Credit Information Quality and Corporate Debt Maturity: Theory and Evidence * Marco Sorge The World Bank and Chendi Zhang University of Sheffield * The authors thank Stijn Claessens, Hans Degryse, Mark Flannery, Steven Ongena, Maria Fabiana Penas, Luc Renneboog and Peer Stein for valuable comments and suggestions. We are grateful to the FSDI team of the World Bank, in particular Stijn Claessens and Konstantinos Tzioumis, for kindly providing us with data. This research was partially funded by the Research Grants program of the Finance, Private Sector and Infrastructure Department, Latin America and the Caribbean Region, World Bank. Corresponding author: Marco Sorge, International Finance Corporation, 2121 Pennsylvania Ave NW, Washington, DC 20433, USA. Phone: +1 202 458 9064. Fax: +1 202 974 4310. Email: msorge@ifc.org 1. Introduction A common problem faced by many firms around the world is the scarce availability of long- term sources of funds. Exclusive reliance on short-term borrowing may expose companies to illiquidity risks and reduce their overall growth potential. To address these issues, many countries have embarked on policies promoting the development of long-term loan or bond markets with mixed results. However, while the negative implications of excessive short-term borrowing on growth and stability are well known (eg. Chang and Velasco, 2001 and Demirguc-Kunt and Maksimovic, 1998), there is no consensus on its underlying determinants and hence the main priorities for reform. Under various assumptions, the decision to borrow at short-term maturities has been modelled in the corporate finance literature as a solution to debt-related agency problems (Barnea, Haugen and Senbet, 1980), or justified as a disciplinary tool to limit moral hazard (Rey and Stiglitz, 1993), as the result of coordination failures among banks (Dewatripont and Maskin, 1995), driven by the fear of early project termination by uninformed investors (Von Thadden, 1995), or as the consequence of illiquidity problems and inadequate regulation and institutions (Diamond and Rajan, 2000). In a signalling framework under asymmetric information, firms with favorable insider information may distinguish their quality by issuing short-term debt and roll it over, provided issuing costs are sufficiently high (Flannery, 1986, and Diamond, 1991). This paper asks the question whether short-term lending is preferred by banks in credit markets where information about borrowers is relatively more opaque and there is a higher dispersion in the credit qualities of obligors. Most models in corporate finance analyze the choice of debt maturity from the perspective of the borrower, which chooses its debt maturity structure and optimizes from a variety of capital sources available in a developed country paradigm. Bearing in mind the credit constraints faced by firms in developing countries, this paper focuses on optimal debt maturity choice from the perspective of the lender. We look into the conditions under which banks would lend short or long term under imperfect information about heterogeneous borrowers. The value of short-term lending as a hedge against uncertainty is analyzed in a dynamic model both under incomplete but symmetric information and with asymmetric information. As a complement to the hidden-action model by Rey and Stiglitz (1993) in "Short-term contracts as a monitoring device", the hidden- 1 information framework proposed in this paper emphasizes the benefits of short-term contracts as a screening device in high-risk credit markets. Building on Von Thadden (1995), we do not assume that monitoring is perfect and the lender learns the firm's credit quality with certainty. Our paper models information acquisition as a screening process, through repeated short-term lending relationships, which may be more or less effective depending on the level of uncertainty in the micro-, macroeconomic as well as institutional environment. The results of the model are then tested on empirical data of both developed economies and developing countries around the world. Although most existing empirical studies on corporate debt maturity focus on individual countries (mainly the US), there is a growing literature on how institutional differences across countries influence maturity choices (see, Dermirguc-Kunt