International Review of Financial Analysis 25 (2012) 2863

Contents lists available at SciVerse ScienceDirect

International Review of Financial Analysis

Credit market conditions and the impact of access to the public debt market on corporate leverage
Amrit Judge, Anna Korzhenitskaya
Economics and Statistics Department, Middlesex University Business School, The Burroughs, Hendon, London NW4 4BT, UK

a r t i c l e

i n f o

a b s t r a c t
This study examines the role played by credit ratings in explaining corporate capital structure choice during a period characterised by a major adverse loan supply shock. Recent literature has argued that supply-side factors are potentially as important as demand-side forces in determining corporate leverage. This is based on the premise that debt markets are segmented and that those rms that have access to the private debt markets do not necessarily have access to the public debt markets. The question of access to debt nance has become a major issue for public policy makers in several developed economies during the 20072009 nancial crisis. The UK economy has been subjected to a period of severe tightening of credit market conditions resulting in a signicant reduction in the availability of bank credit to the corporate sector. An important question is whether the contraction in the ow of bank credit to rms has affected rms equally or whether rms with access to alternative sources of debt nance have been able to mitigate the effect of adverse changes to the cost and availability of bank credit. To investigate this issue, this study employs data over a 20 year period that includes two recessions and three noticeable periods of credit market tightening. Despite the fact that a severe recession has accompanied the 20072009 nancial crisis we argue that the underlying forces driving the weakness in bank lending to the corporate sector are mainly supply side rather than demand side factors. In this study we use the possession of a credit rating as an indicator of access to the public debt markets. Our results provide support for the notion that having a rating is associated with higher leverage ratios, even after controlling for demand-side leverage determinants and macroeconomic conditions. More importantly, the study nds that the impact on leverage of having a credit rating varies over our sample period with the effect being greatest in those years when credit market conditions were tightest. The results are robust to the use of an alternative measure for public debt market access, different proxies for measuring the tightness of the credit markets, alternative econometric specications and various sub-periods within our overall sample period. 2012 Elsevier Inc. All rights reserved.

Available online 2 October 2012 JEL classication: G3 G32 Keywords: Capital structure Credit ratings Bond market access Financial crisis

1. Introduction The seminal work of Modigliani and Miller (1958) assumes that in the absence of market imperfections supply of capital is perfectly elastic and capital structure decision of a rm depends entirely on its demand-side considerations (Lemmon & Roberts, 2007). The key assumption is that rms can borrow as much as they wish at the same cost of capital and a rm's capital structure is purely a function
We are very grateful to Standard and Poor's and Fitch Ratings for providing us with credit rating data. Thanks also to Ian Byrne, Bridget Gandy, John Grout, John Redwood and Ian Stewart for useful comments and suggestions. We thank seminar participants at the University of Santiago de Compostella, University of Porto, ISCTE Business School, Standard and Poor's, Joint Seminar at the Department for Business, Innovation and Skills and HM Treasury, GdRE Symposium on Money, Banking and Finance at Reading University, Money Macro and Finance Research Group 43rd Annual International Conference at the University of Birmingham, European Conference on Banking and the Economy at the University of Southampton, and Middlesex University for helpful comments and suggestions. The usual disclaimer applies. Corresponding author. E-mail address: A.Korzhenitskaya@mdx.ac.uk (A. Korzhenitskaya). 1057-5219/$ see front matter 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.irfa.2012.09.003

of rm's characteristics, such as, size, protability, asset tangibility, and growth opportunities, that inuence its demand for debt. In the real world market frictions, such as information asymmetry, imply that the supply of capital is inelastic and rms can be rationed by their lenders in terms of both pricing and debt availability. Most of the previous empirical literature concentrates on the demand-side determinants of capital structure while paying little attention to the supply of capital (see Frank & Goyal, 2007; Rajan & Zingales, 1995). Recently researchers have recognised the importance of the supplyside factors as a potential driver of the capital structure decision. For example, Faulkender and Petersen (2006) argue that in the presence of information asymmetry rms that can access the public debt capital markets face less nancial constraints and are able to borrow more. Conversely, they suggest that rms that desire to raise funds but are constrained by lack of access to capital markets might be signicantly under-levered. The importance of supply side factors has come to the fore since the onset of the current nancial crisis in the latter half of 2007. During the last three years, banks have attempted to repair their balance sheets

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

29

12 10 8 6 4

% 2
0 -2 -4 -6 -8 Annual Real GDP Growth Source: Office for National Statistics
Fig. 1. Real GDP growth 19622010.

and consequently have signicantly cut back on their lending commitments to the corporate sector. 1 The Bank of England's (BOE) credit conditions surveys have reported that between the fourth quarter of 2007 through to the end of 2008 nancial market turbulence reduced signicantly UK banks' capacity to extend credit to the corporate sector. 2 The credit conditions surveys report that during this period there was a signicant tightening of price and non-price terms on loans to the corporate sector. Banks widened their spreads and raised the fees and commissions they charged on loans to rms. In addition banks imposed stricter covenants, raised collateral requirements and reduced maximum credit lines. This has made raising bank loan nance extremely difcult for creditworthy rms since the fourth quarter of 2007 and consequently has limited the availability of debt-based nance for rms that are heavily reliant on banks for their debt capital. Post the nancial crisis the future level of bank lending could be subject to greater restrictions as the new Basel capital requirements, which will more than double the core Tier 1 capital ratio from 2% to 4.5%, come into force. Some have argued that the Basel guidelines do not go far enough. For example, David Miles, an external member of the Bank of England monetary policy committee, has suggested a target capital ratio of between 15 and 20% (Mallaby, 2011). Kernan, Wade, and Watters (2010) argue that the forthcoming Basel III requirements will increase the amount of capital banks need to hold to support their corporate lending operations which will hike lending costs and lead to a reduction in lending capacity within the banking system. They anticipate that this will be likely to result in a longer term structural impetus for rising bond issuance over bank loans (Kernan et al., 2010, page 9). The drying up of the ow of bank credit could have serious consequences for the UK economy's ability to pull itself out of recession and therefore prolonging the economic downturn slowing down economic recovery. The problem is potentially more acute in the UK because banks have traditionally been the main source of capital for the private sector with 76% of debt being currently provided by banks (Kernan et al., 2010). The situation is however likely to change according to Kernan et al. (2010) who point out that since the events of September 2008, corporate bond issuance by U.K. businesses with a credit rating had increased by 22.1 billion, while U.K. nancial institutions have reduced their net lending (both in sterling and foreign currencies) to U.K. companies by 59.1 billion. Kernan et al. (2010) suggest that this makes corporate bond issuance the main provider of new debt
1 Balance sheet repairs may also take the form of injection of new equity capital and selling assets. 2 See Bank of England, 2007a,b and 2008a,b,c,d Credit Conditions Survey.

1962 Q1 1963 Q4 1965 Q3 1967 Q2 1969 Q1 1970 Q4 1972 Q3 1974 Q2 1976 Q1 1977 Q4 1979 Q3 1981 Q2 1983 Q1 1984 Q4 1986 Q3 1988 Q2 1990 Q1 1991 Q4 1993 Q3 1995 Q2 1997 Q1 1998 Q4 2000 Q3 2002 Q2 2004 Q1 2005 Q4 2007 Q3 2009 Q2 Quarterly Real GDP Growth

nancing on a net basis since the third-quarter of 2008. Bacon, Grout, and O'Donovan (2009) survey chief nancial ofcers and treasurers of UK rms and nd that the possession of a credit rating and the resulting access to public debt markets it offers has become especially important during the 20072009 nancial crisis. Recent trends in lending data from the Bank of England (2009d) points to rated rms raising capital market debt to pay back bank loans and issuing bonds rather accessing new bank loan facilities. The BOE suggest that access to the debt capital markets has enabled rated rms to mitigate the impact of a shortening in the maturity of bank lending available since the onset of the nancial crisis (Bank of England, 2009d). The BOE in its August 2009 Trends in Lending report suggested that while companies with bond market access had turned to arm's length sources of nance, smaller businesses without access still remained severely nancially constrained. Bacon et al. (2009) report that many rms that did not have a rating during the crisis were seeking to obtain one. A credit rating by providing access to the public debt markets can offer considerable benets to a rm. Not only does it widen the investor base and improves debt pricing but also provides an opportunity to enter foreign bond markets and gain international visibility, thereby reducing the reliance on local banks. Faulkender and Petersen (2006), Mitto and Zhang (2008), Kisgen (2009) nd that companies with a rating have access to broader sources of debt nance, and as a result have higher leverage ratios compared to unrated rms. There is also evidence that rated companies suffer less during adverse economic conditions. For example, Chava and Purnanandam (2009) nd that in the US, bank-dependent rms suffered larger valuation losses and greater subsequent decline in their capital expenditure during and after the Russian crisis of 1998 as compared to their rated counterparts. Similarly, Campello, Giambona, Graham, and Harvey (2009a) nd that the majority of US rms have been adversely affected by the 2008 credit supply shock but the impact has been greatest for nancially constrained rms. Adverse economic conditions and distortions in the supply of capital can severely affect rms' leverage and especially those rms that do not have access to alternative sources of nance, such as the public debt markets. The last two decades provide periods when the macroeconomic environment was stable and turbulent together with periods that experienced signicant movements in credit market conditions. This study investigates the role played by access to public debt markets over a twenty year period during which there were three episodes of credit market tightening with the most recent being the severest. Signicantly, the second of these periods (20002003) of tight credit was not associated with an economic downturn whereas the rst (19911992) and the third period (20072009) were. Although we

30

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

Percentage changes on a year earlier (%)

50 40 30 20 10 0 -10 -20 -30 Nominal Real Annual Real GDP Growth

01-Dec-80 0

01-Dec-69

01-Dec-91

01-Dec-02

01-Aug-73 3

01-Aug-84

01-Aug-95

Source: Bank of England and Office for National Statistics

Fig. 2. Sterling lending to UK PNFCs year-on-year growth 19642010.

demonstrate that the recent weakness in bank lending is largely supply driven this additional fact helps to further downplay any potential effects of credit demand on our results. During the 20072009 crisis we have witnessed the largest reduction of bank lending to the UK corporate sector in recent economic history. At the peak of the nancial crisis several major banks experienced nancial distress resulting in a severe lack of liquidity in the banking system forcing all banks to change considerably the terms of their lending to the corporate sector. Commitment fees and interest spreads went up, while debt maturities went down. The nancial crisis therefore provides a very unique opportunity to investigate the role of access to public debt markets in determining rms' leverage during a period of reduced bank loan supply. To the best of our knowledge, this study is the rst to investigate the impact of access to public bond markets on corporate leverage during a period characterised by a major tightening in credit market conditions in a UK context. The remainder of the paper is organised as follows. Section 2 presents an overview of the literature on credit market conditions and capital structure. Section 3 presents an analysis of the conditions of the UK credit markets between 1998 and 2010. In this section we also examine whether the weakness in bank lending during the nancial

crisis reects a reduction in the supply of credit or weaker demand for funds from rms as their investment opportunities have dried up during the recession. We present a robust analysis of the underlying forces driving the reduction in the ow of credit to the corporate sector. Section 4 describes the rating characteristics of our sample. Sections 5 and 6 present our empirical analysis and the results from robustness tests, respectively. Finally, Section 7 concludes. 2. Access to public debt markets, credit market conditions and capital structure: overview of the empirical literature Following Modigliani and Miller (1958) theorem three competing theories emerged that attempt to explain what determines capital structure choice (Cole, 2008): the trade-off theory, pecking order theory, and market timing theory. Under the trade-off theory of capital structure, a rm chooses its leverage ratio by balancing the costs and benets of using debt. The primary gains of debt are the tax-shields, which arise from the deductibility of interest on debt on the prot and loss account, whereas, the costs of debt are principally direct and indirect nancial distress costs (Frydenberg, 2011). Cole (2008) points out that the trade-off theory is often set up as a competitor theory to the pecking

20000 15000

Net Monthly Flows (millions)

10000 5000 0 01-Mar-99 01-Dec-00 01-Sep-02 01-Apr-03 01-Mar-06 01-Dec-07 01-Sep-09 01-Jan-98 01-Aug-98 01-Feb-02 01-Nov-03 01-Jun-04 01-Jan-05 01-Aug-05 01-May-07 01-Feb-09 01-Apr-10 01-Oct-99 01-Oct-06 01-May-00 -5000 -10000 -15000 -20000 Net Monthly Flows (millions) Source: Bank of England
Fig. 3. Net monthly ow of lending to UK rms.

01-Aug-06

01-Jun-08 8

01-Jun-64

01-Feb-68

01-Jun-75

01-Feb-79

01-Jun-86

01-Apr-88

01-Feb-90

01-Jun-97

01-Feb-01

01-Nov-10

01-Jul-01

01-Jul-08

0 01-Apr-10

01-Apr-66

01-Apr-77

01-Oct-82

01-Apr-99

01-Oct-71

01-Oct-93

01-Oct-04

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

31

40 30
3 month growth rate

Growth rate %

20 10 0 -10 -20
12 month growth rate 3 month growth rate
12 month growth rate

01-Jan-99

01-Jul-99

01-Jul-00

01-Jan-01

01-Jul-02

01-Jul-03

01-Jul-04

01-Jul-05

01-Jan-06

01-Jul-06

01-Jan-07

01-Jul-07

01-Jul-08

01-Jul-09

01-Jan-10

01-Jul-10

01-Jan-00

01-Jan-02

01-Jan-03

01-Jan-04

01-Jan-05

01-Jan-08

Source: Bank of England

Fig. 4. Growth rate of monetary nancial institutions' loans to private non-nancial corporations.

order theory of capital structure (e.g. Frydenberg, 2011; Myers, 1984) Pecking order theory was introduced by Myers (1984) and Myers and Majluf (1984), which states that there is a nancing hierarchy of retained earnings, debt and then equity. Myers (1984) and Myers and Majluf (1984) argue that due to the information asymmetry between managers and investors, rms will use their retained earning whenever is possible, then issue bonds for external capital, and raise equity only as a last resort. Another theory that has gained prominence in recent capital structure literature is the market-timing theory. This theory, proposed by Baker and Wurgler (2002), argues that capital structure is the cumulative outcome of past attempts to time the market.3 The theory suggests that managers issue equity when they believe it is overvalued (as measured by market-to-book ratio) and repurchase equity or issue debt when they believe it is undervalued, i.e. managers time the market and make their nancing decisions according to favourable conditions in the debt or equity markets (Baker & Wurgler, 2002). Each of these theories identies key factors in determining capital structure choice, such as rm size, tangibility, market-to-book and profitability, but according to Frank and Goyal (2007) neither of them can fully explain capital structure choice (Frank & Goyal, 2007). Frank and Goyal (2007) argue that currently there is no unied model of leverage available that can simultaneously account for all the stylised facts. They claim that different theories apply to rms under different circumstances. Frydenberg (2011) in his overview of capital structure theories declares that capital structure is a too complex fabric to t into a single model.4 Factors suggested by the capital structure theories and widely examined in the previous literature largely represent demand-side determinants of leverage. In the world with information asymmetry and loan supply frictions, rms can nd it difcult to raise the desired amount of debt and can be signicantly under-levered (Faulkender & Petersen, 2006). This issue becomes especially relevant during the periods when credit market conditions are tight and the supply of capital becomes an important determinant of capital structure choice. Recently there has been a move in the empirical literature towards examining the link between credit market conditions and rms' capital structure decisions. Lemmon and Roberts (2007) explore the relationship between the loan supply shock of 1989 and rms' nancing decisions. Their ndings underline that even large rms with access to public debt market are affected by capital supply shocks. Chava and Purnanandam (2009) examine the shock to the US banking system during the Russian crisis of 1998 using full sample analysis and matching sample techniques. They nd that bank-dependent rms lost
3 4

Baker and Wurgler (2002), p.23. Frydenberg (2011), p.25.

disproportionally higher market value and suffered larger declines in capital investments and growth rates following the crisis as compared to rms with access to the public debt market (Chava & Purnanandam, 2009, p.30). Leary (2009) in a study of the relevance of capital market supply frictions for corporate capital structure decisions following the 1966 credit crunch in the United States nds that larger rms with access to public debt market were less affected by contraction in bank loan supply due to their greater ability to substitute toward arm's length debt nancing. Leary (2009) nds that the use of bond debt by rms with access to public debt markets increased, relative to that of small, bank-dependent rms. As a result the leverage of bank-dependent rms signicantly declined compared to rms with access to public bond markets. By using rm size as a proxy for debt market access, he nds that following the 1966 loan supply contraction, leverage ratios of small, bank-dependent rms signicantly decreased relative to large rms with bond market access. When Leary (2009) expands his sample period to cover the 35 years from 1965 to 2000, he nds the leverage difference between rms with and without public debt market access becomes greater in periods of reduced loan supply and tighter credit markets. Voutsinas and Werner (2011) examine how nancial constraints and uctuations in the supply of credit affect rm capital structure. They investigate the impact of asset bubble in the 1980s and the credit crunch of the late 1990s on corporate capital structure decisions of publicly listed Japanese rms. They nd that both eventsthe asset bubble burst of the 1980s and the credit crunch in the 1990swere followed by severe reductions in leverage (total leverage was reduced by 0.0239 when the bubble burst). Voutsinas and Werner (2011) conclude that uctuations in the supply of credit and changes in monetary conditions have a serious impact on rms' capital structures. During the Japanese credit crunch all rms experienced a severe reduction in their leverage levels, but especially smaller sized bank-dependent ones. The question of access has become a major issue since the latter half of 2007, when the banking systems around the world experienced major liquidity problems, resulting in a severe tightening of credit market conditions leading to signicant falls in lending to the corporate sector. Ivashina and Scharfstein (2009) indicate that in the US bank lending dropped considerably across all loans types during the 20072009 crisis. They nd that new bank loans fell by 47% during the peak of the nancial crisis (fourth quarter of 2008) relative to the third quarter. When compared to the peak of the credit boom (second quarter of 2007) they nd that bank loans dropped by 79% in the fourth quarter of 2008. The terms and conditions of bank lending have also worsened. Campello, Giambona, Graham, and Harvey (2009a) report that the tightening of US credit markets during the 20072009 nancial

01-Jan-09

01-Jan-11

01-Jul-01

32

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

25.0 20.0 15.0 10.0

%
5.0 0.0 Jan 2003 Jan 2007 Jan 2004 Jan 2005 Jan 2006 Jan 2008 Apr 2003 Apr 2004 Apr 2005 Apr 2006 Apr 2007 Jul 2003 Oct 2003 Jul 2004 Oct 2004 Jul 2005 Oct 2005 Jul 2006 Jul 2007 Oct 2007 Apr 2008 -5.0 -10.0 Major UK lenders Source: Bank of England
Fig. 5. Contributions to growth in lending to UK businesses (12 month growth rate (%) in the stock of lending).

Foreign lenders

Other monetary financial institutions

crisis has manifested itself in the form of an increase of commitment fees by 14 basis points, mark-ups over LIBOR/Prime rate by 69 basis points and decline in maturity by 2.6 months from 30 months on average. Empirical evidence suggests that debt market segmentation has resulted in differential sensitivity to the recent credit market shock. For example, Campello, Graham, and Harvey (2009b) provide evidence on how nancially constrained and unconstrained US rms manage their investment expenditure during the 2007 nancial crisis. In particular they nd that the nancial crisis has had a severe impact on credit constrained rms, leading to deeper cuts in planned R&D (by 22%), employment (by 11%), and capital spending (by 9%). Furthermore, the inability of these rms to borrow externally has caused many rms to cancel or postpone attractive investment projects, with 86% of CFOs in the US stating that they had to restrict investments in attractive projects during the credit crisis. Similarly, Kisgen (2007) suggests that having access to alternative sources of debt capital can help rms raise funds during adverse economic conditions and prevent underinvestment in positive-NPV projects. All rms have suffered from the credit supply shock but the impact has been greatest on those rms heavily reliant on the banking sector for their funding. The UK's corporate sector has also been adversely affected by the reduction in the bank lending during the 20072009 crisis. Bacon et al. (2009) in their survey of chief nancial ofcers and corporate treasurers on the impact of changing banking and credit market conditions on corporate funding plans during the 20072009 nancial crisis, found that with increased borrowing margins and reduced maturity periods, the availability of funds from the banking sector fell signicantly. In addition, they point out that if banking market capacity is reduced in the foreseeable future, bond markets are likely to become an alternative source of capital. Several studies have looked at the effect of macroeconomic conditions on rms' capital structure decisions. Cantillo and Wright (2000) suggest that macroeconomic conditions have a powerful effect on how rms choose their lenders. They nd that less constrained companies tend to issue more debt during favorable economic conditions (Cantillo & Wright, 2000; Korajczyk & Levy, 2003; Levy, 2000). Levy (2000) investigates how rms' capital structure choice varies with macroeconomic conditions in the presence of agency problems. He nds counter-cyclical patterns for debt issues for rms that access public capital markets. This nding is supported in the later work of Korajczyk and Levy (2003) who point out that capital structure choice varies over time and across rms. By splitting their sample into nancial constrained and nancial unconstrained rms, they also nd that leverage of nancially unconstrained rms varies counter-cyclically

with macroeconomic conditions. In other words, unconstrained rms time their issue choice to coincide with periods of favorable macroeconomic conditions, while constrained rms do not. In a similar manner, Bougheas, Mizen, and Yalcin (2006) evaluate the inuence of rmspecic characteristics on the response of corporate nance to monetary policy. They examine UK manufacturing rms over two periods: the rst, 19901992, a period of tight monetary policy that coincided with a recession and a harsh environment for existing and new corporate borrowers. The second, 19931999, was a period of loose monetary policya time of sustained economic growth, falling unemployment and ination, relatively low interest rates, and less constrained borrowing conditions. Their results show that there was a marked difference in the response to rm specic characteristics when interacted with monetary policy. In particular, they found that small, young and risky rms were more noticeably affected by monetary tightening than large, old and secure rms.

3. Credit market conditions in the U.K. 19982010 In this section we present an analysis of the conditions of the UK credit markets before, during and after the 20072009 nancial crisis. We are interested in establishing whether the period under study provides a good setting for identifying the effect of supply frictions on the capital structure of UK rms. Leary (2009), in the context of the 1966 US credit crunch, suggests that there are three elements to this. Firstly, did the 20072009 nancial crisis and any earlier credit crisis represent a change in credit supply? Secondly, if these periods were subject to credit supply shocks were they bank-specic or shocks to total capital supply? And, thirdly, could the effects on capital structure be driven by simultaneous changes in credit demand? To nd answers to these questions we investigate three aspects of the UK credit markets: the ow and stock of bank lending to the corporate sector, capital market issuance and the pricing of bank loans. We will utilise evidence from these areas to examine whether the observed weakening in bank lending to the corporate sector is an indicator of a reduction in the supply of credit due to a tightening in bank's credit provision or weaker demand for nance from rms as a result of the recession. Fig. 1 shows that between 2008 and 2009 the UK has been mired in the deepest as well as longest postwar recession since the 1930s with seven quarters of negative growth. 5 The recession began in the second quarter of 2008 and by the end of 2009 the annual rate of
5

A recession is dened as two or more consecutive quarters of falling real GDP.

Jan 2009

Oct 2006

Jul 2008

Oct 2008

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

33

70.0 60.0 50.0 40.0

billions

30.0 20.0 10.0 0.0 -10.0 2003 2003 2004 2004 2005 2005 2006 2006 2007 2007 2008 2008 2009 2009 2010 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 H2 H1 -20.0 -30.0 -40.0 Loans Source: Bank of England
Fig. 6. Net capital market and net bank loan funds raised by UK rms 20032010.

