Transaction Costs and Capital-Structure Decisions: Evidence from International Comparisons

Transaction Costs and Capital-Structure Decisions: Evidence from International Comparisons Abstract This study examines the effect of transaction costs and information asymmetry on firms capital-structure decisions in 40 countries. The findings indicate that transaction costs affect both capital-market timing and capital-structure rebalancing. Past market-timing activity has a significantly negative impact on the current debt ratio, and this impact is stronger for firms facing lower transaction costs of external financing, as defined by legal origin, capital-market development, and securities rules in their home countries. Further analysis indicates that firms in countries with lower transaction costs also rebalance their capital structure more quickly after a deviation from the target, but the rebalancing does not eliminate the markettiming effect on capital structure completely. JEL code: G30, G32 Key word: Capital structure, transaction cost, capital market development, market timing, rebalance speed 0

1. Introduction The objective of this study is to empirically examine capital-structure decisions in an international context. Previous studies in this area generally focus on the transaction cost perspective the trade-off theory of capital structure, which states that the choice of financing method is determined by the relative costs (direct and indirect) of different financing options. Giannetti (2003) and Fan et al. (2008) find that firms in countries where stock markets are more developed have lower leverage and that firms in countries offering better protection for creditors have more long-term debt in their capital structure. Qian and Strahan (2007) find that in countries with stronger creditor protection, bank loans have more concentrated ownership, are longer in maturities, and have lower credit spreads, while Bae and Goyal (2009) argue that enforcement, rather than the existence of rights, matters for loan size and maturity. Other studies examine trade-off theory by comparing the debt ratio s sensitivity to traditional variables for testing this theory (De Jong et al., 2008, Gonzalez and Gonzalez, 2008, Fan et al., 2009) across countries. The basic idea is that if those variables are good proxies for transaction costs and agency problems, then their impact on debt ratios should differ across countries with different standards of investor protection, creditor protection, and capital market development. This paper examines the effects of transaction costs on financing decisions from another perspective. We argue that transaction costs affect a firm s ability to both time the equity market and rebalance its capital structure after a deviation. Several studies have examined the impact of equity market timing on debt ratio in the U.S. market but they disagree about whether the effect is persistent. While Baker and Wurgler (2002), and Huang and Ritter (2009) show that firms past market timing activity has a persistent impact on the 1

capital structure, Hovakimian (2006), and Kayhan and Titman (2007) find that the markettiming effect is at most temporary. Leary and Roberts (2005) argue that a dynamic trade-off model with adjustment costs can explain the observed effect of market timing on capital structure. Internationally, the evidence of the market-timing effect on debt ratio is also mixed. While Chang et al. (2009) show that the market-timing effect exists in Japan, especially among group-affiliated firms, Mahajan and Tartaroglu (2008) find that this effect is shortlived in other G-7 countries. None of the studies, however, explores the market-timing effect in emerging markets and cross-sectional differences in this effect across countries. This paper aims to fill the above gap in the literature. Lower transaction costs, however, may also result in a higher adjustment speed towards the target after a deviation, which causes the market-timing effect to be less persistent. This argument follows that of Hovakimian et al. (2004), that the preference for internal financing and the temptation to time the market by selling new equity when the share price is relatively high interfere with the tendency to maintain the firm s debt ratio close to its target. However, the net effect of stronger market timing ability and higher adjustment speed is unknown and an open empirical question. We examine a sample of 16,857 firms in 40 countries between 1992 and 2006. We regress the debt ratio on market-timing measures together with standard control variables and expect a negative relation between the debt ratio and market-timing measures. We adopt two measures from the literature to measure a firm s past market-timing activity. The first one is Baker and Wurgler s (2002) external finance weighted average market-to-book ratio. This measure takes a higher value if a firm s past external financing was more positively related to its market valuation. Baker and Wurgler find that it predicts negatively a firm s financial leverage ratio. The second measure comes from Kayhan and Titman (2007), who argue that 2

Baker and Wurgler s timing measure could not effectively isolate equity issuance for growth opportunity from issuance for market timing. To address the problem, they decompose Baker and Wurgler s measure into two parts, one measuring past investment opportunities and the other measuring timing behavior. In our analysis, we use both measures to examine the market-timing effect on the debt ratio. We perform a regression analysis to test whether transaction costs and information asymmetry affect market timing s effect on capital structure. We adopt four proxies for transaction costs from previous studies. The first one is a dummy for common-law countries, which is motivated by La Porta, Lopez-de-Silanes, Shleifer and Vishny s (LLSV hereafter, 1997) finding that common-law countries have larger stock markets than civil-law countries. The second one is a market-development index formed by a principal component analysis of five indicators of capital market size, investor protection, and enforcement of securities law, as discussed in Section 3.3. Therefore, we expect firms in common-law countries and firms in more developed markets to face lower transaction costs of external financing. The last two are related to specific securities rules. One of them is existing shareholders pre-emptive right in equity offerings. We argue that the existence of a pre-emptive right makes market timing more costly, if still possible, because firms are not allowed selling shares directly to outside investors. The other proxy is permission for stock repurchase. If firms are allowed buy back their shares, they can rebalance their capital structure more easily after timing the equity market, and their past market-timing activity will have a weaker effect on the current debt ratio. Our results indicate that firms in common-law countries, more developed markets, and more open markets (those with no pre-emptive rights and stock repurchase allowed) have lower debt ratios. This is consistent with findings from previous studies indicating that firms 3

