselves implying no

capital structure behavior.

Case 3a: Permanent change to the firm that does not affect the probability of bankruptcy (e.g.,

changes in relative tax rates that do not affect after tax cash flows), such that the functional form of

C(L) or B(L) changes (or both), but the levels of Lr do not change (I denote the new functions as

CC(L) and BB(L), respectively).

TT implication: The firm will have a new target leverage, L*T. Firms will undertake capital

structure behavior to move to the new target if (BB(L*T)-CC(L*

T)) – (BB(LT)-CC(LT))>AC. For

example, if the new cost of debt function is CC(L)=zC(L) with z>1 and B(L) is unchanged, the

firm’s new target leverage will solve B’(L*T)=zC’(L*T), implying L*T< LT. The firm will decrease

leverage to L*T if (B(L*

T)-zC(L*T)) – (B(LT)-zC(LT))>AC. To the extent that the functional forms

of C(L) and B(L) are measurable, changes in those measures will imply changes in capital

structure. This has been implicitly examined in previous literature (Marsh (1982), Mackie-Mason 19 The new optimal leverage might also have changed to the leverage level required to maintain the higher rating, LLr*+1, in which case the firm would increase leverage only if (B(LLr*+1)-C(LLr*+1)) – (B(Lr*)-C(Lr*)) >AC, to obtain the tradeoff benefits of higher leverage at the same rating. A firm might also not receive an upgrade from a more lenient policy. If LLr* is less than LT, then V(LLr*) is greater than V(Lr*) since B’(L)-C’(L)>0 ∀ L< LT. The firm will increase leverage to LLr* if (B(LLr*)-C(LLr*)) – (B(Lr*)-C(Lr*)) >AC. A firm might also reduce leverage from a more lenient policy with no upgrade since LLr*+1 could now maximize firm value. This could occur if the more lenient credit rating standards reduce the cost of departing from the tradeoff optimum enough for the firm to try to achieve the higher rating. The firm will reduce leverage if (B(LLr*+1)-C(LLr*+1)+ Φr*+1) – (B(Lr*)-C(Lr*)) >AC. LLr* could also be greater than or equal to LT, in which case LT or LLr*+1 could maximize firm value, and LLr* would not maximize firm value. The firm would either increase leverage to LT or decrease leverage to LLr*+1, if the benefits of doing so outweighed the adjustment costs. Since no upgrades occur in any of these instances, no capital structure behavior is implied following rating changes.

35

(1990) and Graham (1996)), and I include several measures from these papers in the empirical

tests of this paper.

TTCR implication: Since the levels of Lr have not changed, the firm would not have received an

upgrade or downgrade in this case. Therefore no specific capital structure activity is implied from

this case following changes in credit ratings. Unlike TT however, in some cases TTCR will imply

no capital structure activity even if AC=0 simply because firm value could still be maximized at

Lr* (even though LT has changed). Alternatively, the shift in the curve can also imply a different

level of optimal leverage under TTCR, but not due to any change in rating. For these cases, the

implications are similar in nature to TT.20

Case 3b: Permanent change in the firm affecting firm fundamentals (e.g., change in firm business

risk), such that the functional form of C(L) changes to CC(L) and the levels of Lr change to LLr. (I

assume for simplicity that changes in B(L) would not affect the levels of Lr).

TT implication: The implications are identical to Case 3a.

TTCR implication: This case can be interpreted as a combination of cases 2 and 3a. As one

example, consider a firm that has an increase in business risk or probability of distress such that

CC(L)=zC(L) with z>1 and LLr < Lr ∀ r. This implies that the firm will be downgraded at its

current level of leverage, Lr*, and will likely have a different optimal level of leverage. This new

20 In cases where the change is significant enough, capital structure activity is implied if for some L, (BB(L)-CC(L)+ Φ (L)) – (BB(L r*)-CC(L r*) + Φ (L r*))>AC. As an example, if B(L) has not changed and CC(L)=zC(L), this inequality requires minimally that (B(L)-zC(L)+ Φ (L)) > (B(L r*)-zC(L r*) + Φ (L r*)) for some L. Since L r* maximized firm value before the change, we know that for any L, (B(L)-C(L)+ Φ (L)) < (B(L r*)-C(L r*) + Φ (L r*)). These two inequalities together imply that for capital structure to be implied, minimally, (1-z)[C(L)-C(L r*)]>0. If z is greater than 1, then L must be less than L r* for capital structure behavior to be implied, and if z is less than 1, L must be greater than L r*.