Bonds & CP

Equity

decline in GDP reached nearly 5.6%the biggest fall since records began in 1955. 6

3.1. Bank lending to the UK corporate sector Fig. 2 provides a historical perspective on the annual growth in bank sterling lending to UK rms over the period 19642010. The shaded areas indicate periods when the UK economy was in recession (1974 to 1975, 1980 to 1981, 1990 to 1991 and 2008 to 2009). The gure shows that recessions are always associated with decreases in the growth of lending and that growth has become negative in real terms during the last four recessions. The 20082009 recession has seen the deepest real terms contraction in bank lending. For example, in July 2009 sterling-only net lending to private non-nancial corporations was 8.4 billion which was the weakest ow since the series began in 1963.7 However, it is worth pointing out that a slowdown in bank lending does not always take place in the midst of a recession. Fig. 2 shows that there have been three occasions in the last fteen years where lending has weakened during periods of positive economic growth, these being around 1994, 19971999 and 20022003. We believe that this fact makes it less likely that any credit rating or access effects on capital structure we observe will be due to simultaneous changes in credit demand. Fig. 3 presents Bank of England (BOE) data on net monthly bank lending ows to the UK corporate sector for the period 1998 to 2011 and includes lending in both sterling and foreign currency (expressed in sterling millions).8 The chart shows that from 1999 through to the end of 2003 net monthly lending ows made several forays into negative territory indicating a net repayment of bank loans in those months. Net monthly lending ows were negative in 17 out of 60 months (28% of months) during this period with an average ow of 1559 million. In stark contrast, during the four year period from 2004 to 2007 lending ows were negative in only 2 months out of 48 (4% of months) and the average net monthly lending ow had nearly tripled to 4634 million. This was then followed by the nancial crisis which resulted in the biggest reduction in bank lending to corporates since the BOE started collecting this data in 1998. For the three years between January 2008 and January 2011 the net monthly bank lending ow was negative in 25 out of 37 months (68% of months) with an average net monthly lending ow

equal to 698 million.9 Over these 37 months the stock of bank lending to non-nancial rms shrank by 25.8 billion. Fig. 4 shows the twelve-month and three-month growth rates of the stock of lending over the period January 1999 to January 2011. 10 There was a dip in both growth rates towards the end of 1999, during the rst half of 2002 and then again over the second half of 2003 which is consistent with the tightening of credit markets experienced during these years. From the middle of 2004 to the middle of 2005 we see a rapid increase in both growth rates. For a two year period the twelve month growth rate of the stock of lending averages around 15% and then it picks up again in the middle of 2007. Twelve month lending growth peaks at 24% in April 2008 after which it commenced a rapid decline hitting negative growth in May 2009 for the rst time since the monthly series began in 1999. 11 The annual rate of contraction (negative growth rate) of the stock of loans peaked in the rst quarter of 2010 and since then the rate of contraction has slowed. The strong ow and growing stock of lending to the corporate sector shown in Figs. 3 and 4 during the pre-crisis years of 2004 to 2007 is consistent with the notion that during this period the banks were very proactive in encouraging rms to take on higher levels of debt and most borrowers could not resist the cheap nancing facilities available. The BOE points out that during this period the macroeconomic environment was very favorable, asset prices were rising and interest rates around the world were relatively low which facilitated an increase in the amount of lending to companies in the UK and the rest of the world. Furthermore, the BOE suggests that before the credit crisis borrowing margins were on the whole at historically low levels and at the peak of the boom in the latter part of 2006 banks were competing aggressively to provide credit on favorable terms (Bank of England, 2009a). Fig. 5 presents a breakdown of the growth of UK lending by geographical source. During the pre-crisis years 2005 to 2007 there was an increasing contribution by foreign lenders to the growth of UK lending with about half of the growth in lending to private non-nancial rms in 2007 being attributed to the activities of foreign lenders. However, with the onset of the nancial crisis the balance sheets of banks globally came under severe pressure during the latter part of 2007 and consequently the contribution of those foreign lenders began to fall. This decline gathered momentum in the second half of 2008, as foreign banks cut back on new lending abroad. Fig. 5 shows that the growth in lending

Sourced from the Ofce For National Statistics website. 7 See www.bankofengland.co.uk/statistics/fm4/2009/jul/FM4.pdf or see Bank of England (2009f) Trends in lending, September, page 4. 8 Monthly changes of monetary nancial institutions' sterling and all foreign currency loans (excluding securitisations) to private non-nancial corporations (in sterling millions) seasonally adjusted.

9 Bank of England data covering lending by all UK-resident banks and building societies showed that there was a net repayment of loans in 21 of the months over the period January 2009 to January 2011. 10 Monthly 12 and 3 month growth rate of monetary nancial institutions' sterling and all foreign currency loans (excluding securitisations) to private non-nancial corporations (in percent) seasonally adjusted 11 The three month growth rate peaked at 36.5% in October 2007.

34

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

350 300

Sep 09 (327bp)

Basis points

250
Sep 08 (190bp)

200 150 100 50

Sep 07 (42bp)

Aug 2008

Jul 2006

Jun 2004

Jun 2009

Apr 2005

Aug 2003

Jan 2004

Jan 2009

Nov 2004

Nov 2009

Mar 2003

Oct 2007

Mar 2008

Apr 2010

Jul 2011

Dec 2006

May 2007

Sep 2005

Sep 2010

Source: Bank of England

Fig. 7. Average estimated spreads on investment-grade syndicated loans.

Net % indicating tight credit conditions

100 80 60 40 20 0 -20 -40 US EUROPE UK 1990Q2 1991Q1 1991Q4 1992Q3 1993Q2 1994Q1 1994Q4 1995Q3 1996Q2 1997Q1 1997Q4 1998Q3 1999Q2 2000Q1 2000Q4 2001Q3 2002Q2 2003Q1 2003Q4 2004Q3 2005Q2 2006Q1 2006Q4 2007Q3 2008Q2 2009Q1 2009Q4 2010Q3

Source: US Federal Reserve Board, European Central Bank and Bank of England
Fig. 8. Credit market conditions in the US, Europe and the UK (Positive net% implies tight credit conditions for corporates).

by UK lenders has also decreased, but their relative contribution to overall growth is now greater than in 2007.

3.2. Capital market issuance by UK rms 20032010 In this section we examine whether the credit supply shocks described above affected the ow of credit from both the bank and capital markets or were isolated to the banking sector. This is important for this study for two reasons. Firstly, if the nancial crisis resulted in a reduction in total capital supply then we would expect to observe a zero impact of access on leverage during these periods. Secondly, if rms reduce their borrowing from banks during the nancial crisis but are able to switch to another source of nance, such as capital market debt, then we would expect to observe a positive access effect. Furthermore, in this case the resulting weakness in bank lending is more likely to reect tighter credit supply than weaker demand. Rated rms in the UK raise debt funds from public debt capital markets as well as by borrowing from banks. This allows them to diversify their sources of debt nance though in the UK bank lending tends to be the dominant source of debt funds for private non-nancial rms (Kernan et al. (2010)). The Bank of England (2009c) reported that the average maturity of new lending during the crisis fell relative to before the crisis. The BOE noted that before the crisis the term premium

associated with borrowing over longer periods had been relatively small, and so loan facilities were mainly arranged with maturities of ve to seven years. However during the crisis, banks found it very difcult and costly to raise longer-term funding and this was reected in the prices they charged for longer-term facilities. The BOE suggest that a reluctance by companies to lock in those higher costs over a long period, given uncertainty about future demand contributed to a decline in the maturity of new bank lending to two to three years. The BOE suggested that large investment-grade companies requiring longer-term nancing were able to borrow in the capital markets given improved public debt market conditions in early 2009. Bond issuance by investmentgrade companies was relatively strong in the early months of 2009, allowing these companies to mitigate the impact of a shortening in the maturity of bank lending available (see Bank of England, 2008c). There were indications that during the nancial crisis large rms were issuing capital market debt and using the proceeds to repay bank debt. Fig. 6 shows that during 2009 large non-nancial rms were accessing capital markets reected by higher public debt and equity issuance, with some using equity proceeds to repay bank debt in order to reduce leverage.12 It is clear to see from Fig. 6 that there was greater equity issuance during the crisis years as rms were seeking
12 The BOE suggest that rms were also repaying bank debt from other forms of funding such as internally generated funds.

Dec 2011

Feb 2006

Feb 2011

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

35

80 70 25%25% 27% 28%29%

35% 27%27% 30% 25%

Number of Firms

60 50 40 30 20 4% 10 0 8% 8% 9% 7% 5% 6% 10%11%11% 20% 18%

23%

Percentage

20% 15% 10% 5% 0%

10 14 17 20 24 24 28 35 39 41 63 67 65 67 68 69 67 63 55 51

Fig. 9. Number and proportion of rms with S&P rating between 1989 and 2008.

60 50

23% 20% 17% 15% 20%

25% 20%

Number of Firms

40 30 20 10 42 0 1998 1999 2000 2001 2002 47 44 36 35 12% 13% 12% 12% 12%

15%

Percentage

15% 10% 5% 43 2003 43 2004 45 2005 53 2006 43 2007 39 0% 2008

Fig. 10. Number and proportion of rms with Fitch rating between 1998 and 2008.

to reduce their leverage given the severe economic conditions. Fig. 6 also shows that non-nancial rms' net equity and bond issuance was considerably higher in 2009 than its average over the 20032008 period. Furthermore, the BOE reports that between January and October of 2010 gross bond issuance in the UK was greater than its annual average over the 20002007 period which was partly the result of strong issuance by non-investment grade rms (Bank of England, 2010b, Quarter 4, page 252). Survey evidence also points to a preference for capital market issuance during this time. For example, Deloitte Chief Financial Ofcer (CFO) Survey, 2009a, reported that sentiment among chief nancial ofcers (CFOs) about equity and corporate bond issuance rose in June 2009, to its highest level since the survey started in 2007 and for the rst time there was a preference for bond and equity issuance over bank borrowing. The Deloitte Chief Financial Ofcer (CFO) Survey, 2009b reported that a net balance of CFOs perceived UK companies to be overleveraged and expected a long-term shift in the corporate funding mix towards capital market issuance and away from bank borrowing. The proportion of UK rms issuing bonds for the rst time has been increasing since the second half of 2008.13 Bank of England, 2010b, suggests that the majority of the new issuers in the UK have used the proceeds to repay maturing bank loans which they argue is consistent with ongoing disintermediation of banks by UK rms. Bank of England (2010a) points out that there is evidence that since 2010 a wide range of companies have been accessing the public debt markets instead of borrowing from the banks. The report indicates that around 40% of those rms that have issued corporate bonds in 2010 did so for the

rst time and suggests the fall in the cost of bond nance (corporate bonds yields have fallen) might be the factor behind such issuance. The Bank of England (2009h) provides a number of explanations for the strength of capital market issuance, relative to bank lending during the nancial crisis years. Firstly, it is suggested that increased equity issuance may have reected a desire by some rms to reduce their leverage in light of the weaker economic environment and the belief that corporate leverage had risen too far. Secondly, the reduced availability of bank lending, and its increased cost relative to reference rates such as three-month Libor, particularly for loans of longer maturities, might have encouraged companies to raise funds from capital markets. Thirdly, it is suggested that bank debt is considered to have a greater adverse affect on rms' credit ratings than capital market debt, due to a belief of greater renancing risk associated with shorter debt maturities. The BOE suggest that this may have led some rms to prefer bond issuance over bank borrowing. In addition to BOE data and corporate surveys there has been a large volume of anecdotal evidence pointing to increasing bond issuance by non-nancial rms during the nancial crisis. For example, In the US and even more so in Europe, it is the small to medium-sized companies that will most drastically be affected by a signicant contraction in bank lending because they do not have access to bond markets. (Davies, 2008, FT.com) Companies have been turning to the bond markets to renance debt as bank lending has become constrained. (Sakoui & Lee, 2009, FT.com) With banking market capacity much reduced for the foreseeable future the capital markets are seen as a replacement funding source even for those that have traditionally not made use of bonds. Some unrated

13

Bank of England (2010b), Quarter 4, page 252.

36

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

160 140 13%

15% 13% 11%

14%

16% 14% 12%

Number of Firms

120 100 8% 80 60 4% 40 20 34 0 27 27 3% 3%

Percentage

10% 8% 5% 3% 2% 15 1% 1% 1% 0% 0% 25 7 10 6 1 2 6% 2% 70 119 134 118 97 128 47 20 4% 2% 0%

Fig. 11. Frequency and proportion of rating grades assigned by S&P amongst UK non-nancial rms (19892008).

corporates are expecting to seek their rst rating in order to gain access to new sources of funding. (Bacon et al., 2009, page 3). New investment grade corporate bond issuance in Euros for the rst 4 weeks of January 2009 reached the same level as the rst 15 weeks last year (Bacon et al., 2009, page 5) Businesses, frozen out by the world's biggest banks, have ocked to the corporate bond market to raise new funds, triggering a 160 per cent surge in debt issuance since the beginning of the year. About $331 billion (229 billion) has been raised through corporate bond issues since January in Europe, America and Britain, compared with $127 billion for the same period last year. Financiers say that the rise has happened because banks effectively stopped lending competitively to business after the collapse of Lehman Brothers, the Wall Street investment bank, last September. (Jagger & Power, 2009) since the events of September 2008, U.K. nancial institutions have reduced their net lending (both in sterling and foreign currencies) to U.K. companies by 59.1 billion, while corporate bond issuance by U.K. businesses had increased by 22.1 billion, according to Bank of England gures. (Kernan et al., 2010, page 7) According to Dealogic, European bond issuance reached 557.2 billion in 2009 compared with 338.4 billion of loans, the rst time bond market volumes have exceeded loans. (Churchill, 2010, http://www.risk. net/credit/) British companies, including Cadbury, Vodafone, Tesco and National Grid, have issued a swath of bonds in recent months, dipping into a pool of pension fund money while the banks were closed for lending. (Mortished, 2009, page 41) Bond markets virtually closed following Lehman Brothers' collapse last September but co-ordinated state bail-outs have since reassured debt investors, with the number of global issues recovering from 133 in October to 315 last month. The UK also enjoyed its best January in at least three years, with pounds 15.7 bn placed across 15 issues. Bond issuance is vital if companies are to fund future investment and is particularly critical now that the banks are making less credit available. (Aldrick & Ebrahimi, 2009, page 3) The Financial Times (31st August 2010) reports that The effect of the nancial crisis on bank lending is prompting companies to develop other funding channels. In 2009 companies around the world with investment-grade ratings raised record volumes of debt in capital markets, much of it to renance bank loans. In the US, investment-

grade bond issuance by non-nancial companies totalled $512bn (402bn, 330 bn). In Europewhich has a much smaller bond investor baseit hit 218 bn and in the UK, 47 bn was raised, according to data from Citigroup. Peter Goves, credit strategist at Citi, says subinvestment grade bond issuance in the rst half of this year was also at a record, with companies seeking to renance the bank loans they used for leveraged buy-outs. (Cohen & Goff, 2010, page 7). Our analysis would seem to suggest that the UK provides a good setting for identifying the effect of supply frictions on capital structure. The preceding analysis has demonstrated that large UK corporates have been reducing their borrowing from banks and switching to alternative sources of nance such as capital market debt. It follows that the resulting weakness in bank lending is more likely to reect tighter credit supply than weaker demand. The evidence we have presented also indicates that the credit supply shocks that have taken place during our sample period were specic to bank credit and not shocks to total debt capital supply.

3.3. Corporate loan pricing and interest rate spreads In the sections above we have shown that there has been a signicant weakening in bank lending to the UK corporate sector during the 20072009 nancial crisis. We have argued that an important question is whether this weakening in lending is due to a reduction in the supply of credit, as banks have restricted the ow of lending to rms, or subdued demand for funding from rms as growth prospects have declined during the recession. The evidence we have presented so far points to a decrease in the supply of bank credit and hence a credit market tightening. However, this analysis is incomplete without an examination of the behaviour of interest rate spreads during this period. If the lending slowdown is largely driven by a fall in demand then, all else being equal, we would expect spreads charged on lending to decrease. However, if a tightening of supply is the dominant factor, spreads would be expected to increase (Bank of England, 2009a). Therefore, in order to more fully disentangle the demand and supply effects on bank lending we need to look at the behaviour of interest rate spreads before and during the nancial crisis. The cost of bank loan nance to a rm can be broken down into the fees charged by a bank to provide loan facilities, the spread over a given reference rate14 at which loans are provided, and the prevailing level of that reference rate in the markets (Bank of England, 2009g). The BOE

14 Usually three-month Libor or the Bank Rate. The Bank Rate is the ofcial rate paid on commercial bank reserves by the Bank of England.

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

37

120 22% 100

25%

20% 17% 17%

Number of Firms

80

Percentage

14%

15%

60 9% 40 5% 20 1% 6 0 AAA AA+ AA AAA+ A ABBB+ BBB BBB- BB+ BB BB22 2% 10 19 41 79 66 80 104 4% 3% 12 4% 17 1% 6 2% 8 0% 5% 10%

Fig. 12. Frequency and proportion of rating grades assigned By Fitch amongst UK non-nancial rms (19982008).

reported that the extra cost associated with borrowing over longer periods had been relatively small in the years leading up to the crisis (Bank of England, 2009c). Furthermore, in communications with the BOE, UK banks suggested that spreads and fees had fallen to unsustainably low levels by early 2007 (Bank of England, 2009a). The BOE Credit Conditions surveys in 2008 and early 2009 indicated that since onset of the crisis spreads over reference rates increased substantially across all types of lending. Fig. 7 shows that the period from September 2007 to September 2008 witnessed a signicant increase in the cost of syndicated bank credit as loan spreads widened. Spreads over reference rates on new-investment-grade syndicated lending jumped by 148 basis points (bp) during this 12 month period and peaked at 327 basis points a year later in September 2009. During this two year period syndicated loan spreads went up by nearly eight times (42 bp to 327 bp). The Credit Condition surveys also found that the net percentage balances of lenders were reporting increased fees on secured lending and that fees or commissions on loans to rms also went up. As the weakness in bank lending is associated with higher spreads this would suggest that a tightening in credit supply is most probably the key driver for weak bank lending during the nancial crisis. In order to shed further light on this issue we can examine the reasons put forward to explain why loan spreads increased during the nancial crisis. Firstly, UK banks reported to the BOE that spreads over reference rates had increased to better reect the higher costs of longer-term funding. From the start of the crisis banks found it extremely difcult and costly to raise longer-term funding and consequently this was reected in the prices they charged for long-term loans. At this time UK banks were under pressure to lengthen the term structure of wholesale funding to better match that of assets, as a result using Libor as a reference rate was no longer appropriate since longer-term funding rates were a better reection of the banks marginal cost of funding. Banks indicated that the spreads charged on longer-term facilities had been increased to better account for the risks of funding over the lifetime of the loan (Bank of England, 2009b). Secondly, the BOE reported that increases in spreads might also have reected in part a re-pricing of risk due to increased perceptions of credit risk, following a lengthy period earlier in the decade when corporate credit risks were underpriced. Banks suggested that deteriorating credit quality of borrowers had led to higher capital requirements and costs, which had acted to fuel the hike in spreads (Bank of England, 2009c). Thirdly, higher capital requirements under the new Basel II capital adequacy framework explained in part the increased charges for unused loan facilities (Bank of England, 2009a). As these reasons are strongly linked with credit supply factors this evidence also points to an independent tightening in the supply of credit during the nancial crisis.

3.4. Measuring credit market conditions: loan ofcer surveys The US Federal Reserve Board, European Central Bank (ECB) and the Bank of England (BOE) conduct a quarterly survey of commercial banks under their jurisdiction to measure the extent to which banks are willing to provide loans to the corporate sector. Of the three surveys the US Federal Reserve's Senior Loan Ofcer Opinion Survey is the oldest established survey running since 1967, although it was suspended from the rst quarter of 1984 through to the second quarter of 1990. 15 The ECB survey started in 2003 and is conducted in each member country by the respective national central bank, and the results are then collated and analysed at the aggregate level. The ECB credit conditions survey is sent to senior loan ofcers of a sample of euro area banks. The banks participating in the survey comprises around 90 banks from all euro area countries and takes into account the characteristics of their respective national banking systems. The questionnaire examines issues relating to both loan demand and loan supply. The loan supply questions address issues relating to credit standards and credit conditions and terms, as well as to the various factors that may be behind any loan supply changes.16 The BOE ran their survey for the rst time in the second quarter of 2007.17 The BOE survey asks questions about both how bank lending trends have changed over the past three months (relative to the previous three months), and how they are expected to change over the next three months (relative to the latest three months). The survey also asks about changes in the amount of credit lenders are willing to supply and about how both price and non-price terms are changing such as collateral requirements and loan covenants. The latter gives an indication of whether the terms and conditions on which banks are willing to lend have improved or worsened. To calculate aggregate survey results, the BOE assigns to each lender a score based on their response. Lenders who report that credit conditions have changed a lot are assigned twice the score of those who report that conditions have changed a little. These scores are then weighted by lenders' market shares. The results are analysed by calculating net percentage balancesthe difference between the weighted balance of lenders reporting that, for example, terms and conditions were looser/tighter. The net percentage balances are scaled to lie between 100. Negative balances indicate

15 The US Federal Reserve Loan Ofcer data is obtained from http://www.federalreserve. gov/boarddocs/snloansurvey/. 16 The ECB Lending Survey data is obtained from http://www.ecb.int/stats/money/ surveys/lend/html/index.en.html. 17 The BOE Credit Conditions Survey data is obtained from http://www.bankofengland. co.uk/publications/Pages/other/monetary/creditconditions.aspx.

38

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

that lenders, on balance, reported/expected credit availability to be lower than over the previous/current three-month period, or that the terms and conditions on which credit was provided became expensive or tighter respectively. This is opposite to the Federal Reserve and ECB surveys where a positive net percentage balance indicates a credit market tightening. Therefore, in this study we multiply the BOE net percentage balances by minus 1 to make them consistent with those of the Federal Reserve and ECB. Thus, for all three surveys a positive (negative) net percentage balance indicates an overall tightening (loosening) in the supply of credit. In this study we use the net percentage balance of respondents to the question How has the availability of credit provided to the corporate sector overall changed? to get an overall assessment of the conditions of the UK credit markets. We use the results to the corresponding question in the Federal Reserve and ECB surveys to measure the conditions of the credit markets in the US and Europe, respectively. Fig. 8 presents net percentage balance data for the US, European and UK credit condition surveys for the period from 1990 to 2010. Fig. 8 indicates that since the middle of 2005 there appears to be some degree of synchronicity in the state of the credit markets in the US, Europe and the UK. Simple correlation analysis conrms that credit market conditions are highly correlated with a correlation coefcient around +0.8, although credit markets in Europe and the UK are slightly more closely aligned with each other than with the US. Fig. 8 also suggests that the severity of the tightening in credit conditions during the nancial crisis was greatest in the US.

