The Determinants of Corporate Debt Ownership Structure: Evidence from Market-Based and Bank-Based Economies

This paper offers a comparative analysis of the determinants of the corporate debt ownership structure in a bank-oriented economy (Germany) and market-oriented economy (the UK). Using GMM estimators, we control for the problems of endogeneity, heteroscedasticity, normality, simultaneity and measurement errors that are common in firm-level panel data. The results show that the firms in both countries adjust their debt ownership structure towards their target levels - British firms being the swiftest. The findings confirm the validity of the liquidation and renegotiation hypothesis and the flotation cost hypothesis in both countries. However, the moral hazard and adverse selection hypothesis receives strong support in the UK but not in Germany. Moreover, the role of control factors (market related variables) in determining the choice of the lender is country dependent. Therefore, the debt ownership structure of a firm is influenced by both the firm-specific factors and the financial systems and corporate governance traditions in which the firm operates.


INTRODUCTION
Whilst the choice between public and private debts is known to have different implications on the value of the firms, observations show that the firms raise loans form both sources. However, the evidence on what determines the mix is sparse. James (1987) reports that the announcement of bank loan issuance results in positive abnormal stock returns while the issue of non-bank private debt or straight debt to repay bank debt affect the firms' stock prices inversely. Similarly, Datta et al. (2000) find a significantly negative share price response to public debt announcements. On the other hand, James (1996) shows that a mix of public and private debt allows distressed firms to alter their capital structure through non-court restructurings and hence the choice of mix is a complex phenomenon.
The literature offers three possible explanations. First, the ' liquidation and renegotiation' hypothesis postulates that the renegotiation of public debt is difficult, costly and is more likely to lead to a liquidation of distressed firms (Chemmanur and Fulghieri (1994)). Thus, firms with such risks are likely to avoid public debt. Second, the 'moral hazard and adverse selection' hypothesis suggests that the bank (monitored) loans are different from public debt as the banks have cost advantage in lending and hold more information about the prospects of the borrowing firms (Diamond, 1984). Fama (1985) argues that bank debt is like inside debt that may mitigate underinvestment problems due to information asymmetries. Therefore, the firms with potential agency conflicts are contended to benefit from issuing private (e.g. bank) debt rather than public debt. Finally, the 'flotation cost' hypothesis states that there are economies of scale in issuing substantial amount of public debt and hence only the larger firms are likely to benefit from the cost-advantageous public debt (Blackwell and Kidwell (1988)). This implies that the option of issuing public debt or bank debt is limited to the larger firms only. This paper investigates the potential determinants of issuing bank debt in the frameworks of the three hypotheses outlined above and extends the literature in several ways. First, these hypotheses were originally developed and tested in the US market. Very few studies on international experience allow for the possible implications of the country specific factors and environment on the choice of the source of debt 1 . This study offers a unique comparison of the evidence from a bank-based economy (Germany) and a market-based economy (UK) that should have direct implications on the choice between the bank debt and public debt 2 .
Confirming the bank-based tradition, Schmidt et al. (1999) find no orientation to capital markets in Germany. Mayer (1994) argues that dispersed corporate ownership in the UK is an obstacle to have a long-term relationship between firms and banks causing firms to rely on stock markets for external financing. Modigliani and Perotti (2000) argue that strong shareholders' protection makes equity markets greater and reduces the dominance of bank lending. These arguments offer further relevance to examining the corporate debt ownership structure in these two countries. We also control for the possible effects of relevant economic and financial factors of these countries. Therefore, this study offers an out of sample tests of the propositions developed in the US by providing unique comparative evidence from a bank-based economy and a market-based economy.
Second, factors affecting the firms' debt composition structure change overtime. Firms with long run optimal debt ownership target attain it through an adjustment process. We adopt an autoregressive-distributed lag model that allows us to examine the determinants of bank debt use, the speed of adjustment process to desired optimal bank-debt ratio and to provide the static long-run relationship between maturity and firm-specific factors. To our knowledge, this is the first debt ownership study to consider all three issues.
Finally, it is likely that random shocks affect both dependent and explanatory variables at the same time. It is possible that the observed relation between debt-mix and its potential determinants indicate the effects of debt-mix on the latter rather than vice-versa. We control for this important issue of endogeneity by using a GMM procedure. This also overcomes the problems of heteroscedasticity, normality, simultaneity and measurement errors that are known to be common in firm-level data. We are not aware of any other study on debt ownership that controls for endogeneity.

The Dependent Variables:
We measure the debt ownership structure by the ratio of bank debt to total debt (bank-debt ratio) 3 . We also examine the determinants of maturity structure of bank loans. They are defined as (a) short-term bank debt (payable within one year) to total debt (b) long-term bank debts (payable after one and after five years) to total debt.
2 See Antoniou et al. (2002) for a detailed discussion on the institutional and financial traditions of these countries. 3 Houston and James (1996), among many others, use this definition.

Explanatory Variables
We model the debt ownership structure as a function of the arguments representing the predictions of optimality of debt-mix structure, liquidation and renegotiation, moral hazard and adverse selection, and flotation (transaction) costs. We also control for several factors that represent the market conditions.