36

optimum could be at several different potential levels, including LLr*, LLr*-1 or L*T. Assuming a

one rating downgrade for simplicity, TTCR implies the firm will reduce leverage to LLr* to regain

the higher rating level if ((B(LLr*)-zC(LLr*) + Φr*)) – (B(Lr*)-zC(Lr*))>AC (note if z=1 this

reduces to Case 2). Although L*T is lower than LT in this case, if L*

T is still higher than Lr*, the

quantity (B(LLr*)-zC(LLr*)) – ((B(Lr*)-zC(Lr*)) is negative (but less negative than if z=1).

Rearranging, the firm will reduce leverage to regain the higher rating if Φr*>AC+((B(Lr*)-zC(Lr*))

- (B(LLr*)-zC(LLr*)). The firm will reduce leverage if the credit rating benefits outweigh the

adjustment costs and the tradeoff benefits of remaining at the higher leverage.21 If the cost of the

downgrade is large, TTCR implies a firm will reduce leverage following a downgrade as in Case 2.

The upgrade implications follow similarly, where upgrades decrease the likelihood of subsequent

capital structure behavior.22

REFERENCES

Arellano, M., 1987, Computing robust standard errors for within-groups estimators, Oxford Bulletin of Economics and Statistics 49, 431-434. Baker, Malcolm and Jeffrey Wurgler, 2002, Market timing and capital structure, Journal of Finance 57, 1-32. Berger, Philip G., Eli Ofek, and David L. Yermack, 1997, Managerial entrenchment and capital structure decisions, Journal of Finance 52, 1411-1438. Blume, Marshall E., Felix Lim, and A. Craig Mackinlay, 1998, The declining quality of U.S. corporate debt: myth or reality?, Journal of Finance 53, 1389-1413. Boot, Arnoud W. A., Todd T. Milbourn, and Anjolein Schmeits, 2005, Credit ratings as coordination mechanisms, Review of Financial Studies, forthcoming. 21 If, due to the change in C(L), L*

T is lower than Lr* but greater than LLr*, the firm will achieve first positive and then negative tradeoff benefits if it moved to LLr*. In this case the firm will again reduce leverage to regain the higher rating if Φr*>AC+((B(Lr*)-zC(Lr*)) - (B(LLr*)-zC(LLr*)), but the firm might be more likely to reduce leverage than in the previous case given that the tradeoff effects are indeterminate. 22 For example, if the firm reduced its business risk such that CC(L)=zC(L) with z<1 and LLr > Lr ∀ r, the firm could attain the same rating with a higher level of leverage. Tradeoff effects will imply a higher leverage under TT or TTCR. If the firm also received an upgrade and LLr* is less than L*

T, the firm will increase leverage to LLr* if (B(LLr*)-zC(LLr*)) – (B(Lr*)-zC(Lr*)) >AC+ Φr*+1. The left hand side is positive since B’(L r*)-C’(L r*)>0, implying tradeoff benefits of increasing leverage. The upgrade itself decreases the likelihood of an increase in leverage.

37

Cantor, Richard and Frank Packer, 1994, The credit rating industry, Federal Reserve Bank of New York Quarterly Review 19, 1-26. Crabbe, Leland, and Mitchell A. Post, 1994, The effect of a rating downgrade on outstanding commercial paper, Journal of Finance 49, 39-56. Doherty, Neil A., and Richard D. Phillips, 2002, Keeping up with the joneses: changing rating standards and the buildup of capital by U.S. property-liability insurers, Journal of Financial Services Research 22, 55-78. Fama, Eugene F., and Kenneth R. French, 2002, Testing tradeoff and pecking order predictions about dividends and debt, Review of Financial Studies 15, 1-33. Fama, Eugene F., and James MacBeth, 1973, Risk, return and equilibrium: Empirical tests, Journal of Political Economy 81, 607-636. Faulkender, Michael, and Mitchell Petersen, 2006, Does the source of capital affect capital structure?, Review of Financial Studies, forthcoming. Financial Times, 2004, Downgrade offers exit for GE: News of US Airways’ lower credit rating gives creditors a chance to withdraw financing for its regional jets, May 6, pg. 15. Financial Times, 2005, Top tier rating is losing its appeal, March 9, pg. 25. Fischer, Edwin O., Robert Heinkel, and Josef Zechner, 1989, Dynamic capital structure choice: theory and tests, Journal of Finance 44, 19-40. Flannery, Mark J., and Kasturi P. Rangan, 2005, Partial adjustment toward target capital structures, Journal of Financial Economics, forthcoming. Graham, John R., 1996, Debt and the marginal tax rate, Journal of Financial Economics 41, 41-74. Graham, John R., and Campbell R. Harvey, 2001, The theory and practice of corporate finance: evidence from the field, Journal of Financial Economics 60, 187-243. Hahn, Thomas K., 1993, Commercial paper, Federal Reserve Bank of Richmond Economic Quarterly 79, 45-67. Hovakimian, Armen, Tim Opler, and Sheridan Titman, 2001, The debt-equity choice, Journal of Financial and Quantitative Analysis 36, 1-24. Johnson, Richard, 2003, An examination of rating agencies’ actions around the investment-grade boundary, Federal Reserve Bank of Kansas City Research Working Papers. Kisgen, Darren J., 2006, Credit Ratings and Capital Structure, Journal of Finance, forthcoming. Leary, Mark T., and Michael R. Roberts, 2005, Do firms rebalance their capital structures?, Journal of Finance 60, 2575-2619.