All nancial data is sourced from DataStream. Credit rating data is sourced directly from Standard and Poor's (S&P) and Fitch. For credit rating we use a company's long-term credit rating. Credit rating data from S&P covers the 20 years from 1989 to 2008 and Fitch credit ratings are available for the 11 years from 1998 to 2008. In our sample 1010 or 15.4% of rm-year observations possess a rating (either S&P or Fitch). This percentage is similar to Mitto and Zhang's (2010) 15% for a sample of Canadian rms and to Faulkender and Petersen's (2006) 19% for US rms. Fig. 9 illustrates the frequency of rms with a S&P credit rating for the period from 1989 to 2008 and the proportion of rms in our sample that possess a S&P rating. Fig. 9 shows that only 4% (10 rms) of rms in our sample were rated by S&P in 1989. The percentage of S&P rated rms peaked at 29% in 2006 and stood at 27% in 2008. Fig. 10 shows the frequency and percentage of rms with a Fitch credit rating for the period from 1998 to 2008. Twelve percent of our sample possessed a Fitch rating in 1998 and this had reached 20% in 2008.

4.1. Rating categories The ratings assigned to the rms in our sample range from the highest AAA to the lowest CC, indicating each rm's individual credit quality (see Appendix, Table A). All ratings above and including BBBfall into the category of investment grade ratings and ratings below and including BB+ are considered to be non-investment or speculative grade ratings. In our sample of 1010 rm-years with a rating, 909 possess an investment grade rating and 101 have a non-investment grade rating. Figs. 11 and 12 illustrate frequencies of the rating categories assigned by S&P and Fitch credit rating agencies to UK non-nancial rms. Fig. 11 presents frequencies of the rating grades assigned by S&P to UK rms. Fig. 11 shows that the highest concentration of S&P rated rm-years is observed within the A+ to BBB rating interval with 596 out of 887 rm-years observations (66% of S&P rated rm-year observations). The rating category with the highest frequency is A with 134 rm-years observations (15% of the S&P rated rm-years). Fig. 12 presents frequencies of the rating grades assigned by Fitch to UK nonnancial rms. Similar to the S&P ratings data Fig. 12 indicates that most of the Fitch rated rm-year observations are concentrated within the A to BBB credit rating interval with 70% of the rated rm-years possessing a Fitch rating between these grades (329 out of 470 of the rm-year observations). The highest frequency is observed in the BBB category with 22% of the rm-years in this rating category (104 of the rm-year observations).

4. Sample description Our sample employs data for the top 500 UK listed non-nancial rms by market capitalization for the period from 1989 through to 2008. Following previous studies on capital structure (Byoun, 2008; Faulkender & Petersen, 2006; Hovakimian, Kayhan, & Titman, 2008; Kisgen, 2006; Kisgen, 2009) we exclude all nancial rms from the sample, resulting in a panel of 7258 rm-year observations. According to Faulkender and Petersen (2006) and Chava and Purnanandam (2009) rms with no debt might either not be able to access debt markets or they might simply not want to have access and prefer to nance themselves with equity. If rms in this category qualify to have a rating but do not want to obtain it, they will be misclassied as rms without access. There are 707 of rms (around 10%) with zero debt in our sample. To avoid any misclassication bias in our analysis we follow Faulkender and Petersen (2006) and Chava and Purnanandam (2009) and exclude all zero-debt rms from the analysis. This leaves us with a panel of 6551 rm-year observations.

25%

23% 19% 15%

20%

15%

13% 9% 6% 10% 9%

11%

11%

10% 8% 5%

10%

5%

4% 3%

2%

3% 2%

0%

Fig. 13. Proportion of European non-nancial rms with a S&P rating in 2010. (Source: Author's calculation using Standard and Poor's data).

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863 Table 1 Differences in leverage between rms with and without access to public bond market. The table reports tests for differences in the mean and median of rms' leverage ratios for rms with and without access to public bond market. The sample is based on listed non-nancial rms for the period between 1989 and 2008 and contains rm-year observations with positive debt only. Access to the public debt markets is measured by: 1) the possession of long-term corporate credit rating (Panel A); 2) rm size, where Top20% are classied as having access and Remaining80% are classied as not having access (Panel B); 3) rm size, where Top20% are classied as rms with access and Bottom20% as rms without access (Panel C); 4) rms with rating as having access and rms in the bottom 20% as not having access (Panel D). Leverage is measured by debt-to-asset ratio. The rst two columns measure leverage on a market value of assets basis, the last two columns on a book value of assets basis. Columns I and III display results for gross value of leverage and columns II and IV for net leverage (total debt less cash and equivalents). ***, **, * indicate signicance at 1%, 5% and 10% level respectively. I II III IV Table 1 (continued) I II III IV

39

Gross leverage Net leverage MV MV

Gross leverage Net leverage BV BV

Panel C: Access proxied by rm size (Top20Bottom20) Firms with access vs. rms without access (mean difference test) Mean difference 0.1530*** 0.1703*** 0.1181*** T-Stat 19.3147 19.6379 17.5575 Signicance 0.0000 0.0000 0.0000 Firms with access vs. rms without access (median test) Median 0.1929*** 0.1761*** 0.1493*** difference Chi-squared 529.824 486.072 376.025 Signicance 0.0000 0.0000 0.0000

0.2176*** 20.7819 0.0000 0.1955*** 357.441 0.0000

Gross leverage Net leverage MV MV

Gross leverage Net leverage BV BV

Panel A: Access proxied by credit rating Firms with access N 1007 1007 999 Mean 0.2794 0.1858 0.3047 Percentiles 25 0.1465 0.0554 0.1750 Median 0.2474 0.1670 0.2876 Percentiles 75 0.3896 0.2940 0.4083 Firms without access N 5455 5455 5470 Mean 0.2283 0.1272 0.2228 Percentiles 25 0.0708 0.0058 0.1007 Median 0.1724 0.0945 0.1990 Percentiles 75 0.3316 0.2493 0.3097 Firms with access vs. rms without access (mean difference test) Mean difference 0.0511*** 0.0586*** 0.0818*** T-Stat 8.0039 8.6906 14.5352 Signicance 0.0000 0.0000 0.0000 Firms with access vs. rms without access (median test) Median 0.0750*** 0.0725*** 0.0886*** difference Chi-squared 124.254 99.616 151.007 Signicance 0.0000 0.0000 0.0000 Panel B: Access proxied by rm size (Top20Remaining 80) Firms with access N 1361 1361 1359 Mean 0.2930 0.1947 0.2816 Percentiles 25 0.1540 0.0669 0.1709 Median 0.2533 0.1752 0.2622 Percentiles 75 0.3985 0.3101 0.3657 Firms without access N 5088 5088 5087 Mean 0.1649 0.0875 0.1968 Percentiles 25 0.0649 0.0097 0.0958 Median 0.1649 0.0875 0.1968 Percentiles 75 0.3250 0.2418 0.3135 Firms with access vs. rms without access (mean difference test) Mean difference 0.0715*** 0.0736*** 0.0582*** T-Stat 12.2208 11.5828 12.3625 Signicance 0.0000 0.0000 0.0000 Firms with access vs. rms without access (median test) Median 0.0884 0.0877 0.0654 difference Chi-squared 206.785 189.604 151.433 Signicance 0.0000 0.0000 0.0000 Panel C: Access proxied by rm size (Top20Bottom20) Firms with access N 1361 1361 1359 Mean 0.2930 0.1947 0.2816 Percentiles 25 0.1540 0.0669 0.1709 Median 0.2533 0.1752 0.2622 Percentiles 75 0.3985 0.3101 0.3657 Firms without access N 1109 1109 1112 Mean 0.1399 0.0244 0.1635 Percentiles 25 0.0129 0.0763 0.0335 Median 0.0604 0.0009 0.1129 Percentiles 75 0.1663 0.0943 0.2235

999 0.2072 0.0729 0.2075 0.3266 5470 0.1057 0.0119 0.1160 0.2430 0.1015*** 14.2398 0.0000 0.0915*** 108.796 0.0000

Panel D: Access proxied by the possession of credit rating and rm size (RatedSmall20)a Firms with access N 906 906 899 899 Mean 0.2651 0.1787 0.2959 0.2008 Percentiles 25 0.1430 0.0549 0.1740 0.0723 Median 0.2413 0.1600 0.2800 0.2037 Percentiles 75 0.3605 0.2794 0.3933 0.3146 Firms without access N 1097 1097 1100 1100 Mean 0.1388 0.0230 0.1624 0.0295 Percentiles 25 0.0134 0.0763 0.0339 0.2037 Median 0.0609 0.0001 0.1133 0.0007 Percentiles 75 0.1663 0.0943 0.2233 0.1414 Firms with access vs. rms without access (mean difference test) Mean difference 0.1263*** 0.1557*** 0.1335*** 0.2303*** T-Stat 14.9524 16.8948 17.4246 20.4751 Signicance 0.0000 0.0000 0.0000 0.0000 Firms with access vs. rms without access (median test) Median 0.1804*** 0.1601*** 0.1667*** 0.2044*** difference Chi-squared 427.283 356.270 342.211 294.127 Signicance 0.0000 0.0000 0.0000 0.0000
a Note that some rms that have a rating fall into the bottom 20% of rms by size distribution. This may be due to the fact that our sample consists of the top 500 UK rms by market capitalisation, and therefore, while these rms are considered to be small, in our sample they are large enough to have a rating.

1359 0.1886 0.0766 0.1928 0.2924 5087 0.1106 0.0194 0.1106 0.2439 0.0851*** 14.0578 0.0000 0.0822 136.775 0.0000

4.2. Credit ratings in Europe Fig. 13 shows the percentage of non-nancial rms in European countries that possess a long-term S&P credit rating in 2010. For the three largest European economies Germany, France and the UK we restrict our sample to around the largest 300 non-nancial listed rms. On this basis, the UK has the highest proportion of rated rms with 19% of listed non-nancial rms possessing a rating followed by France with 15% and Germany with 13%. This suggests that the UK is a good setting to examine the role of ratings in determining capital structure decisions.18 5. Empirical analysis 5.1. Methodology The empirical analysis that follows presents results from univariate and multivariate analyses. Bond market access is proxied by the possession of a credit rating (Faulkender & Petersen, 2006; Mitto & Zhang, 2008). Following Faulkender and Petersen (2006) and Leary (2009) leverage is measured as a ratio of gross total debt to market value of assets. The univariate analysis examines differences in leverage and other rm characteristics between rms with and without a rating in our sample and other alternative measures of debt capital market access. The multivariate analysis shows the impact of credit rating on leverage while controlling for other rm characteristics and macroeconomic conditions. Following Chava and Purnanandam (2009) all variables
18 In our sample Luxembourg has a small number of listed rms (26 rms) which explains the relatively large percentage of rated rms.

1359 0.1886 0.0766 0.1928 0.2924 1112 0.0290 0.2046 0.0027 0.1428

40

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

Table 2 Mean differences in rm characteristics between rms with and without access. The table presents the results from Univariate independent sample t-tests for the sample of UK listed non-nancial rms with and without access to public bond market for the period from 1989 to 2008 and contains rm-year observations with positive debt only. Access to the public debt markets is measured by: 1) the possession of long-term corporate credit rating (Panel A); 2) rm size, where Top20% are classied as having access and Remaining80% are classied as not having access (Panel B); 3) rm size, where Top20% are classied as rms with access and Bottom20% as rms without access (Panel C); 4) rms with rating as having access and rms the bottom 20% as not having access (Panel D). Mean values are reported. Fifth column reports mean differences between the rms with and without access. *** indicates statistical signicance at 1% level, **at 5% level, *at 10% level. Access Panel A: Access is measured by credit rating Size 15.5565 Age 3.5237 Protability 0.1269 Asset tangibility 0.6543 Market-to-book 1.4841 R&D expenditure 0.0112 Asset volatility 0.2275 Equity return 0.0106 Short-term debt 0.2529 Tax paid 0.2494 N 1003 1007 1002 1007 999 1007 993 997 1007 1000 No access 12.8937 3.2751 0.1017 0.5347 1.6349 0.0149 0.2442 0.0071 0.3713 0.2520 N 5447 5438 5415 5454 5444 5454 5288 5296 5455 5441 Mean difference 2.6628 0.2486 0.0252 0.1196 0.1508 0.0038 0.0167 0.0036 0.1184 0.0027 t-stat 61.5640 7.5004 3.6673 16.0639 2.9988 3.8371 4.0735 0.2514 14.4944 0.2649 P-value 0.0000 0.0000 0.0002 0.0000 0.0028 0.0001 0.0000 0.8015 0.0000 0.7911 R vs. NR R > NR R > NR R > NR R > NR R b NR R b NR R b NR R > NR R b NR R b NR

Panel B: Access is measured by rm size (Top20Remaining80) Size 15.6249 1361 Age 3.5127 1356 Protability 0.1016 1351 Asset tangibility 0.6898 1361 Market-to-book 1.1872 1361 R&D expenditure 0.0093 1361 Asset volatility 0.2070 1348 Equity return 0.0117 1348 Short-term debt 0.2498 1361 Tax paid 0.2666 1361 Panel C: Access is measured by rm size (Top20Bottom20) Size 15.6249 1361 Age 3.5127 1356 Protability 0.1016 1351 Asset tangibility 0.6898 1361 Market-to-book 1.1872 1361 R&D expenditure 0.0093 1361 Asset volatility 0.2070 1348 Equity return 0.0117 1348 Short-term debt 0.2498 1361 Tax paid 0.2666 1361

12.6831 3.2637 0.1066 0.5171 1.7255 0.0156 0.2505 0.0068 0.3801 0.2477

5080 5076 5054 5088 5078 5088 4924 4932 5088 5074

2.9418 0.2491 0.0050 0.1727 0.5383 0.0064 0.0435 0.0049 0.1302 0.0189

96.1117 8.5686 1.1583 25.8194 15.6919 7.6340 12.8165 0.4020 17.8304 2.1258

0.0000 0.0000 0.2468 0.0000 0.0000 0.0000 0.0000 0.6877 0.0000 0.0336

R > NR R > NR R b NR R > NR R b NR R b NR R b NR R > NR R b NR R > NR

11.0939 2.8152 0.0883 0.4059 2.8684 0.0347 0.3110 0.0607 0.5526 0.2041

1105 1109 1102 1109 1107 1109 1023 1027 1109 1106

4.5310 0.6975 0.0133 0.2839 1.6811 0.0254 0.1040 0.0490 0.3028 0.0625

105.1458 18.0569 1.2721 31.5714 18.3599 12.3439 15.7111 2.1932 26.5421 5.4052

0.0000 0.0000 0.2036 0.0000 0.0000 0.0000 0.0000 0.0284 0.0000 0.0000

R > NR R > NR R > NR R > NR R b NR R b NR R b NR R b NR R b NR R > NR

Panel D: Access is measured by the possession of credit rating and rm size (RatedSmall20) Size 15.6774 905 11.0919 1093 Age 3.5524 906 2.8287 1097 Protability 0.1386 902 0.0867 1090 Asset tangibility 0.6606 906 0.4065 1097 Market-to-book 1.5044 898 2.8619 1095 R&D expenditure 0.0112 906 0.0345 1097 Asset volatility 0.2228 895 0.3094 1015 Equity return 0.0245 896 0.0619 1019 Short-term debt 0.2614 906 0.5554 1097 Tax paid 0.2615 899 0.2037 1094

4.5855 0.7238 0.0519 0.2541 1.3575 0.0233 0.0867 0.0374 0.2940 0.0578

87.8429 16.8188 4.5496 26.0053 13.5655 10.7972 12.3808 1.6205 24.1281 4.3555

0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.1053 0.0000 0.0000

R > NR R > NR R > NR R > NR R b NR R b NR R b NR R b NR R b NR R > NR

are winsorised at 1% level to eliminate outliers that could inuence the results. The rst stage of our analysis employs univariate tests to identify if rated rms possess different characteristics compared to those without a rating. We begin by comparing the leverage of rated and non-rated rms. We then compare rm characteristics of rated and non-rated rms that capital structure theories predict to have an impact on rms' leverage. We follow the prior literature in our choice of variables (Faulkender & Petersen, 2006; Leary, 2009). These characteristics include: rm size, rm age, protability, asset tangibility, market-tobook ratio, R&D expenditure, asset volatility, equity return, a portion of short-term debt and tax paid. To verify that our results are not driven by the way we dene access (possession of a credit rating), we follow Leary (2009) and Voutsinas and Werner (2011) and create alternative measures of access based on rm size. Leary (2009) indicates that while this may not be a perfect proxy, size is clearly highly correlated with public debt market access (Leary, 2009, page 1160). In addition, Leary (2009) points out

that according to Johnson (1997), and Krishnaswami, Spindt, and Subramaniam (1999), the proportion of outstanding debt from public sources is strongly correlated with rm size. He denes rms with access based on the upper two (three) deciles of book assets, while those without access are those in the lower two (three) deciles. In Leary's sample the upper two deciles contain large rms with assets greater than $100 million, whilst the lower two deciles contain small rms with assets between $1 million and $10 million (Leary, 2009, page 1161).19 Following Leary's (2009) and Voutsinas and Werner's (2011) approach, we create three sized-based measures of access: 1. Firms with access are dened as being in the top 20% (30%) of the distribution by book value of assets, and rms without access are

19 The exception could be large multinational companies. According to Aggarwal and Kyaw (2010) multinational companies, while being large and diversied, have signicantly lower debt ratios compared to domestic companies, with such debt ratios decreasing with the degree of multinationality.

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

41

Table 3 Access and capital structure, access is measured by the possession of a credit rating: Pooled OLS. The table presents estimates of Eq. (1) using annual data of UK listed non-nancial rms for the period from 1989 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by the possession of a credit rating. Credit rating is interacted with the year dummies to measure the variation in the effect of credit rating over time. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website.a Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables Credit rating Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Tax paid Rating_1989 Rating_1991 Rating_1992 Rating_1993 Rating_1994 Rating_1995 Rating_1996 Rating_1997 Rating_1998 Rating_1999 Rating_2000 Rating_2001 Rating_2002 Rating_2003 Rating_2004 Rating_2005 Rating_2006 Rating_2007 Rating_2008 Stock market return 0.0494 (0.032) 0.0173 (0.012) 0.0758*** (0.025) 0.0267 (0.024) 0.0100 (0.024) 0.0425* (0.025) 0.0230 (0.029) 0.0726** (0.031) 0.0877*** (0.029) 0.1021*** (0.029) 0.1271*** (0.029) 0.1109*** (0.029) 0.1013*** (0.029) 0.1456*** (0.029) 0.0894*** (0.030) 0.0842*** (0.030) 0.0665** (0.029) 0.0785*** (0.030) 0.1372*** (0.032) 0.0592*** 0.0570* (0.029) 0.0105 (0.012) 0.0742*** (0.024) 0.0312 (0.023) 0.0231 (0.023) 0.0490** (0.025) 0.0310 (0.028) 0.0771** (0.031) 0.0883*** (0.028) 0.1052*** (0.028) 0.1340*** (0.028) 0.1169*** (0.028) 0.1051*** (0.028) 0.1444*** (0.028) 0.0874*** (0.029) 0.0786*** (0.029) 0.0666** (0.027) 0.0785*** (0.029) 0.1390*** (0.031) 0.0501*** (1) 0.0524* (0.027) 0.0028 (0.004) 0.0073 (0.005) 0.2338*** (0.019) 0.1119*** (0.024) 0.0114*** (0.002) 0.2108* (0.116) 0.4458*** (0.033) 0.1159*** (0.006) (2) 0.0534** (0.026) 0.0008 (0.004) 0.0085 (0.005) 0.2284*** (0.018) 0.0834*** (0.024) 0.0103*** (0.002) 0.2573** (0.111) 0.4295*** (0.033) 0.1162*** (0.006) 0.0925*** (0.011) (3) 0.0537** (0.026) 0.0000 (0.004) 0.0086 (0.005) 0.2211*** (0.018) 0.0823*** (0.024) 0.0105*** (0.002) 0.2636** (0.111) 0.4338*** (0.033) 0.1151*** (0.006) 0.0908*** (0.011) 0.0405*** (0.009) 0.0569* (0.029) 0.0091 (0.012) 0.0689*** (0.024) 0.0298 (0.023) 0.0219 (0.022) 0.0482** (0.024) 0.0306 (0.028) 0.0759** (0.030) 0.0876*** (0.028) 0.1047*** (0.028) 0.1331*** (0.028) 0.1123*** (0.028) 0.1051*** (0.028) 0.1462*** (0.027) 0.0833*** (0.029) 0.0755*** (0.029) 0.0654** (0.027) 0.0760*** (0.029) 0.1393*** (0.031) 0.0507*** (4) 0.0467* (0.026) (5) 0.0548** (0.025) (6) 0.0537** (0.025)

0.0080 (0.005) 0.2316*** (0.018) 0.1165*** (0.024) 0.0112*** (0.002) 0.2083* (0.115) 0.4477*** (0.034) 0.1149*** (0.006)

0.0083* (0.005) 0.2290*** (0.017) 0.0824*** (0.024) 0.0104*** (0.002) 0.2576** (0.111) 0.4291*** (0.033) 0.1165*** (0.006) 0.0920*** (0.010)

0.0498 (0.032) 0.0186 (0.012) 0.0770*** (0.025) 0.0274 (0.024) 0.0095 (0.024) 0.0428* (0.025) 0.0232 (0.029) 0.0730** (0.031) 0.0871*** (0.029) 0.1018*** (0.029) 0.1266*** (0.029) 0.1111*** (0.029) 0.1013*** (0.029) 0.1457*** (0.029) 0.0895*** (0.030) 0.0848*** (0.030) 0.0673** (0.029) 0.0796*** (0.030) 0.1388*** (0.032) 0.0608***

0.0568* (0.029) 0.0103 (0.012) 0.0739*** (0.024) 0.0310 (0.023) 0.0231 (0.023) 0.0489** (0.025) 0.0309 (0.028) 0.0770** (0.031) 0.0884*** (0.028) 0.1052*** (0.028) 0.1341*** (0.028) 0.1168*** (0.028) 0.1051*** (0.028) 0.1444*** (0.028) 0.0874*** (0.029) 0.0784*** (0.029) 0.0664** (0.027) 0.0782*** (0.029) 0.1385*** (0.031) 0.0498***

0.0086* (0.005) 0.2211*** (0.017) 0.0823*** (0.024) 0.0105*** (0.002) 0.2636** (0.111) 0.4338*** (0.033) 0.1151*** (0.006) 0.0908*** (0.010) 0.0405*** (0.009) 0.0569* (0.029) 0.0091 (0.012) 0.0689*** (0.024) 0.0298 (0.023) 0.0219 (0.022) 0.0482** (0.024) 0.0306 (0.028) 0.0759** (0.030) 0.0876*** (0.028) 0.1047*** (0.028) 0.1331*** (0.028) 0.1123*** (0.028) 0.1051*** (0.028) 0.1462*** (0.027) 0.0833*** (0.029) 0.0755*** (0.029) 0.0654** (0.027) 0.0760*** (0.029) 0.1393*** (0.031) 0.0507*** (continued on next page)

42 Table 3 (continued) Variables GDP growth Constant Observations R-squared F test Prob > F
a

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

(1) (0.013) 0.0079*** (0.002) 0.2466*** (0.046) 6214 0.4851 45.1230 0.0000

(2) (0.013) 0.0090*** (0.002) 0.3364*** (0.047) 6214 0.5031 47.9731 0.0000

(3) (0.013) 0.0091*** (0.002) 0.3381*** (0.046) 6211 0.5065 46.1024 0.0000

(4) (0.013) 0.0076*** (0.002) 0.2788*** (0.026) 6214 0.4848 46.4853 0.0000

(5) (0.013) 0.0090*** (0.002) 0.3275*** (0.026) 6214 0.5031 48.8269 0.0000

(6) (0.013) 0.0091*** (0.002) 0.3382*** (0.026) 6211 0.5065 46.9419 0.0000

Available at the ofcial IMF website at http://www.imf.org/external/pubs/ft/weo/2010/02/weodata/index.aspx.

those in the remaining 80% (70%) of the size distributionreferred to as Top20% Remaining80% (Top30%Remaining70%) thereafter. This classication includes all rms in the sample. 2. Firms with access are dened as those in the top 20% (30%) of the distribution by book value of assets, and rms without access are dened as those in the bottom 20% (30%) of the size distribution referred to as Top20%Bottom20% (Top30%Bottom30%)thereafter. In this classication we drop the middle 60% (40%) of rms in the size distribution, thus keeping 40% (60%) of rms in the sample. 3. Firms with access are those with a credit rating and rms without access are those in the bottom 20% (30%) of the size distribution referred to as RatedBottom20% (RatedBottom30%) thereafter. In this case we exclude all unrated rms that are not included in the bottom 20% (30%) of rms by book value of assets. 5.2. Univariate analysis: differences between rms with and without a credit rating 5.2.1. Differences in leverage ratios In this section we examine if leverage of rms possessing a credit rating is different to those without a rating. Leverage is measured by debt-to-asset ratio, which is dened as book value of total debt (long-term debt plus short-term debt) divided by market value of assets (MV), where market value of assets is dened as total assets minus book value of equity plus market value of equity. Following Faulkender and Petersen (2006) and Voutsinas and Werner (2011) we also employ book value (BV) of assets in the denominator to measure leverage. Finally, we use net debt (total debt minus cash) as an alternative to total debt to measure net leverage (for full details see variable denitions in Appendix, Table B). Table 1 reports the results of tests for differences in the mean and median leverage between rms with and without a credit rating. Panel A of Table 1 shows that rms with a rating are clearly more highly levered than their unrated counterparts whether we measure leverage by MV or BV of assets or by total or net debt. When we employ MV measures of gross and net leverage (columns I and II), the univariate t-test shows that companies with rating have on average 56% more leverage. This difference increases to between 8 and 10% when we employ BV measures of gross and net leverage (columns III and IV) (p-value b 0.01). These results imply that the possession of a credit rating increases rm's leverage by between 22% [5.11/22.83] and 46% [5.86/ 12.72] when we compare MV leverage ratios and by between 37% and 96% when we compare BV leverage ratios. The results are robust throughout the whole distribution. Whether we look at differences in leverage at the 25th, 50th or 75th percentiles of the distribution, we observe that rms possessing a credit rating are more highly levered. For the median rm, credit rating increases market gross leverage by 7.5% and book gross leverage by almost 9%. Our result is slightly lower than that of Faulkender and Petersen (2006) results who nd that having a debt rating raises MV debt ratio by 13.7% and BV ratio by 15.7%. The results are similar when we use the alternative measures of public debt market access in panels B, C and D. The univariate test shows that rms with a rating have higher leverage ratios than those without a rating.

