Asymmetric Partial Adjustment towards Target Leverage: International Evidence 1

Transcription

1 Asymmetric Partial Adjustment towards Target Leverage: International Evidence 1 Viet Dang, 2 Ian Garrett, 3 and Cuong Nguyen 4 Manchester Business School Abstract Employing asymmetric partial adjustment models of leverage that take into account both costs of deviations from target leverage and costs of leverage adjustment and financial constraints, we find evidence that for firms in France, Germany, Japan, the UK and US there is asymmetry in the speed wh which they adjust toward their target leverage. The speed is different according to whether firms have a financing surplus or defic and whether they are under- or over-levered. In particular, we find that firms that have a financing defic and are over-levered adjust towards their target leverage fastest. We also find that firms that tend to adjust more quickly towards their target leverage have lower profabily and growth opportunies, fewer tangible assets and are smaller in size. 1 We would like to thank participants at the Manchester Business School Doctoral Conference 2010 and especially Arif Khurshed, Maria Marchica, Roberto Mura, Norman Strong and Martin Walker for helpful comments and suggestions on earlier drafts of this paper. Any remaining errors are our own. 2 Vietanh.Dang@mbs.ac.uk. Manchester Business School, Universy of Manchester, M15 6PB, UK. 3 Ian.Garret@mbs.ac.uk. Manchester Business School, Universy of Manchester, M15 6PB, UK. 4 Corresponding author. Cuong.Nguyen@postgrad.mbs.ac.uk. Manchester Business School, Universy of Manchester, M15 6PB, UK. 1

2 I. Introduction A recent theme in the capal structure lerature has been the extent to which, and the rate at which, firms adjust their actual leverage towards their target leverage (see Fama and French (2002), Flannery and Rangan (2006), Antoniou, Guney, and Paudyal (2008), Faulkender, Flannery, Hankins, and Smh (2010), among others). Testing the dynamic tradeoff theory, which states that firms reach their optimal levels of leverage at the point at which the marginal costs of debt financing (such as financial distress costs and agency costs) equal the marginal benefs that accrue from interest tax shields, these studies find that firms do not adjust their leverage continuously toward their target. Rather, they adjust partially due to the presence of costs arising from deviations from target leverage and costs associated wh adjusting leverage towards the target. Despe significant efforts in the current lerature, ltle consensus so far has been reached wh regard to the rate at which firms adjust towards their target leverage. For example, Flannery and Rangan (2006) report rather a quick adjustment speed for US firms, wh about one third of the deviations from target leverage filled whin one year. This is much faster than the 7 18% documented by Fama and French (2002). Such apparent inconsistency could result from differences in the choice of data, model specifications, and econometric methods used. More importantly, these studies assume that firms adjust towards their target at a homogenous rate whether they have a financing surplus or defic or whether they are over or under-levered. For instance, Hovakimian Opler, and Tman (2001) find that deviations from target leverage largely determine firms choices of securies and hence their adjustment speeds. However, they do not allow for the fact that there may be an asymmetric impact on firms target adjustment. To overcome this, Byoun (2008) proposes an asymmetric partial adjustment model which explicly accounts for costs of deviations from target 2

3 leverage and adjustment costs i.e. over-levered firms move toward their target leverage at faster speeds due to relatively higher costs of deviation from target leverage. In addion to issues surrounding asymmetry in the speed of adjustment, Antoniou et al. (2008) argue that most capal structure studies do not consider how the financial orientation of the economy in which firms operate impacts their financing decisions since the major body of the current lerature keeps self busy wh US data. This issue is crical for impacts sources of financing available to firms (Ball, Kothari, and Robin, 2000) and hence may affect their financing choices. In this paper, we bring these two very recent strands of the lerature together and examine whether firms target adjustment behavior varies not only asymmetrically, depending on whether they are in a posion of a financing surplus/defic and whether they are over-/under-levered but also whether varies across market-based and bank-based economies. In particular, we examine the capal structure behavior for firms in five countries France, Germany, Japan (bank-oriented economies), and the UK and the US (marketoriented economies) using asymmetric partial adjustment models which incorporate costs of deviations from target leverage, adjustment costs, and firms financial constraints. In line wh the current lerature, we find evidence that firms move toward their target leverage at reasonably fast speeds, wh the estimated speed of adjustment coefficients ranging from (Japanese firms) to (French firms). Further, we find an asymmetric pattern in firms adjustment speeds. In the case of firms that are over-levered, the estimated speed of adjustment coeffcients range from (Japanese firms) to (French firms). However, when firms are under-levered, these figures range from (US firms) to (French firms). This finding is consistent wh the current lerature that over-levered firms have relatively more reason to be worried about deviations from their target leverage (Byoun,