38

Leland, Hayne E., 1994, Corporate debt value, bond covenants, and optimal capital structure, Journal of Finance 49, 1213-1252. Mackie-Mason, Jeffrey K., 1990, Do taxes affect corporate financing decisions?, Journal of Finance 45, 1471-1493. Marsh, Paul, 1982, The choice between equity and debt: an empirical study, Journal of Finance 37, 121-144. Nayar, Nandkumar, and Michael S. Rozeff, 1994, Ratings, commercial paper, and equity returns, Journal of Finance 49, 1431-1449. Patel, Jayen, Dorla Evans, and John Burnett, 1998, The junk bond behavior with daily returns and business cycles, The Journal of Financial Research 21, 408-418. Petersen, Mitchell A., 2005, Estimating standard errors in finance panel data sets: comparing approaches, working paper, Northwestern University. Rogers, William, 1993, Regression standard errors in clustered samples, Stata Technical Bulletin 13, 19-23. SEC, 2003, “Report on the Role and Function of Credit Rating Agencies in the Operation of the Securities Markets”, January 24. Shyam-Sunder, Lakshmi, and Stewart C. Myers, 1999, Testing static tradeoff against pecking order models of capital structure, Journal of Financial Economics 51, 219-244. Standard and Poor’s, 2001, Corporate Ratings Criteria, (McGraw-Hill, New York, NY). Stern Stewart Roundtable on Capital Structure and Stock Repurchase, 2001, Journal of Applied Corporate Finance 14, 8-41. Taggart, Robert A., 1977, A model of corporate financing decisions, Journal of Finance 32, 1467-1484. The Economist, 2005, Credit-rating agencies: Three is no crowd, March 26 - April 1, page 15. Titman, Sheridan and Roberto Wessels, 1988, The determinates of capital structure choice, Journal of Finance 43, 1-19. Wall Street Journal, 2004, EDS aims to save debt rating, May 11, B10.

Welch, Ivo, 2004, Capital structure and stock returns, Journal of Political Economy 112, 106-131. Wooldridge, Jeffrey M., 2002, Econometric Analysis of Cross Section and Panel Data (The MIT Press, Cambridge).

39

Table I

Credit Rating Upgrades and Downgrades by Rating Level

Number of firm years in the sample in which the firm was downgraded or upgraded to the indicated rating level. Percentages are shown for the number of downgrades to a rating as a percentage of the number of firm years with that rating. The sample is all Compustat firms from 1987 to 2003 with a credit rating for 3 consecutive years and 3 years of non-missing variables required for conducting the tests of the paper. The credit rating is a firm’s long-term domestic issuer credit rating.

AA+ AA AA- A+ A A- BBB+ BBB BBB-

#Downgraded to 8 20 46 71 118 149 166 203 141 (Pct. of firm years) 6.1% 4.4% 9.5% 8.3% 8.5% 13.9% 14.5% 15.4% 14.0%

#Upgraded to 8 18 25 58 100 122 116 105 125

(Pct. of firm years) 6.1% 4.0% 5.2% 6.8% 7.2% 11.4% 10.1% 8.0% 12.4%

Total Firm Years 131 453 485 858 1387 1073 1146 1317 1007

BB+ BB BB- B+ B B- CCC+ CCC CCC-

#Downgraded to 106 118 126 126 142 79 63 29 16 (Pct. of firm years) 16.2% 13.3% 13.6% 11.6% 28.8% 34.1% 56.3% 46.0% 47.1%