5.2.2. Differences in rm characteristics between rated and non-rated rms The previous analysis shows that rms perceived to have access to the debt capital markets have higher leverage ratios than those without. In this section we examine whether rated rms exhibit the characteristics normally associated with highly levered rms. Capital structure theories predict that rm characteristics such as size, age, protability, asset tangibility, growth opportunities, business risk, and tax shields are related to the level of a rm's indebtedness. Here we test whether these characteristics are related in a similar fashion to rms with access to the capital markets. The results in panel A of Table 2 show that rated rms are considerably larger than non-rated rms (about 266% on averagedifferences in natural logarithms). This result is consistent with the notion that since the average size of issues of bonds is higher than borrowing from the private sources and public bonds issues are associated with higher xed costs hence it is only large rms that borrow from public debt market (Faulkender & Petersen, 2006). The difference is statistically signicant at 1% level. We nd that rms with a rating are signicantly older (the difference is 25%) and more protable. These results are similar to Faulkender and Petersen (2006). Consistent with Faulkender and Petersen (2006) we nd that rms with and without a rating differ in the type of assets they possess: rms with a rating have more tangible assets (the difference is 12%), but spend less on research and development, suggesting less intangible assets and have lower growth opportunities. Firms with a rating also have less volatile assets and have higher equity returns. Public bond markets tend to provide debt with longer maturities (Faulkender & Petersen, 2006). As expected, rms with rating have 12 % less short-term debt than those without one. We also compare average tax rates ratios, that proxy for marginal tax rates, of rms with and without a rating. Firms with access are expected to have higher tax paid ratios due to the higher tax shields effect. In Panel A the difference in tax ratios is insignicant. In panels B, C and D rms with access have higher average tax rates than their unrated counterparts. Our results are robust to whether we use credit rating as a measure of access or rm size (see Panel B, C and D, Table 2). The only exceptions are: protability appears to be greater for rms without access in Panel Bwhen we use rm size Top 20% vs. Remaining 80% as a measure of accessbut it is also not statistically signicant. Equity returns are smaller for rms without access but again these gures are not statistically signicant. By comparing rms with and without access we nd that they differ substantially in the rm characteristics. Firms with access possess characteristics that according to capital structure theories would motivate them to increase their debt ratios (Faulkender & Petersen, 2006). To separate the effect of having a rating on leverage levels we have to control for rm characteristics that can be associated with a rm's demand for debt. 5.3. Regression analysis In this paper we are interested in examining whether the effect of having access to the public debt markets on rms' leverage is inuenced

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

43

Table 4 Access and capital structure, access is measured by rm size (Top 20% vs. Remaining 80% and (Top 30% vs. Remaining 70%): Pooled OLS. The table presents estimates of Eq. (1) using annual data of UK listed non-nancial rms for the period from 1989 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by a dummy, which is equal to 1 if a rm is in the top 20% in models 13 (30% in models 46) of the distribution by book value of total assets and 0 if a rm is in the remaining 80% in models 13 (70% in models 46) of size distribution. Size dummy is interacted with the year dummies to measure the variation in the effect of access over time. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Top20 Remaining80 0.0206 (0.021) 0.0084 (0.005) 0.2280*** (0.018) 0.1134*** (0.024) 0.0107*** (0.002) 0.1980* (0.114) 0.4462*** (0.033) 0.1156*** (0.006) (2) Top20 Remaining80 0.0225 (0.021) 0.0087* (0.005) 0.2261*** (0.018) 0.0795*** (0.024) 0.0100*** (0.002) 0.2487** (0.110) 0.4282*** (0.033) 0.1170*** (0.006) 0.0921*** (0.010) (3) Top20 Remaining80 0.0213 (0.021) 0.0090* (0.005) 0.2178*** (0.017) 0.0790*** (0.023) 0.0100*** (0.002) 0.2560** (0.110) 0.4330*** (0.033) 0.1157*** (0.006) 0.0907*** (0.010) 0.0425*** (0.009) 0.0500*** (0.013) 0.0003 (0.013) 0.0335** (0.014) 0.0147 (0.014) 0.0161 (0.014) 0.0056 (0.015) 0.0033 (0.015) 0.0168 (0.018) 0.0564*** (0.019) 0.0816*** (0.019) 0.1151*** (0.020) 0.0741*** (0.020) 0.0918*** (0.020) 0.1272*** (0.022) 0.0463** (0.021) 0.0404* (0.022) 0.0274 (0.020) 0.0209 (0.021) 0.0977*** (0.022) (4) Top30 Remaining70 0.0090 (0.019) 0.0062 (0.005) 0.2345*** (0.018) 0.1024*** (0.023) 0.0101*** (0.002) 0.1854 (0.114) 0.4647*** (0.034) 0.1168*** (0.006) (5) Top30 Remaining70 0.0128 (0.019) 0.0070 (0.005) 0.2318*** (0.018) 0.0725*** (0.023) 0.0095*** (0.002) 0.2331** (0.110) 0.4471*** (0.033) 0.1180*** (0.006) 0.0869*** (0.010) (6) Top30 Remaining70 0.0118 (0.019) 0.0073 (0.005) 0.2235*** (0.017) 0.0721*** (0.023) 0.0096*** (0.002) 0.2399** (0.110) 0.4520*** (0.033) 0.1167*** (0.006) 0.0855*** (0.010) 0.0430*** (0.009) 0.0532*** (0.013) 0.0160 (0.011) 0.0577*** (0.014) 0.0101 (0.014) 0.0201 (0.015) 0.0089 (0.016) 0.0077 (0.017) 0.0272 (0.020) 0.0598*** (0.018) 0.0994*** (0.020) 0.1219*** (0.020) 0.0795*** (0.020) 0.0752*** (0.021) 0.1099*** (0.022) 0.0352* (0.021) 0.0302 (0.021) 0.0205 (0.020) 0.0230 (0.020) 0.1084*** (0.020) (continued on next page)

Access Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Tax paid Access_1989 Access _1991 Access _1992 Access _1993 Access _1994 Access _1995 Access _1996 Access _1997 Access _1998 Access _1999 Access _2000 Access _2001 Access _2002 Access _2003 Access _2004 Access _2005 Access _2006 Access _2007 Access _2008

0.0469*** (0.013) 0.0073 (0.013) 0.0384*** (0.014) 0.0157 (0.015) 0.0179 (0.015) 0.0060 (0.015) 0.0083 (0.016) 0.0130 (0.018) 0.0570*** (0.020) 0.0804*** (0.020) 0.1110*** (0.020) 0.0786*** (0.021) 0.0926*** (0.020) 0.1314*** (0.023) 0.0562*** (0.022) 0.0537** (0.022) 0.0328 (0.020) 0.0316 (0.021) 0.1088*** (0.022)

0.0500*** (0.013) 0.0006 (0.013) 0.0335** (0.014) 0.0139 (0.014) 0.0159 (0.014) 0.0050 (0.015) 0.0059 (0.016) 0.0157 (0.018) 0.0570*** (0.019) 0.0804*** (0.019) 0.1133*** (0.020) 0.0791*** (0.020) 0.0905*** (0.020) 0.1242*** (0.022) 0.0471** (0.021) 0.0415* (0.021) 0.0282 (0.020) 0.0252 (0.021) 0.0972*** (0.023)

0.0529*** (0.013) 0.0240** (0.011) 0.0632*** (0.014) 0.0130 (0.014) 0.0202 (0.016) 0.0072 (0.016) 0.0093 (0.017) 0.0270 (0.020) 0.0619*** (0.018) 0.1019*** (0.020) 0.1209*** (0.021) 0.0838*** (0.020) 0.0795*** (0.021) 0.1162*** (0.022) 0.0453** (0.021) 0.0424** (0.021) 0.0265 (0.020) 0.0346* (0.020) 0.1183*** (0.021)

0.0530*** (0.013) 0.0165 (0.011) 0.0576*** (0.014) 0.0090 (0.014) 0.0202 (0.015) 0.0087 (0.016) 0.0102 (0.017) 0.0265 (0.019) 0.0593*** (0.018) 0.0986*** (0.020) 0.1210*** (0.020) 0.0819*** (0.020) 0.0736*** (0.021) 0.1084*** (0.022) 0.0368* (0.021) 0.0311 (0.021) 0.0211 (0.020) 0.0269 (0.020) 0.1075*** (0.020)

44 Table 4 (continued) Variables

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

(1) Top20 Remaining80 0.0540*** (0.013) 0.0069*** (0.002) 0.2754*** (0.026) 6214 0.4873 50.8876 0.0000

(2) Top20 Remaining80 0.0447*** (0.013) 0.0086*** (0.002) 0.3252*** (0.026) 6214 0.5057 50.9954 0.0000

(3) Top20 Remaining80 0.0459*** (0.013) 0.0086*** (0.002) 0.3363*** (0.026) 6211 0.5094 49.8481 0.0000

(4) Top30 Remaining70 0.0449*** (0.014) 0.0068*** (0.002) 0.2848*** (0.026) 6214 0.4981 52.9758 0.0000

(5) Top30 Remaining70 0.0367** (0.014) 0.0084*** (0.002) 0.3308*** (0.026) 6214 0.5142 52.8924 0.0000

(6) Top30 Remaining70 0.0375*** (0.014) 0.0085*** (0.002) 0.3419*** (0.026) 6211 0.5180 51.6025 0.0000

Stock market return GDP growth Constant Observations R-squared F test Prob > F

by credit market conditions. This is important for our study because our sample period includes the years before and during the 20072009 nancial crisis and recession. Our sample period also incorporates the recession of 19911992 and several episodes of credit market tightening, 1991 to 1992, 1999 to 2000 and 2002 to 2003. To examine whether credit market conditions are important in inuencing the impact of having access to the public debt markets on rm leverage we control for changing credit market conditions. We employ two different approaches. Firstly, we interact our access variable with our year dummies to create a separate access coefcient for each year of our sample period. We examine whether the access coefcient varies over our sample period in a manner consistent with the tightening and/or loosening of credit markets witnessed during this period. Secondly, we use two alternate proxies for changes in credit market conditions. These are data from the Loan Ofcer Surveys from the Bank of England, European Central Bank and the US Federal Reserve and the average spread on investment-grade syndicated loans. 5.3.1. The effect of public debt market access over time In this section we examine whether the impact on leverage of having access to the public debt markets credit rating varies over our sample period. We do this by interacting our access variables (credit rating and rm size dummies) with the year dummies [see Eq. (1)]. This allows the coefcient on access to public debt markets to vary by year. Our analysis will show whether there is any variation in the effect of access over time and if this variation is consistent with tightening or loosening of UK credit market conditions. We would expect the access coefcient to rise when credit markets are tight because rms dependent on bank credit have limited access to alternative sources of debt capital and thus would nd themselves debt constrained relative to rms with access to the public debt markets. Conversely we would expect the access coefcient to fall when credit markets are loose. To examine whether the effect of access on leverage varies over our sample period we estimate the following model:
Leverageit 0 1 Access 1 X it 2 GDP t 3 FTSEALLSHt 4 IDi 2t Access Year t it

1 Leverage is the response variable that measures the amount of total debt in the capital structure of a rm;CreditRating is a dummy which equals to one if a rm possesses a S&P or Fitch credit rating and zero otherwise; is a vector of rm-specic control variables; Xit IDi is a vector of industry-specic control variables; GDPt and FTALLSHt macroeconomic control variables. The Xit includes rm characteristics (rm size, protability, asset tangibility etc.) to control for demand-side factors. To control for macroeconomic conditions and time period effects that may inuence capital structure demand we follow Bougheas et al. (2006), Leary (2009) and Kayo and Kimura (2011) and include GDP annual growth

rate and stock market return. In our study we use 11 industry dummies to control for industry-specic effects. Kayo and Kimura (2011) in their study of nancial behavior of rms in developed and emerging countries nd that industry-level characteristics account for nearly 12% of leverage variance. We follow Mitto and Zhang (2008) and Leary (2009) and employ pooled OLS regression methods with standard errors adjusted for clustering at the rm level. The latter controls for the residual correlation across years for a given rm and is an alternative method to the xed-effects demeaning method to control for each rm's individual characteristics.20 The regression results are presented in Tables 3 to 6. Table 3 presents results from estimating Eq. (1) where access is measured by the possession of a credit rating. We also estimate Eq. (1) using rm size as an alternative measure of public debt market access. We use several variants of our rm size access variable. In Table 4 the largest 20% (30%) of rms measured by total assets are dened as having access and the remaining 80% (70%) are deemed not to have access. In Table 5 the largest 20% (30%) of rms (top 2 (3) deciles of the total asset distribution) measured by total assets are dened as having access and the smallest 20% (30%) of rms (bottom 2 (3) deciles) are classied as not having access. In Table 6 rms with a credit rating are dened as having access and the bottom 2 (3) deciles of the total asset distribution are classied as not having access. The key variables of interest are the interaction terms between our variable measuring public debt market access and our year dummies. 21 To illustrate the variation over time in the effect on leverage of having public debt market access we plot the coefcients on bond market access obtained from estimating Eq. (1) against time. We use the results presented in Table 3 (model 3) and Tables 46 (model 6). 22 Fig. 14 presents a graph of our access coefcient over time where access is measured by the possession of a credit rating and our rm size dummies. However, our results seem to exhibit greater variation over our sample period 19912008 and during the years 19912000 which are common to both our study and Faulkender and Petersen's (2006). For example, Faulkender and Petersen (2006) nd that the rating coefcient in their study is always signicantly greater than zero and varies from a low of 5.3% in 1991 to a high of 8.6% in 1998. They suggest that their results are not driven by a few years and nd no systematic variation in the rating coefcient during their sample period. Our results in Table 3 show that the rating coefcient was

20 Wooldridge, 2002, Econometric Analysis of Cross-section and Panel Data., Chapter 11, page 330. 21 The coefcient on our credit rating dummy variable has a different meaning in a model that includes credit rating-year dummy interaction terms. Without interaction terms the coefcient on the credit rating dummy would measure the average difference between rated and non-rated rms over our sample period. In our estimations the credit rating coefcient represents the difference between rated and non-rated rms for the year 1990 (the year dummy that we have excluded from the regression as a basis for comparison). 22 Our ndings are very similar across all the specications in Tables 3, 4, 5 and 6.

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

45

Table 5 Access and capital structure, access is measured by rm size (Top 20% vs. Bottom 20% and Top 30% vs Bottom 30%): Pooled OLS. The table presents estimates of Eq. (1) using annual data of UK listed non-nancial rms for the period from 1989 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by a size dummy, which is equal to 1 if a rm is in the top 20% in models 13 (30% in models 46) of the distribution by book value of total assets and 0 if a rm is in the bottom 20% in models 13 (30% in models 46) of size distribution . Size dummy is interacted with the year dummies to measure the variation in the effect of access over time. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables Access Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Tax paid Access_1989 Access _1991 Access _1992 Access _1993 Access _1994 Access _1995 Access _1996 Access _1997 Access _1998 Access _1999 Access _2000 Access _2001 Access _2002 Access _2003 Access _2004 Access _2005 Access _2006 Access _2007 Access _2008 Stock market return 0.0131 (0.017) 0.0394** (0.015) 0.0132 (0.017) 0.0459** (0.018) 0.0336* (0.017) 0.0582*** (0.019) 0.0549*** (0.019) 0.0489** (0.021) 0.0014 (0.022) 0.0123 (0.023) 0.0775*** (0.022) 0.0666*** (0.022) 0.1012*** (0.022) 0.0781*** (0.025) 0.0142 (0.023) 0.0007 (0.024) 0.0175 (0.022) 0.0025 (0.022) 0.1278*** (0.023) 0.0760*** 0.0117 (0.016) 0.0404*** (0.015) 0.0136 (0.016) 0.0469*** (0.018) 0.0355** (0.017) 0.0588*** (0.018) 0.0545*** (0.018) 0.0483** (0.021) 0.0007 (0.022) 0.0118 (0.022) 0.0775*** (0.022) 0.0664*** (0.022) 0.0992*** (0.021) 0.0738*** (0.024) 0.0082 (0.023) 0.0061 (0.023) 0.0205 (0.022) 0.0065 (0.021) 0.1222*** (0.023) 0.0789*** (1) Top20Bottom20 0.0777*** (0.024) 0.0003 (0.008) 0.2271*** (0.024) 0.1144*** (0.035) 0.0044* (0.002) 0.2105 (0.150) 0.3665*** (0.054) 0.1030*** (0.009) (2) Top20Bottom20 0.0666*** (0.024) 0.0007 (0.008) 0.2247*** (0.023) 0.0992*** (0.035) 0.0044* (0.002) 0.2333 (0.147) 0.3647*** (0.054) 0.1046*** (0.009) 0.0510*** (0.017) (3) Top20Bottom20 0.0698*** (0.024) 0.0010 (0.008) 0.2164*** (0.023) 0.0966*** (0.034) 0.0045* (0.002) 0.2303 (0.147) 0.3707*** (0.054) 0.1022*** (0.009) 0.0482*** (0.017) 0.0469*** (0.014) 0.0112 (0.016) 0.0398** (0.016) 0.0134 (0.016) 0.0456** (0.018) 0.0355** (0.017) 0.0589*** (0.019) 0.0513*** (0.018) 0.0463** (0.021) 0.0007 (0.022) 0.0140 (0.022) 0.0799*** (0.021) 0.0609*** (0.021) 0.1005*** (0.021) 0.0775*** (0.024) 0.0076 (0.023) 0.0068 (0.023) 0.0210 (0.022) 0.0107 (0.021) 0.1228*** (0.023) 0.0755*** (4) Top30Bottom30 0.0510** (0.023) 0.0058 (0.006) 0.2140*** (0.020) 0.0997*** (0.027) 0.0063*** (0.002) 0.1585 (0.123) 0.4112*** (0.042) 0.1093*** (0.008) (5) Top30Bottom30 0.0435* (0.022) 0.0069 (0.006) 0.2118*** (0.020) 0.0810*** (0.027) 0.0063*** (0.002) 0.1923 (0.121) 0.4041*** (0.041) 0.1105*** (0.008) 0.0588*** (0.013) (6) Top30Bottom30 0.0458** (0.022) 0.0069 (0.006) 0.2043*** (0.019) 0.0798*** (0.027) 0.0063*** (0.002) 0.1939 (0.121) 0.4102*** (0.042) 0.1094*** (0.008) 0.0570*** (0.013) 0.0421*** (0.012) 0.0128 (0.015) 0.0111 (0.013) 0.0273* (0.015) 0.0256 (0.016) 0.0256 (0.017) 0.0396** (0.018) 0.0339* (0.019) 0.0088 (0.022) 0.0289 (0.020) 0.0610*** (0.022) 0.1047*** (0.022) 0.0740*** (0.021) 0.0852*** (0.022) 0.0768*** (0.024) 0.0111 (0.023) 0.0011 (0.022) 0.0114 (0.022) 0.0036 (0.021) 0.1288*** (0.022) 0.0511** (continued on next page)

0.0134 (0.016) 0.0077 (0.013) 0.0297* (0.015) 0.0235 (0.016) 0.0229 (0.017) 0.0367** (0.019) 0.0343* (0.019) 0.0077 (0.022) 0.0316 (0.020) 0.0638*** (0.022) 0.1052*** (0.022) 0.0784*** (0.021) 0.0883*** (0.022) 0.0815*** (0.024) 0.0198 (0.023) 0.0081 (0.023) 0.0061 (0.022) 0.0137 (0.021) 0.1349*** (0.022) 0.0468**

0.0123 (0.016) 0.0108 (0.013) 0.0270* (0.015) 0.0270* (0.016) 0.0256 (0.017) 0.0394** (0.018) 0.0364* (0.019) 0.0096 (0.022) 0.0283 (0.020) 0.0600*** (0.022) 0.1036*** (0.022) 0.0763*** (0.021) 0.0836*** (0.022) 0.0752*** (0.024) 0.0125 (0.023) 0.0004 (0.022) 0.0110 (0.022) 0.0073 (0.021) 0.1280*** (0.022) 0.0522**

46 Table 5 (continued) Variables

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

(1) Top20Bottom20 (0.027) 0.0069** (0.003) 0.2143*** (0.037) 2355 0.5694 35.0829 0.0000

(2) Top20Bottom20 (0.028) 0.0068** (0.003) 0.2479*** (0.038) 2355 0.5736 35.1206 0.0000

(3) Top20Bottom20 (0.027) 0.0069** (0.003) 0.2597*** (0.037) 2354 0.5779 34.7663 0.0000

(4) Top30Bottom30 (0.022) 0.0084*** (0.003) 0.2336*** (0.030) 3625 0.5513 42.5476 0.0000