rely more on equity financing when property rights are better protected. More important for our purpose, past market-timing activity has a negative impact on the current debt ratio, and the negative impact is stronger in common-law countries and more developed markets than in civil-law countries and less developed markets. The negative impact is also stronger in countries where pre-emptive rights do not exist and stock repurchase is not allowed. Although we report results based on Fama-Macbeth estimation procedure, our results are qualitatively unchanged when we estimate the models using pooled ordinary-least-squares (OLS) regressions. To confirm that transaction costs affect a firm s ability to time the equity market, we also run a probit model to predict the debt-equity financing decision. Consistent with our argument that transaction costs affect a firm s ability to time the market, we find that a firm is more likely to issue equity (versus debt) when firm value is higher if it is in a common-law country or a more developed capital market, but it is less likely to do so if pre-emptive rights are applied. In contrast, firms that have the right to repurchase stocks are less likely to time their equity issuance, inconsistent with our expectation that they can be more aggressive in timing the market because they face a lower cost to rebalance their capital structure. To examine the impact of transaction costs on the speed of capital structure rebalancing, following Lemmon et al. (2008), we use the system generalized method of moments (GMM) by Blundell and Bond (1998) to estimate the speed. We also report the results from OLS regressions for comparisons and robustness, although Huang and Ritter (2009) argue that using pooled OLS regressions ignoring firm fixed effects tends to overestimate the speed of adjustment. Our analysis shows that the speed of capital-structure rebalancing is faster in common-law countries than in civil-law countries, faster in more developed markets than in less developed markets, faster in countries where pre-emptive 4

rights do not exist than where such rights exist, and faster in countries where stock repurchase is allowed than where repurchase is not allowed. All findings are consistent with our prediction that firms facing lower transaction costs rebalance their capital structure towards the target more quickly than firms facing higher transaction costs. Together with the findings on market timing, however, our results indicate that in countries with lower transaction costs, the additional market-timing effect is not completely eliminated by the higher rebalancing speed. Our study contributes to the literature in several ways. First, it fills a gap in the literature by testing the market-timing effect of capital structure in the international context. Among the studies that examine capital structure in the international context, few study the capital structure s market-timing effect. An exception is Mahajan and Tartaroglu (2008), who examine the effect of market-timing activity on capital structure in G-7 countries and find that the market-timing effect on the debt ratio is short-lived, except for Japan. They argue that their results are more in line with the dynamic trade-off model. Their sample, however, consists of only firms in developed markets, which are more homogenous in economic development. Our sample is more diversified in terms of economic development, capital market development, and investor protections. This heterogeneity allows us to examine how country-level and firm-level variables affect the capital-structure decisions of firms in different countries and explore the cross-country and cross-firm differences in the markettiming effect of capital structure. Second, our study relates the market-timing effect on capital structure to transaction costs. It confirms McLean et al. s (2009) finding that firms in countries facing lower transaction costs are more capable of timing the market. The difference in ability to time the 5

market also creates a divergence in the debt ratio over time. We confirm this by utilizing both Baker and Wurgler s (2002) and Kayhan and Titman s (2007) timing measures. Third, our study examines the impact of transaction costs on the speed of the debt ratio s adjustment using a system GMM approach. Although several studies use similar methods to estimate the speed of adjustment (Lemmon et al., 2008 and Gonzalez and Gonzalez, 2008), they do not further investigate the determinants of the speed of adjustment. An exception is Antoniou et al. s (2008) study, which compares the determinants of capital structure in market-oriented and bank-based economies. This study, however, examines only five developed markets and its findings are inconclusive. Our study is among the first to use the system GMM approach to examine the impact of transaction costs on the speed of adjustment, and we find evidence supporting our hypotheses. The remainder of the paper is organized as following. Section 2 reviews the capitalstructure literature and develops hypotheses. Section 3 discusses the construction of key variables, including the market-timing measure, and the legal and market-development measures. Section 4 introduces sample selection and reports sample properties. Section 5 provides results from the regression analysis. Section 6 concludes. 2. Literature Review and Hypotheses Development Traditional corporate finance theory, in general, assumes that firms have no transaction costs associated with altering their capital structure. Static trade-off theory suggests that in the absence of transaction costs, companies should rest on their respective optimal leverage ratios that balance marginal benefits (e.g., tax shield) and marginal costs (e.g., financial distress, 6