Lagged Dependent Variable
There are some potential costs (hold-up problems, monitoring costs, occurrence of inefficient liquidations) benefits (low moral hazard and adverse selection costs, flexible renegotiations) of bank financing. Berlin and Loeys (1988) argue that firms can obtain an optimal debt ownership structure by trading-off the inefficiencies of harsh bond covenants of public debt and the agency costs of hiring a delegated monitor for bank debt. In modelling the determinants of the outstanding debt ownership structure it is, implicitly, assumed that the firms have target bankdebt ratio. Inclusion of lagged bank-debt ratio in the model could be a benchmark to decide whether firms have an optimal debt ownership structure, and if any, the degree of divergence (or convergence) from (to) the target level may potentially be detected in the framework of adjustment costs. If the firms optimally determine their debt ownership structure the coefficient of lagged dependent variable should be (statistically) between 0 and 1. Considering the relatively lower cost of raising bank debt and lack of well-developed bond market in Germany, the German firms should have more motivation to adjust their debt ownerships structure more swiftly than the British firms who should incur higher transaction costs in adjusting public debt.

Liquidation and renegotiation:
The 'liquidation and renegotiation' hypothesis postulates that the renegotiation of public debt is difficult, costly and is likely to lead the liquidation of distressed firms (Chemmanur and Fulghieri (1994)). Thus, firms with such risks avoid public debt. A number of variables measure the potential risk of liquidity and the need for renegotiation. Schuhmacher (1998) shows that the choice of financing source depends on maturity need. Bank (public) debts are suitable for financing the short (long) term needs. Kanatas and Qi (2001) contend that bank debt is more suitable for financing assets with shorter economic lives. This is consistent with the notion of maturity matching hypothesis that firms match the maturity of liabilities with assets to prevent liquidation. Hence, we expect an inverse relation between debt maturity and bank-debt ratio. We measure the debt maturity by the ratio of long-term debt (maturing after one year) to total debt.

Leverage
It is argued that bank monitoring generates a public good that reduces the costs of public debt (see, e.g. Fama (1985)). As a result, higher bank debt may imply higher leverage due to complementary effect of bank debt on public debt. Berlin and Loeys (1988) argue that private debt (especially bank debt) provides more emphasis on monitoring than the public debt does.
Similarly, Hoshi et al. (1993) argue that firms with low leverage use public debt as they have higher incentive to take optimal investment decisions as they risk their net-worth while investing.
Thus, we expect a positive association between the bank-debt ratio and the leverage of the firm.
On the other hand, firms with higher debt ratios may restrict their bank borrowings to avoid their liquidations (Diamond, 1993). Banks' motivation to monitor on behalf of junior public debtholders may decline with the increase in public to private debts ratio. Hooks and Opler (1993) find that bank borrowing is highest among firms employing relatively little debt in their capital structure.
Hence, a negative relationship between bank debt and leverage is expected. We measure leverage by the ratio of book value of total debt to book value of total assets 4 .

Liquidity
Conventional wisdom suggests that the firms are likely to borrow from banks to mitigate their short-term liquidity problems, implying an inverse relation between liquidity and bank loan. Given the close association between firms and the banks in Germany, the liquidity factor should be less important for the German firms. The current ratio (current assets to current liabilities) represents the liquidity in our model.

Interest Coverage Ratio
Firms may get into financial distress if they failed to adjust themselves to adverse shocks.
Interest coverage ratio can be a proxy of the severity of financial distress (James (1996)). Chemmanur and Fulghieri (1994) theorise that firms with lower financial distress probably opt for public debt against bank debt since the lower interest cost of public debt outweighs the benefits of flexible renegotiations in bank debt. We thus expect a negative relationship between interest coverage ratio (a proxy for financial distress) and bank-debt ratio. The ratio of EBITD to total interest expense represents this variable in our model.

Moral hazard and adverse selection:
This hypothesis postulates that the banks have more information about the prospects of the borrowing firms and cost advantage in lending. Bank loan is like inside debt that may mitigate underinvestment problems due to information asymmetries. Therefore, the firms with potential 4 Alternative measure of leverage we use is the ratio of book value of total debt to market value of equity plus book value of total debt. agency conflicts should benefit more from issuing bank debt instead of public debt. The variables representing this argument are discussed below.