#Upgraded to 97 118 103 78 34 18 8 7 4

(Pct. of firm years) 14.8% 13.3% 11.1% 7.2% 6.9% 7.8% 7.1% 11.1% 11.8%

Total Firm Years 656 885 929 1088 493 232 112 63 34

40

Table II

Debt and Equity Decisions Following Rating Changes Percentage of firm years in which the firm undertakes the indicated capital market activity, given the change in rating the previous year. Issuance and reduction is defined as a net issuance or reduction greater than 5% of beginning of year assets. The indicated change in rating is as of the year prior to the capital structure decision. “Downgrade to Speculative” indicates a downgrade from an investment grade rating at time t-2 to a speculative grade rating at time t-1, and “Upgrade to Investment Grade” indicates the opposite. The sample is all Compustat firms from 1987 to 2003 with a credit rating for 3 consecutive years and 3 years of non-missing variables required for conducting the tests of the paper.

Rating Change % Firms % Firms % Firms % Firms (Previous Year) Issue Debt Red. Debt Issue Equity Red. Equity

Downgrades No Downgrade 24.7% 13.0% 5.9% 6.7% Downgraded 14.7% 24.2% 6.5% 2.9% Down to Speculative 13.3% 36.3% 7.4% 2.3%

Upgrades No Upgrade 22.4% 14.9% 5.9% 6.1% Upgraded 31.8% 11.6% 7.3% 6.7% Upg. to Inv. Grade 28.6% 9.1% 8.6% 6.9%

41

Table III

Capital Structure Decisions Following Credit Rating Changes Coefficients and standard errors from pooled time-series cross-section regressions of net debt raised for the year minus net equity raised for the year divided by beginning of year total book (columns 1, 3, 5 and 7) or market (columns 2, 4, 6 and 8) assets on credit rating dummy variables and on various explanatory variables. Downgrade and upgrade are dummy variables with a value of 1 if the firm was downgraded or upgraded the previous year, respectively. “Fixed effects” refers to the inclusion of dummy variables for each firm in the sample. Errors are White’s consistent standard errors or clustered errors. The sample is all Compustat firms from 1987 to 2003 with a credit rating for 3 consecutive years and 3 years of non-missing data for computing the variables. I use ***, **, and * to denote significance at the 1, 5 and 10 percent levels, respectively.

1 2 3 4 5 6 7 8

Downgrade -0.0408*** -0.0343*** -0.0205*** -0.0220*** -0.0205*** -0.0220*** -0.0288*** -0.0272*** (0.0035) (0.0028) (0.0040) (0.0029) (0.0039) (0.0028) (0.0039) (0.0029) Upgrade 0.0099** 0.0077*** 0.0057 0.0057* 0.0057 0.0057* 0.0105** 0.0084** (0.0049) (0.0033) (0.0048) (0.0033) (0.0048) (0.0034) (0.0053) (0.0038) ∆Lev. (Book) -0.0325* -0.0325* -0.0922*** (0.0204) (0.0189) (0.0238) ∆Lev. (Mkt) -0.0295** -0.0295** -0.0465*** (0.0143) (0.0142) (0.0136) ln(Sales) -0.0052*** -0.0040*** -0.0052*** -0.0040*** (0.0012) (0.0008) (0.0013) (0.0008) ∆Sales 0.0338*** 0.0321*** 0.0338*** 0.0321*** 0.0130* 0.0135** (0.0096) (0.0063) (0.0099) (0.0066) (0.0093) (0.0062) EBITDA 0.1923*** 0.1002*** 0.1923*** 0.1002*** (0.0317) (0.0161) (0.0328) (0.0165) ∆EBITDA -0.0055 -0.0091 -0.0055 -0.0091 0.0929** 0.0370 (0.0441) (0.0239) (0.0431) (0.0237) (0.0486) (0.0258) Market/Book 0.0055** -0.0020** 0.0055** -0.0020** (0.0024) (0.0008) (0.0024) (0.0009) ∆Market/Bk -0.0049** 0.0016* -0.0049** 0.0016* -0.0002 -0.0009** (0.0022) (0.0008) (0.0023) (0.0009) (0.0005) (0.0004) ∆Z-Score 0.0024 0.0015 0.0024 0.0015 -0.0003 0.0042* (0.0040) (0.0020) (0.0039) (0.0020) (0.0057) (0.0025) Rating Level -0.0027*** -0.0021*** -0.0027*** -0.0021*** (0.0004) (0.0003) (0.0005) (0.0003) Intercept 0.0267*** 0.0171*** 0.0530*** 0.0533*** 0.0530*** 0.0533*** N/M N/M