4 2008) while under-levered firms tend to be relatively more relaxed wh their target adjustment decisions (Goldstein, Ju, and Leland (2001), and Lemmon and Zender (2004)). There is also evidence that firms wh a financing defic tend to move faster toward their target leverage. While adjustment costs are relatively less significant in the case of large financing gaps (Faulkender et al., 2010), firms wh a financing defic will likely find more compulsory to vis capal markets to cover their financing gaps. The presence of abovetarget leverage makes s impact even more pronounced. Contrary to Byoun (2008), we find that adjustment speeds are fastest when firms have both a financing defic and above-target leverage (from (Japanese firms) to (German firms)) and generally lowest when they experience both a financing surplus and below-target leverage. When financial constraints are considered, low-profabily/low-growth/low-asset tangibily/small/high-earnings volatily firms are found to have relatively faster adjustment speeds. This evidence has important implication about costs of deviations from target leverage and adjustment costs. For example, low-profabily firms are more likely to violate debt covenants, implying relatively more significant costs of deviations from target leverage. Low profabily is also likely to result in a more substantial financing defic. More importantly, we realize that these transion variables affect firms adjustment speeds more (less) significantly in the presence of a financing surplus and/or below-target leverage (a financing defic and/or above-target leverage). Therefore, firms target leverage adjustment behaviors are determined first by costs of deviations from target leverage and adjustment costs before financial constraints are taken into account. The remaining of the paper proceeds as follows. Section II presents the empirical models and methodology. Section III describes our data and sample selection. Section IV interprets the empirical findings. Section V offers some concluding remarks.

5 II. Empirical Models and Methodology A. Symmetric Partial Adjustment Models When firms adjust their leverage, they need to take into account two major costs costs of deviations from target leverage and adjustments costs. Their ultimate goal is to minimize the sum of these two costs. Wh the assumption that these two costs are both quadratic and addive, the total costs related to leverage adjustments can be expressed as: * 2 (1) ( ) ( ) 2 C t = a D D + b D D 1 where C t is the total costs of leverage adjustments,, * D is the unobserved target leverage ratio, D and D -1 are the leverage ratios for firm i at time t and t 1, respectively. and a and b are the respecive weights for the costs of deviations from target leverage and adjustment costs. To minimize C t, we have to choose D such that: C D * (2) = a( D D ) 2b( D D ) = 0, or ( D D ) 2 1 = a ( ) ( * D ). D 1 a + b 1 Hence, we arrive at the partial adjustment model of leverage: * (3) ( D D ) = θ ( D D ). 1 1 In other words, a partial adjustment model of leverage is typically used to model firms attempts to close out their deviations from target leverage over time. Its full form can be expressed as follows: * (4) D D = θ ( D D 1) + δ 1, where D and D -1 are the leverage ratios for firm i at time t and t 1, respectively. θ = a/(a+b) and is the proportion of the actual leverage change to the desired movement towards target change. It measures the speed at which firms undertake partial adjustment towards their target leverage. In the presence of adjustment costs, firms are unable to make continuous and full

6 adjustment such that θ should lie between 0 and 1 wh a higher value indicating a higher speed of adjustment. δ is the well-behaved error term. ratio that can be specified as a function of firms characteristics as follows: (5) D * = ψ' x * D is the unobserved target leverage To implement this, we regress actual leverage on firm characteristics using the following static model of leverage: (6) D = ψ' x + ε where x represents the vector of independent variables including profabily, growth opportunies, asset tangibily, firm size, effective tax rates, earnings volatily, dividend payout ratios, non-debt tax shields, and share price performance. ψ is a vector of the estimated coefficients. ε is the well-behaved error term. The fted value for firms target leverage ratios for the current accounting period is constructed as: (7) D = ' x * ψ. In this paper, we employ a two-step approach. In the first step, we estimate firms target leverage, as specified by Equation (5). Firms symmetric adjustment speeds then are estimated according to Equation (4). Although other conventional estimators such as pooled OLS, fixed-effects and DIF-GMM (Blundell and Bond, 1998) have some mer, we use the SYS-GMM method to control individual heterogeney and partially retain variations among firms. This method therefore is likely to yield more asymptotically consistent estimations in presence of heteroscedasticy and serial correlation than others do. It is derived from the estimation of a system of two equations, one in levels wh lagged first differences as instruments and the other in first differences wh lagged levels as instruments. In dynamic panel models, is claimed to perform well when series are persistent i.e. θ is close to uny

7 (Blundell et al. 2000). In addion, there is a significant fall in the fine sample bias as a result of the exploation of addional moment condions. To specify x, we follow the lerature and consider the following variables: Market leverage. This is the ratio of book value of total debt (Datastream em WC03255) to market value of total assets (market capalization (WC08001) + total debt). Book leverage. The ratio of book value of total debt to book value of total assets (WC02999). Profabily. Both the pecking-order and dynamic trade-off theories suggest that profable firms tend to depend relatively less on leverage to avoid information asymmetries (Myers and Majluf, 1984). Rajan and Zingales (1995) and Dejong and Veld (2001) document empirical evidence to support this. This variable is measured by operating income (WC01250) scaled by book value of total assets. Growth opportunies. The free cash flow hypothesis states that low-growth firms should employ more leverage to migate the free cash flow problem (Jensen, 1986). This view has been empirically supported by Hovakimian (2004) and Flannery and Rangan (2006). On the contrary, the trade-off theory suggests that relatively lower levels of financial distress costs should allow them to be relatively more levered. The variable is proxied by the market-to-book ratio - market value of total assets (market capalization plus book value of total debt) scaled by book value of total assets. Asset tangibily. As suggested by Jensen and Meckling (1976), better collaterals are likely to alleviate agency costs of debt, thus allowing firms to employ relatively more leverage. The trade-off theory therefore proposes a posive relation between these and firms leverage. The variable is measured by tangible assets (WC02501) scaled by book value of total assets (WC02999)