and Maksimovic, 1999, Giannetti, 2003, and Fan, Titman and Twite, 2006). All of the above papers mainly focus on legal institutions of countries. A recent study by Djankov, McLiesh, and Shleifer (2006) analyzes the benefits of credit bureaus for the growth of private credit. Our study contributes to this line of research by investigating the impact of institutions aiming at reducing credit information asymmetries on the structure of corporate debt maturity. In particular, we focus on public and private credit bureaus and accounting standards as determinants of debt maturity structure after controlling for the impact of legal institutions, financial development and other macro and micro factors. The main findings of the paper can be summarized as follows. From a theoretical standpoint, unlike in Hart and Tirole (1988), we find that it is possible for lenders to separate in equilibrium borrowers of different risk levels, under the assumption that higher-risk borrowers are more myopic, i.e. have higher time discount factors. In this case, repeated short-term contracting becomes an important mechanism for lenders to "learn" about the credit quality of borrowers. Our model shows that a high degree of information asymmetries in credit markets makes short-term lending the choice preferred by banks in equilibrium, although it may not be the socially optimal outcome. In the empirical part of the paper, we assemble a novel cross-country database to test the main predictions arising from the model. Our main finding is that better credit information (as proxied by the existence and coverage of private and public credit registries as well as by improvements in accounting standards) is associated with a higher share of long-term debt as a proportion of total corporate debt in both developed and developing countries. We also find that countries with more uncertain legal frameworks are characterized by higher short-term 2 debt. This suggests that short-term lending may be a valuable hedge against uncertainty from the lender's perspective. Furthermore, countries with a lower dispersion of firms' default probabilities are characterized by a higher ratio of short-term to total corporate debt. This is consistent with our theory of short-term lending being used by banks as a screening device when differences in default risks among firms are more opaque. Overall, our findings suggest that promoting institutions and policies to improve the quality of credit information around the world is an important prerequisite for increasing access of firms to long-term finance. The paper is organized as follows. Section 2 reviews the literature on the determinants of the optimal choice of debt maturity. Section 3 presents a two-period stylized model of a bank choosing loan maturity under incomplete but symmetric information between lender and borrower. Section 4 extends the analysis to a more general dynamic set-up with asymmetric information between lender and borrower, and discusses our main testable hypothesis. Section 5 introduces the data and empirical results, and Section 6 concludes and draws some policy implications. 2. The literature on debt maturity structure Since Stiglitz (1974) extended Modigliani and Miller's (1958) contribution to formally establish debt maturity irrelevance in perfect markets, the literature in corporate finance on debt maturity choices has identified a variety of imperfections in capital markets that can explain why the choice of maturity in fact matters.1 A number of theoretical studies explain why risky firms with long-term projects might borrow on a short-term basis in the presence of asymmetric information. Using a signalling framework, Flannery (1986) shows that firms with favorable insider information may distinguish their quality by issuing short-term debt and roll it over, provided issuing costs are sufficiently high. The model predicts that debt maturity is shorter when there are more information asymmetries and less risk. By incorporating liquidity risk into a framework similar to that in Flannery's model, Diamond (1991) shows that debt maturity is a non- monotonic function of risk ratings: the shortest maturity for both the lowest and highest risk ratings. Rajan (1992) analyzes how information asymmetries and bargaining power affect the 1For instance, theories of debt maturity have focused on the role of agency costs (see, Myers, 1977, and Barnea, Haugen and Senbet, 1980), tax (see, Brick and Ravid, 1985, and Lewis, 1990), and coordination failures among banks (Dewatripont and Maskin, 1995). Ravid (1996) provides a comprehensive survey. 