(0.0016) (0.0011) (0.0117) (0.0077) (0.0129) (0.0077) N/M N/M

β1 > -β2 (p-value)

0.0001 0.0001 0.0015 0.0001 0.0015 0.0001 0.0084 0.0002

Fixed Effect? No No No No No No Yes Yes Clustering? No No No No Yes Yes Yes Yes N 12851 12851 12851 12851 12851 12851 12425 12425 R2 0.8% 1.2% 3.1% 2.9% 3.1% 2.9% N/M N/M

42

Table IV

Capital Structure Decisions Following Rating Changes – 2 Years Lagged Leverage Changes Coefficients and standard errors from pooled time-series cross-section regressions of net debt raised for the year minus net equity raised for the year divided by beginning of year total book (columns 1-3) or market (columns 4-6) assets on credit rating dummy variables and on various explanatory variables. Downgrade is a dummy variable with a value of 1 if the firm was downgraded the previous year and Upgrade is a dummy variable equal to 1 if the firm was upgraded the previous year. Errors are White’s consistent standard errors. The sample is all Compustat firms from 1988 to 2003 with a credit rating for 3 consecutive years and non-missing data for computing the variables. I use ***, **, and * to denote significance at the 1, 5 and 10 percent levels, respectively.

1 2 3 4 5 6

Downgradet-1 -0.0197*** -0.0195*** (0.0040) (0.0029) Upgradet-1 0.0023 0.0011 (0.0053) (0.0035) ∆Lev. (Book) t-1 -0.0668*** -0.0291* -0.0202 (0.0193) (0.0215) (0.0220) ∆Lev. (Book) t-2 -0.0589** -0.0489** -0.0446** (0.0236) (0.0245) (0.0247) ∆Lev. (Mkt) t-1 -0.0593*** -0.0404*** -0.0287*** (0.0127) (0.0126) (0.0129) ∆Lev. (Mkt) t-2 -0.0826*** -0.0611*** -0.0496*** (0.0122) (0.0124) (0.0127) ln(Sales) t-1 -0.0057*** -0.0051*** -0.0042*** -0.0037*** (0.0014) (0.0013) (0.0008) (0.0008) ∆Sales t-1 0.0325*** 0.0302** 0.0311*** 0.0288*** (0.0114) (0.0114) (0.0070) (0.0069) EBITDA t-1 0.2108*** 0.1979*** 0.1121*** 0.1015*** (0.0399) (0.0407) (0.0193) (0.0193) ∆EBITDA t-1 0.0359 0.0361 -0.0112 -0.0052 (0.0555) (0.0553) (0.0277) (0.0277) Market/Book t-1 0.0083** 0.0083** -0.0017* -0.0018* (0.0033) (0.0033) (0.0011) (0.0011) ∆Market/Book t-1 -0.0078** -0.0078** 0.0014 0.0015 (0.0031) (0.0031) (0.0011) (0.0011) ∆Z-Score t-1 0.0020 0.0014 0.0019 0.0010 (0.0039) (0.0039) (0.0020) (0.0020) Rating Level t-1 -0.0027*** -0.0024*** -0.0022*** -0.0019*** (0.0004) (0.0004) (0.0003) (0.0003) Intercept 0.0232*** 0.0492*** 0.0465*** 0.0142*** 0.0515*** 0.0487***

(0.0015) (0.0132) (0.0131) (0.0010) (0.0078) (0.0079)

β1 > -β2 (p-value)

N/A N/A 0.0065 N/A N/A 0.0001

N 10931 10931 10931 10931 10931 10931 R2 0.4% 3.2% 3.4% 0.8% 2.7% 3.0%

43

Table V

Capital Structure Decisions Following a Downgrade by Individual Rating

Coefficients, in percentages, on Downgrade, a dummy variable indicating the firm was downgraded the previous period, in regressions of net debt minus net equity divided by book or market assets on Downgrade and various control variables. ***, **, and * below the coefficient denotes significance at the 1, 5 and 10 percent levels, respectively. Bold columns in Panel A are where at least 2 out of 4 coefficients for that rating are significant at 5%. Rows (1) and (2) are tests using book measures of leverage and (3) and (4) are tests with market measures. Rows (1) and (3) include control variables of changes in leverage, EBITDA, Sales, z-score, M/B and levels of EBITDA, Sales and M/B. Rows (2) and (4) exclude control variables of changes in EBITDA, Sales, z-score and M/B. Panel A shows all ratings levels, and Panel B shows regressions for ratings bolded in Panel A, but with the Downgrade dummy variable separated into two dummy variables indicating downgrades moving one category (e.g., A to A-) and Downgrades of more than one category (e.g., A to BBB+). (in percentages) Panel A: Coefficients on “Downgrade” in Individual Rating Tests