asset substitution, etc.) of debt financing. Firms will also adjust back to the optimal leverage ratio in a short time if there is a deviation from the target. Transaction costs vary substantially across countries. A series of papers by LLSV (1997, 1998, 2002) compares investor protection in countries of different legal origins and examines how legal origins affect capital-market development and financing activity. LLSV (1998) show that common-law countries offer better protection for investors than civil-law countries do. LLSV (2002) find that protection for minority shareholders is positively associated with firm valuation, LLSV (1997) conclude that common-law countries have more developed stock markets than civil-law countries. In addition, stock-market development is positively related to rule of law and investor protection. The above studies generally suggest that firms belonging to more developed markets that offer better investor protection face lower transaction costs. Transaction costs have significant impacts on capital-market timing. Comparing preand post-1970 periods of the U.S. market, Pontiff and Woodgate (2008) find the size of equity issuance is negatively related to future stock return in the latter but not in the former. They contend that the development of the capital market, which lowers transaction costs, drives the negative issuance effect on stock return in the post-1970 period. McLean et al. (2009) further argue that transaction costs are higher when the market is less developed or legal enforcement is weak. They study 41 countries all over the world. Using country-level marketdevelopment variables to proxy for transaction costs, they show that the share-issuance effect on stock returns is stronger when transaction costs are lower (i.e., in more developed) markets. Following their results, we expect that equity issuances of firms in more developed markets are more sensitive to market conditions. In addition, if market-timing activity has a persistent effect on capital structure, then the capital structure of companies in more developed markets 7

will be more affected by past market-timing activity. We use two variables to proxy for market development: a dummy variable for common-law countries and a composite index generated by LLSV s investor protection and market development variables. Besides capital market development, we identify two securities rules may also affect transaction costs, which in turn affect a firm s ability to time the market or to rebalance its debt ratio towards the target after market timing. The first such rule is existing shareholders pre-emptive right in equity offerings. A pre-emptive right gives existing shareholders of an equity issuer the first priority to purchase new shares on a pro-rata basis. In other words, a rights offering is the default method of equity offering if the pre-emptive right exists. While the pre-emptive right can prevent existing shareholding from being diluted, it also stops companies from timing the equity market by selling shares to outside investors. The second rule is permission for stock repurchase. If a firm is allowed to buy back its shares, it can do so to rebalance its debt ratio to the target after timing the equity market. As a result, past markettiming activity will have a weak effect on its current debt ratio. Following our above discussions, we develop the following hypothesis: H1a: The market-timing effect of capital structure is stronger in common-law countries than in civil-law countries, and in more developed markets than in less developed markets. H1b: The market-timing effect of capital structure is weaker in countries where a pre-emptive right exists and stock repurchase is allowed. While market timing will cause a firm s debt ratio to deviate from the optimal level, the impact should be short-term if the cost of rebalancing is zero such that the firm can continuously rebalance its capital structure. Transaction costs inhibit capital-structure rebalancing, however. Recent studies on the speed of capital-structure adjustment focus on 8

the U.S. market, and they disagree about the speed of adjustment towards the target debt ratio after a deviation. Fama and French (2002) estimate an adjustment speed of 7-18% per year. Flannery and Rangan (2006) estimate an even faster speed of adjustment of about 35% per year. Leary and Roberts (2005) and Alti (2006) also document a high speed of adjustment. Huang and Ritter (2009) estimate the adjustment speed using the long differencing estimator for a dynamic panel first proposed by Hahn et al. (2007) and find an adjustment speed of 17.0% per year for book leverage and 23.2% per year for market leverage. Lemmon et al. (2008) use a system GMM technique and estimate a speed of adjustment of about 25% per year. We examine the difference in speed of adjustment across countries using the system GMM approach. A recent paper by Antoniou et al. (2008) use a similar technique to examine the speed of adjustment of firms in G-5 countries (France, Germany, Japan, U.K, and U.S.) and finds that the speed of adjustment is the fastest in France and slowest in Germany and Japan. They argue that because Japanese and German firms have close ties with their creditors, they face low cost of being off the target. Unlike Antoniou et al. (2008), instead of viewing the situation in terms of the cost of deviation, we argue that firms in different countries face different transaction costs of financing and that firms in more developed markets can rebalance their capital structure more quickly because they face lower transaction costs. Following our discussions above, we have the following hypothesis: H2a: The speed of adjustment is faster in common-law countries than in civil-law countries, and faster in more developed markets than in less developed markets, H2b: The speed of adjustment is slower in countries where a pre-emptive right exists than in countries without a pre-emptive right, and faster in countries where stock repurchase is allowed than in countries where buy-back is not allowed. 9