Growth Opportunities
Owing to information asymmetry outside investors are weakly informed of the firms' growth options and are concerned about agency problems. Hence, they demand for higher premium.
Negotiable bank debt is preferred to public debt in order to mitigate asset substitution and underinvestment problems (Berlin and Loeys (1988)). Blackwell and Kidwell (1988) argue that less risky firms are likely to issue public debt that contains less detailed restrictive covenants.
MacKie- Mason (1990) argues that R&D-intensive firms should avoid raising public debt due to information asymmetries between managers and outside investors. Yosha (1995) contends that firms with potentially valuable future growth projects will not borrow from public debt markets due to high disclosure costs of revealing sensitive information 5 . Thus, a positive association between growth opportunity (market-to-book ratio) and bank-debt ratio is anticipated.
On the other hand, Hoshi et al. (1993) show that firms with value-enhancing investment opportunities will use low cost public debt as it will be costly for such firms to forego positive-NPV projects because of high financing cost of bank debt. Houston and James (1996) argue that hold-up problems together with bank information monopolies may lead to a negative relationship between market-to-book ratio and the reliance on bank debt. Banks can hold up borrowing firms especially if the firms do not have alternative financing sources 6 . This implies an inverse relation between bank-debt ratio and market-to-book ratio. We measure growth opportunities (market-to-book ratio) as the ratio of (book value of total assets less book value of equity plus market value of equity) to (book value of total assets). This ratio can be viewed as a proxy of project quality perceived by the market (Johnson (1997)) 7 . Diamond (1991) showed that highly-rated firms issue public debt while medium and low quality firms issue bank debt 8 . The supply side argues that banks ration credits if they cannot distinguish between good and bad firms (Stiglitz and Weiss (1981)). Then, firms with high degree of information asymmetries and favourable information should issue private debt. On 5 Carey et al. (1993) note that US firms with takeover plans rely on private placement to protect the confidentiality of their transactions. 6 Houston and James (1996) document that high-growth US firms relying on a single bank and having no public debt issue have relatively low bank debt. This implies the presence of the hold-up problem and highlights the importance of diversification of debt financing sources for such firms. 7 Alternative measures of growth opportunities used are depreciation expense to total assets (Krishnaswami et al. (1999)) and the ratio of intangible assets to total assets. 8 Datta et al. (1999) provide empirical support to Diamond's reputation hypothesis. They find that the length of firmbank relationship significantly reduces the at-issue yield spread, thus, the cost of external debt.

Firm Quality
demand side, hidden-information view contends that firms will seek better-informed financier when the advantage of hidden information is substantial (MacKie-Mason (1990)). If private lenders are better informed than the public lenders it might be cost effective for firms with potential information asymmetries to borrow from private sources. James (1987) argues that firms disclose the terms of private financing to signal their true value to the market. Thus, the adverse selection hypothesis expects a positive association between firm quality and bank-debt ratio. We measure firm quality by abnormal earnings. Following Stohs and Mauer (1996) this is defined as the difference between earnings per share in years (t+1) and (t) divided by share price in year (t).

Dividend Payout Ratio
Dividend policy of firm is known to reveal information and hidden-information problem may be exacerbated for non-dividend paying firms. Therefore, the non-dividend paying firms are likely to avoid issuing public debt. Low et al. (2001) show that investors regard small firms' dividend decision as a function of bank monitoring. They show that market reaction to dividend omissions by small firms with high levels of bank debt is much less negative than that to by the firms with little or no bank debt. Thus, we anticipate an inverse relation between payout ratio and bankdebt ratio. We use the payout ratio as a proxy of dividend policy.

Earnings Volatility
Earnings are difficult to forecast and more difficult when its volatility is high. MacKie-Mason (1990) argues that managers are likely to have advantageous hidden information in while the earnings are volatile. Then, issuing public debt will be costlier as investors will stipulate high 'lemons' premia. This variable is also a proxy for observable credit risk and probability of financial distress (Johnson (1997)). Sy (1999) demonstrates that high credit risk firms' managers issue private debt due to the benefits of renegotiating tighter restrictions. On the other hand, low credit risk firms' managers will issue public debt due to benefits from increased flexibility. Thus, a positive relation between earnings volatility (proxy for potential information asymmetries) and the bank debt is expected. We measure earnings volatility as absolute annual percentage change in earnings minus average of this percentage change in the whole period.

Firm Size
Size of the firm can have several implications on the choice between public and private debts.
The costs of issuing public debt are considerably higher for small firms. Coase (1937) argues that deterrent transaction and contracting costs discourage small firms from raising external funding forcing them to rely on their retained earnings. Small firms have higher risk of getting into financial distressed and higher information asymmetries. Hence, they are likely to borrow short-term bank debts in order to avoid diseconomies of scale and the costs of financial distress.
Large firms are generally mature, less risky and have relatively low growth opportunities, thus low potential agency problems. Moreover, large firms tend to have better reputation and public information leading them to issue cost-efficient public debt (Diamond (1991)). Mayer and Alexander (1990) show that larger firms in the UK raise less bank loan. The size effect is more important in market-based countries because of higher cost of financial distress and lack of strong support from their banks. Thus, an inverse relation between bank-debt ratio and firm size is expected. We measure firm size in three ways, the natural logarithms of: i) total sales, ii) total assets, and iii) total assets minus book value of equity plus market value of equity.

Stock Return Volatility
It is likely that under uncertain market conditions firms may face difficulties in raising public debt.
Hadlock and James (2002) state that firms with high stock return volatility tend to have substantial information asymmetries between outsider and insiders. They find that undervalued firms tend to use bank debt. Thus, a direct relationship between bank-debt ratio and stock return volatility is expected. We measure the volatility by the standard deviation of weekly stock returns over the previous year, matched to the month of firms' fiscal year-end.