AA+ AA AA- A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B B- CCC+ CCC CCC- (1) 5.5 4.9 -1.5 0.1 -1.4 -0.1 -1.1 -2.6 -1.4 -1.5 -4.4 -1.8 -0.3 -1.5 -2.4 -13.1 -7.2 -10.3

* * *** ** *** ** ** (2) 5.1 4.3 -2.0 -0.3 -1.5 -0.2 -1.1 -2.8 -1.8 -1.7 -4.3 -3.4 -1.1 -2.9 -2.8 -13.3 -4.6 -11.4

** * *** ** * * *** ** (3) 2.0 1.6 -1.1 0.3 -1.7 -0.3 -0.8 -2.2 -2.0 -2.1 -4.3 -2.2 -0.9 -1.9 -1.1 -12.5 -7.0 -9.6

* ** *** ** ** *** * ** ** ** (4) 1.5 1.4 -1.3 0.2 -1.6 -0.7 -0.9 -2.3 -2.2 -2.5 -4.5 -3.4 -1.8 -3.2 -1.3 -11.0 -6.8 -8.7

** ** *** ** ** *** ** ** ** ** ** Panel B: “Downgrade” separated into two dummy variables for most significant ratings levels of Panel A AA- A BBB BBB- BB+ BB CCC+ CCC CCC- (1) Downgrade1 -1.5* -0.7 -2.3** -0.8 -0.6 -2.7 -7.3 -3.0 -0.5 Downgrade2+ -2.2 -2.2 -3.0** -2.4 -2.9 -5.9** -16.9*** -7.8*** -11.4** (2) Downgrade1 -2.0** -1.2 -2.5** -1.2 -0.6 -2.5 -8.5* -1.7 -0.6 Downgrade2+ -1.9 -2.5 -3.3** -2.9 -3.3* -5.7** -16.3*** -5.0 -12.3** (3) Downgrade1 -1.2* -1.3** -2.1*** -1.7* -1.3 -3.2** -6.4 -6.9 1.8 Downgrade2+ -1.1 -2.3* -2.3** -2.5* -3.3** -5.2*** -16.7*** -7.3** -11.4** (4) Downgrade1 -1.3** -1.4** -2.2*** -1.9* -1.6 -3.3** -5.6 -6.3 -0.9 Downgrade2+ -1.3 -2.2* -2.5** -2.8* -3.8** -5.4*** -14.3*** -6.9** -9.6**

44

Table VI

Incremental Effects of Particular Rating Downgrades on Leverage Decisions Coefficients and standard errors on credit rating dummy variables and on various explanatory variables from pooled time-series cross-section regressions of net debt raised for the year minus net equity raised for the year divided by beginning of year total book (columns 1 and 3) and market (columns 2 and 4) assets and logit regressions with a dependent variable indicating debt issuance or debt reduction. Downgrade is a dummy variable with a value of 1 if the firm was downgraded the previous year, Down:IG/SG and Down:B/CCC are dummy variables indicating a downgrade to speculative grade or CCC, respectively, and DownCP: A2/A3 and DownCP:A1/A2 are dummy variables indicating a commercial paper rating downgrade to A3 or A2, respectively. Errors are White’s consistent standard errors. The sample is all Compustat firms from 1987 to 2003 with a credit rating for 3 consecutive years and 3 years of non-missing data for computing the variables. I use ***, **, and * to denote significance at the 1, 5 and 10 percent levels, respectively.

OLS Regressions Logit Regressions Book Market Book Market Issue Debt Reduce Debt 1 2 3 4 5 6 7 8