8 Firm size. Due to their better transparency and more stable asset bases, big firms are likely to employ more leverage, as suggested by the trade-off theory (Myers, 1977). Altinkilic and Hansen (2000) and Antoniou et al. (2008) also document empirical evidence that large firms are usually well-diversified and therefore have relatively lower bankruptcy costs. The variable is measured by the natural logarhm of book value of total assets (in US$ value) (WC07230) in 1980 dollars. Effective tax rates. The trade-off theory suggests that the presence of tax benefs of debt financing encourages firms to depend relatively more on leverage (DeAngelo and Masulis, 1980). This variable is measured by the ratio of income taxes (WC01451) to total pre-tax income (WC01401). Earnings volatily. According to the trade-off theory, firms wh high earnings volatily are more likely to face wh higher costs of debt financing and bankruptcy risk due to a relatively higher likelihood of not meeting debt obligations resulting from the cyclical nature of their earnings. Hence, these firms are expected to have less leverage as empirically documented by Antoniou et al. (2008). This variable is measured by the first-difference of annual earnings (changes in net income used to calculate basic earnings per share (WC01706) (%) minus the average of their first differences). Dividend payout. There has been contradictory evidence on the impact of firms dividend payout ratios on their leverage. Rozeff (1982) finds that firms dividend payout ratios are negatively correlated wh their leverage due to both agency and transaction costs. In contrast, Chang and Rhee (1990) argue that there may be a posive relation between them when dividend tax rates are higher than those on capal gains. This variable is measured by the ratio of common cash dividends (WC05376) to net income used to calculate basic earnings per share (WC01706).

9 Non-debt tax shields. The trade-off theory suggests that firms motivation to enjoy tax benefs from interest payments will encourage them to employ relatively more leverage. However, DeAngelo and Masulis (1980) argue that non-debt tax shields can be substutes for tax benefs from debt financing, implying firms wh higher non-debt tax shields should be relatively less levered. This variable is measured by the ratio of annual total depreciation (WC01148) to book value of total assets. Share price performance. Both the market-timing and inertia hypotheses agree that firms capal structures are generally the cumulative results of their managers abily to time the equy market. As a result, new equy is likely to be issued due to overvaluation (Baker and Wurgler, 2002). Therefore, they both conclude that there exists a negative relation between firms share price performance and their leverage. This variable is measured by the annual change in share price (WC05001). Countries macroeconomic orientation on firms target leverage include equy premium, term structure of interest rates, GDP growth, role of laws, credor rights, ownership concentration, and anti-director rights. High equy premium will encourage firms to use more debt (Antoniou et al., 2008). However, if is the result of investors overconfidence, the market-timing and inertia hypotheses suggest that firms are likely to depend more on equy. An increase in interest rates will make debt more costly for firms, thus discouraging them from employing leverage (Cook and Tang, 2010). Cook and Tang also document a posive relation between firms leverage levels and their countries economic growth. From firms point of view, a strict legal system leads to higher financial distress costs, implying a negative relation between legal systems level of strictness and firms leverage. However, a strict legal system ensures that lenders are better protected,

10 leading them to be willing to lend at relatively lower rates. 1 Firms in economies wh higher level of credor rights tend to be more levered since a higher level of credor rights may imply relatively lower costs associated wh debt financing i.e. better-protected credors are likely to be willing to lend at relatively lower interest rates. 2 To avoid any ownership dilution potentials, firms wh higher levels of ownership concentration tend to be relatively less levered. However, higher levels of ownership concentration imply fewer information asymmetries, enabling them to have more leverage as a relatively cheaper source of financing (Antoniou et al., 2008). Finally, a higher level of anti-director rights will reduce information asymmetries, suggesting a posive relation between them. 3 B. Asymmetric Partial Adjustment Models wh Deviations from Target Leverage As suggested by Hovakimian et al. (2001), deviations from target leverage may largely determine firms choice of securies and hence adjustment speeds since equy issuance costs on average are higher than those for debt issuance due to relatively higher information asymmetries (Altinkilic and Hansen, 2000). In addion, the fact that costs of deviations from target leverage vary in accordance wh the magnude of these deviations (Byoun, 2008) also places the assumption that there should be a particular adjustment speed regardless of firms leverage levels at odd. Put differently, the problem wh Equation (4) is that is does not take into account the fact that over-levered firms may face relatively higher costs of deviations from target leverage due to relatively higher financial distress costs than under-levered firms. Similarly, firms which increase leverage are not likely to face the same level of adjustment costs as firms which reduce. To account for these, C t can be rewrten as: 1 The level of law enforcement is ranged from 0 (least effective) to 10 (most effective) (La Porta et al. (1997 and 1998)). 2 The index is ranged from 0 (least effective) to 4 (most effective) (La Porta et al., 1997 and 1998). 3 The index is ranged from 0 (least effective) to 6 (most effective) (La Porta et al., 1997 and 1998).