(5) Top30Bottom30 (0.022) 0.0086*** (0.003) 0.2676*** (0.030) 3625 0.5576 42.7392 0.0000

(6) Top30Bottom30 (0.021) 0.0086*** (0.003) 0.2782*** (0.030) 3624 0.5611 42.0772 0.0000

GDP growth Constant Observations R-squared F test Prob > F

very small and not signicantly different from zero in the earlier years of our sample period, relatively high and statistically signicant in the years corresponding to the credit market tightening (1992, 19992003, and 2008), and relatively low during the years before and after the these periods (1991, 19931998, and 20042007). When we employ other measures of access we nd that the access coefcient was negative in some years within our sample period (these years generally correspond to periods of loose credit conditions). In model 3 of Table 3 the rating coefcient varies from a low of 0.9% in 1991 to a high of 14% in 2003 and 2008, with the latter coefcient being statistically signicant. These differences in the characteristics of the time varying access coefcient between our study and that of Faulkender and Petersen (2006) might be due to the fact that in the UK the period studied in our paper includes three episodes of signicant credit market tightening either side of major credit booms, whereas in the US the period investigated by Faulkender and Petersen (2006) (19862000) was for the most part relatively neutral in credit condition terms (see Fig. 7).23 Furthermore, US rated rms have a longer history and tradition of using the public debt markets leading to less switching into or out of public debt and therefore more stable debt sourcing structures. In the UK rms that have access to the public debt markets have tended to rely heavily on bank sourced debt despite having the option to issue public debt. As pointed out above for UK private non-nancial rms around three quarters of debt capital is sourced from banks. This has been largely due to the relatively attractive terms banks have provided credit to the corporate sector. For example, in the years leading up to the start of the nancial crisis in Summer 2007, loan fees and spreads fell to historically low levels, driven by strong competition among lenders. The coefcients on the demand-side factors are generally consistent with those predicted by capital structure theories. The sign of the coefcient on the variable measuring rm size varies across specications and is not statistically signicant. Recent studies report mixed results for rm size. For example, Mitto and Zhang (2008) and Voutsinas and Werner (2011) report a signicant positive coefcient whereas Faulkender and Petersen (2006) and Leary (2009) nd a negative coefcient on rm size. Faulkender and Petersen (2006) suggest that this could be due to the high correlation between rm size and the credit rating variable, since only large rms obtain a credit rating because of the high xed costs of issuing public debt. We nd some evidence that age is positively related to leverage. Protability is negatively and signicantly related to leverage, which is consistent with the pecking order theory prediction that rms prefer to nance their operations using retained earnings rst, followed by debt and nally external

equity. The coefcient on asset tangibility is positive and highly statistically signicant. This is in line with the previous ndings by Faulkender and Petersen (2006) and Mitto and Zhang (2008), Leary (2009) and Voutsinas and Werner (2011) and with the notion that tangible assets serve as collateral when raising debt. We nd that growth opportunities, measured by R&D expenditure and market-to-book ratio, are negatively and signicantly related to leverage. Riskiness of business operations, measured by asset volatility, is inversely related to leverage, which is consistent with the notion that asset volatility increases a rm's probability of default leading to lower leverage. Consistent with Faulkender and Petersen's (2006) and Leary's (2009) results, equity return is negatively related to leverage, suggesting that an increase in rm's equity over the previous year decreases its leverage ratio. Following Faulkender and Petersen (2006) we also include the portion of short-term debt that is due in one year including current portion of long-term debt due within one year. Faulkender and Petersen (2006) indicate that rms with access (rating) are able to issue public bonds that tend to be of longer maturities than private debt, whereas rms without access (rating) might be restricted to raising their nancing from private investors. As expected the coefcient on short-term debt is negative and highly statistically signicant. The trade-off theory states that one of the advantages of having more debt in capital structure is the increased tax shield effect. Faulkender and Petersen (2006) predict that rms with higher marginal tax rates should experience higher interest tax shields effect and therefore have more debt. However they nd a negative coefcient on their simulated marginal tax rate variable. We use average tax rates to proxy for marginal tax rates. We also nd a negative and signicant coefcient on our tax paid variable. In Fig. 14 we present a graph of the interaction coefcients on bond market access and year dummies from 1991 to 2008. We plot the interaction coefcients from model 3 of Table 3, where access is measured by the possession of a credit rating and model 6 of Tables 4 to 6, where access is measured by our rm size dummies. The coefcient on our credit rating variable reaches a small peak in 1992 (about 7%, p-value> 0.01, model 3 Table 3) a little after the 1991 recession during which bank lending to UK private non-nancial rms weakened signicantly (see Fig. 2) but this period was not characterised by a nancial crisis.24 Surprisingly, Faulkender and Petersen (2006) nd for their sample of US rms that the effect of possessing a rating is at its lowest during the 19901991 US recession, when, according to the US Federal Reserve Loan Ofcer Survey, credit markets were relatively tight.25 The results in Table 3 show that the credit rating coefcient is relatively low and

23 The US Federal Reserve Loan Ofcer Survey shows that US credit markets were tight during the 19901991 period followed by a seven year period (19921998) when they were slightly loose and then returning to tight credit markets by the end of the decade (19992000).

24 Bougheas et al. (2006) point out that UK credit markets were tight during 19901992 and suggest that during this period interest rates were kept high in order to support the value of sterling. They also suggest that at this time external debt nance was difcult to come by. 25 Faulkender and Petersen (2006) suggest that his period was not subjected to a fully edged banking credit crunch.

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

47

Table 6 Access and capital structure, access is measured rated rms versus rms in the bottom 20% (30%) of total assets: Pooled OLS. The table presents estimates of Eq. (1) using annual data of UK listed non-nancial rms for the period from 1989 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by a dummy, which is equal to 1 if a rm has a rating and equal to 0 if a rm is in the bottom 20% in models 13 (30% in models 46) of the distribution by book value of total assets. Access is interacted with the year dummies to measure the variation in the effect of public bond market access over time. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables Access Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Tax paid ratedsmall_89 ratedsmall_91 ratedsmall_92 ratedsmall_93 ratedsmall_94 ratedsmall_95 ratedsmall_96 ratedsmall_97 ratedsmall_98 ratedsmall_99 ratedsmall_00 ratedsmall_01 ratedsmall_02 ratedsmall_03 ratedsmall_04 ratedsmall_05 ratedsmall_06 ratedsmall_07 ratedsmall_08 Stock market return 0.0068 (0.032) 0.0161 (0.015) 0.0336 (0.025) 0.0238 (0.027) 0.0014 (0.027) 0.0031 (0.028) 0.0216 (0.031) 0.0067 (0.033) 0.0233 (0.032) 0.0270 (0.033) 0.0879*** (0.033) 0.0859*** (0.032) 0.0950*** (0.031) 0.0686** (0.032) 0.0265 (0.032) 0.0055 (0.031) 0.0005 (0.032) 0.0270 (0.032) 0.1420*** (0.036) 0.0541* 0.0023 (0.030) 0.0183 (0.015) 0.0339 (0.025) 0.0208 (0.025) 0.0079 (0.025) 0.0011 (0.028) 0.0158 (0.030) 0.0093 (0.033) 0.0247 (0.032) 0.0302 (0.032) 0.0929*** (0.032) 0.0899*** (0.031) 0.0959*** (0.030) 0.0677** (0.031) 0.0250 (0.031) 0.0024 (0.030) 0.0015 (0.031) 0.0271 (0.031) 0.1424*** (0.035) 0.0592** (1) RatedSmall20 0.0514 (0.033) 0.0043 (0.009) 0.1839*** (0.024) 0.0895*** (0.032) 0.0071*** (0.002) 0.2049 (0.147) 0.2774*** (0.054) 0.0882*** (0.009) (2) RatedSmall20 0.0315 (0.032) 0.0056 (0.008) 0.1808*** (0.023) 0.0721** (0.031) 0.0072*** (0.002) 0.2332 (0.144) 0.2718*** (0.054) 0.0903*** (0.009) 0.0657*** (0.017) (3) RatedSmall20 0.0336 (0.032) 0.0056 (0.008) 0.1732*** (0.023) 0.0718** (0.030) 0.0074*** (0.002) 0.2280 (0.143) 0.2781*** (0.053) 0.0883*** (0.009) 0.0627*** (0.017) 0.0454*** (0.014) 0.0019 (0.030) 0.0193 (0.015) 0.0287 (0.025) 0.0216 (0.025) 0.0067 (0.025) 0.0009 (0.027) 0.0155 (0.030) 0.0089 (0.033) 0.0248 (0.032) 0.0308 (0.032) 0.0936*** (0.032) 0.0865*** (0.031) 0.0972*** (0.030) 0.0715** (0.031) 0.0211 (0.031) 0.0000 (0.030) 0.0022 (0.031) 0.0244 (0.031) 0.1447*** (0.035) 0.0570** (4) RatedSmall30 0.0273 (0.030) 0.0111 (0.007) 0.1723*** (0.021) 0.0940*** (0.029) 0.0069*** (0.002) 0.1782 (0.119) 0.2885*** (0.040) 0.0891*** (0.008) (5) RatedSmall30 0.0141 (0.030) 0.0122* (0.007) 0.1702*** (0.021) 0.0724*** (0.028) 0.0071*** (0.002) 0.2180* (0.116) 0.2781*** (0.039) 0.0906*** (0.008) 0.0650*** (0.014) (6) RatedSmall30 0.0172 (0.030) 0.0118* (0.007) 0.1604*** (0.020) 0.0722*** (0.027) 0.0073*** (0.002) 0.2206* (0.116) 0.2841*** (0.039) 0.0890*** (0.008) 0.0623*** (0.013) 0.0544*** (0.013) 0.0098 (0.029) 0.0189 (0.014) 0.0347 (0.024) 0.0065 (0.024) 0.0203 (0.024) 0.0164 (0.026) 0.0031 (0.029) 0.0248 (0.032) 0.0399 (0.031) 0.0476 (0.031) 0.1050*** (0.030) 0.0909*** (0.030) 0.0979*** (0.029) 0.0855*** (0.029) 0.0318 (0.030) 0.0099 (0.029) 0.0131 (0.030) 0.0309 (0.030) 0.1360*** (0.035) 0.0314 (continued on next page)

0.0054 (0.031) 0.0134 (0.014) 0.0424* (0.025) 0.0077 (0.026) 0.0145 (0.026) 0.0131 (0.027) 0.0086 (0.030) 0.0236 (0.033) 0.0397 (0.031) 0.0453 (0.032) 0.0998*** (0.032) 0.0915*** (0.031) 0.0957*** (0.030) 0.0828*** (0.031) 0.0384 (0.031) 0.0168 (0.030) 0.0120 (0.031) 0.0347 (0.031) 0.1337*** (0.035) 0.0256

0.0094 (0.030) 0.0173 (0.014) 0.0413* (0.024) 0.0051 (0.024) 0.0222 (0.024) 0.0171 (0.027) 0.0030 (0.029) 0.0259 (0.032) 0.0406 (0.031) 0.0478 (0.031) 0.1048*** (0.031) 0.0955*** (0.030) 0.0967*** (0.029) 0.0816*** (0.030) 0.0370 (0.030) 0.0134 (0.029) 0.0129 (0.030) 0.0347 (0.030) 0.1337*** (0.034) 0.0327

48 Table 6 (continued) Variables

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

(1) RatedSmall20 (0.028) 0.0083*** (0.003) 0.1943*** (0.043) 1898 0.5178 23.6723 0.0000

(2) RatedSmall20 (0.028) 0.0080*** (0.003) 0.2368*** (0.044) 1898 0.5272 25.3179 0.0000

(3) RatedSmall20 (0.028) 0.0081*** (0.003) 0.2498*** (0.043) 1897 0.5319 24.3821 0.0000

(4) RatedSmall30 (0.022) 0.0112*** (0.003) 0.1980*** (0.037) 2503 0.4977 23.3137 0.0000

(5) RatedSmall30 (0.022) 0.0114*** (0.003) 0.2356*** (0.037) 2503 0.5083 25.6012 0.0000

(6) RatedSmall30 (0.022) 0.0114*** (0.003) 0.2506*** (0.036) 2502 0.5155 24.4444 0.0000

GDP growth Constant Observations R-squared F test Prob > F

0.20

0.15

0.10

0.05

0.00
2000 2001 1991 1992 1993 1995 1997 1998 1999 2002 2003 2004 2005 2007 1994 1996 2006 2008

-0.05
Rated Firms vs. Non-Rated Top 30% vs. Bottom 30% Top 30% vs. Remaining 70% Rated Firms vs. Bottom 30%

Fig. 14. Access coefcient over time (19912008).

mainly insignicant during the 19931998 period. It however begins to increase steadily from 1996 and reaches another peak in the year 2000. We get similar results in Tables 4, 5 and 6 when we use our rm size access variables and our restricted sample specications, although in some instances our coefcient on access interaction becomes negative and statistically signicant. Overall, these ndings suggest that there was no discernable difference in leverage levels between rated and non-rated rms over the period covering 19931998. These results are consistent with the fact that credit markets were relatively loose during these post-recession years. Fig. 2 shows that this period witnessed several years of positive year-on-year growth in sterling lending to UK private non-nancial corporations with an average nominal growth rate of 8.8% during the four years between 1995 and 1998.26 Annual real GDP growth averaged 3.2% per annum during this period and also Bougheas et al. (2006) note that interest rates fell to relatively low levels over these years. Around the period covering 19992003 the coefcient is relatively high and statistically signicant and it is at its highest during years 2000 and 2003 (Tables 3 and 4). In Tables 5 and 6, when we dene rms with and without public debt access as those being in the top 3 and bottom 3 deciles of the rm size distribution, respectively, the coefcient peaks in 2000 and 2002. Although these years did not experience a nancial crisis on the scale of the 20072009 crisis they did coincide with the bursting of the dot-com bubble around the time of the millennium, the terrorist's acts of 9/11 in 2001, a string of audit scandals in 2001 and 2002, the Iraq war and rising commodity prices all leading to a signicant tightening of UK credit markets. The combined effect of these events had a major adverse impact on the availability of credit for

26

Calculated using Bank of England data.

UK rms from the banking sector. The Treasurer (2001) interviewed Fenton Burgin (Director, Close Brothers Debt Advisory Group) who suggested that post 11 September 2001 those borrowers in nondefensive sectors or without signicant ancillary business were experiencing the greatest increase in the cost of bank loans. The Treasurer (2001) also quotes Declan McGrath (Head of Syndications, UK and Europe, The Royal Bank of Scotland) who said that prior to 11 September 2001 the global economic situation had been deteriorating, and what has occurred since 11 September is an acceleration of that (page 59). Another interviewee of The Treasurer (2001) Chris Shawyer (Managing Director, Capital Markets, Lloyds TSB) suggested that banks that were already more focused on credit issues and tighter covenant structures before 9/11 imposed more severe restrictions to the availability of credit post 9/11. The Treasurer (2002a) interviewed Pippa Mason (Managing Director, Head of UK and Ireland Corporate Debt capital Markets, Citigroup) who argued that the Enron scandal in October 2001 and WorldCom's bankruptcy in July 2002 contributed to a decline in corporate credit quality and difcult credit market conditions during 2002 and 2003. The Treasurer (2002b) also interviewed Ian Fitzgerald (Director and Head of Distribution and Syndication at Lloyds TSB Capital Markets) who suggested that around this time banks became increasingly more selective about the types of structures in which they would invest. He pointed out that the damaged condence of lending banks and the ight to quality led to the credit market being only open to good quality well-rated companies, while other companies became credit-constrained. The Treasurer (2002b) quotes Ian Fitzgerald who said that The market is open and good quality well-rated companies are raising funds successfullybut there is no doubt that the market has become more nervous (page 67). This would seem to suggest that companies that suffered the most were those that could not access the public debt market and, as a result, became capital constrained.

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

49

Table 7 Capital structure and bank lending standards in the EU over the period from 2003 to 2008, access is measured by the possession of a credit rating: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by possession of a credit rating. Credit conditions are measured by bank lending standards in the EU, which are measured quarterly. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 20032008 Credit rating Rating EUQ1 Rating EUQ2 Rating EUQ3 Rating EUQ4 Rating EUAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0130** (0.0061) 0.0009 (0.0078) 0.2262*** (0.0312) 0.0987*** (0.0325) 0.0240 (0.0146) 0.0473 (0.1698) 0.5903*** (0.0651) 0.1287*** (0.0114) 0.0974*** (0.0160) 0.1509*** (0.0369) 0.0068 (0.0045) 0.5945*** (0.0865) 1714 0.5672 52.53 0.0000 0.0130** (0.006) 0.0008 (0.008) 0.2210*** (0.031) 0.0973*** (0.033) 0.0236 (0.015) 0.0520 (0.169) 0.6004*** (0.065) 0.1307*** (0.011) 0.0979*** (0.016) 0.1375*** (0.037) 0.0074 (0.005) 0.5972*** (0.087) 1714 0.5717 57.8638 0.0000 0.0131** (0.006) 0.0007 (0.008) 0.2208*** (0.031) 0.0977*** (0.033) 0.0237 (0.015) 0.0512 (0.169) 0.5977*** (0.065) 0.1314*** (0.011) 0.0983*** (0.016) 0.1255*** (0.038) 0.0076* (0.005) 0.5978*** (0.087) 1714 0.5708 58.9326 0.0000 0.0134** (0.006) 0.0007 (0.008) 0.2221*** (0.031) 0.0977*** (0.033) 0.0237 (0.015) 0.0516 (0.169) 0.5987*** (0.065) 0.1312*** (0.011) 0.0988*** (0.016) 0.1190*** (0.038) 0.0085* (0.004) 0.6043*** (0.087) 1714 0.5700 54.8366 0.0000 0.0134** (0.006) 0.0008 (0.008) 0.2250*** (0.031) 0.0976*** (0.033) 0.0238 (0.015) 0.0499 (0.169) 0.5957*** (0.065) 0.1298*** (0.011) 0.0984*** (0.016) 0.1300*** (0.036) 0.0073 (0.004) 0.6025*** (0.087) 1714 0.5683 50.4589 0.0000 0.0468** (0.020) (2) Q1 20032008 0.0441** (0.020) 0.0276*** (0.004) (3) Q2 20032008 0.0454** (0.020) (4) Q3 20032008 0.0501** (0.020) (5) Q4 20032008 0.0466** (0.020) (6) Ave 20032008 0.0463** (0.020)

0.0254*** (0.004) 0.0252*** (0.004) 0.0166*** (0.006) 0.0280*** (0.004) 0.0133** (0.006) 0.0007 (0.008) 0.2214*** (0.031) 0.0972*** (0.033) 0.0237 (0.015) 0.0522 (0.169) 0.6001*** (0.065) 0.1313*** (0.011) 0.0987*** (0.016) 0.1212*** (0.037) 0.0079* (0.004) 0.6026*** (0.087) 1714 0.5709 56.9906 0.0000

During the pre-crisis years (20042007) the coefcient on the access interaction term is lower than in 2008 and during the period covering 19992003, indicating that possession of a credit rating had a lower material impact on rms leverage levels in these years. Fig. 4 shows that the 3 month growth rate in the stock of lending was increasing at a rapid rate during 20042007 peaking in October of 2007. Furthermore, Fig. 7 shows that during the 20042007 period credit market conditions in the UK, EU and US were relatively loose implying that bank credit was readily available at attractive rates of interest. Since the 20072009 nancial crisis is associated with a severe banking credit crunch resulting in a substantial reduction in loan supply to the corporate sector, we would expect our access coefcient to rise in 2008. Fig. 14 conrms this prediction, irrespective of how we measure

access the coefcient experiences a substantial increase in 2008 over the preceding year. When we measure access by credit rating and employ our full sample the coefcient in model 3 of Table 3 increases by 6%, going from 7.6% in 2007 to nearly 14% in 2008. When we compare rms in the top 2 deciles and the bottom 2 deciles in model 3 of Table 5, the increase is much larger with the coefcient increasing by 13% from a negative 1% in 2007 to a positive 12% in 2008. We get a similar result when we compare the top 3 deciles and the bottom 3 deciles in model 6. In model 6 of Table 6 we compare rated rms with the smallest 30% of rms and nd that the coefcient on access in 2008 increases to 13.6% from 3.1% in 2007, which again represents a substantial change as credit markets went from being very loose to extremely tight.

50

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

5.3.2. Credit market conditions proxies In this section we employ the following two measures for changes in bank loan supply: 1) The Loan Ofcer Survey results from the US Federal Reserve Board, European Central Bank (ECB) and the Bank of England (BOE). 2) The spread on investment grade syndicated loans to UK rms. 5.3.2.1. Credit conditions proxy: loan ofcer survey (bank lending standards). According to Leary (2009) the Federal Reserve loan ofcer survey results provide a direct measure of the willingness of banks to extend loans (Leary, 2009, p. 1177). As pointed out in Section 3.3

this survey data is available for the US economy on a quarterly basis from 1967 to 1984 and from the second quarter of 1990. The ECB loan survey data is available from April 2003 whereas the BOE commenced their survey in the second quarter of 2007. We therefore initially limit our analysis to the period 2003 to 2008. As noted in Section 3.3 the correlation coefcient between the ECB and UK survey series is high at + 0.8. Given this strong correlation and the fact that the BOE credit conditions survey data is only available from the second quarter of 2007 we combine the ECB and BOE data using ECB survey data from 2003 through to the 2006 and BOE data for 2007 and 2008. We use this combined survey data to proxy for UK credit market conditions. We also employ the ECB survey data stand alone and the US Federal Reserve loan ofcer survey data.