Downgradet-1 -0.0175*** -0.0175*** -0.0226*** -0.0237*** -0.4096*** -0.4878*** 0.4592*** 0.4636*** (0.0038) (0.0027) (0.0041) (0.0031) (0.0793) (0.0590) (0.0755) (0.0725) Down: IG/SGt-1 -0.0090 -0.0117* -0.0234 0.3896*** (0.0089) (0.0073) (0.0136) (0.1498) Down: B/CCCt-1 -0.0465*** -0.0563*** -0.7231*** -0.5562*** (0.0183) (0.0181) (0.3021) (0.2063) DownCP: A2/A3 t-1 -0.0346*** -0.0287*** -1.1434** 0.6687*** (0.0128) (0.0109) (0.5186) (0.2491) DownCP: A1/A2 t-1 0.0211* 0.0207*** 0.5492*** -0.1570 (0.0109) (0.0061) (0.1971) (0.2258) ∆Lev. (Book) t-1 -0.0281 -0.0323* (0.0204) (0.0201) ∆Lev. (Mkt) t-1 -0.0312** -0.0321** -0.7789*** -0.7705*** 0.8236*** 0.8443*** (0.0141) (0.0141) (0.2310) (0.2308) (0.2569) (0.2583) ln(Sales) t-1 -0.0050*** -0.0038*** -0.0052*** -0.0040*** -0.1399*** -0.1367*** 0.0613*** 0.0664*** (0.0012) (0.0008) (0.0012) (0.0008) (0.0167) (0.0171) (0.0206) (0.0209) ∆Sales t-1 0.0310*** 0.0300*** 0.0323*** 0.0312*** 1.0599*** 1.0538*** -0.5466*** -0.5588*** (0.0096) (0.0062) (0.0095) (0.0061) (0.0917) (0.0917) (0.0976) (0.0982) EBITDA t-1 0.1945*** 0.1007*** 0.1941*** 0.1008*** 2.3271*** 2.3213*** 2.5852*** 2.5950*** (0.0313) (0.0159) (0.0313) (0.0159) (0.3245) (0.3237) (0.3736) (0.3750) ∆EBITDA t-1 0.0003 -0.0030 -0.0042 -0.0088 -1.1734** -1.2199*** -1.1727** -1.1503** (0.0437) (0.0239) (0.0436) (0.0237) (0.4729) (0.4700) (0.4787) (0.4794) Market/Book t-1 0.0054** -0.0018** 0.0052** -0.0020** 0.0809*** 0.0802*** -0.0232 -0.0260 (0.0023) (0.0008) (0.0023) (0.0008) (0.0231) (0.0230) (0.0332) (0.0337) ∆Market/Book t-1 -0.0048** 0.0014* -0.0046** 0.0015* -0.0771*** -0.0763*** 0.0147 0.0161 (0.0022) (0.0008) (0.0022) (0.0008) (0.0237) (0.0237) (0.0375) (0.0390) ∆Z-Scoret-1 0.0018 0.0003 0.0023 0.0013 -0.0546 -0.0449 0.0192 0.0217 (0.0040) (0.0020) (0.0039) (0.0019) (0.0479) (0.0473) (0.0460) (0.0460) Rating Levelt-1 -0.0021*** -0.0015*** -0.0021*** -0.0017*** 0.0140** 0.0098 0.1588*** 0.1505*** (0.0004) (0.0003) (0.0005) (0.0003) (0.0066) (0.0075) (0.0085) (0.0094) CPRatingt-1 0.0032 0.0015 -0.0448 -0.0643 (0.0030) (0.0021) (0.0590) (0.0770) Intercept 0.0507*** 0.0494*** 0.0518*** 0.0522*** 0.7863*** 0.7432*** 4.4744*** 4.3876***

(0.0122) (0.0081) (0.0122) (0.0081) (0.1808) (0.1831) (0.2348) (0.2376) N 12851 12851 12851 12851 12851 12851 12851 12851 R2 3.1 3.1 3.1 2.9 N/A N/A N/A N/A

45

Table VII

Impact of Credit Rating and Leverage Changes on Capital Structure Decisions by Year

Coefficients and standard errors from cross-sectional regressions by year of net debt raised for the year minus net equity raised for the year divided by beginning of year total market assets on a constant, a dummy variable for if a firm was downgraded the period before, the firms change in market leverage the previous period, and control variables measured at the beginning of each year. The control variables (not shown) are the firms credit rating level, z-score and levels and changes in EBITDA/A, EBITDA divided by total assets, and ln(Sales), the natural log of total sales, and the market to book ratio. The samples exclude firm years with missing values for any of the variables. Errors are White’s consistent standard errors. I use ***, **, and * to denote significance at the 1, 5 and 10 percent levels, respectively. 1987 1988 1989 1990 1991 1992 Downgrade -0.0538*** -0.0118 -0.0371*** -0.0320*** -0.0250*** -0.0170** (0.0198) (0.0155) (0.0157) (0.0103) (0.0093) (0.0091) ∆Lev. (Mkt) -0.0186 -0.0871* -0.0382 -0.0080 -0.0787* -0.1026*** (0.0889) (0.0598) (0.0555) (0.0395) (0.0542) (0.0426) 1993 1994 1995 1996 1997 1998 Downgrade -0.0275*** -0.0083 -0.0186** -0.0221 -0.0156 -0.0035 (0.0101) (0.0101) (0.0115) (0.0183) (0.0131) (0.0171) ∆Lev. (Mkt) 0.0122 -0.0561 -0.0878** -0.2070*** 0.0871 -0.1997*** (0.0517) (0.0440) (0.0558) (0.0822) (0.0524) (0.0647) 1999 2000 2001 2002 2003 Downgrade -0.0161 -0.0301*** -0.0185** -0.0150** -0.0180*** (0.0126) (0.0121) (0.0087) (0.0076) (0.0072) ∆Lev. (Mkt) -0.1686*** -0.0121 -0.0831*** -0.0165 -0.0158 (0.0409) (0.0387) (0.0275) (0.0288) (0.0304)