11 * 2 * * 2 * (8) C a ( D D ) 1( D D < 0) + a ( D D ) 1( D D 0) t = ( D D ) ( D D < 0) + b ( D D ) 1( D D 0) + b where 1(.) is the indicator function. In Equation (8), 1 ( D * D < 0) and 1( 1 < 0) D can D be substutes for each other since over-levered firms are likely to find relatively more compulsory to reduce leverage to avoid financial distress costs i.e. relatively higher costs of deviations from target leverage and vice versa. Hence we have: * 2 * * 2 * (9) C a ( D D ) 1( D D < 0) + a ( D D ) 1( D D 0) Minimize C t, we have: t = * 2 * ( D D ) 1( D D < 0) + b ( D D ) 1( D D 0) + b C * * * * (10) = 2a ( D D ) 1( D D < 0) + 2a ( D D ) 1( D D 0) D 1 2 2b * * ( D D ) 1( D D < 0) 2b ( D D ) 1( D D 0) = which, after some arrangements, can be wrten as (11) ( D D ) a ( ) ( ) 1 ( 0) ( ) ( * ) 1( * 0) 1 * * a2 D D 1 D D < + D D 1 D D a + b a + b 1 = Hence, we arrive at the asymmetric, partial adjustment model, as follows: * a * b (12) ( D ) α ( D D ) D α ( D D ) D = D where α 3 = a 1 /(a 1 +b 1 ) and α 4 = a 2 /(a 2 +b 2 ). α 3 and α 4 are the proportion of the actual leverage + change to the desired movement towards target change. a D indicates above-target leverage and b D indicates below-target leverage. The full form of Equations(4) and (12) can be wrten as: TD TD 1 (13) ΔD = D D 1 = = α 0 + α1tde + λ, A A 1 (14) ΔD = D D TD = TD a b = α + α TDE D + α TDE D + ν, A A 1

12 where TDE represents the difference between firms current target leverage ratios and their leverage ratios for the previous accounting period. It is defined as: * * TD 1 (15) TDE = D D 1 = D, A 1 where TD and TD -1 are firms total debt for the current and previous accounting period; A and A -1 are firms total assets for the current and previous accounting period; D is a dummy variable equal to 1 in case of above-target leverage and 0 otherwise for firm i at time t; a b D is a dummy variable equal to 1 in case of below-target leverage and 0 otherwise for firm i at time t; α 0 and α 2 are the constant terms; α 1, α 3 and α 4 are estimated coefficients of adjustment speeds; and λ and ν are the well-behaved error terms. In the absence of adjustment costs, α 1, α 3 and α 4 should be equal to 1. However, as perfect market condions cannot be achieved in practice, they should be in the 0-1 viciny. The consensus in the current lerature has been that over-levered firms should be more concerned about their leverage posion due to relatively higher costs of deviations from target leverage while under-levered firms may be more reluctant to adjust their leverage i.e. to save their debt financing capacy for future needs (Byoun, 2008). Consequently, can be expected that α 3 > α 4. 4 C. Asymmetric Partial Adjustment Models wh Deviations from Target Leverage and Financing Gaps In the spir of Faulkender et al. (2010), firms financing gaps can tell a lot about their target adjustment behaviors since they determine the level of adjustment costs. Specifically, due to the presence of sunk costs associated wh large financing gaps, firms wh a 4 This assumption is similar to Byoun (2008) but our rationale is contrary to his argument. Specifically, Byoun argues that if adjustment costs are higher for equy than for debt, then α 3 > α 4. However, this is in contrast to the deviations from target leverage argument since over-levered firms are more likely to issue equy than debt (Hovakimian et al., 2001).

13 significant financing defic or surplus may be found wh faster adjustment speeds toward their target leverage. A financing gap (FG) is defined as: (16) FG = DIV + I + ΔW OCF where OCF stands for operating cash flows after interest and taxes (WC WC01151 WC01451); I is firms net investments (WC04870); ΔW stands for the change in net working capal; and DIV represents firms dividend payments (WC05376). Equation (16) can be equivalently rewrten as: (17) FG = NCF CDIV OSUF where NCF stands for net cash flow - financing (WC04890); CDIV is cash dividends (WC04551); and OSUF represents other sources/uses - financing (WC04448). Whenever values on CDIV and OSUF are missing, we set them to 0. 5 When we let Equation (13) interact wh financing gaps, becomes: s d (18) ΔD = β 0 + β1tded + β2tded + τ Similarly, Equation (14) can be rewrten as: s d a s d b ΔD = β + β D + β D TDE D + β D + β7d TDE D + (19) 3 ( 4 5 ) ( 6 ) where s D is a dummy variable equal to 1 in case of a financing surplus and 0 otherwise for firm i at time t. D is a dummy variable equal to 1 in case of a financing defic and 0 d otherwise for firm i at time t. Our hypothesis is that when firms experience a financing defic, they will find relatively more compulsory to vis capal markets i.e. to eher issue new debt or equy than when they have a financing surplus which should make them ξ relatively more relaxed wh their target adjustment behaviors. Therefore, on average, β 1 should be less than β 2. In Equation (19), deviations from target leverage may make the impact 5 The definion of financing gaps by Equation (17) here is more suable wh Datastream data, given s availabily and account structure.