3 choice between short- and long-term debts from arm's length lenders, and Diamond (1993) links the choice of maturity with the choice of seniority of debt contracts under asymmetric information. Using a hidden-action model, Rey and Stiglitz (1993) have further demonstrated the disciplinary role of short-term lending to resolve moral hazard problems. They show that short-term lenders have desirable incentives to exert control and invest in monitoring, and the possibility of not rolling over loans is an effective threat over firms. Furthermore, Von Thadden (1995) argues that monitoring by lenders helps overcome the short-term bias of corporate investment under asymmetric information. 2 In this paper, the choice of debt maturity is analyzed from the perspective of the lender and not from that of the borrower as common in the literature (see, e.g., Flannery, 1986). This is especially applicable to developing countries where the bargaining power in deciding maturity choice is more likely to be on the side of the banks than the firms which are credit constrained. As a complement to the hidden action model by Rey and Stiglitz (1993), the hidden-information framework proposed in this paper emphasizes the benefits of short-term contracts as a screening device in high-risk credit markets. Building on Von Thadden (1995), we do not assume that monitoring is perfect and the lender learns the firm's credit quality with certainty. Our paper models information acquisition as a screening process, through repeated short-term lending relationships, which may be more or less effective depending on the level of uncertainty in the micro-, macroeconomic as well as institutional environment. A number of empirical studies have focused on the impact of information asymmetries on the choice of debt maturity by firms within individual countries,3 or across countries.4 Using loan-level data for the US, Berger, Espinosa-Vega, Frame and Miller (2005) investigate the importance of information asymmetries and credit risk ratings for loan maturity choices. They find that information asymmetries reduce loan maturities and, consistent with Diamond (1991), the relationship between debt maturity and risk ratings is found to be stronger when 2 While earlier models have been primarily concerned with financing choices in closed economies, recent financial turmoil in emerging markets has stimulated research into the linkages between debt maturity, the term structure of interest rates and the possibility of self-fulfilling currency and banking crises (see, e.g., Kaminsky and Reinhart, 1999, Demirguc-Kunt and Detrgiache, 1998). Short-term debt has often been criticized as a source of financial instability. On the one hand, a number of theoretical studies show that the accumulation of short- term external debt is important in the generation of self-fulfilling financial crises (see Chang and Velasco, 2001, Rodrik and Velasco, 1999). On the other hand, Diamond and Rajan (2000) argue that the build-up of short-term debt in emerging markets is the consequence of the illiquidity and poor quality of investments in countries lacking adequate regulation and institutions. Detragiache and Spilimbergo (2002) find empirical evidence supporting this theory. 3See, Berger, Espinosa-Vega, Frame and Miller (2006) and Ortiz-Molina and Penas (2006), Barclay and Smith (1995), Guedes and Opler (1996) and Stoh and Mauer (1996) for the US studies. 4 See, Dermirguc-Kunt and Maksimovic (1999), Giannetti (2003), Fan, Titman and Twite (2006), and Schmukler and Vesperoni (2006). 