46

Table VIII. Target Minimum Credit Rating Cases. Capital structure activity for selected downgraded firms in the sample. The rating downgrade, subsequent capital market activity and subsequent upgrade are shown. NetDIss is net debt issuance minus net equity issuance during the year divided by beginning of year total assets. Previous leverage activity is shown, including a firm’s average NetDIss for the two years prior to the subsequent NetDIss indicated, and the change in the firm’s book leverage during the year of the downgrade.

Leverage Reduction After Downgrade Previous Leverage Activity

Company

Rating Downgrade (Year)

NetDIss Subsequent Year

Subsequent Upgrade (Year)

Average NetDIss Prior 2 yrs

∆L in year of downgrade

AMC Entertainment B+ to CCC+

(2000) -41.4% CCC+ to B-

(2001) +10.0% +6.9%

Avon Products BBB- to BB (1988)

-13.0% BB to BBB+ (1991)

+6.7% +3.0%

BP PLC AA to AA- (1992)

-5.4% AA- to AA (1996)

+1.7% +3.4%

CBS Corp. BBB- to BB (1995)

-15.8% BB to BBB- (1998)

+2.4% -6.9%

Colonial Gas Co. A- to BBB+ (1989)

-11.8% BBB+ to A- (1990)

+6.2% +1.4%

Envirosource Inc. B- to CCC+ (1992)

-10.5% CCC+ to B- (1993)

+0.4% -1.7%

Fruit of the Loom BB to CCC (1990)

-18.6% CCC to BB+ (1991)

+5.4% -3.2%

GoodYear Tire BBB- to BB (1991)

-8.4% BB to BBB- (1993)

-8.1% +3.2%

Honeywell Int’l A to A- (1991)

-6.8% A- to A (1992)

+3.5% +2.0%

Northwestern Corp. A+ to A (1995)

-25.9% A to A+ (1996)

+11.7% +0.2%

Occidental Petroleum BBB to BBB- (1990)

-11.2% BBB- to BBB (1991)

-1.3% +0.6%

Panenergy Corp. BBB- to BB+ (1989)

-6.1% BB+ to BBB- (1990)

-1.0% +8.5%

Panenergy Corp. BBB- to BB+ (1992)

-14.0% BB+ to BBB- (1993)

-1.7% -5.7%

Pioneer Natural Res. BBB- to BB+ (1998)

-12.6% BB+ to BBB- (2003)

+9.9% +15.0%

Rymer Foods B+ to CCC+ (1988)

-18.8% CCC+ to B- (1990)

-10.3% -2.2%

Toro Corporation BBB- to BB+ (1993)

-9.7% BB+ to BBB- (1995)

-2.6% +6.9%

47

Figure 1. Firm Capital Structure Behavior Given Previous Year’s Change in Leverage and Credit Rating. Grouping firms by previous year’s leverage change, the figure shows capital structure decisions of downgraded firms (Panel A) and upgraded firms (Panel B) compared to firms with no rating change. Panel A. Downgraded firms versus firms with no rating change

0%

10%

20%

30%

40%

50%

60%

70%

<0% 0% to 1% 1% to 5% 5% to 10% >10%Change in Leverage Previous Year

% o

f firm

s w

ith le

vera

ge re

duci

ng c

apita

l mar

ket

activ

ity (N

etD

Iss<

0)

Downgraded previous year No change in rating

Panel B. Upgraded firms versus firms with no rating change

0%

10%

20%

30%

40%

50%

60%

70%

>0% 0% to -1% -1% to -5% -5% to -10% <-10%Change in Leverage Previous Year

% o

f firm

s w

ith le

vera

ge in

crea

sing

cap

ital m

arke

tac

tivity

(Net

DIs

s>0)

Upgraded previous year No change in rating