14 of financing gaps more pronounced for the reason mentioned earlier. Specifically, we expect that β 4 < β 5 and β 6 < β 7. In addion, is also likely that β 4 > β 6 and β 5 > β 7 since costs of deviations from target leverage are generally higher when firms experience above-target leverage. A question arising from Equation (19) is that in the presence of both deviations from target leverage and financing gaps, which factor will be relatively more important and therefore needs to be considered first when firms are making their target adjustment decisions? In light of that question, we expect that deviations from target leverage will be generally considered first. In the spir of the trade-off theory, the magnude of deviations from target leverage directly determines how easy is for firms to access external capal markets. Over-levered firms may therefore find relatively harder to vis capal markets and vice versa. However, financing gaps only determine whether is necessary to vis these markets or not. D. Asymmetric Partial Adjustment Models wh Deviations from Target Leverage, Financing Gaps and Transion Variables While the evidence on the partial adjustment strand has been contentious, there has been noticeable consensus on major determinants of firms target leverage. Followers of the trade-off theory have documented profabily, growth opportunies, asset tangibily, earnings volatily, and firm size as key determinants of firms target leverage (Long and Malz (1985), Smh and Watts (1992), Easterbrook (1984), Tman and Wessels (1988) and so on). In the mean time, the pecking-order theory derives firms financing behaviors directly from their financing needs and consequently concludes that profabily ultimately determines their decisions to vis capal markets (Donaldson (1961), Myers (1984), Myers and Majluf (1984), and Shyam-Sunder and Myers (1999)). We argue that these variables can act well as proxies for firms financial constraints. Flannery and Hankins (2007) show that financial constraint proxies such as firms cash outflow of dividends and investments,

15 profabily, and asset sales may impact their adjustment speeds. In particular, those wh high dividend payout ratios tend to adjust toward their target leverage more slowly due to reduced financial flexibily. Similarly, firms wh fewer tangible assets should have more reasons to be concerned about their leverage posion than otherwise since they will generally have less value in case of liquidation (Flannery and Rangan (2006) and Mao (2003)). As a result, they are likely to move faster toward their target leverage. That is a similar story for low-profabily/low-growth/small/high-earnings volatily firms. Therefore, if we allow Equations (14) and (18) to interact wh these transion variables, they can accordingly be rewrten as: (20) ΔD L H = β + β TDE D + β TDE D + μ, L H a L H b (21) ΔD = β + ( β D + β D ) TDE D + ( β D + β D ) TDE D + η, L H s L H d (22) Δ D = β + ( β D + β D ) TDE D + ( β D + β D ) TDE D + υ, where L D is a dummy variable equal to 1 in case of low-profabily/low-growth/low-asset tangibily/small/low-earnings volatily firms and 0 otherwise for firm i at time t. H D is a dummy variable equal to 1 in case of high-profabily/high-growth/high-asset tangibily/big/high-earnings volatily firms and 0 otherwise for firm i at time t. Profabily. The pecking-order theory and dynamic trade-off theory suggest that profable firms tends to depend less on external financing due to the accumulation of retained earnings and hence are likely to have lower leverage levels than otherwise. In addion, high profabily may also lead to their better abily to meet debt obligations. As a result, they are expected to be relatively more relaxed wh their target adjustment decisions. That in turn implies relatively slower adjustment speeds for these firms. In other words, we generally expect to see that β 1 > β 2 in (20) and β 4 > β 5 and β 6 > β 7 in both Equations (21) and (22). It is also likely that β 4 > β 6 and β 5 > β 7 in Equation (21) and β 4 < β 6 and β 5 < β 7 in

16 Equation (22) for the reason mentioned earlier when deviations from target leverage and financing gaps are considered. Growth opportunies. Growth opportunies tend to reduce leverage for the trade-off theory suggests a posive relation between growth opportunies and the level of financial distress costs since high-growth firms are more likely to fail (Rajan and Zingales (1995), Mao (2003), Hovakimian (2004), and Flannery and Rangan (2006)). Moreover, due to information asymmetries, firms tend to issue equy in the first instance when overvaluation eventually results in higher expected growth. This is consistent wh Myers (1977) that high-growth firms should depend less on external debt financing to avoid the sub-investment problem, implying lower levels of target leverage for these firms. Our hypothesis is that low-growth firms may be found adjusting faster toward their target leverage. This is so because lowgrowth firms are likely to be over-levered which then makes them subject to relatively higher costs of deviations from target leverage. Tangibily. Tangible assets enable firms to have more access to debt markets as they serve as collaterals better (Hovakimian et al. (2004) and Leary and Roberts (2005)), especially in bank-oriented economies where collateral requirements are more stringent and firms generally prefer debt financing to equy financing. This is in line wh Sibilkov (2009) that costs of managerial discretion increases wh asset liquidy and Benmelech, Garmaise, and Moskowz (2005) that greater asset liquidy is associated wh greater loan size. However, this does not necessarily imply that firms wh high asset tangibily should be more concerned about their leverage posion. Instead, firms wh low asset tangibily should do that since according to Flannery and Rangan (2006) and Mao (2003), they are likely to have less value in case of liquidation. Moreover, due to their relatively lower collateral qualy, lenders are likely to apply higher interest rates on them, making debt financing