Change in Stock Price
MacKie-Mason (1990) contends that rising share price of a company implies that the investors are convinced about the improvement in the firm's prospects. In this situation, firms will be in advantageous position to borrow public debt at favourable terms. Thus, a negative relation is expected between bank-debt ratio and change in share price. We measure this variable as the first difference of log of annual share price, with a six-month lag 9 , matched to the month of firms' fiscal year-end.

Term Structure of Interest Rates
Kashyap et al. (1993) argue that tight monetary policies increase the cost of banks' capital, which in turn discourages firms from borrowing from banks. Oliner and Rudebusch (1996) contend that lenders would not be funding low-quality firms under such conditions. UK firms rely on capital markets while German f irms are argued to rely on bank financing. Thus, as highlighted by Bolton and Freixas (2000), the effect of monetary policies on corporate sector may not be similar across the sample countries. Mayer (1994) argues that the credit constraints 9 The lag of six months is to allow for a time gap between the decision making process and the issuance of debt.
have much more pronounced impact on real sector in bank-based economies than in marketbased economies. Our models, incorporating term-structure of interest rates, are expected to shed light on these issues. This is measured as the difference between the month-end yields on long-term government bond and three-month treasury-bills, with a six-month lag, matched to the month of firms' fiscal year-end.

Profitability
Hoshi et al. (1993) showed that firms with good performance are likely to issue public debt. If private debt mitigates agency costs of free cash flow, then, bank-debt ratio and profitability should not be significantly associated. Free cash flow problem should be inconsiderable in countries where share ownership is concentrated. On the other hand, since UK firms have the chance to tap the developed bond market they should not have significant financial constraints caused by fluctuations in internal resources. This implies that profitability and bank-debt ratio should not be significantly correlated in the UK either. According to screening theory of banking closely held or family owned firms are less likely to borrow from intermediaries (see Cantillo and Wright (2000)). Thus, the firms' profitability and bank-debt ratio should be negatively associated especially in Germany where ownership concentration is quite high. We measure firm profitability by the ratio of EBITD (earnings before interest, tax and depreciation) to total assets.

Fixed Assets Ratio
Boot et al (1991) predicts that the firms with potential collateral are likely to issue bank loan.
James (1996) emphasises that almost all bank debts of financially distressed firms are secured while public debts are rarely secured. In addition, Berger and Udell (1995) argue that some banks specialise in lending to the firms with substantial asymmetric information problems. These can be reflected in the nature of loan contract terms such as the rate of interest rate and collateral. Edwards and Fischer (1994) state that collateral seems to be one of the requirements for majority of bank loans in Germany and the UK. Then, the relation between fixed assets ratio and bank debt should positive. We use the ratio of net tangible assets to total assets as a proxy for asset collateral.

III. THE SAMPLE AND DATA
Motivated by the objective of comparing the determinants of debt ownership structure in a market-oriented economy and a bank-oriented economy we analyse the cases of British and German firms. It is generally accepted in the literature that the UK system is market-oriented while German system is bank-oriented. The sample includes all non-financial firms (dead or alive) traded in their domestic stock exchanges. The sample period, guided by the (un)availability of data, starts from 1969 for the UK, from 1987 for Germany and ends in 2000.
To allow for dynamic model estimation firms with less than three consecutive observations are excluded from the sample 10 . Data are obtained from Datastream International.
The descriptive statistics (table 1), as expected in a bank-oriented economy, show that German firms have substantially higher bank-debt ratio (BANK-TOTAL) (94.5 %) than that of the UK firms (60.4%). In Germany, the short-term bank debt payable within 1 year is 45.6% of the total debt (BANK-SHORT) while the long-term bank debt payable after five years (BANK-LONG>5) is only 15.8% of the total debt. This suggests that the German firms rely on banks primarily for short-term loan. On the other hand, the UK firms' average bank-debt ratio is 60.4 %. Comparing the ratios of long-term bank debt payable after one year to total debt (BANK-LONG>1) in Germany and the UK, one can see the discernible difference between the ratios as it is 48.9 % for the former and 14.7 % for the latter. This implies relatively lower contribution of the banks in financing the corporate sector in the UK. However, their importance in short term finance is noticeable.
Analysis of annual bank-debt ratios (not reported) reveals stability (1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000) in Germany and it ranges between 90% and 96%. However, this ratio has been declining since 1996 and takes its lowest value in 2000. The reduction is reflected in long-term bank debt ratios. On the contrary, short-term bank debt ratio is increasing. This trend implies that the German firms are switching from long-term debts to short-term debts but not at the same magnitude.
In the UK, there is an apparent increase in the long-term bank-debt ratio.

IV. THE MODEL AND METHODOLOGY
Unlike most previous studies on corporate debt ownership, this study is based on dynamic panel data. Advantages in using panel data, relative to period average cross-sectional data, include increase in the degrees of freedom with large number of observations, more variability and reduction in colinearity among regressors. These advantages provide more efficient estimations.
10 See Appendix for further details on sample selection.