4 information asymmetries are higher. Recent studies show that institutions aiming at reducing information asymmetries (such as public and private credit registries) foster the development of private credit markets (Jappelli and Pagano, 2002 and Djankov, McLiesh, and Shleifer, 2006), increase access to credit (Barron and Staten, 2003) and firm performance (Kallberg and Udell, 2003) and reduce non-performing loans and the costs of firm financing (Brown, Jappelli and Pagano, 2006). Galindo and Miller (2001) find that, in countries with more developed credit bureaus, firms face less severe financial constraints. This applies to large firms listed on the stock market as well as for small and medium-size companies. This paper contributes to the empirical literature by investigating the impact of institutions aiming at reducing credit information asymmetries on the structure of corporate debt maturity. In particular, we focus on public and private credit bureaus and accounting standards as determinants of debt maturity structure after controlling for the impact of the legal framework, financial development and other macro and micro factors. 3. Short-term lending as a hedge against uncertainty In order to understand how uncertainty about borrowers' credit quality may affect optimal debt maturity structure, we begin in this section with a two-period model of a bank choosing loan maturity under incomplete but symmetric information. This stylized framework will then be generalized in Section 4 to a multi-period case with asymmetric information. The intuition of the model in this section can be briefly summarized as follows. A bank with incomplete information about borrowers' credit quality compares two lending options: (i) committing to a non-renegotiable long-term contract versus (ii) retaining the flexibility of rolling over renegotiable short-term contracts. The latter option carries higher transaction costs but entails the possibility to exit the investment, thus limiting losses, should the borrower default in the first period. Intuitively, the more the uncertainty, the higher will be in equilibrium the bank's preference for renegotiable short-term contracts. From the lender's perspective, short-term lending acts as a hedge against uncertainty. 3.1 General set-up and payoff structure Assume a two-period risk-neutral setting with perfect competition. A bank faces the choice of lending its initial endowment of $1 to either a long-term project or a short-term project as 5 described below5. Alternatively, the bank could simply buy Treasury-bills yielding a risk-free interest rate normalized, without loss of generality, to equal zero. The long-term project lasts for two periods. At the beginning of period 1, it is common knowledge that the long-term project requires $1 bank finance to get started and, if successful, will yield a return equal to (1+ +)2 only at the end of the second period. At the outset, the bank does not know but knows that is a random number that will be drawn at the end of period 1 from a zero-mean uniform distribution U (-- , + ) with >0 and 0< < 1. Thus, is the expected per-period return of the project. At the beginning of period 1, the bank will lend $1 at the long-term per-period interest rate rL (0 rS the bank will Bank collects principal at interest rate rS. receive full principal and interest (1+ rS) and interest due (1+ rS)2 and reinvest in the project at rate rS. or recovers only (1-LGD) Otherwise it will recover only 1-LGD, in the event of default. and invest in T-bills for the 2nd period. Thus, the bank's expected payoffs from lending either long-term or short-term are:7 L = h( +rL)(1- LGD)2 +(1-h( +rL))(1+rL)2 (1) S = h( +rS )(1- LGD) +(1-h( +rS ))(1+rS )2 -c (2) where h = 1 2( +) and we have assumed no time discounting. The bank incurs a transaction cost c in the case of repeated short-term lending contracts. 7The default probability on the long-term project is: Prob { (1+ +)2 < (1 + r L )2 } = Prob { < r L - } = h ( +rL), where h is the probability density function. Similarly, the default probability on the short-term project is h ( +r S ). 7 3.2 The impact of higher volatility on the optimal debt maturity choice Based on the assumptions and pay-off structures illustrated in the previous subsection, a few preliminary conclusions can be drawn comparing the profits that the bank can expect from lending long-term vs. short-term. On the one hand, short-term lending appears more flexible in adjusting to extreme project outcomes. In case of unexpectedly low revenues from the projects (i.e. very low values of ), short-term loans are preferable as they allow the bank to exit the investment early and limit the losses. On the other hand, though, in a low-risk environment financing long-term projects can be more productive (in fact it is reasonable to assume that r L > r S) and will reduce transaction costs. In order to analyze the impact of return volatility on the bank's expected profits, we differentiate (1) and (2) with respect to , obtaining: L ( - rL )[(1 - LGD )2 - (1 + rL )2 ] (3) = 2 ( + )2 ( - rS )[(1 - LGD ) - (1 + rS )2 ] S (4) = 2 ( + )2 Under the restriction that r < and r < (i.e. debt service is below the expected returns of L S the projects)8, it is straightforward to show that both partial derivatives (3) and (4) are negative. For given interest rates9, a higher volatility () increases projects' default probabilities and consequently reduces lenders' expected profits for both long-term and short- term lending. Furthermore, given that (1-LGD)2 <1-LGD, partial derivative (3) will be higher in absolute value compared to (4). This is because the possibility of exiting the investment after a first-period default limits the losses in case of short-term lending. Summarizing, in a low-risk environment long-term lending can be more productive and associated with lower transaction costs. As increases, however, there is a higher probability of very bad states of nature (i.e. a higher probability mass gets shifted to the tails of the distribution of ). In this case, short-term lending becomes the superior choice as it allows banks to exit the investment early instead of having to continue lending to an unprofitable 8This restriction is easily satisfied by most realistic projects whose default probabilities (h(+rL) or h(+rS)) are lower than 50%. 9We are concerned here with only the short-term impact of volatility on lenders' profits. In the long run, a higher volatility of project returns will lead to a reallocation of banks' portfolio from long-term to short-term lending until a new equilibrium interest rate is reached. 8 project. What is driving the bank's preference for short-term lending as volatility increases is the option-like nature of short-term contracting (i.e. capping the downside risk by cashing out 1-LGD at the end of period 1). In this sense, short-term lending can be a valuable hedge against uncertainty. 4. Short-term lending as a screening device This section extends the previous analysis to a more general dynamic set-up with asymmetric information. When the probability of default is private information of the firms, banks may initially use short-term lending as a screening device until more information is known about the credit quality of the borrowers and it becomes safer to commit to long-term loans. The higher the degree of asymmetric information, the more screening will be necessary through short-term contracting. 4.1 Dynamic contracting with asymmetric information Consider a multi-period competitive screening game under asymmetric information. Banks are all identical whereas firms can be of two types: high default risk (qH) and low default risk (qL) with qH > qL.10 At the outset, the bank does not know which type of firm it is facing. However, it has prior beliefs that the firm is high risk with a probability H and is low risk with a probability 1- H. At the beginning of the period, each bank makes a take-it-or-leave-it offer to a firm of lending $1 at a given per-period interest rate rt. The firm will decide whether to accept or reject the offer, based on its private information about its own default risk q. In particular, in a discrete-time framework we can simply express ex-ante expected payoff at time t0 both for the bank (B) and the firm (F) as: B ,F = EQ0 t t t T B,F (5) =1 10In an arbitrage-free risk-neutral framework in which all securities are valued based on the risk-free interest rate i and the equivalent martingale measure Q (see e.g. Harrison and Kreps, 1979), we assume for simplicity that default is an exogenous stochastic event and define default as an unpredictable jump in a Poisson process with intensity q (see Duffie and Singleton, 2003). In other words, this implies that the probability that the firm defaults over an interval of time (t, t+), conditional on the firm not defaulting prior to t, is approximately q (for very small ) under the equivalent martingale measure Q. In addition, we assume that the risk-free interest rate i and default risk q are time-invariant (i.e. the firm's "type" remains the same throughout the game; see Besanko and Kanatas, 1996). 