17 relatively more expensive. These together suggest that firms low asset tangibily will likely experience faster adjustment speeds. Firm size. When firm size acts as a proxy for financial constraints, is likely that small firms will experience relatively faster adjustment speeds. Big firms tend to have more access to capal markets and hence be more relaxed wh their target adjustment decisions (Tman and Wessels (1988) and Johnson (1998)). They also tend to depend relatively more on leverage due to their better abily to negotiate wh lenders and better cred ratings when banks consider firms size as one of the proxies for their credabily. The question is why do not these firms adjust faster to toward their target leverage levels while they have better abily to do that? The answer lies in the presence of adjustment costs. 6 In the presence of these costs, firms are likely to adjust more slowly if costs of deviations from target leverage are relatively lower for them. Hence, small firms need to be relatively more concerned if they stay away from their target leverage, thus having to move faster to these levels. Earnings volatily. Firms wh high earnings volatily are hypothesized to be less levered since there is always a chance that they cannot meet their debt obligations. Costs of deviations from target leverage therefore are also relatively higher for them. III. Data and Sample Selection Our sample includes firm-year data from 1980 to 2007 collected from Datastream for five countries, namely France, Germany, Japan, the UK and the US. Since we estimate a series of dynamic panel data models using SYS-GMM, we require that firms have at least five consecutive annual observations (Arellano and Bond (1991). We exclude financial firms (wh SIC code I from 6000 to 6999) and utily firms (wh SIC code I from 4000 to 4999) from the sample as these firms are likely to be heavily regulated and hence have different 6 Adjustment costs mainly consist of issuance costs (Fischer, Heinkel, and Zechner (1989) and Strebulaev (2007)).

18 financing behaviors. All variables of interest are winsorized between 0.5 and 99.5% to eliminate any unexpected effects by outliers. Data on countries instutional factors including anti-director rights, credor rights, rule of law, and ownership concentration are from La Porta et al. (1997 and 1998), unless stated otherwise. Macroeconomic variables namely GDP growth, terms structure of interest rates, and equy premium are collected from Datastream. Finally, the final data set consists of 9,034 firms wh 78,108 firm-year observations. Table I summarizes the number of firms and firm-year observations available for each country as well as provides descriptive statistics for the variables of interest. [Table I about here] Table I shows that firms in bank-oriented economies have higher leverage ratios (both book- and market-based) than their counterparts in market-oriented economies wh Japanese firms having the highest leverage ( ), followed by French and Germany firms. This is consistent wh the conclusion by Fukuda and Hirota (1996) that firms wh stronger relationship wh banks tend to be relatively more levered since is relatively easier for them to obtain debt financing from these banks at relatively lower costs. The relatively lower leverage ratios observed for UK and US firms can be explained by the lower level of ownership concentration among these firms and their looser relationships wh their banks (La Porta et al., 1997 and 1998). The market-to-book ratios are highest among UK and US firms, somewhat implying the market orientation for firms in these two countries. This also probably suggests that wh these firms, equy capal markets are generally more important than for those in bankoriented economies i.e. their preference for equy capal. Except for Japanese firms, firms in other bank-oriented economies have relatively lower asset tangibily than do their capal market-based counterparts. Due to their looser banking relationship wh banks, firms in market-oriented economies have to satisfy stricter

19 collateral requirements than those in bank-oriented economies. When total assets serve as a proxy for firms size, firms in bank-oriented economies appear to be larger than their counterparts in market-oriented economies. In the spir of the trade-off theory, this implies a better abily to access debt financing sources for those in bank-oriented economies, partially lending support to their relatively higher levels of leverage on average. Effective tax rates for UK and US firms are lower than these for firms in bank-oriented economies, consistent wh their generally lower levels of profabily. German firms have the highest dividend payout ratios, followed by Japanese, UK, French, and US firms. This is agreeable to the nature of the tax system in each individual country in the sample. For example, the German tax system generally discourages internal equy so imposes relatively lower rates on dividend payments, resulting in firms there finding more beneficial to increase dividend payments. This is almost the same story wh the Japanese system. In the UK, prior to 1997, dividend payments were largely encouraged. However, the system then began to favor firms earnings retention. Meanwhile, in both France and the US, their tax systems have consistently been in favor of earnings retention, explaining why dividend payout ratios are lowest among firms in these countries. In brief, tax considerations have an important role in firms dividend policies, in line wh their ultimate goal of wealth maximization for their shareholders. Earnings volatily is generally higher among firms in market-oriented economies, throwing some light into why they have to be relatively more conservative wh their financing policies and employ less leverage in their capal structures. IV. Empirical Results A. Target Leverage Estimations and Determinants of Target Leverage Table II reports the fixed-effect estimation results for determinants of target leverage, as specified by Equation (6). Our results suggest that firms target leverage behaviors can be

20 better explained by the trade-off theory. There is evidence of a negative relation between profabily and leverage for firms across all countries in the sample. This finding is in line wh Rajan and Zingales (1995) and Strebulaev (2007) that profable firms tend to depend relatively more on internal financing to avoid financial distress costs. The magnude of the impact is more significant among firms in bank-oriented economies, suggesting that firms profabily tends to has more influence on their decisions to vis debt markets and implying their preference for debt financing in these countries. Since firms in these countries are generally of bigger size, also lends some support to the claim that bigger firms tend to have a relatively more stable history of profabily (Kayhan and Tman, 2007). The impact of profabily is less pronounced among UK and US firms. This finding has an interesting implication about the relatively looser relationship between them and their banks which results in profabily having relatively less impact on their decisions to vis debt markets. It may also suggest that wh firms in these countries, equy markets are relatively more important financing sources. [Table II about here] Growth opportunies penalize firms leverage in bank-oriented economies more significantly, suggesting that the trade-off theory works better wh them (Hovakimian (2004), Miao (2005) and Flannery and Rangan (2006)). This is contrary to our inial expectation that growth opportunies should have relatively more pronounced impact on firms in market-oriented economies due to disclosure practices and lender-borrower relations in these countries (Antoniou et al., 2008). The impact of growth opportunies is most pronounced among Japanese firms, further confirming the nature of the relationship between them and their banks and their preference for debt financing. Asset tangibily and firm size have both statistically and economically significant impacts on firms across all the countries in the sample, in line wh the prediction by the