The Model
We employ a dynamic model to investigate the potential determinants of corporate debt ownership structure (bank-debt ratio). In equation (1), a partial adjustment model, α (the constant), β, γ and δ are estimable unknown parameters; η i are firm-specific factors which do not change overtime (e.g. firm reputation), ω t are time-varying factors which do not change across firms (e.g. economic recession), ε it is the disturbance term which is assumed to be serially uncorrelated with mean zero and variance s 2 . (1)

Estimation Techniques
Recent developments show the superiority of the Generalised Method of Moments (GMM) technique in estimating the models such as equation (1). We apply two versions of the GMM technique (a) difference-GMM (GMM-DIF) and (b) system-GMM (GMM-SYS). In GMM-DIF technique, the model is estimated in first-differences using level regressors as instruments to control for unobservable firm heterogeneity. In GMM-SYS, the model is estimated in both levels and first-differences, i.e., level-equations are simultaneously estimated using differenced lagged regressors as instruments. In this way, apart from controlling for individual heterogeneity, variations between firms could partially be retained. Furthermore, for comparative purposes, we also employ fixed-effects (within groups) estimator 12 .
The primary motivation for analysing panel data is to control for unobservable firm heterogeneity. It is also rather difficult to establish exogeneity between the regressors and error term especially in company financial data. Thus, the direction of causality between variables might be ambiguous because of the potential endogeneity. Consequently, using the contemporaneous data for both dependent variable and its determinants may cause spurious results. In financing mix literature the endogeneity problem is either largely ignored or corrected for only using fixed effects or control variables approach. We control for this important problem by employing a GMM technique to avoid significant bias in estimates.

Optimal Debt Ownership Structure and the Speed of Adjustment
Static panel data models imply that the firms are able to adjust their financing structure without any delay 13 . However, due to adjustment costs a delay is highly likely. We investigate this possibility by adopting a partial adjustment process. Assume that the target debt ownership structure ( ) is a function of k explanatory variables as in equation (1). To search for the existence of target debt ownership structure in the framework of adjustment costs, the following procedure is followed. Assume that desired target level, where ? it is disturbance term serially correlated with mean zero and possibly heteroscedastic, and ? k 's are estimable unknown parameters which are common to each firms. The model assumes that firms adjust their current ratios, BANK DEBT it , with the degree of adjustment coefficient "?" to attain the target debt ownership structure.
A unit ? coefficient will suggest that the actual change in bank-debt ratio is equal to the desired change and firms have complete adjustment without transaction costs. However, ? = 0 implies no adjustment due to unaffordable transaction costs and firms will set their current bank debtratios to its past value. Thus, ? is inversely proportional to transaction costs. Substituting (2) into (3), we obtain the following equation: This adjustment model (equation 4) assumes that the adjustment coefficient ? lies between zero and unity because of the presence of transaction costs. If the cost of being in disequilibrium is higher (lower) than the cost of adjustment, ? tends to unity (zero). 13 We also account for this possibility in tables 6 and 7.

V. EMPIRICAL RESULTS
We estimate the above models on a number of dependent variables. For Germany, we have four different dependent variables viz. (a) total bank-debt ratio (b) short-term bank-debt ratio (c) long-term bank-debt ratio (payable after one year) and (d) long-term bank-debt ratio (payable after five years). The three dependent variables in the UK are (a) total bank-debt ratio (b) shortterm bank-debt ratio and (c) long-term bank-debt ratio. This difference is due to the unavailability of the breakdown of long-term debt by maturity in the UK. We also split UK data into two subsample periods : 1969-2000 and 1983-2000. The comparison of results between the UK and Germany is primarily based on the dependent variable of total bank-debt ratio.

Dynamic Debt Ownership Structure:
The GMM results in panel A of tables 2 to 5 reveal that our model captures the dynamics in firms' debt ownership decisions. The significant but less than unit coefficients of the lagged dependent variables (LDV) imply costly and non-instantaneous adjustment process 14 , 15 . The GMM estimates show a common pattern in the adjustment speed [?=1-(coefficient of LDV) in (4)] for both countries. The adjustment process is quicker for shorter-term bank debt. The adjustment coefficients of total bank-debt ratio displayed by the GMM-SYS (panel A of tables 4 and 5) indicate that the UK firms are quicker in adjusting the debt ownership structure towards their desired level than their German counter parts. The slower adjustment by German firms could be because of relatively high adjustment cost or the cost of being off the target is insignificant. Overall, the results show that firms in these countries attempt to trade-off between the cost of being off-target and the cost of adjustment.

Debt Maturity
As predicted by the liquidation and renegotiation hypothesis the relationship between debt maturity and bank-debt ratio is negative in both countries. It confirms that firms tend to borrow short-term loans from banking system.