9 where: t (q) = e -(i+q)t (6) and EQ0 denotes the expectation, under the risk-neutral probability measure Q, conditional on the information available at time t=0. Finally, i is the risk-free interest rate. It is important to note that the discount factor ( ) is a function of q. In other words, it is common knowledge t that a high-risk firm will be more myopic, i.e. more impatient about capturing profit opportunities earlier as opposed to later (t (qH) < t (qL) at each time t>0; see Figure 2). From a microeconomic perspective, a key reason why a bank could prefer short-term loans as opposed to long-term commitments lies in the possibility of renegotiating the terms of the loan over time as its information set increases (see also Section 3). In other words, the bank will be able in this case to offer at the beginning of each period t (for t = 1,...,T) a different interest rate r , based on its beliefs which it can update given the entire history of the t t game up to time t. For these reasons, we will not attempt here to characterize the full set of equilibria that can be enforced with short-term contracts.11 We will look for separating equilibria, in which the bank, different from the case of "pooling", is able to extract information about the firm's type in the course of the game. Our interest lies in particular in the subset of so-called "Partial Revelation Equilibria" (PBE). As will be seen more precisely below, in this class of equilibria the firm only partially reveals her type in the early stages of the game and the bank fully discovers the risk level of the firm only at a later stage, which we will denote with F (= Full Revelation Stage).12 As information is revealed sequentially during the game, partial revelation equilibria offer a general set-up to assess the role of short-term contracts as a screening device.13 More precisely, we define the full revelation stage as the period F ( 0 K(qL).16 Putting all the pieces together, the pay-off of the firm (who knows her "type" q) can be written as: F= t xt [ Rt ­ rt ] T (7) t=1 The bank's expected pay-off is: B = T EQ0 { x t r t t - K t d t - F x t c d t } (8) t=1 t=1 where t [ (1-LGD) d t + LGD t ], the risk-free discount factor d t = e ­ i t, and , R, and K are all functions of q. We assume that short-term contracting implies a transaction cost (c) incurred in every period t F. Given that the bank in case of default will be able to recover only a portion (1-LGD) of the principal and interests, its risk is partially reduced. This intuitively explains why it will discount interest payments from the firm at a rate t , which is a weighted average between the riskless (d ) and the risky discount factors ( ) (see Duffie t t and Singleton, 2003). The costs of funding of each loan offer at time t (Kt) and transaction costs (c) incurred during the short-term lending phase (t F) are discounted using the risk- free discount factor (d t) since they are the same in both default and no-default states. 14 Firms with a higher probability of default tend to "gamble for resurrection" (Ed Kane). This assumption is in line with much of the literature that has applied the option pricing framework to characterize the risk-shifting incentives deriving from limited liability in a volatile financial environment. 15 In the symmetric information model presented in Section 3, we had considered a representative long-term project and a sequence of short-term projects, whose risk and returns were driven by a random variable and therefore unknown both to the bank and the firms running the projects. Here instead the risk/return profiles of the two types of firms, (qH, RH) and (qL, RL), are common knowledge. However, a firm's type, i.e. which firm belongs to which risk/return pair, is not observable by the bank as it constitutes private information of the firm. 16 For instance, an important source of funding costs is capital requirements for banks: lending to a higher risk firm is more costly for a bank as it consumes more capital. 