21 trade-off theory and the empirical evidence reported by Flannery and Rangan (2006) and Mao (2003). Similar to profabily and growth opportunies, s impact is particularly more pronounced among firms in bank-oriented economies. The relatively less economically significant coefficient of asset tangibily on UK and US firms may suggest that as a result of the looser relationship between them and their banks, collateral qualy has less important impact on their decisions to vis corporate debt markets. The role of effective tax rates on firms target leverage is not clear. Their coefficient is statistically and economically insignificant across firms in all countries in the sample. A posive coefficient for Japanese firms indicates that firms in this economy are likely to depend more on debt financing when the tax rates for them are higher, agreeable to their highest level of effective tax rates among the five countries. On contrary, the negative coefficient of the variable on firms in other countries remains largely a mystery, probably due to the endogeney problem. Such inconsistency in the impact of this variable has been discussed in details by Ang and Peterson (1986) and Tman and Wessels (1988). 7 The effect of dividend payout ratios on firms in all countries seems to be both statistically and economically insignificant and the evidence on is rather mixed, contrary to the finding documented by Chang and Rhee (1990) and Rozeff (1982). There exists a posive correlation between these ratios and target leverage for firms in all bank-oriented economies. This implies that a higher dividend payout ratio among these firms is possibly observed by their lenders as an indicator of an improvement in their performance, thus enabling them to obtain new debt financing at relatively lower costs. On the other hand, wh firms in capal 7 The data of Antoniou et al. (2008) also supports this finding. Overall, the consensus in the lerature has been that the inconsistency in the effect of effective tax rates on firms target leverage is caused by the invariations in corporate tax rates across firms.

22 market economies, an increase in their payout ratios tends to be viewed as an increase in risk by lenders, resulting in higher costs imposed if they wish to vis debt markets later. 8 The coefficient on non-debt tax shields is posive across all countries, inconsistent wh the idea of the trade-off theory that firms wh higher non-debt tax shields should employ less leverage. However, according to MacKie-Mason (1990), this is probably due to the fact that depreciation of tangible assets composes the major part of their non-debt tax shields, suggesting higher asset tangibily for these firms. The coefficient on share price performance is significantly negative across firms in all countries. This finding is consistent wh the market timing hypothesis (Baker and Wurgler (2002) and Hovakimian, Hovakimian and Tehranian (2004)) that firms leverage is the result of managers attempt to time the equy market. Similar to the finding reported by Antoniou et al. (2008), there is no clear distinction in the effect of this variable between the two groups. Managers attempt to time the equy market is therefore prominent in all countries, regardless of their instutional, economic, and legal backgrounds. Finally, the coefficient on earnings volatily is both economically insignificant across firms in all countries in the sample. Our data seems not to support the evidence reported in the current lerature that firms wh higher earnings volatily should depend less on debt financing due to relatively higher financial distress costs for them (Antoniou et al., 2008). It is however consistent wh the evidence reported by Leary and Roberts (2005) that firms earnings volatily has ltle to do wh their target leverage. 8 Dividend payout ratios are reported to be endogenously determined (Blundell et al., 1992). Therefore, to avoid that problem, Antoniou et al. (2008) lag the variable by two periods. To account for the change of tax on dividends in the UK in 1997, they introduce a dummy variable of 0 before 1997 and 1 after that for the value of dividend payout ratios in their model. Since the coefficient of this variable is both statistically and economically insignificant, we do not follow that approach for may partially capture the time-series effect of macroeconomic variables. For more detailed discussion on the change in the tax treatment of dividends in the UK, refer to Antoniou et al. (2008).

23 The pooled estimation results are generally consistent wh previous results wh profabily, growth opportunies, effective tax rates, and share price performance (asset tangibily, firm size, and non-debt tax shields) having negative (posive) impacts on firms target leverage. The estimated coefficients for all the macroeconomic variables have the expected signs (Colum (6) in Table II). There is evidence of a negative relation between the term structure of interest rates and firms leverage, as contrary to that between and the level of credor rights. Overall, our results suggest that the impacts tradional determinants of capal structures on firms leverage are country-dependent. Both the magnude and direction of these impacts are generally in line wh the trade-off theory and current empirical evidence, wh the exception of effective tax rates, dividend pay-out ratios, and earnings volatily. More importantly, there is evidence of firms well-defined target leverage, as suggested by the trade-off theory. B. Symmetric Adjustments toward Target Leverage Table III reports the SYS-GMM estimation results for the partial adjustment models specified by Equations (13) and (14). The results for the symmetric model (13) contained in Columns (1), (3), (5), (7) and (9) show that, on average, firms in France, Germany, Japan, the UK and the US move towards their target leverage at reasonably fast speeds, ranging from (Japanese firms) to (French firms). This indicates that firms in these five sample countries can close deviations from their target leverage between two and three years. Hence, this finding provides strong evidence for the trade-off theory and is generally consistent wh previous empirical evidence (e.g. Flannery and Rangan (2006), Antoniou et al. (2008) and Byoun (2008)). [Table III about here]