Leverage
In the case of Germany only the lagged leverage in the short-run model is significantly negatively associated with the bank-debt ratio. This offers partial support to Drukarcyk et al. 's. (1985) view that debt-ratio is one of the most important factors in Germany to get bank loan. On the other hand, in the UK only the long-run multiplier is significantly negative for 1969-2000 14 As expected, GMM-DIF produces lower LDV coefficients than GMM-SYS does (see Blundell and Bond, 1998). 15 Cantillo andWright (2000), andHoshi et al. (1993) also report significantly positive estimated coefficients of LDV. (Table 5, panel B). Considering the leverage as a proxy of financial distress, these results are not surprising in terms of prevailing bankruptcy rules. German bankruptcy laws allow for liquidation rather than reorganisation and UK laws may cause premature liquidation. These findings support Diamond's view that firms with high debt ratios may restrict their bank borrowing to avoid liquidation and frequent renegotiations 16 .

Liquidity
The estimates show that the reliance on bank debt is inversely associated with firms' liquidity in the UK (tables 3, 5 and 7; and especially for the period 1969-2000). Hence, it seems that liquid firms avoid borrowing from banks possibly due to hold-up problems (in the UK) or monitoring costs, which are not specific to arm's-length public debt. It is also possible that higher internal liquidity reduces the need for short-term loans which primarily come from banks.

Interest Coverage
The interest coverage ratio does not affect firms' debt ownership decisions in both the UK and Germany 17 at any meaningful level It is possible that the firms in both countries have very high coverage ratio (table 1) and hence this factor is not considered as a significant risk factor by the managers.
Overall, the 'liquidation and renegotiation' hypothesis that predicts the renegotiation of public debt is difficult, costly and is more likely to lead the liquidation of distressed firms and hence firms with such risks are likely to opt for bank debt receives strong support from both the UK and Germany.

Growth Opportunities
The estimates reveal the absence of a non-monotonous relation between market-to-book ratio (MTBR) and bank debt use in Germany. We also find that there is no linear relation between MTBR and the use of bank debt in this country. It seems that costs related to information asymmetries, agency conflicts and monitoring, and hold-up problems are not widespread in Germany. This could be due to their corporate governance structure that largely mitigates agency and asymmetric information problems 18 . On the other hand, a U -shaped relation between bank debt ratio and MTBR is found in the UK.  B (1969-2000) records significant coefficient with correct sign in the cases of 'short-term bank debt ratio'. Hence, one may state that low-growth and high-growth firms borrow from banks due to different reasons while medium-growth firms opt for public debt markets. With respect to the GMM-DIF results in Table 3 (panel B), the coefficient of MTBR 2 is insignificant while MTBR is significantly negative 19 . This strong negative relation implies that banks focus their lending on tangible assets of the firm but not on the firms with intangible growth opportunities (supply side).
Alternatively, firms with profitable growth opportunities restrict their bank borrowings due to potential hold-up problems (demand side) 20 .

Firm Quality
The impact of firm quality on debt replacement decisions seems to differ across countries. For German firms it is insignificant while it has a significant positive effect on the British firms' choice of the source of debt (Table 5,  . The observed positive relation is possible if the undervalued firms with better future prospects borrow from better informed private lenders (banks). This largely mitigates the information asymmetry problem and the firms are revalued to their equilibrium levels. On the other hand, the insignificance effect of firm quality in shaping debt ownership structure of German firms may be due to their existing close relation with the banks and their borrowing tradition.

Dividend Payout Ratio
The association between bank-debt and dividend payout ratios is country-dependent (tables 2 and 3). It is significantly negative in the UK and significantly positive in Germany. This implies that firms in the UK avoid the adverse consequences of issuing public debt when they cut dividend payments. Hence, the information content of paying dividends with respect to firms' growth prospects and future cash stream seems to be prevalent in the UK. On the other hand, the positive relation in Germany remains a puzzle where banks are b oth financiers and shareholders. But why firms pay dividends (essentially to the banks themselves) and then borrow from the banks remains a puzzle.

Earnings Volatility
The GMM-DIF results show that the earnings volatility has a significant negative effect in the cases of UK firms (table 3)  Overall, the discussion above offers mixed support to the moral hazard and adverse selection hypothesis. It receives strong support from the UK but not from Germany. The evidence suggests that the corporate governance, the level of development of public debt market and the degree of information asymmetry affects the choice of the lender.

Firm Size
Existing studies almost uniformly report that small firms borrow from banks. Our results for both Germany (table 4b) and the UK (table 3b) confirm the significant negative relation between firm size and bank loan. Hence, the flotation costs hypothesis that small firms avoid public debt due to high flotation costs receives strong support 21 . Furthermore, the argument that small firms are immature, riskier, and have relatively high growth options, thus, tend not to borrow from public debt market are also confirmed.