11 Given the setup and payoff structure illustrated above, consider now the following strategies: r = dt t (KH + c) for all t = 1, 2, ...., F-1 before the firm's type is revealed (9a) tH r t = dt (KH ) for all t = F+1, F+2, ...., T if the firm has revealed to be high risk (9b) tH r t = dt (KL) for all t = F+1, F+2, ...., T if the firm has revealed to be low risk (9c) tL This means that interest rate offers in equilibrium will satisfy a zero profit condition under perfect competition (i.e. banks' profits in equation (8) equal zero for any t F, if interest rate r follows strategies in Equations (9a), (9b) and (9c)). Similarly, perfect competition among t firms will equalize the returns of high and low risk projects with their respective costs of funding (i.e. Equation (7) equals zero in equilibrium for any t F): RH,t = dt(KH + c)> RL,t for all t = 1, 2, ...., F-1 (10a) t H RH,t = dt (KH ) > RL,t= dt (KL) for all t = F+1, ...., T (10b) t H t L For all periods tF) and (2) being able to discriminate interest rate offers by risk type. 4.2 Main implications of the model and testable hypotheses In the above analysis, the contract between the bank and the firm has two components: an initial short-term screening component under asymmetric information and a long-term lending component under perfect information. In this subsection, we will examine how the timing of information revelation (early revelation vs. late revelation) affects the relative length of these two components. In addition, we will use the results obtained so far to analyze how the bank's optimal choice between long-term and short-term lending might differ from the social optimum depending on the timing of information revelation. Finally, we will discuss which factors influence the timing of information revelation (F) and thus debt maturity structure. This will allow us to draw a number of key implications from the model and outline our main testable hypotheses which will be analyzed in Section 5. As seen in Section 4.1, the later new information is revealed to the bank about the type of the firm it is facing (i.e. the larger is F), the longer the relative length of the period in which banks roll over short-term loans to screen borrowers. In contrast, the earlier new information is revealed to the bank about the firm's type (i.e. the smaller is F), the longer the relative length of the period under perfect information, and more long-term lending should be observed (Figure 3). 17 Admati and Perry (1987) use a similar concept of separating equilibrium, where the degree of impatience of firms determines in equilibrium the time delays in accepting offers by different types of firms. 13 Table 1 summarizes the main results of the models in sections 3 and 4 for various parameter assumptions and indicates whether banks' optimal debt maturity choices correspond also to the social optimum or not. Long-term lending is socially optimal as it maximizes total surplus ex-ante ( EQ0 {B +F }) by both reducing the socially wasteful transaction costs of rolling T over short-term loans and maximizing total availability of finance (i.e. ). In contrast, xt t=1 rolling over short- term lending is not socially optimal as it increases transaction costs of the economy and leads to the possibility of credit rationing.18 Thus, a key implication of our model is that short-term lending acts as a screening device and will be chosen in equilibrium by banks- although socially suboptimal - in case that credit information about the borrower is poor and hence the revelation process takes longer. Even though the asymmetric information model presented in this section has different setup from the symmetric information model introduced in Section 3, both models convey a similar message: uncertainty or the lack of information about borrowers' credit quality leads banks to prefer short-term lending in equilibrium, although it may not be the socially optimal outcome. We now consider a particular set of partial revelation equilibria in order to illustrate the determinants of the speed of the information revelation process (i.e. the determinants of F). We analyze the case in which the firm rejects the short-term loan offer each period (rt = RH,t for t * = R L if H < * (A4) = either R H or R L if = * H where * is the cut-off prior probability H that makes the bank indifferent ex-ante between lending at RH only in the event that the firm is high risk and lending at R L in either case, i.e. whichever the firm's risk level. This extends to the dynamic model of section 4.1 as follows. To support the strategies (9), (10) and (11) in section 4.1 as an equilibrium, the beliefs H t must be such that, at each time t *), the prospect of lending a large dollar amount at a high interest rate RH more than offsets the bank's concern for taking on a higher risk. By contrast, if the bank were to charge RL it would still attract also high-risk types, only that the interest rate charged would now be insufficient to ensure an adequate risk-adjusted return to cover the cost of funding high- risk firms. 