Table 8 Capital structure and bank lending standards in the EU over the period from 2003 to 2008, access is measured by rated rms against the smallest 30% of rms: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Access is measured by rated rms against bottom 30% of total assets distribution. Credit conditions are measured by bank lending standards in the EU, which are measured quarterly. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 20032008 Access (Ratedsmall30) RatedSmall30 EUQ1 RatedSmall30 EUQ2 RatedSmall30 EUQ3 RatedSmall30 EUQ4 RatedSmall30 EUAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0390*** (0.013) 0.0260** (0.012) 0.2096*** (0.038) 0.0935* (0.055) 0.0073 (0.011) 0.1236 (0.204) 0.3868*** (0.092) 0.0988*** (0.017) 0.0899*** (0.026) 0.1308** (0.056) 0.0113 (0.007) 0.7444*** (0.179) 642 0.5526 26.5359 0.0000 0.0378*** (0.013) 0.0265** (0.012) 0.2043*** (0.038) 0.0907 (0.055) 0.0072 (0.011) 0.1140 (0.203) 0.4006*** (0.092) 0.1009*** (0.017) 0.0894*** (0.026) 0.1057* (0.058) 0.0122* (0.007) 0.7356*** (0.179) 642 0.5609 32.6344 0.0000 0.0385*** (0.013) 0.0265** (0.012) 0.2014*** (0.038) 0.0916* (0.055) 0.0071 (0.011) 0.1140 (0.202) 0.4012*** (0.092) 0.1027*** (0.017) 0.0896*** (0.026) 0.0721 (0.059) 0.0123* (0.007) 0.7432*** (0.178) 642 0.5618 33.3118 0.0000 0.0402*** (0.013) 0.0258** (0.012) 0.2006*** (0.038) 0.0914* (0.055) 0.0069 (0.011) 0.1117 (0.201) 0.4102*** (0.093) 0.1028*** (0.017) 0.0910*** (0.026) 0.0334 (0.061) 0.0153** (0.007) 0.7755*** (0.178) 642 0.5633 30.4522 0.0000 0.0429*** (0.014) 0.0251** (0.012) 0.2025*** (0.037) 0.0907 (0.055) 0.0065 (0.011) 0.1142 (0.201) 0.4140*** (0.094) 0.1014*** (0.016) 0.0908*** (0.026) 0.0238 (0.059) 0.0126* (0.007) 0.8054*** (0.184) 642 0.5622 25.9553 0.0000 0.1627*** (0.058) (2) Q1 20032008 0.1546*** (0.059) 0.0241*** (0.004) (3) Q2 20032008 0.1581*** (0.058) (4) Q3 20032008 0.1705*** (0.058) (5) Q4 20032008 0.1761*** (0.059) (6) Ave 20032008 0.1630*** (0.058)

0.0272*** (0.005) 0.0352*** (0.006) 0.0378*** (0.010) 0.0343*** (0.006) 0.0396*** (0.013) 0.0261** (0.012) 0.2008*** (0.038) 0.0905 (0.055) 0.0069 (0.011) 0.1113 (0.201) 0.4096*** (0.092) 0.1025*** (0.017) 0.0901*** (0.026) 0.0512 (0.059) 0.0133* (0.007) 0.7629*** (0.178) 642 0.5640 31.5794 0.0000

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

51

Following Leary (2009) we modify Eq. (1) using these credit condition proxies and estimate the following model: Leverageit 0 1 Access 1 X it 2 GDP t 3 FTSEALLSHt 4 IDi 2t Access Credit Conditionst it

Where the coefcient on Access CreditConditionst is the coefcient of interest. We initially estimate Eq. (2) using data from 2003. We use the bank lending standards survey data for each quarter (Q1, Q2, Q3 and Q4) as well as an average of the four quarters. Access to the public debt markets is measured by the possession of a credit rating and a rm size dummy. In the case of the latter we dene the largest 30% of rms

by book value of assets as having access and the smallest 30% as not having access. Tables 7 to 9 present results from estimating Eq. (2) where access is interacted with the data on bank lending standards from the ECB loan ofcer survey. Tables 10 to 12 present results from Eq. (2) where access is interacted with the combined data from the ECB and the BOE loan ofcer surveys. The US Federal Reserve loan ofcer survey data is available from 1990 and our rm-level and UK macroeconomic data starts in 1989. Since we nd a strong correlation between UK and US bank lending standards as measured by the loan ofcer surveys we have also estimated Eq. (2) using the US survey data for the period from 1990 to 2008 enabling us to make virtually full use of the time series dimension of our data. These results are presented in Tables 1315.

Table 9 Capital structure and bank lending standards in the EU over the period from 2003 to 2008, access is measured by the largest 30% of rms against the smallest 30% of rms: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is measured by the largest 30% of rms against the smallest 30% of rms by book value of total assets. Credit conditions are measured by bank lending standards in the EU, which are measured quarterly. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 20032008 Access (Top30Bottom30) Top30Bottom30 EUQ1 Top30Bottom30 EUQ2 Top30Bottom30 EUQ3 Top30Bottom30 EUQ4 Top30Bottom30 EUAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0298*** (0.010) 0.0060 (0.010) 0.2397*** (0.039) 0.1290*** (0.040) 0.0088 (0.014) 0.0279 (0.191) 0.5248*** (0.086) 0.1197*** (0.013) 0.0856*** (0.025) 0.1079** (0.047) 0.0161*** (0.006) 0.7220*** (0.135) 1,006 0.5957 46.5607 0.0000 0.0291*** (0.010) 0.0066 (0.010) 0.2243*** (0.039) 0.1270*** (0.040) 0.0080 (0.014) 0.0459 (0.191) 0.5459*** (0.086) 0.1237*** (0.013) 0.0856*** (0.025) 0.0638 (0.048) 0.0165*** (0.006) 0.7179*** (0.135) 1,006 0.6097 60.1167 0.0000 0.0296*** (0.010) 0.0065 (0.010) 0.2217*** (0.039) 0.1273*** (0.040) 0.0081 (0.014) 0.0432 (0.188) 0.5446*** (0.086) 0.1250*** (0.013) 0.0857*** (0.025) 0.0133 (0.051) 0.0168*** (0.006) 0.7215*** (0.135) 1,006 0.6094 62.6459 0.0000 0.0308*** (0.010) 0.0060 (0.010) 0.2236*** (0.039) 0.1252*** (0.040) 0.0083 (0.014) 0.0450 (0.187) 0.5579*** (0.086) 0.1247*** (0.013) 0.0877*** (0.025) 0.0447 (0.053) 0.0218*** (0.006) 0.7562*** (0.135) 1,006 0.6099 58.2356 0.0000 0.0319*** (0.010) 0.0057 (0.010) 0.2337*** (0.039) 0.1233*** (0.040) 0.0089 (0.014) 0.0410 (0.187) 0.5584*** (0.088) 0.1173*** (0.013) 0.0864*** (0.025) 0.0220 (0.052) 0.0192*** (0.006) 0.7662*** (0.137) 1,006 0.6034 49.4569 0.0000 0.1228*** (0.040) (2) Q1 20032008 0.1170*** (0.040) 0.0313*** (0.004) (3) Q2 20032008 0.1197*** (0.040) (4) Q3 20032008 0.1301*** (0.040) (5) Q4 20032008 0.1258*** (0.040) (6) Ave 20032008 0.1222*** (0.040)

0.0337*** (0.004) 0.0420*** (0.006) 0.0360*** (0.008) 0.0413*** (0.005) 0.0304*** (0.010) 0.0063 (0.010) 0.2222*** (0.039) 0.1250*** (0.040) 0.0082 (0.014) 0.0479 (0.188) 0.5577*** (0.086) 0.1234*** (0.013) 0.0864*** (0.025) 0.0169 (0.051) 0.0188*** (0.006) 0.7423*** (0.135) 1,006 0.6113 61.6898 0.0000

52

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

Table 10 Capital structure and bank lending standards for the combined loan ofcer survey data in the EU and the UK over the period from 2003 to 2008, access is measured by the possession of a credit rating: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. Loan Ofcer survey data in the UK is only available from 2007 Q2. We therefore combine the survey results from the ECB and BOE for the period 2003 to 2008 as a proxy for credit market conditions in the UK. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by possession of a credit rating. Credit conditions are measured by bank lending standards in the UK, which are measured quarterly. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 20032008 Credit rating Rating EUUKQ1 Rating EUUKQ2 Rating EUUKQ3 Rating EUUKQ4 Rating EUUKAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0130** (0.0061) 0.0009 (0.0078) 0.2262*** (0.0312) 0.0987*** (0.0325) 0.0240 (0.0146) 0.0473 (0.1698) 0.5903*** (0.0651) 0.1287*** (0.0114) 0.0974*** (0.0160) 0.1508*** (0.0369) 0.0068 (0.0045) 0.5945*** (0.0865) 1714 0.5672 52.53 0.0000 0.0131** (0.006) 0.0007 (0.008) 0.2209*** (0.031) 0.0974*** (0.033) 0.0236 (0.015) 0.0521 (0.169) 0.6000*** (0.065) 0.1311*** (0.011) 0.0981*** (0.016) 0.1337*** (0.037) 0.0069 (0.005) 0.5968*** (0.087) 1714 0.5715 58.8642 0.0000 0.0130** (0.006) 0.0007 (0.008) 0.2207*** (0.031) 0.0976*** (0.033) 0.0237 (0.015) 0.0512 (0.169) 0.5982*** (0.065) 0.1312*** (0.011) 0.0982*** (0.016) 0.1279*** (0.038) 0.0082* (0.005) 0.5985*** (0.087) 1714 0.5712 58.3804 0.0000 0.0133** (0.006) 0.0007 (0.008) 0.2219*** (0.031) 0.0974*** (0.033) 0.0237 (0.015) 0.0520 (0.169) 0.6003*** (0.065) 0.1310*** (0.011) 0.0988*** (0.016) 0.1174*** (0.038) 0.0097** (0.005) 0.6075*** (0.087) 1714 0.5705 55.2581 0.0000 0.0135** (0.006) 0.0008 (0.008) 0.2245*** (0.031) 0.0968*** (0.033) 0.0237 (0.015) 0.0518 (0.169) 0.6001*** (0.065) 0.1299*** (0.011) 0.0986*** (0.016) 0.1275*** (0.036) 0.0077* (0.004) 0.6064*** (0.087) 1714 0.5694 51.3409 0.0000 0.0468** (0.020) (2) Q1 20032008 0.0432** (0.020) 0.0262*** (0.004) (3) Q2 20032008 0.0465** (0.020) (4) Q3 20032008 0.0522*** (0.020) (5) Q4 20032008 0.0515*** (0.020) (6) Ave 20032008 0.0479** (0.020)

0.0278*** (0.004) 0.0300*** (0.005) 0.0338*** (0.007) 0.0313*** (0.004) 0.0132** (0.006) 0.0007 (0.008) 0.2212*** (0.031) 0.0971*** (0.033) 0.0236 (0.015) 0.0525 (0.169) 0.6010*** (0.065) 0.1312*** (0.011) 0.0985*** (0.016) 0.1239*** (0.037) 0.0082* (0.005) 0.6026*** (0.087) 1714 0.5713 57.8346 0.0000

We standardise the loan ofcer survey variable to facilitate an assessment of the economic magnitude of the credit condition coefcients. 27 We nd that a one-standard deviation move in the EU bank lending standards proxy increases the leverage difference between rms with and without access by about: a) 1.72.8 percentage points in Table 7 (where access is measured by the possession of a credit rating). This change represents about 3560% of the average leverage difference of 4.68% obtained in model (1) of Table 7.

b) 2.43.8 percentage points in Table 8 (where access is measured by rated rms against those in the bottom 30% of the size distribution by book value of total assets). This change represents about 15 24% of the average leverage difference of 16% estimated in model (1) of Table 8. c) 3.14.2 percentage points in Table 9 (where access is measured by the largest 30% against the smallest 30% of rms by the size distribution). This change represents about 2535% of the average leverage difference of 12% estimated in model (1) of Table 9. We nd similar but slightly higher results for the combined EU and UK bank lending standards proxy. This is what we would expect to nd as the combined EU and UK bank lending standards variable better reects credit conditions in the UK than just EU lending

We standardise our variables by subtracting the sample mean from the actual value of the credit condition variable for each year and dividing it by the standard deviation of the variable.

27

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

53

Table 11 Capital structure and bank lending standards for the combined loan ofcer survey data in the EU and the UK over the period from 2003 to 2008, access is measured by rated rms against the smallest 30% of rms: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. Loan Ofcer survey data in the UK is only available from 2007 Q2. We therefore combine the survey results from the ECB and BOE for the period 2003 to 2008 as a proxy for credit market conditions in the UK. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Access is measured by rated rms against bottom 30% of total assets distribution. Credit Conditions are measured by bank lending standards in the EU and UK, which are measured quarterly. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 20032008 Access (Ratedsmall30) RatedSmall30 EUUKQ1 RatedSmall30 EUUKQ2 RatedSmall30 EUUKQ3 RatedSmall30 EUUKQ4 RatedSmall30 EUUKAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0390*** (0.013) 0.0260** (0.012) 0.2096*** (0.038) 0.0935* (0.055) 0.0073 (0.011) 0.1236 (0.204) 0.3868*** (0.092) 0.0988*** (0.017) 0.0899*** (0.026) 0.1308** (0.056) 0.0113 (0.007) 0.7444*** (0.179) 642 0.5526 26.5359 0.0000 0.0382*** (0.013) 0.0265** (0.012) 0.2032*** (0.038) 0.0909 (0.055) 0.0071 (0.011) 0.1137 (0.203) 0.4021*** (0.092) 0.1016*** (0.017) 0.0895*** (0.026) 0.0955 (0.058) 0.0110 (0.007) 0.7373*** (0.179) 642 0.5616 33.1443 0.0000 0.0381*** (0.013) 0.0265** (0.012) 0.2025*** (0.038) 0.0914* (0.055) 0.0072 (0.011) 0.1142 (0.203) 0.3998*** (0.092) 0.1019*** (0.017) 0.0895*** (0.026) 0.0834 (0.059) 0.0138* (0.007) 0.7411*** (0.179) 642 0.5611 33.2563 0.0000 0.0397*** (0.013) 0.0258** (0.012) 0.2017*** (0.038) 0.0908 (0.055) 0.0070 (0.011) 0.1109 (0.202) 0.4102*** (0.093) 0.1018*** (0.017) 0.0911*** (0.026) 0.0404 (0.060) 0.0184** (0.008) 0.7776*** (0.179) 642 0.5632 31.2138 0.0000 0.0420*** (0.013) 0.0252** (0.012) 0.2042*** (0.037) 0.0892 (0.055) 0.0066 (0.011) 0.1118 (0.202) 0.4169*** (0.094) 0.1001*** (0.017) 0.0906*** (0.026) 0.0487 (0.055) 0.0135* (0.007) 0.7980*** (0.182) 642 0.5633 26.6437 0.0000 0.1627*** (0.058) (2) Q1 20032008 0.1548*** (0.058) 0.0248*** (0.004) (3) Q2 20032008 0.1576*** (0.058) (4) Q3 20032008 0.1706*** (0.058) (5) Q4 20032008 0.1801*** (0.059) (6) Ave 20032008 0.1624*** (0.058)

0.0268*** (0.005) 0.0373*** (0.006) 0.0526*** (0.011) 0.0341*** (0.005) 0.0390*** (0.013) 0.0262** (0.012) 0.2020*** (0.038) 0.0904 (0.055) 0.0070 (0.011) 0.1116 (0.202) 0.4077*** (0.092) 0.1019*** (0.017) 0.0900*** (0.026) 0.0667 (0.058) 0.0139* (0.007) 0.7566*** (0.178) 642 0.5633 32.4324 0.0000

standards. A one-standard deviation move in this variable increases leverage difference between rms with and without access by about: a) 2.63.4 percentage points in Table 10 (where access is measured by the possession of a credit rating). This change represents about 5572% of the average leverage difference of 4.68% obtained in model (1). b) 2.55.3 percentage points in Table 11 (where access is measured by rated rms against those in the bottom 30% of the size distribution by book value of total assets). This change represents about 1533% of the average leverage difference of 16% estimated in model (1).

c) 3.25.4 percentage points in Table 12 (where access is measured by the largest 30% against the smallest 30% of rms by the size distribution). This change represents about 2745% of the average leverage difference of 12% estimated in model (1). We nd slightly lower results for the US bank lending standards proxy. A one-standard deviation move in this variable increases leverage difference between rms with and without access by about: a) 1.21.9 percentage points in Table 13 (where access is measured by the possession of a credit rating). This change represents about 3048% of the average leverage difference of 3.97% obtained in model (1).

54

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

Table 12 Capital structure and bank lending standards for the combined loan ofcer survey data in the EU and the UK over the period from 2003 to 2008, access is measured by the largest 30% of rms against the smallest 30% of rms: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. Loan Ofcer survey data in the UK is only available from 2007 Q2. We therefore combine the survey results from the ECB and BOE for the period 2003 to 2008 as a proxy for credit market conditions in the UK. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is measured by the largest 30% of rms against the smallest 30% of rms by book value of total assets. Credit Conditions are measured by bank lending standards in the EU and UK, which are measured quarterly. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 20032008 Access (Top30vsBottom30) Top30Bottom30 EUUKQ1 Top30Bottom30 EUUKQ2 Top30Bottom30 EUUKQ3 Top30Bottom30 EUUKQ4 Top30Bottom30 EUUKAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0298*** (0.010) 0.0060 (0.010) 0.2397*** (0.039) 0.1290*** (0.040) 0.0088 (0.014) 0.0279 (0.191) 0.5248*** (0.086) 0.1197*** (0.013) 0.0856*** (0.025) 0.1079** (0.047) 0.0161*** (0.006) 0.7220*** (0.135) 1006 0.5957 46.5607 0.0000 0.0294*** (0.010) 0.0065 (0.010) 0.2233*** (0.039) 0.1268*** (0.040) 0.0080 (0.014) 0.0455 (0.190) 0.5474*** (0.086) 0.1241*** (0.013) 0.0856*** (0.025) 0.0485 (0.049) 0.0145** (0.006) 0.7165*** (0.135) 1006 0.6101 61.6209 0.0000 0.0293*** (0.010) 0.0065 (0.010) 0.2225*** (0.039) 0.1276*** (0.040) 0.0081 (0.014) 0.0439 (0.189) 0.5430*** (0.086) 0.1245*** (0.013) 0.0856*** (0.025) 0.0293 (0.050) 0.0191*** (0.006) 0.7232*** (0.135) 1006 0.6091 61.5586 0.0000 0.0306*** (0.010) 0.0060 (0.010) 0.2242*** (0.039) 0.1251*** (0.040) 0.0082 (0.014) 0.0473 (0.188) 0.5587*** (0.086) 0.1241*** (0.013) 0.0878*** (0.025) 0.0363 (0.053) 0.0266*** (0.006) 0.7651*** (0.135) 1006 0.6102 58.3504 0.0000 0.0316*** (0.010) 0.0057 (0.010) 0.2332*** (0.039) 0.1223*** (0.040) 0.0086 (0.014) 0.0471 (0.188) 0.5656*** (0.088) 0.1166*** (0.013) 0.0865*** (0.025) 0.0013 (0.048) 0.0203*** (0.006) 0.7691*** (0.136) 1006 0.6063 52.5365 0.0000 0.1228*** (0.040) (2) Q1 20032008 0.1164*** (0.040) 0.0315*** (0.004) (3) Q2 20032008 0.1204*** (0.040) (4) Q3 20032008 0.1317*** (0.040) (5) Q4 20032008 0.1326*** (0.040) (6) Ave 20032008 0.1234*** (0.040)

0.0338*** (0.005) 0.0449*** (0.006) 0.0543*** (0.008) 0.0417*** (0.005) 0.0300*** (0.010) 0.0063 (0.010) 0.2228*** (0.039) 0.1254*** (0.040) 0.0081 (0.014) 0.0483 (0.189) 0.5554*** (0.086) 0.1234*** (0.013) 0.0863*** (0.025) 0.0047 (0.050) 0.0197*** (0.006) 0.7392*** (0.135) 1006 0.6112 61.7194 0.0000

b) 2.13.3 percentage points in Table 14 (where access is measured by rated rms against those in the bottom 30% of the size distribution). This change represents about 1219% of the average leverage difference of 17% estimated in model (1). c) 2.34.0 percentage points in Table 15 (where access is measured by the largest 30% against the smallest 30% of rms by the size distribution). This change represents about 1324% of the average leverage difference of 17% estimated in model (1). Our results show that the estimated coefcient on the access and bank lending standards interaction terms are positive and signicant. These results suggest that bond market access becomes a more important determinant of leverage during the periods of reduced bank

loan supply. We nd that the leverage difference between rated and non-rated rms is greater when rms with access to the bond markets are benchmarked against smaller rms. In model 1 of Table 7 the average leverage difference between rated and non-rated rms is 4.68%, however, when we restrict our non-access rms to those in the bottom 30% of the rm size distribution the average leverage difference increases to 16.27% (model 1 Table 8). Furthermore, we also nd that changing bank lending standards has a greater impact on the leverage difference between rated rms and small rms. In the overwhelming majority of cases, a one-standard deviation move in bank lending standards increases the leverage difference between rated rms and the smallest 30% of rms by a greater amount than when we model leverage differences between rated and all non-

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

55

Table 13 Capital structure and bank lending standards in the US over the period from 1990 to 2008, access is measured by the possession of a credit rating: Pooled OLS. The table presents estimates of Eq. (1) using annual data of UK listed non-nancial rms for the period from 1990 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by the possession of a credit rating. US Federal Reserve Loan Ofcer Survey results are used as a proxy for credit market conditions, this data is available quarterly. We also use the average survey result for the year. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 19902008 Credit rating Rating USQ1 Rating USQ2 Rating USQ3 Rating USQ4 Rating USAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0014 (0.004) 0.0094* (0.005) 0.2290*** (0.018) 0.0873*** (0.024) 0.0100*** (0.002) 0.2528** (0.110) 0.4158*** (0.032) 0.1136*** (0.006) 0.0948*** (0.011) 0.0699*** (0.012) 0.0083*** (0.002) 0.3325*** (0.048) 5968 0.5014 68.9461 0.0000 0.0014 (0.004) 0.0094* (0.005) 0.2281*** (0.018) 0.0874*** (0.024) 0.0100*** (0.002) 0.2530** (0.110) 0.4169*** (0.032) 0.1142*** (0.006) 0.0951*** (0.011) 0.0612*** (0.012) 0.0080*** (0.002) 0.3318*** (0.048) 5968 0.5019 66.8217 0.0000 0.0014 (0.004) 0.0094* (0.005) 0.2284*** (0.018) 0.0875*** (0.024) 0.0100*** (0.002) 0.2511** (0.110) 0.4190*** (0.032) 0.1145*** (0.006) 0.0953*** (0.011) 0.0567*** (0.012) 0.0079*** (0.002) 0.3328*** (0.048) 5968 0.5024 67.3449 0.0000 0.0015 (0.004) 0.0093* (0.005) 0.2289*** (0.018) 0.0875*** (0.024) 0.0100*** (0.002) 0.2497** (0.110) 0.4203*** (0.032) 0.1146*** (0.006) 0.0954*** (0.011) 0.0539*** (0.012) 0.0080*** (0.002) 0.3339*** (0.048) 5968 0.5026 67.4952 0.0000 0.0015 (0.004) 0.0093* (0.005) 0.2296*** (0.018) 0.0875*** (0.024) 0.0100*** (0.002) 0.2483** (0.110) 0.4210*** (0.033) 0.1143*** (0.006) 0.0951*** (0.011) 0.0568*** (0.012) 0.0081*** (0.002) 0.3344*** (0.048) 5968 0.5025 67.4822 0.0000 0.0397*** (0.014) (2) Q1 19902008 0.0409*** (0.014) 0.0124*** (0.004) (3) Q2 19902008 0.0411*** (0.014) (4) Q3 19902008 0.0404*** (0.014) (5) Q4 19902008 0.0391*** (0.014) (6) Ave 19902008 0.0405*** (0.014)

0.0167*** (0.004) 0.0189*** (0.004) 0.0168*** (0.004) 0.0176*** (0.004) 0.0014 (0.004) 0.0093* (0.005) 0.2287*** (0.018) 0.0875*** (0.024) 0.0100*** (0.002) 0.2504** (0.110) 0.4197*** (0.032) 0.1145*** (0.006) 0.0953*** (0.011) 0.0555*** (0.012) 0.0079*** (0.002) 0.3332*** (0.048) 5968 0.5025 67.5013 0.0000

rated rms. Simply put smaller rms are affected more by changing credit market conditions. This nding is consistent with BOE evidence which reports that lending conditions for small and medium-sized companies (SMEs) were more constrained than for larger companies during the nancial crisis. It was suggested that this was partly due to small rms' relative lack of access to capital markets. The BOE cites an August 2009 survey by the EEF, the manufacturers' organisation, which indicated that SMEs in the manufacturing sector were more likely to have experienced a decline in the availability of debt nance, and an increase in its cost, than large rms. In communications with the BOE UK banks reported that smaller companies faced stricter lending criteria and higher loan spreads than larger companies, because they were less well diversied and consequently less able than larger companies to withstand adverse shocks (Bank of England (2009f). Trends in lending).