The control factors:
The GMM-SYS results (tables 4 and 5) show that the association of bank-debt ratio with stock return volatility varies across countries in a pattern similar to earnings volatility. It is interesting to note that the UK firms with low volatile stock prices are more likely to issue bank debt. An increase in share price (Change in stock price) may represent firms' quality and convince the public debt-holders about their future prospects. However, our findings reveal that the association of bank-debt ratio with share performance seems to be insignificant in both Germany and the UK. The estimates show that the association of bank debt use with term structure of interest rates is country-dependent. In Germany, it has a significant positive coefficient (Table 4. panel A) while it is negative in the UK (tables 3 and 5). Hence, one can conclude that when longterm interest rates are relatively higher the UK firms are reluctant to raise bank debt while German firms tend to issue bank debt. The relation between profitability and bank-debt ratio is insignificant in both Germany and the UK. Similarly, the static long-run results show that asset collateral does not seem to be important in choosing debt provider in any of the sample countries. However, some differences in short-run are detected. The coefficient on lagged fixed- 20 The quality results remain the same with alternative measures of growth opportunities (depreciation ratio and intangible-assets ratio). 21 When the firm size is measured by Ln(Total Sales) the quality of results remain the same. assets ratio is significantly positive in the UK. The current fixed-assets ratio coefficient is positive in Germany (Table 4, panel A) and negative in the UK (panel A tables 3 and 5, 1969-2000). The positive impact of lagged collateral in the UK and current collateral in Germany could be due to banks' requirements for collateral while lending 22 .
In general, estimates above provide strong empirical support to the liquidation and reorganisation hypothesis and the flotation costs hypotheses in both the UK and Germany.
However, the moral hazard and adverse selection hypothesis receives support only from the UK evidence.

Maturity Structure of Bank Debt
We replaced the dependent variable total bank-debt ratio by short-term and long-term bank debt ratios and re-estimated the models 23 . For Germany, the long-run results (panel B, tables 2 and 4) show that the choice of dependent variable is important. The models using the dependent variables BANK-SHORT and BANK-LONG>1 have much stronger explanatory power of determinants of the debt ownership structure. Another notable finding is that the coefficient on share performance is significantly positive under both BANK-LONG>1 and BANK-LONG>5.
Furthermore, as predicted, the coefficient of interest coverage is significantly negative for BANK-LONG>1 and BANK-LONG>5 (panel B Table 4).
The estimates of static long-run model for the UK (panel B tables 3 and 5) reveal that the results are less sensitive to the choice of dependent variable. The robustness of the results to the choice of the dependent variable is especially noticeable in table 5 panel B (1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000). The expected negative relation between bank-debt and share price change can be seen under BANK-SHORT for the period 1969-2000. In addition, a positive association between bank-debt ratio and earnings volatility is found (Table 3, panel B, 1983-2000). Finally, the implications of term structure of interest rates, leverage, interest coverage, liquidity, payout ratio and firm size remain the do not change. Overall, the estimates for the UK are more robust than for Germany. This shows that the validity of the theories of debt ownership developed from the US market are sustainable in the UK than in Germany indicating a prominent role of financial structure of the country.

Static debt ownership structure
The estimates of static debt ownership structure model are presented in tables 6 and 8 for Germany and in tables 7 and 9 for the UK. Recall that this model assumes the target debt ownership structure is adjusted instantaneously in response to random changes in the macroeconomic environment and conditions of the firm. In other words, we stipulate that there is no lag in adjustment process toward an optimal debt ownership structure. In general, the findings based on static and dynamic models for Germany tend not to conflict with each other and the implications seem to be similar. However, one should be cautious about the results for the UK as there are substantial differences between the implications of static and dynamic models especially in terms of the number of the significant variables. This implies the need for dynamic modelling. Overall, as indicated by R 2 and Wald Statistics of joint significance of variables dynamic models have better explanatory power than the static models.

Fixed-effects Estimates (WG)
Probably the most appropriate way to compare GMM and WG estimates using the dependent variable BANK-TOTAL is to examine static GMM-SYS and fixed-effects results 24 . For Germany, the results are similar except the significant negative coefficient of quality of the firm in WG estimate. In the UK, the results of both estimation procedures with respect to the variables 'maturity, firm size, profitability, fixed-assets ratio, coverage ratio, and term structure of interest rates' are the same. In each case, however, WG estimator produces more significant variables than the GMM estimator does. These findings highlight the importance of controlling for endogeneity, as the results are sensitive to econometric specifications.