24 Our results further reveal that firms in bank-oriented economies (except for Japan) tend to adjust toward their target leverage at relatively faster speeds than those in marketbased economies. Since the former firms rely relatively more on leverage, they are potentially more concerned about costs of deviations from target leverage than do the latter firms. Consequently, they are under greater pressures to adjust toward their target leverage. Further, German and French firms have close relationships wh banks, which make easier and less costly for them to undertake leverage adjustments via new debt issues or debt reductions, which should then interpret into faster adjustment speeds. C. Asymmetric Adjustments toward Target Leverage Firms wh Above- or Below-Target Leverage The results for the asymmetric partial adjustment model (14) reported in Columns (2), (4), (6), (8) and (10) of Table III support our prediction that over-levered firms adjust significantly faster toward their target leverage than under-levered firms (except for Japanese firms). Specifically, firms wh above-target leverage in France, Germany, the UK and the US have adjustment speeds of 0.572, 0.459, and 0.467, respectively, while those wh below-target leverage have their adjustment speeds of 0.469, 0.382, and 0.337, respectively. The difference between the adjustment speeds for under- and over-levered firms is both statistically and economically significant. For instance, German and UK firms adjust toward their target leverage at speeds of nearly 20% faster when they are over-levered than when they are under-levered. There are three possible explanations on why over-levered firms have quicker adjustment speeds than under-levered firms do. First, financial distress costs for over-levered firms are possibly higher since they bear a higher possibily of breaching debt covenants when they exist, thus forcing them to revert back to their target leverage faster (Byoun, 2008). Second, unreported results show that firms wh overabove-target leverage are likely

25 to have a financing defic, which makes more compulsory for them to adjust. This issue is further discussed in subsequent sections. Third, over-levered firms may face lower adjustment costs than under-levered firms since they are likely to revert back to their target leverage levels via debt retirements which are arguably less costly than debt issues. Overall, facing wh potentially higher costs of deviations from target leverage and lower adjustment costs, over-levered firms move toward their target leverage faster than their under-levered counterparts. D. Asymmetric Adjustments toward Target Leverage Firms wh a Financing Surplus or Defic Table IV reports the estimation results for Equations (18) and (19) that model asymmetric adjustments condional on firms financing gaps. The results in Columns (1), (3), (5), (7) and (9) show that firms wh a financing defic have significantly faster adjustment speeds (from (US firms) to (French firms)) than those wh a financing surplus. This finding is in stark contrast wh Byoun s recent evidence on US firms (2008) and lends support to our prediction that firms wh a financing defic may to find relatively more compulsory to cover their financing gaps while those wh a financing surplus tend to be relatively more relaxed. That is why the latter firms may adjust relatively more slowly toward their target leverage (from (Japanese firms) to (French firms)). 9 [Table IV about here] In Columns (2), (4), (6), (8) and (10), we examine how firms wh different financing gaps undertake leverage adjustments, controlling for their posions in relation to their target leverage. The results show that the presence of above-target leverage makes the impact of a 9 Byoun s results may be driven by his choice of model specifications. In addion, he also remains largely silent on why firms wh a financing surplus should adjust faster toward their target leverage in the presence of adjustment costs and lower costs of deviations from target leverage since unreported results on the correlation among firms characteristics indicate that the presence of a financing surplus tends to be associated wh below-target leverage. See Appendix I for details.

26 financing defic on firms adjustment speeds more pronounced. Importantly, when firms experience both above-target leverage and a financing defic, their adjustment speeds become the quickest, ranging from (Japanese firms) to (German firms). These estimated adjustment speeds are statistically faster than firms wh above-target leverage and a financing surplus or those wh below-target leverage and a financing defic. Condional on having below-target leverage, firms do not have statistically different adjustment speeds when they have a financing defic or surplus (except for Japanese firms and US firms). One possible explanation for this finding is that in the presence of belowtarget leverage, firms are relatively more relaxed wh their target adjustment decisions and hence financing gaps have relatively less impact on their adjustment mechanisms. Similarly, condional on having a financing surplus, their adjustment speeds do not statistically differ between over-levered and under-levered firms (wh the exception of Japanese and US firms). Overall, our results for Equations (13), (14), (18), and (19) support the hypothesis that firms wh above-target leverage and/or a financing defic eher have relatively more incentives or find more compulsory to adjust quickly toward their target leverage than do the remaining firms. Our results wh regard to deviations from target leverage are generally consistent wh the current lerature (e.g., Byoun, 2008). However, those wh regard to financing gaps are contrary to Byoun s - firms wh below-target leverage and/or a financing surplus tend to be more relaxed wh their target adjustment decisions and hence undertake slower leverage adjustments. E. Asymmetric Adjustments toward Target Leverage Firms wh Low or High Profabily Table V reports the results for three asymmetric adjustment models based on Equations (20), (21), and (22) wh the transion variable being profabily. First, the results in Columns (1), (4), (7), (10), and (13) support our prediction that across all the five sample