The models in Tables 8, 11, 14 and 15, where access is an indicator variable equal to one for rated rms or rms in top 30% of the size (total asset) distribution and zero for rms in the bottom 30%, are closest in construction to the models estimated by Leary (2009) and Voutsinas and Werner (2011). A comparison of our results with those of Leary (2009) shows that the size of our access and interaction coefcients are much higher than those reported by Leary (2009). For example, the coefcient on access in model 1 of Table 14 is 0.1698, which indicates an average leverage difference of nearly 17% between rated rms and those in the bottom 30% of the size distribution for our sample over the period 1990 to 2008. Similarly, for the period 2003 to 2008 we get an access coefcient of 0.1627 (model 1, Table 11). The corresponding gure for Leary's (2009) sample over the period 1965 to 2000 is 0.028 or 2.8%. When we compare our results in model 1 of Table 15 with the corresponding results for

56

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

Table 14 Capital structure and bank lending standards in the US over the period from 1990 to 2008, access is measured by rated rms against the smallest 30% of rms: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 1990 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Access is measured by rated rms against bottom 30% of total assets distribution. Public bond market access is proxied by a possession of credit rating. US Federal Reserve Loan Ofcer Survey results are used as a proxy for credit market conditions, this data is available quarterly. We also use the average survey result for the year. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 19902008 Access (Ratedsmall30) RatedSmall30 USQ1 RatedSmall30 USQ2 RatedSmall30 USQ3 RatedSmall30 USQ4 RatedSmall30 USAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0282*** (0.007) 0.0171** (0.008) 0.1516*** (0.020) 0.1002*** (0.027) 0.0021 (0.002) 0.1659 (0.113) 0.2802*** (0.038) 0.0751*** (0.008) 0.0697*** (0.013) 0.0644*** (0.019) 0.0077*** (0.003) 0.5303*** (0.083) 2399 0.5179 32.6167 0.0000 0.0282*** (0.007) 0.0171** (0.007) 0.1504*** (0.020) 0.0997*** (0.027) 0.0022 (0.002) 0.1648 (0.113) 0.2847*** (0.038) 0.0783*** (0.008) 0.0705*** (0.013) 0.0281 (0.020) 0.0063** (0.003) 0.5270*** (0.082) 2399 0.5215 31.6921 0.0000 0.0285*** (0.007) 0.0170** (0.007) 0.1516*** (0.020) 0.0996*** (0.027) 0.0022 (0.002) 0.1610 (0.113) 0.2913*** (0.038) 0.0797*** (0.008) 0.0713*** (0.013) 0.0085 (0.020) 0.0060** (0.003) 0.5316*** (0.082) 2399 0.5242 32.8836 0.0000 0.0287*** (0.007) 0.0167** (0.007) 0.1532*** (0.020) 0.0994*** (0.027) 0.0021 (0.002) 0.1580 (0.113) 0.2954*** (0.039) 0.0806*** (0.008) 0.0715*** (0.013) 0.0045 (0.020) 0.0062** (0.003) 0.5370*** (0.082) 2399 0.5257 33.4479 0.0000 0.0286*** (0.007) 0.0166** (0.007) 0.1545*** (0.020) 0.0999*** (0.027) 0.0021 (0.002) 0.1563 (0.113) 0.2960*** (0.039) 0.0789*** (0.008) 0.0707*** (0.013) 0.0133 (0.019) 0.0066** (0.003) 0.5382*** (0.082) 2399 0.5239 33.1992 0.0000 0.1698*** (0.031) (2) Q1 19902008 0.1723*** (0.031) 0.0214*** (0.004) (3) Q2 19902008 0.1738*** (0.031) (4) Q3 19902008 0.1741*** (0.031) (5) Q4 19902008 0.1715*** (0.031) (6) Ave 19902008 0.1734*** (0.031)

0.0291*** (0.004) 0.0334*** (0.005) 0.0267*** (0.005) 0.0307*** (0.005) 0.0285*** (0.007) 0.0168** (0.007) 0.1524*** (0.020) 0.0996*** (0.027) 0.0022 (0.002) 0.1592 (0.113) 0.2935*** (0.039) 0.0801*** (0.008) 0.0712*** (0.013) 0.0030 (0.020) 0.0060** (0.003) 0.5335*** (0.082) 2399 0.5247 33.0491 0.0000

Japanese rms reported by Voutsinas and Werner (2011) we again nd that UK rms leverage is more sensitive to bond market access. Voutsinas and Werner (2011) nd that non-bank dependent rms have 10.5% higher leverage than bank dependent rms during the period 1990 to 2007 whereas our result for a similar period is 17.08%. For models that include interaction terms the coefcient on our access lending standards interaction variable is 0.0341 in model 6 in Table 11 and 0.0307 in Table 14 which are more than two times that of the corresponding coefcient of 0.015 reported by Leary (2009). These differences in the economic signicance of access to public debt markets might simply be due to the fact that the period under investigation in our study is characterised by greater variation in

credit market conditions which include the unprecedented deterioration in bank lending during the 20072009 nancial crisis. However, notwithstanding the fact that our results and those of Leary (2009) and to a lesser extent those of Voutsinas and Werner (2011) are not fully comparable as our periods only partially overlap and the variable denitions and model specications are not identical, our ndings would seem to suggest that UK rms leverage is more sensitive to whether rms have access to public debt markets and the condition of the credit markets. 28 This result is consistent with our earlier

28 We could draw a similar conclusion comparing Japanese rms with US rms based on Voutsinas and Werner's (2011) results.

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

57

Table 15 Capital structure and bank lending standards in the US over the period from 1990 to 2008, access is measured by the largest 30% of rms against the smallest 30% of rms: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 1990 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of asset. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Access is measured by the largest 30% of rms against the smallest 30% of rms measured by the book value of total assets. US Federal Reserve Loan Ofcer Survey results are used as a proxy for credit market conditions, this data is available quarterly. We also use the average survey result for the year. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) Original 19902008 Access (Top30vsBottom30) Top30Bottom30 USQ1 Top30Bottom30 USQ2 Top30Bottom30 USQ3 Top30Bottom30 USQ4 Top30Bottom30 USAve Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0289*** (0.006) 0.0110 (0.007) 0.1866*** (0.019) 0.1175*** (0.026) 0.0001 (0.002) 0.1522 (0.117) 0.3799*** (0.039) 0.0974*** (0.008) 0.0615*** (0.013) 0.0720*** (0.015) 0.0060*** (0.002) 0.5618*** (0.069) 3473 0.5510 61.0569 0.0000 0.0288*** (0.006) 0.0115* (0.007) 0.1850*** (0.019) 0.1161*** (0.026) 0.0003 (0.002) 0.1518 (0.117) 0.3853*** (0.039) 0.1006*** (0.008) 0.0640*** (0.013) 0.0220 (0.018) 0.0029 (0.002) 0.5542*** (0.069) 3473 0.5556 62.7444 0.0000 0.0290*** (0.006) 0.0113* (0.007) 0.1878*** (0.019) 0.1144*** (0.026) 0.0005 (0.002) 0.1460 (0.117) 0.3942*** (0.039) 0.1015*** (0.008) 0.0649*** (0.013) 0.0144 (0.018) 0.0023 (0.002) 0.5579*** (0.069) 3473 0.5597 66.1038 0.0000 0.0294*** (0.006) 0.0108 (0.007) 0.1908*** (0.019) 0.1130*** (0.026) 0.0006 (0.002) 0.1375 (0.116) 0.4017*** (0.039) 0.1019*** (0.008) 0.0645*** (0.013) 0.0423** (0.018) 0.0026 (0.002) 0.5668*** (0.069) 3473 0.5629 67.8071 0.0000 0.0298*** (0.006) 0.0104 (0.007) 0.1931*** (0.019) 0.1129*** (0.026) 0.0006 (0.002) 0.1325 (0.116) 0.4062*** (0.040) 0.0996*** (0.008) 0.0633*** (0.013) 0.0189 (0.017) 0.0032 (0.002) 0.5760*** (0.069) 3473 0.5620 66.5924 0.0000 0.1708*** (0.021) (2) Q1 19902008 0.1727*** (0.021) 0.0227*** (0.003) (3) Q2 19902008 0.1743*** (0.021) (4) Q3 19902008 0.1761*** (0.021) (5) Q4 19902008 0.1753*** (0.021) (6) Ave 19902008 0.1752*** (0.021)

0.0327*** (0.004) 0.0402*** (0.004) 0.0342*** (0.004) 0.0367*** (0.004) 0.0293*** (0.006) 0.0111* (0.007) 0.1892*** (0.019) 0.1136*** (0.026) 0.0006 (0.002) 0.1410 (0.117) 0.3993*** (0.039) 0.1017*** (0.008) 0.0648*** (0.013) 0.0278 (0.018) 0.0020 (0.002) 0.5629*** (0.069) 3473 0.5615 66.9972 0.0000

assertion that UK rms have historically made less use of the public debt markets than their US counterparts and relied relatively more on the banks for their debt nancing leading to greater concentration of debt funding sources and therefore a higher level of sensitivity of leverage to bond market access and importantly to credit market conditions. 5.3.2.2. Credit conditions proxy: interest rate spreads. For our second measure of credit market conditions we use spreads on syndicated bank lending to UK rms. Campello, Lin, Ma, and Zhou (2011) point out that the syndicated loan market has become the largest source

of debt funding for companies in the last twenty years growing from $150 billion in 1987 to $1.7 trillion in 2007. The BOE indicate that syndicated bank lending in the form of term loans and revolving credit facilities has become a key source of debt nance for large UK rms (Bank of England, 2009e). For our measure of syndicated loan spread we use the average disclosed spread over references rates in the currency in which loan tranches are denominated, weighted by tranche size. This data is sourced from the BOE and is available from 2003. Fig. 7 in Section 3.3 shows a three to four year period before the nancial crisis during which spreads on investment-grade syndicated loans were relatively low, reecting in part the strong supply of

58

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

Table 16 Capital structure and investment grade syndicated loan spread, access is measured by the possession of a credit rating and by rm size: Pooled OLS. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 2003 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Access is measured by possession of a credit rating; by rated rms against the smallest 30% of rms; by the top 30% of rms against the smallest 30% of rms; and by the top 30% of rms against the remaining 70% of rms. The rst model presents the results from the original regression and is identical to model 1 in Table 7. Credit market conditions are measured by spreads on syndicated bank lending to UK rms. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all specications to control for industry-specic effects. Variables denitions are presented in Appendix, Table B. Standard errors (in parenthesis) are adjusted for heteroskedasticity and clustering by rms. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables Credit rating Rating Syndsp RatedSmall30 RatedSmall Syndsp Top30Bottom30 Top30Bottom30 Syndsp Top30Remaining70 Top30Remaining70 Syndsp Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market return GDP growth Constant Observations R-squared F test Prob > F 0.0130** (0.006) 0.0009 (0.008) 0.2262*** (0.031) 0.0987*** (0.033) 0.0240 (0.015) 0.0473 (0.170) 0.5903*** (0.065) 0.1287*** (0.011) 0.0974*** (0.016) 0.1509*** (0.037) 0.0068 (0.004) 0.5945*** (0.087) 1714 0.5672 52.5350 0.0000 0.0133** (0.006) 0.0007 (0.008) 0.2219*** (0.031) 0.0973*** (0.033) 0.0237 (0.015) 0.0509 (0.169) 0.5974*** (0.065) 0.1311*** (0.011) 0.0985*** (0.016) 0.1209*** (0.037) 0.0080* (0.004) 0.6014*** (0.087) 1714 0.5702 55.7519 0.0000 0.0399*** (0.013) 0.0262** (0.012) 0.2000*** (0.038) 0.0908 (0.055) 0.0069 (0.011) 0.1129 (0.201) 0.4061*** (0.092) 0.1028*** (0.017) 0.0894*** (0.026) 0.0425 (0.060) 0.0139* (0.007) 0.7650*** (0.178) 642 0.5632 30.3356 0.0000 0.0303*** (0.010) 0.0063 (0.010) 0.2232*** (0.039) 0.1255*** (0.040) 0.0084 (0.014) 0.0452 (0.188) 0.5508*** (0.086) 0.1212*** (0.013) 0.0850*** (0.026) 0.0194 (0.051) 0.0200*** (0.006) 0.7407*** (0.135) 1006 0.6086 61.6574 0.0000 (1) 20032008 0.0468** (0.020) (2) 20032008 0.0467** (0.020) 0.0244*** (0.004) 0.1647*** (0.058) 0.0328*** (0.006) 0.1227*** (0.040) 0.0370*** (0.005) 0.0572*** (0.019) 0.0296*** (0.004) 0.0194*** (0.007) 0.0011 (0.008) 0.2127*** (0.030) 0.0986*** (0.033) 0.0211 (0.014) 0.0683 (0.168) 0.6092*** (0.065) 0.1279*** (0.012) 0.0943*** (0.016) 0.0834** (0.039) 0.0082* (0.005) 0.6742*** (0.095) 1714 0.5756 71.7579 0.0000 (3) 20032008 (4) 20032008 (5) 20032008

credit and favourable macroeconomic conditions. From the end of the third quarter of 2007 we see the start of a sharp rise in spreads as the nancial crisis began. In this section we estimate Eq. (2) but this time we interact access with the syndicated lending spread. Similar to the loan ofcer survey variable, we standardise the spread variable. We estimate the model rstly using all rated and non-rated observations, secondly we restrict non-rated rms to those in the bottom 30% of our sample by the book value of total assets and thirdly we use measures of access based on rm size. The results are presented in Table 16. The results show

that the coefcient on the interaction term between access and the syndicated lending spread is positive and signicant, providing further evidence that demonstrates that access to bond markets as a determinant of leverage becomes more important during periods of credit market tightening. We nd that a one-standard deviation increase in the syndicated bank loan spread increases the difference in leverage ratios between rms with and without rating by 2.4% (model 2). When rated rms are benchmarked against the smallest 30% of rms, we nd that a one-standard deviation increase leads to a greater leverage difference of 3.3% (model 3). These results are

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

59

Table 17 Access and capital structure, access is measured by the possession of a credit rating or rm size: Panel RE and FE. The table presents estimates of Eq. (1) using annual data of UK listed non-nancial rms for the period from 1989 to 2008. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by the possession of a credit rating or size dummies (Top 30 against the remaining 70% of rms; Top 30 against bottom 30% of rms; rated rms against smallest 30% of rms). Access variable is interacted with the year dummies to measure the variation in the effect of credit rating over time. Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all RE specications to control for industry-specic effects. Variables denitions are presented in Appendix A. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1) RE Rated non-rated Rating Top30Remaining70 Top30Bottom30 RatedSmall 30 Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Tax paid Access_1989 Access _1991 Access _1992 Access _1993 Access _1994 Access _1995 Access _1996 Access _1997 Access _1998 Access _1999 Access _2000 Access _2001 Access _2002 Access _2003 Access _2004 Access _2005 0.0048** (0.002) 0.0333*** (0.004) 0.1906*** (0.012) 0.1127*** (0.014) 0.0033** (0.002) 0.1687** (0.080) 0.3636*** (0.018) 0.1119*** (0.005) 0.0640*** (0.007) 0.0233*** (0.006) 0.0585* (0.032) 0.0061 (0.026) 0.0457 (0.030) 0.0149 (0.028) 0.0117 (0.024) 0.0262 (0.023) 0.0084 (0.023) 0.0450* (0.025) 0.0439* (0.024) 0.0612*** (0.023) 0.0854*** (0.024) 0.0648*** (0.024) 0.0651*** (0.024) 0.1061*** (0.024) 0.0486** (0.024) 0.0359 0.0156*** (0.003) 0.0795*** (0.007) 0.1861*** (0.012) 0.1217*** (0.016) 0.0009 (0.002) 0.1833** (0.091) 0.3493*** (0.019) 0.1124*** (0.005) 0.0573*** (0.007) 0.0204*** (0.006) 0.0548 (0.034) 0.0003 (0.028) 0.0382 (0.030) 0.0067 (0.028) 0.0059 (0.024) 0.0150 (0.024) 0.0069 (0.024) 0.0285 (0.025) 0.0223 (0.025) 0.0380 (0.024) 0.0642*** (0.024) 0.0426* (0.024) 0.0425* (0.024) 0.0806*** (0.025) 0.0243 (0.025) 0.0087 0.0075*** (0.002) 0.0317*** (0.004) 0.1917*** (0.012) 0.1058*** (0.014) 0.0026* (0.002) 0.1395* (0.080) 0.3900*** (0.019) 0.1116*** (0.005) 0.0627*** (0.007) 0.0245*** (0.006) 0.0535*** (0.019) 0.0165 (0.020) 0.0553*** (0.019) 0.0116 (0.018) 0.0117 (0.016) 0.0002 (0.016) 0.0030 (0.016) 0.0233 (0.017) 0.0590*** (0.016) 0.0961*** (0.017) 0.1170*** (0.017) 0.0759*** (0.016) 0.0713*** (0.017) 0.1110*** (0.018) 0.0434*** (0.017) 0.0312* 0.0173*** (0.003) 0.0768*** (0.007) 0.1875*** (0.012) 0.1152*** (0.016) 0.0012 (0.002) 0.1591* (0.091) 0.3761*** (0.019) 0.1120*** (0.005) 0.0567*** (0.007) 0.0218*** (0.006) 0.0500*** (0.019) 0.0122 (0.019) 0.0504*** (0.018) 0.0067 (0.017) 0.0149 (0.015) 0.0060 (0.016) 0.0111 (0.015) 0.0127 (0.016) 0.0466*** (0.016) 0.0819*** (0.017) 0.1045*** (0.016) 0.0623*** (0.016) 0.0564*** (0.016) 0.0939*** (0.017) 0.0279* (0.016) 0.0138 0.0295*** (0.004) 0.0339*** (0.005) 0.1614*** (0.014) 0.0990*** (0.017) 0.0039** (0.002) 0.0625 (0.085) 0.3642*** (0.023) 0.0983*** (0.006) 0.0476*** (0.008) 0.0164** (0.007) 0.0175 (0.021) 0.0012 (0.021) 0.0354* (0.020) 0.0147 (0.019) 0.0176 (0.017) 0.0216 (0.018) 0.0201 (0.017) 0.0008 (0.018) 0.0430** (0.018) 0.0738*** (0.019) 0.1165*** (0.017) 0.0887*** (0.017) 0.0974*** (0.017) 0.1015*** (0.019) 0.0377** (0.018) 0.0241 0.0394*** (0.005) 0.0910*** (0.010) 0.1605*** (0.014) 0.0835*** (0.021) 0.0073*** (0.002) 0.1465 (0.099) 0.3391*** (0.024) 0.0988*** (0.006) 0.0450*** (0.009) 0.0147* (0.008) 0.0154 (0.021) 0.0035 (0.020) 0.0298 (0.019) 0.0186 (0.019) 0.0204 (0.017) 0.0270 (0.017) 0.0285* (0.017) 0.0107 (0.018) 0.0290* (0.017) 0.0581*** (0.018) 0.1030*** (0.017) 0.0716*** (0.016) 0.0772*** (0.016) 0.0828*** (0.018) 0.0194 (0.017) 0.0025 0.0103 (0.022) (2) FE Rating non-rated 0.0071 (0.023) 0.0161 (0.015) 0.0095 (0.015) 0.1063*** (0.021) 0.0900*** (0.026) 0.1267*** (0.029) 0.0277*** (0.004) 0.0276*** (0.006) 0.1323*** (0.014) 0.1090*** (0.019) 0.0017 (0.002) 0.1433* (0.084) 0.2415*** (0.023) 0.0790*** (0.006) 0.0413*** (0.009) 0.0200** (0.009) 0.0183 (0.034) 0.0063 (0.030) 0.0278 (0.033) 0.0090 (0.030) 0.0123 (0.027) 0.0147 (0.026) 0.0054 (0.026) 0.0144 (0.027) 0.0212 (0.027) 0.0338 (0.026) 0.0782*** (0.026) 0.0672*** (0.025) 0.0734*** (0.025) 0.0767*** (0.026) 0.0277 (0.025) 0.0100 0.0994** (0.044) 0.0338*** (0.005) 0.0587*** (0.012) 0.1343*** (0.015) 0.0977*** (0.025) 0.0040** (0.002) 0.1911* (0.103) 0.2208*** (0.024) 0.0801*** (0.006) 0.0343*** (0.009) 0.0176* (0.009) 0.0179 (0.036) 0.0093 (0.032) 0.0233 (0.035) 0.0111 (0.031) 0.0115 (0.027) 0.0111 (0.027) 0.0115 (0.027) 0.0071 (0.028) 0.0110 (0.028) 0.0220 (0.028) 0.0637** (0.027) 0.0498* (0.026) 0.0570** (0.026) 0.0633** (0.027) 0.0152 (0.026) 0.0033 (3) RE Top30 Remain 70 (4) FE Top30 Remain 70 (5) RE Top30 Bottom30 (6) FE Top30 Bottom30 (7) RE Rated Small30 (8) FE Rated Small30

(continued on next page)

60 Table 17 (continued) Variables (1) RE Rated non-rated (0.024) 0.0231 (0.023) 0.0346 (0.024) 0.0886*** (0.025) 0.0541*** (0.011) 0.0105*** (0.001) 0.2530*** (0.032) 6,211 0.4856 474 2934.4769 0.0000

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

(2) FE Rating non-rated (0.025) 0.0042 (0.024) 0.0055 (0.024) 0.0576** (0.026) 0.0467*** (0.011) 0.0111*** (0.001) 0.2606*** (0.032) 6,211 0.4165 474

(3) RE Top30 Remain 70 (0.017) 0.0232 (0.016) 0.0298* (0.016) 0.1093*** (0.018) 0.0440*** (0.013) 0.0097*** (0.001) 0.3000*** (0.034) 6,211 0.5000 474 3103.2498 0.0000

(4) FE Top30 Remain 70 (0.016) 0.0065 (0.016) 0.0127 (0.016) 0.0905*** (0.018) 0.0375*** (0.013) 0.0102*** (0.001) 0.2962*** (0.033) 6,211 0.4317 474

(5) RE Top30 Bottom30 (0.018) 0.0143 (0.018) 0.0326* (0.017) 0.1498*** (0.019) 0.0231 (0.020) 0.0077*** (0.002) 0.4988*** (0.047) 3,624 0.5564 411 2241.3301 0.0000

(6) FE Top30 Bottom30 (0.017) 0.0045 (0.017) 0.0127 (0.017) 0.1255*** (0.019) 0.0261 (0.020) 0.0088*** (0.002) 0.4750*** (0.047) 3,624 0.4765 411

(7) RE Rated Small30 (0.025) 0.0100 (0.025) 0.0280 (0.025) 0.1208*** (0.028) 0.0023 (0.020) 0.0097*** (0.002) 0.4577*** (0.047) 2,502 0.5192 332 1447.0146 0.0000

(8) FE Rated Small30 (0.026) 0.0038 (0.027) 0.0103 (0.027) 0.0970*** (0.030) 0.0016 (0.020) 0.0106*** (0.002) 0.4550*** (0.048) 2,502 0.4211 332

Access _2006 Access _2007 Access _2008 Stock market Return GDP growth Constant Observations R-squared Number of FIRMID Chi2 test Prob > Chi2 F test Prob > F

73.0831 0.0000

77.5994 0.0000

47.5970 0.0000

26.0109 0.0000

0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 -0.02 -0.04 Rated Firms vs. Non-Rated Top 30% vs. Bottom 30% Top 30% vs. Remaining 70% Rated Firms vs. Bottom 30%

Fig. 15. Access coefcient over time: xed effects results (19912008).