VI. CONCLUSION
The factors affecting the corporate debt ownership structure in Germany and the UK are Third, we find a U-shaped relation between the use of bank-debt and market-to-book ratio in the UK suggesting only the firms with very low-growth and high-growth opportunity prefer to borrow from banks. Firms with medium growth opportunities prefer public debt. Fourth, the moral hazard and adverse selection hypothesis receives strong support in the UK but not in Germany. Firm quality and bank-debt use are positively correlated in the UK while this association is insignificant in Germany. Hence, the concerns about obtaining a better-informed lender to mitigate adverse selection problems are not significant for German firms but very important for the UK firms. In addition, UK non-dividend paying firms or firms cutting their dividends avoid issuing public debt.
Similarly, earnings volatility does not affect debt decisions of German firms but it inversely affects bank-debt ratio of the UK firms. Fifth, confirming the predictions of the flotation cost hypothesis the evidence suggests that the smaller firms prefer bank loan against public debt in both countries. This also confirms the related arguments that small firms are riskier, immature and benefit more from monitored debt. Finally, the estimates reveal importance of several market related control factors in the model. The stock return volatility reduces the amount of bank debt used by UK firms, relatively high (low) long-term interest rates inspires the German (British) managers to borrow from bank. Overall, the findings confirm that the debt ownership decision of listed firms is not only the result of their own characteristics but also the outcome of legal and financial environment and corporate governance traditions in which they operate. Notes: BANK-TOTAL is the ratio of total bank debt to total debt. BANK-SHORT is bank debt payable within one year; BANK-LONG>1 is bank debt payable after one year; BANK-LONG>5 is bank debt payable after 5 years; all scaled by total debt. MATURITY is the ratio of debt that matures in more than one year to total debt. LEVERAGE1 is the ratio of book value of total debt to book value of total assets. LEVERAGE2 is the ratio of book value of total debt to market value of equity plus book value of total debt. MARKET-TO-BOOK is the ratio of book value of total assets less book value of equity plus market value of equity to book value of total assets, matched to the month of firms' fiscal year-end. DEPRECIATION is the ratio of depreciation expenses to total assets. INTANGIBLES is the ratio of intangible assets to total assets. SIZE1 (SIZE2) is the natural logarithm of total sales (total assets). SIZE3 is the natural logarithm of total assets minus book value of equity plus market value of equity. LIQUIDITY is the ratio of current assets to current liabilities. QUALITY is the difference between earnings per share in years (t+1) and (t) divided by share price in (t), matched to the month of firms' fiscal year-end. DIVIDEND is the dividend payout ratio; dividends to net earnings. PROFITABILITY is the ratio of EBITD to total assets. FIXED-ASSETS is the ratio of net tangible assets to total assets. COVERAGE is the ratio of EBITD to total interest expense. EARNINGS VOLATILITY is absolute annual % change in earnings minus average of this change. RETURN VOLATILITY is the stock return volatility measured by the standard deviation of weekly stock returns over the previous year, matched to the month of firms' fiscal year-end. SHARE PERFORMANCE is the first difference of log of annual share prices, with a six-month lag, matched to the month of firms' fiscal year-end. TERM is term structure of interest rates measured as the difference between the monthend yields on long-term (10 years or more) government bond and three-months treasury-bills, with a sixmonth lag, matched to the month of firms' fiscal year-end.  1989-1999 1990-1999 1990-1999 1989-1999 Table 1 for variable definitions. Correlation 1 and 2 are first and second order autocorrelation of residuals, respectively; which are asymptotically distributed as N(0,1) under the null of no serial correlation. Sargan Test is test of the overidentifying restrictions, asymptotically distributed as ? 2 (df) under the null of instruments' validity. Wald Test-1 tests the joint significance of estimated coefficients; asymptotically distributed as ? 2 (df) under the null of no relationship. (*), (**) and (***) indicates that coefficients are significant or the relevant null is rejected at 10, 5 and 1 percent level, respectively.  1971-1999 1971-1999 1971-1999 1985-1999 1985-1999 1985-1999 1989-1999 1989-1999 1989-1999 1989-1999 Table 1 for variable definitions. Industry dummies are included in all models. Correlation 1 and 2 are first and second order autocorrelation of residuals, respectively; which are asymptotically distributed as N(0,1) under the null of no serial correlation. Sargan Test is test of the overidentifying restrictions, asymptotically distributed as ? 2 (df) under the null of instruments' validity. Wald Tests 1 and 2 test the joint significance of estimated coefficients, and of industry dummies, respectively; asymptotically distributed as ? (df) under the null of instruments' validity. Wald Tests 1 and 2 test the joint significance of estimated coefficients, and of industry dummies, respectively; asymptotically distributed as ? -6.528*** -16.7*** -5.073*** -7.638*** -14.06*** -11.33*** Correlation2 -5.903*** -7.461*** -5.371*** -5.755*** -6.089*** -6.895*** Sargan Test (df) 442.2 (433) 442.4 (434) 320.1 (294) 349.8 (354) 324.5 (314) 329.4 (314) Wald Test-1 (df) 685.6 (15)*** 154.3 (14)*** 248.2 (14)*** 105.1 (15)*** 83.7(14)*** 111.6 (14)*** Wald Test-2 (df) 133.6 (15)*** 328.6 (15)*** 22.7 (15)* 242.5 (15)*** 308.3 (15)*** 37.61 (15) 1971-1999 1971-1999 1971-1999 1984-1999 1984-1999 1984-1999 See notes in Table 1 for variable definitions. Industry dummies are included in all models. Correlation 1 and 2 are first and second order autocorrelation of residuals, respectively; which are asymptotically distributed as N(0,1) under the null of no serial correlation. Sargan Test is test of the overidentifying restrictions, asymptotically distributed as ? 2 (df) under the null of instruments' validity. Wald Tests 1 and 2 test the joint significance of estimated coefficients, and of industry dummies, respectively; asymptotically distributed as ?  1988-1999 1988-1999 1988-1999 1988-1999 See notes in Table 1 for variable definitions. Wald statistics tests the joint significance of estimated coefficient; asymptotically distributed as ?