36 To illustrate further the above restriction on beliefs, we focus on the case where the bank sees its offer at time t-1 rejected by the firm ( i.e. x t-1 = 0 ). With what (posterior) probability will it still believe at time t that -despite the rejection- the firm is high risk ? From Bayes' rule: t H ( xt 1 - t -1 (A6) H H t -1 -1= 0) = (1 - t -1) H + 1 - t -1 - H H t -1 L 1 t -1 Substituting (A6) back into (A5), we obtain the following restriction on beliefs: 1 - t -1 H (1 for all t RL . (A10) Given the expression for R* in (A9), Equation (A10) requires that: T -F t <1H (A11) t=1 ie. separation is possible provided that discount factors for high-risk firms are sufficiently low (or equivalently their risk level qH sufficiently high). 38 Table A1: Sample characteristics (average for the 1994-2004 period) Number Short-Term Private Credit Public Credit GDP Country of Years Debt /TD Bureau Cov. Registry Cov. Growth Inflation Argentina 11 48.3 49.8 15.3 2.0 4.7 Australia 11 32.4 74.3 0.0 2.9 2.6 Austria 11 51.4 31.5 0.8 1.9 1.8 Belgium 11 44.6 53.3 11.0 2.2 1.8 Brazil 11 47.7 43.7 3.4 3.1 44.8 Canada 11 30.8 82.4 0.0 3.4 2.0 Chile 11 38.8 22.6 21.5 4.8 4.4 China 11 69.9 0.0 0.1 9.0 3.9 Czech Republic 6 51.7 14.6 0.3 2.6 5.3 Denmark 11 41.2 5.9 0.0 2.0 2.2 Finland 11 31.7 10.1 0.0 3.4 1.4 France 11 46.3 0.0 1.2 2.2 1.6 Germany 11 45.2 70.8 0.4 1.3 1.5 Greece 11 60.6 8.8 0.0 3.0 5.2 Hong Kong, China 11 56.9 23.8 0.0 3.8 1.2 Hungary 7 59.3 1.2 0.0 3.5 12.8 India 11 38.8 0.0 0.0 5.0 6.3 Indonesia 11 52.1 0.0 0.2 3.3 13.6 Ireland 11 34.6 75.5 0.0 6.4 3.0 Israel 11 45.9 0.7 0.0 1.3 5.6 Italy 11 52.9 43.0 5.6 1.7 2.7 Japan 11 56.0 76.2 0.0 1.3 0.0 Korea, Rep. 11 56.5 80.7 0.0 5.6 3.9 Luxembourg 1 33.9 50.0 0.0 3.5 1.9 Malaysia 11 61.8 46.1 12.6 5.6 2.5 Mexico 11 36.4 34.7 0.0 2.9 14.6 Netherlands 11 40.9 54.0 0.0 2.3 2.3 New Zealand 11 27.5 83.3 0.0 2.3 2.1 Norway 11 25.4 94.9 0.0 2.6 2.1 Pakistan 11 54.1 0.3 0.0 1.3 7.1 Peru 7 51.8 14.2 9.6 4.2 5.7 Philippines 11 49.3 0.9 0.0 4.0 6.1 Poland 5 59.1 38.0 0.0 4.3 10.7 Portugal 11 44.5 2.9 50.9 1.9 3.3 Russia 6 53.3 0.0 0.0 2.2 48.9 Singapore 11 55.3 49.6 0.0 5.3 1.1 South Africa 11 48.0 48.4 0.0 3.1 6.4 Spain 11 46.9 5.0 31.3 2.5 3.2 Sweden 11 31.2 53.4 0.0 2.9 1.2 Switzerland 11 34.6 18.3 0.0 1.3 0.8 Taiwan, China 11 57.6 3.5 0.0 4.8 1.3 Thailand 11 54.1 10.2 0.0 3.3 3.4 Turkey 11 61.0 19.3 0.4 4.0 60.8 United Kingdom 11 42.5 68.4 0.0 2.9 2.7 United States 11 21.6 82.7 0.0 3.4 2.5 Average 10.2 46.1 33.3 3.7 3.2 7.9 39 Figure A1: Short-term debt and country risk This figure presents the average short-term debt to total debt ratios (STD/TD) of firms in each country for the 1994-2004 period. The Y-axis is STD/TD. On the X-axes are variables capturing country risk. 70 CHN MYSTUR 60 GRC POLHUN TWNHKG JPN KOR SGP RUS THA PAK ITA ) AUT CZE PER IDN %( 50 PHL ZAF BRA ARG DT/D FRA DEU ISR ESP PRT BEL GBR DNK ST NLD 40 CHL IND MEX CHE LUX IRL FIN SWEAUS CAN 30 NZL NOR USA 20 0 2 4 6 8 Corruption 70 CHN MYS 60 GRC POL HUN HKG TWN JPN KOR SGP THA PAK ITA ) AUT CZE PER IDN %( 50 ARG PHL ZAF BRA DT/D FRA ESP DEU ISR BEL PRT GBR DNK NLD ST 40 CHL IND MEX CHE LUX IRL AUS SWECAN FIN 30 NZL NOR USA 20 0 5 10 15 Inflation (%) 40 Figure A2: Short-term debt and microeconomic risk This figure presents the average short-term debt to total debt ratios (STD/TD) of firms in each country for the 1994-2004 period. The Y-axis is STD/TD. On the X-axes are variables capturing microeconomic risk. 70 CHN TUR MYS 60 GRC POL TWN JPN HKG KOR SGP THA PAK ITA ) AUT PER IDNCZE %( 50 PHL ZAF ARG BRA DT/D FRA ESP ISR BEL DEU PRT GBR DNK ST 40 NLD CHL IND MEX IRL CHE FINSWE AUS 30 CAN NZL NOR USA 20 5 10 15 20 25 30 Firm Default Prob. (%) 70 CHN TUR MYS 60 HUN POL JPNKOR HKG SGP RUS THA ITA ) PER CZE IDN (% 50 PHL ZAF BRA ARG DT/D FRA BEL DEU GBR ST 40 NLD CHL IND MEX CHE CAN SWE 30 USA 20 0 10 20 30 Non-Performing Loans (%) 41