similar in magnitude to those estimated using the bank lending standards proxy for credit market conditions. 29

6. Robustness checks In this section, we test the robustness of our inferences to the use of alternative estimation methods. We do so by re-estimating our model using panel data estimation techniques, which allow us to control for the unobserved individual heterogeneity by using random or xed effects specications. This will allow us to check whether our results are robust to the inclusion of controls for time-invariant rm characteristics. In this way we can overcome any potential endogeneity problem, when there is an unobservable or missing variable that is correlated with the explanatory variable. We estimate models (1) and (2) using both random and xed effects and employ our credit rating and rm size proxies for access

29 In unreported analysis we also follow Bougheas et al. (2006) and Leary (2009) and use the spread between long-term and short-term interest rates. We calculate this interest rate spread using two measures of short-term interest rates, these being the base rate and the 3 month treasury bill rate, and for the long-term rate we use the monthly yield to maturity on 10 year zero bonds. Our results, available upon request, are qualitatively similar to those found using the syndicated loan spread.

to the debt capital markets. Table 17 presents results from estimating Eq. (1), where access is interacted with year dummies in order to determine whether the effect of access on leverage varies over time. The panel regression results are consistent with those in Tables 3 to 6 obtained from our pooled OLS regressions. In Fig. 15 we plot the coefcient on the access year-dummy interaction terms from the xed effect regressions and nd a very similar time series variation in our access coefcient to that seen from our OLS regressions in Fig. 14. The coefcients are slightly lower than those in the original OLS but remain economically and statistically signicant. In line with our earlier results the xed effect access coefcient peaks in 1992, 2000, 2003 and in 2008, all of which are years associated with relatively high levels of credit market tightening and hence reduced bank loan supply. Our second set of robustness checks are shown in Table 18, which presents the random and xed effects results from estimating Eq. (2), where credit rating is interacted with the loan ofcer survey lending standards indicator of credit conditions in the UK, EU and the US. The results from the panel regressions are again in line with the results obtained from the OLS regressions. The coefcients on the interaction terms between credit rating and the lending standards variables are positive and signicant across all specications. In models (7) and (8) of Table 18 we nd that a one-standard deviation increase in the

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

61

Table 18 Capital structure, bank lending standards in the US, EU,UK and syndicated loan spread, access is measured by the possession of a credit rating: Panel RE and FE. The table presents estimates of Eq. (2) using annual data of UK listed non-nancial rms for the period from 1990 to 2008 for models 14 and from 2003 to 2008 for models 510. The dependent variable is the ratio of total debt to the market value (MV) of assets. Total debt incorporates short-term debt and long-term debt. MV of assets is the sum of MV of equity and BV of total debt. Public bond market access is proxied by the possession of a credit rating. Credit Conditions are measured by average bank lending standards in the US, EU, and the UK (models 38) and by spreads on syndicated bank lending to UK rms (models 9 and 10). Firm size is the natural logarithm of MV of assets. Age is the natural log of rm age plus one. Protability is measured as return on invested capital (ROIC) which calculated as the sum of pre-tax prots and total interest charges divided by invested capital. Asset tangibility is calculated as the difference of total assets minus current assets divided by total assets. Market-to-book ratio of assets is MV of assets over BV of assets. Firm's spending on R&D is expressed as natural logarithm of the ratio of one plus R&D expenditure scaled by total assets. Riskiness of operations is calculated as equity volatility multiplied by the equity-to-asset ratio. Equity volatility is expressed as square root of number of trading days multiplied by standard deviation of natural log of the daily price growth rate. Past equity returns are calculated as natural log of share price at the end of the year over share price at the beginning of the year. Portion of short-term debt is calculated as the sum of short-term debt and current portion of long-term debt due to within one year divided by total debt. Average tax paid is calculated as income taxes over pre-tax income (taxable income). All variables are winzorised at 1% level in order to prevent potential outliers driving the results. All specications include annual stock market return and annual GDP growth rate to control for macroeconomic conditions. Annual stock market return is calculated as natural logarithm of FTSE at the end of the year over FTSE at the beginning of the year. Annual GDP growth rate is sourced from the IMF ofcial website. Industry dummies are included across all RE specications to control for industry-specic effects. Variables denitions are presented in Appendix A. ***, **, * indicate statistical signicance at 1%, 5%, and 10% levels respectively. Variables (1)Original RE 19902008 Rating Rating USAverage Rating EUAverage Rating EUUKAverage Rating Syndsp Size Age Protability Asset tangibility Market-to-book R&D spending Asset volatility Equity return Short-term debt Stock market Return GDP growth Constant 0.0062*** (0.002) 0.0367*** (0.004) 0.1975*** (0.012) 0.1226*** (0.014) 0.0026* (0.002) 0.1497* (0.081) 0.3478*** (0.018) 0.1097*** (0.005) 0.0635*** (0.007) 0.0702*** (0.011) 0.0103*** (0.001) 0.2399*** (0.033) 5,968 0.4782 474 2798.2699 0.0000 0.0179*** (0.003) 0.0877*** (0.007) 0.1930*** (0.012) 0.1337*** (0.017) 0.0020 (0.002) 0.1580* (0.093) 0.3339*** (0.018) 0.1098*** (0.005) 0.0560*** (0.007) 0.0634*** (0.010) 0.0109*** (0.001) 0.2447*** (0.033) 5,968 0.4166 474 0.0062*** (0.002) 0.0368*** (0.004) 0.1970*** (0.012) 0.1224*** (0.014) 0.0026* (0.002) 0.1488* (0.081) 0.3521*** (0.018) 0.1103*** (0.005) 0.0641*** (0.007) 0.0598*** (0.011) 0.0100*** (0.001) 0.2401*** (0.033) 5,968 0.4794 474 2821.9635 0.0000 0.0179*** (0.003) 0.0883*** (0.007) 0.1923*** (0.012) 0.1336*** (0.017) 0.0020 (0.002) 0.1593* (0.093) 0.3386*** (0.019) 0.1104*** (0.005) 0.0566*** (0.007) 0.0533*** (0.011) 0.0107*** (0.001) 0.2437*** (0.033) 5,968 0.4177 474 0.0121** (0.006) 0.0076 (0.009) 0.1712*** (0.022) 0.1385*** (0.030) 0.0324*** (0.006) 0.1042 (0.179) 0.3984*** (0.041) 0.1065*** (0.009) 0.0545*** (0.013) 0.1066*** (0.026) 0.0143*** (0.005) 0.5381*** (0.089) 1,714 0.5479 349 1333.7325 0.0000 0.0231*** (0.009) 0.1305*** (0.041) 0.1543*** (0.021) 0.1510*** (0.049) 0.0272*** (0.006) 0.3640 (0.229) 0.3438*** (0.041) 0.1041*** (0.009) 0.0425*** (0.013) 0.1282*** (0.026) 0.0171*** (0.005) 1.1393*** (0.173) 1,714 0.5749 349 0.0118** (0.006) 0.0074 (0.009) 0.1707*** (0.022) 0.1375*** (0.030) 0.0323*** (0.006) 0.1047 (0.179) 0.4002*** (0.041) 0.1064*** (0.009) 0.0543*** (0.013) 0.1076*** (0.026) 0.0148*** (0.005) 0.5351*** (0.089) 1,714 0.5485 349 1342.2845 0.0000 0.0225*** (0.009) 0.1274*** (0.041) 0.1538*** (0.021) 0.1493*** (0.049) 0.0272*** (0.006) 0.3635 (0.229) 0.3457*** (0.041) 0.1040*** (0.009) 0.0423*** (0.013) 0.1285*** (0.026) 0.0175*** (0.005) 1.1226*** (0.173) 1,714 0.5762 349 0.0511*** (0.006) (2)Original FE 19902008 0.0458*** (0.007) (3) RE 19902008 0.0512*** (0.006) 0.0127*** (0.004) (4) FE 19902008 0.0457*** (0.007) 0.0122*** (0.003) (5) RE 20032008 0.0376** (0.019) (6) FE 20032008 0.0228 (0.049) (7) RE 20032008 0.0380** (0.019) (8) FE 20032008 0.0227 (0.049) (9) RE 20032008 0.0365* (0.019) (10) FE 20032008 0.0235 (0.049)

0.0237*** (0.005)

0.0193*** (0.005) 0.0276*** (0.005) 0.0229*** (0.005) 0.0199*** (0.004) 0.0115** (0.006) 0.0072 (0.009) 0.1699*** (0.022) 0.1390*** (0.030) 0.0322*** (0.006) 0.1018 (0.179) 0.3954*** (0.041) 0.1075*** (0.009) 0.0538*** (0.013) 0.1384*** (0.027) 0.0088** (0.004) 0.5133*** (0.087) 1,714 0.5481 349 1354.7661 0.0000 0.0163*** (0.004) 0.0214** (0.009) 0.1205*** (0.039) 0.1531*** (0.022) 0.1518*** (0.049) 0.0274*** (0.006) 0.3561 (0.229) 0.3395*** (0.041) 0.1061*** (0.009) 0.0419*** (0.013) 0.1654*** (0.028) 0.0090** (0.004) 1.0606*** (0.165) 1,714 0.5743 349

Observations R-squared Number of FIRMID Chi2 test Prob > Chi2 F test Prob > F

185.9606 0.0000

174.3568 0.0000

73.6705 0.0000

74.3344 0.0000

74.5429 0.0000

combined EU and UK lending standards variable increases the difference in leverage between rated and non-rated rms by about 23%. This result is very close to the 3.1% we observe in the original OLS regression. Finally, we estimate our access and syndicated loan spread interaction regressions using panel estimation techniques. The results in models (9) and (10) of Table 18 show that we continue to nd that the coefcient on the interaction term between access and loan spread is positive, statistically signicant and not too dissimilar to our OLS results. In unreported regressions when access is measured by our various rm size and rating specications we get qualitatively similar results.30

Overall, the results from the panel regressions conrm our original nding that leverage differences between rms with and without access (rated and non-rated rms) increase (decrease) during periods of credit market tightening (loosening). Therefore, the conclusion that access to bond markets has a greater impact on leverage during periods of reduced bank loan supply is robust to the use of random and rm-xed effects estimation.

7. Conclusion This paper examines the factors that are important in determining capital structure choices for a large sample of UK rms over a twenty

30

These results are available upon request.

62

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863 Table B Variable denitions and predicted signs with leverage. Variables Denition Expected sign

year period 1989 to 2008. To the best of our knowledge, this study is the rst to consider the role bond market access has on inuencing leverage levels in a UK setting. Such access is potentially more important during a contraction in bank loan supply when rms without access to bond markets might nd themselves constrained in the amount of debt capital they can raise. The 20072009 nancial crisis and the resulting contraction in the ow of bank credit to the UK corporate sector demonstrated in unambiguous terms that access to diversied sources of debt funding had never been more important. The period we investigate is interesting as it exhibits signicant variation in macroeconomic and credit market conditions and more so than the periods examined in previous studies. We nd that access to bond markets as measured by the possession of a credit rating has an economically and statistically signicant effect on the level of leverage for UK rms over our sample period. More importantly, we nd that the leverage difference between rms with and without access is greater during periods of credit market tightening and smaller when credit conditions are loose. Our results show that debt market segmentation did not matter when credit conditions were loose, such as the period 2004 to 2007 prior to the nancial crisis, as banks were willingly supplying the UK corporate sector with the credit needed. However, in 2008 at the height of the nancial crisis, rms that could not switch from bank to bond market debt found themselves most severely nancially constrained. Consistent with Bank of England evidence, which reports that smaller rms have suffered disproportionately in terms of the cost and availability of bank credit during the nancial crisis, we nd that positive access effects on leverage are greatest when we compare rms with access against the smallest rms in our sample. Our results also suggest that the economic importance of possessing access to bond markets is greater in the UK than that seen in the US which is consistent with the notion that UK rms are over reliant on bank credit for their debt nancing needs. Survey and anecdotal evidence suggests that those rms that were heavily dependent on the banks for funding and who therefore found themselves nancially constrained during the 20072009 credit crisis were not been able to fund all their protable investment projects. However, those companies that were able to successfully source debt funding during the credit crisis, such as those with access to the bond markets, were in a strong position to fully exploit their investment opportunities. It follows that the credit crisis might have affected the values of these two groups of rms differently, with nancially constrained rms being forced to cut back on value enhancing investment whereas the investment plans of those with access to the bond markets not being affected by the contraction in bank lending. Future research should attempt to further our knowledge of the value implications of access to debt capital markets. Appendix A

Response variable: Market value gross leverage Total debt/(MV of equity + Total debt) Market value net leverage (Total debt Cash and equivalents)/(MV of equity+Total debt) Book value gross leverage Total debt /(BV of assets) Book value net leverage (Total debt Cash and equivalents) /(BV of assets) Predictors: Proxies for access Credit rating dummy Firm size dummy 1

Firm size dummy 2

Access dummy

Equal to 1 for rms with rating, and 0 otherwise Equal to 1 for rms in the top 20% (30%) of distribution by total BV of assets, and 0if in the rest 80% (70%) Equal to 1 for rms in the top 20% (30%) of distribution by total BV of assets, and 0if in the bottom 20% (30%) Equal to 1 for rated rms and 0 for rms in the bottom 20% (30%) of distribution by total BV of assets

+ +

Firm characteristics Firm size Firm size Firm age Firm age Protability Return on invested capital (ROIC) Pre-tax margin Asset tangibility Asset tangibility Growth opportunities Market-to-book of assets Firm's R&D expenditure Business risk Asset volatility Equity volatility

Natural Log (BV of total debt + MV of equity) Natural Log (rm age + 1)

(Pre-tax prot+ Total interest charges) / Invested capital Pretax income as percentage of total sales Total assets Current assets / Total assets MV of assets / BV of assets Natural Log of (1 + Research & Development Expense/Total assets) +

Equity volatility (Equity/MV of assets) No. of trading days Standard Deviation of [Natural Log of (Price day (t) / Price day (t 1)] Natural Log of (Share price end year / Share price beginning year) (Short term debt + Current portion of long term debt due to within 1 year) / Total debt Income tax / Pre-tax income

Past equity return Equity return Debt maturity Short-term debt

Taxes Tax paid Macroeconomic variables Annual GDP growth rate Annual stock market return

LN (Nominal GDPt / Nominal GDPt 1) % LN (FTSE at the end of the year / FTSE at the beginning of the year) %

Table A S&P credit rating categories. Investment grade High credit quality AAA AA+ AA AAMedium credit quality A+ A ABBB+ BBB BBBNon-investment grade BB+ BB BBCCC+ CCC CCCCC+ CC CCC+ C C-

Model includes 11 industries: 1. 2. 3. 4. 5. 6. 7. 8. Oil, mining, chemical and other commodities Telecom xed line Transport, shipping, freight Business support Engineering, general Medical supplies and pharmaceuticals Retail, soft goods, foods and drugs Media and leisure facilities

A. Judge, A. Korzhenitskaya / International Review of Financial Analysis 25 (2012) 2863

63

9. Water and electricity 10. Real estate 11. Software and computer services

References
Aggarwal, R., & Kyaw, N. A. (2010). Capital structure, dividend policy, and multinationality: Theory versus empirical evidence. International Review of Financial Analysis, 19, 140150. Aldrick, P., & Ebrahimi, H. (February 4, 2009). Corporate bond issues surge to boom levels Green shoots emerge as buyers are found for $246 bn of company debt. The Daily Telegraph (pp. 3). Bacon, G., Grout, J., & O'Donovan, M. (2009). Credit crisis and corporatesFunding and beyond. Report by The Association of Corporate Treasurers. [Electronic copy available at www.treasurers.org/creditcrisisimpact] Baker, M., & Wurgler, J. (2002). Market timing and capital structure. Journal of Finance, 57(1), 132. Bank of England (2007a). Credit conditions survey. [Quarter 3]. Bank of England (2007b). Credit conditions survey. [Quarter 4]. Bank of England (2008a). Credit conditions survey. [Quarter 1]. Bank of England (2008b). Credit conditions survey. [Quarter 2]. Bank of England (2008c). Credit conditions survey. [Quarter 3]. Bank of England (2008d). Credit conditions survey. [Quarter 4]. Bank of England (2009a). Trends in lending. Bank of England (2009b). Trends in lending. Bank of England (2009c). Trends in lending. Bank of England (2009d). Trends in lending. Bank of England (2009e). Trends in lending. Bank of England (2009f). Trends in lending. Bank of England (2009g). Trends in lending. Bank of England (2009h). Trends in lending. Bank of England (2010a). Markets and Operations. Quarterly Bulletin, Quarter 4 (pp. 242256). Bank of England (2010b). Section 1 Money and asset prices. Ination report. Bougheas, S., Mizen, P., & Yalcin, C. (2006). Access to external nance: Theory and evidence on the impact of monetary policy and rm-specic characteristics. Journal of Banking and Finance, 30, 199227. Byoun, S. (2008). Financial exibility and capital structure decision. Working paper. Baylor University [Electronic copy available at: http://papers.ssrn.com/abstract_id=1108850] Campello, M., Giambona, E., Graham, J. R., & Harvey, C. R. (2009). Liquidity management and corporate investment during a nancial crisis. Working paper. University of Illinois, University of Amsterdam and Duke University [Electronic copy available at: http://ssrn.com/abstract=1444009] Campello, M., Graham, J. R., & Harvey, C. R. (2009). The real effects of nancial constraints: Evidence from a nancial crisis. Working Paper. University of Illinois and Duke University [Electronic copy available at: http://ssrn.com/abstract=1318355] Campello, M., Lin, C., Ma, Y., & Zhou, H. (2011). The real and nancial implications of corporate hedging. Journal of Finance, 66(5), 16151647. Cantillo, M., & Wright, J. (2000). How do rms choose their lenders? An empirical investigation. Review of Financial Studies, 13, 155189. Chava, S., & Purnanandam, A. (2009). The effect of banking crisis on bank-dependent borrowers. Working Paper Series. College of Management and Stephen M. Ross School of Business [available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=821804] Churchill, N. (February 12). Loan market still has advantages, says LMA director. Credit. [available on-line at http://www.risk.net/credit] Cohen, N., & Goff, S. (August 31). The manager says no. Financial Times. Cole, R. (2008). What do we know about the capital structure of privately held rms? Evidence from the surveys of small business nance. Munich Personal RePEc Archive (MPRA) Paper No. 8086. [Electronic copy available at: http://mpra.ub.uni-muenchen.de/8086/] Davies, P. J. (August 11). Corporate funding getting scarce. : FT.com [available on-line at http://search.proquest.com/docview/229170211?accountid=12441] Deloitte Chief Financial Ofcer (CFO) Survey (2009a). Quarter 2, Available on-line at:. http://www.deloitte.com/view/en_GB/uk/research-and-intelligence/deloitte-researchuk/the-deloitte-cfo-survey/4d8b047c13872210VgnVCM200000bb42f00aRCRD.htm

Deloitte Chief Financial Ofcer (CFO) Survey (2009b). Quarter 3, Available on-line at:. http://www.deloitte.com/view/en_GB/uk/research-and-intelligence/deloitte-researchuk/the-deloitte-cfo-survey/9da69cfbbdd54210VgnVCM200000bb42f00aRCRD.htm Faulkender, M. W., & Petersen, M. A. (2006). Does the source of capital affect capital structure? The Review of Financial Studies, 19(1), 4579. Frank, M. Z., & Goyal, V. K. (2007). Capital structure decisions: which factors are reliably important? Working Paper. University of Minnesota [Electronic copy available at: http://papers.ssrn.com/abstract_id=567650] Frydenberg, S. (2011). Theory of capital structureA review, freedom and plurality Honorary thesis for odd. G. Arntzen, Lars Fallan and Ove Gustafsson, eds., Tapir Academic Press: NO-7005 Trondheim, Norway. [Electronic copy available at: http://papers.ssrn.com/abstract_id=556631] Hovakimian, A. G., Kayhan, A., & Titman, S. (2008). How do managers target their credit ratings? A study of credit ratings and managerial discretion. Working paper. Zicklin School of Business [Electronic copy available at: http://papers.ssrn.com/abstract_ id=1098351] Ivashina, V., & Scharfstein, D. (2009). Bank lending during the nancial crisis of 2008. Working paper. Harvard Business School [Electronic copy available at: http://ssrn. com/abstract=1297337] Jagger, S., & Power, H. (March 23). Corporate bonds step where banks fear to tread. The Times [Electronic copy available at: http://business.timesonline.co.uk/tol/business/ industry_sectors/banking_and_nance/article5956291.ece] Kayo, E., & Kimura, H. (2011). Hierarchical determinants of capital structure. Journal of Banking and Finance, 35, 358371. Kernan, P., Wade, T., & Watters, P. (September 21). Amid the changing banking landscape, U.K. corporates are embracing the capital markets. Standard and Poor's. Kisgen, D. J. (2006). Credit ratings and capital structure. Journal of Finance, 61(3), 10351072. Kisgen, D. J. (2007). The inuence of credit ratings on corporate capital structure decision. Journal of Applied Corporate Finance, 19(3), 6573. Kisgen, D. J. (2009). Do rms target credit ratings or leverage levels? Journal of Financial and Quantitative Analysis, 44(6), 13231344. Korajczyk, R. A., & Levy, A. (2003). Capital structure choice: Macroeconomic conditions and nancial constraints. Journal of Financial Economics, 68, 75109. Leary, M. T. (2009). Bank loan supply, lender choice, and corporate capital structure. Journal of Finance, 64, 11431185. Lemmon, M. L., & Roberts, M. R. (2007). The response of corporate nancing and investment to changes in the supply of credit. HKUST Business School Research Paper No. 0729. [Electronic copy available at: http://ssrn.com/abstract=961616] Levy, A. (2000). Why does capital structure choice vary with macroeconomic conditions? NYU Working Paper No. S-CDM-00-12. [Electronic copy available at: http://ssrn. com/abstract_id=1295832] Mallaby, S. (June 8). The radicals are right to take on the banks. Financial Times. Mitto, U. R., & Zhang, Z. (2008). The capital structure of multinational corporations: Canadian versus U.S. evidence. Journal of Corporate Finance, 14, 706720. Mitto, U. R., & Zhang, Z. (2010). Bond market access, credit quality, and capital structure: Canadian evidence. Financial Review, 45, 579602. Modigliani, F., & Miller, M. (1958). The cost of capital, corporation nance and the theory of investment. The American Economic Review, 48, 261297. Mortished, C. (22nd April). Corporate bonds make a comeback to ll gap. The Times. Myers, S. C. (1984). The capital structure puzzle. Journal of Finance, 39, 575592. Myers, S. C., & Majluf, N. S. (1984). Corporate nancing and investment decisions when rms have information that investors do not have. Journal of Financial Economics, 13, 187221. Rajan, R., & Zingales, L. (1995). What do we know about capital structure? Some evidence from international data. Journal of Finance, 50(5), 14211460. Sakoui, A., & Lee, A. (2009). European companies in record bond sales. [FT.com, 18 January 2009, available on-line at http://www.ft.com/cms/s/0/c72d296e-e58111dd-afe4-0000779fd2ac.html] The Treasurer (2001). The road ahead (pp. 5861). The Treasurer (2002a). Where will the opportunity knock next (pp. 6365). The Treasurer (2002b). What does future hold for loans (pp. 6669). Voutsinas, K., & Werner, R. A. (2011). Credit supply and corporate capital structure: Evidence from Japan. International Review of Financial Analysis, 20(5), 320334. Wooldridge, J. (2002). Econometric analysis of cross-section and panel data.