3. Data and estimate procedure 3.1. Debt ratios and control variables This paper adopts control variables from Rajan and Zingales (1995). In particular, we control for firm size (SIZE) and asset tangibility (TA G) for the cost of financial distress. Bigger firms and firms with more tangible assets could use assets as collateral to issue debt, and therefore they are positively associated with the debt ratio. Moreover, we control market-tobook ratio (MB) and profitability (ROA) for growth opportunities and earning capacity. The two variables are negatively related to the leverage ratio, because firms with higher growth opportunity and greater earning capacity rely more on equity and internal generated profit, respectively. The above four variables are extensively used in later studies on capital structures for both the US and international markets. In particular, Booth et al. (2001) examine capital structures in 10 developing countries and find that capital structures in those countries are affected by the same variables as in the developed countries. Therefore, we use the same set of variables as control variables for our regressions. Our variables are defined as follows (the code in parentheses refers to the data item in Worldscope): 1) Book leverage ratio (TDB) is the ratio of total debt, which is equivalent to total assets (WC02999) minus book equity (WC03501) divided by total assets. 1 2) Firm size (SIZE) is defined as the natural logarithm of net sales (WC01001), translated into the US dollars. 1 We use book leverage rather than market leverage for two reasons. First, book leverage is relatively unaffected by stock price movement. If a firm times the external equity market, its stock price will drop after it issues equity, and its debt ratio will increase as a result. Therefore, using market leverage will weaken the timing effect, if it exists. Second, if we believe that mispricing can exist and persist for a period, market leverage will not reflect the actual financial leverage because market price does not equal the fundamental value. 10

3) Market-to-book ratio (MB) is defined as (total assets book equity + market capitalization) divided by total assets. 4) Asset tangibility (TA G) is defined as net PPE (WC02501) divided by total assets. 5) Profitability (ROA) is defined as return on assets, which is operating income (WC01250) divided by total assets. 3.2. Market-timing variables We adopt the Baker and Wurgler s (2002) external finance weighted-average market-to-book ratio (BWMB hereafter) as our timing measure. For a particular firm-year, the variable is defined as follows: BWMB e + d M FD s s s = ( ) ( ) t 1 s = s s= 0 B s= 0 B er + dr FDr r= 0 r= 0 M, (1) where summations are taken starting from the first year of available data where stock price data are not missing, and e and d denote the net equity and net debt raised, respectively. Specifically, e is estimated as the change in the number of shares outstanding (adjusted for stock splits and distribution) multiplied by stock price at the end of the year, and d is defined as the change in total debt on the balance sheet (WC03255). 2 The sum of e and d, FD, is the firm s total external financing in a year. 3 BWMB captures the firm s historical market-timing behavior because it takes a high value if a firm raised external funds where the market-to- 2 We also define net equity and net debt financing using balance sheet data, as in Baker and Wurgler (2002), and the results are qualitatively the same. 3 To ensure the weight on market-to-book ratio is positive, we follow Baker and Wurgler (2002) to set FD equal to zero when it is negative. Therefore, the timing measure focuses only on securities offerings and not securities repurchases. 11

book ratio was high and held up the financing plan when the market-to-book ratio was low. BWMB may overestimate the significance of timing behavior, however. As Kayhan and Titman (2007) pointed out, the BW s timing variable will have a high (low) value when firms are expected to have high (low) growth opportunities. Kayhan and Titman present their timing variable as follows: M FDs s= 0 B s M M Year Timing ( YTt ) = FD = cov FD, t B B 1, Long - ) M term Timing ( YT t 1 = FD. B Kayhan and Titman (2007) argue that the long-term timing measure (LT) makes it possible to test whether managers fund their financing deficits more with equity when the cost of equity financing is lower (the market-to-book ratio is higher). Baker and Wurgler s (2002) timing intuition is captured more efficiently by the yearly timing measure (YT) than BWMB. Indeed, Kayhan and Titman s timing measure is a decomposition of Baker and Wurgler s timing measure by dividing BWMB by average financing deficit in the firm s history. Using simple algebra manipulations, we find that these two timing measures are closely related: BWMB YT LT cov ( FD, M ) t 1 = + = + = KTCOVt 1 + KTMBt 1 B FD FD FD B M. (2) For our regression analysis, we use KTCOV as a proxy for market timing, because it captures the covariance between firm value and past financing activity. We expect KTCOV to be negatively related with a firm s current debt ratio. KTMB, the historical average of the market-to-book ratio up to t-1, is included to further control for the effect of past investment 12

opportunities on the debt ratio. Mahajan and Tartaroglu (2008) find that KTMB is negatively related to the debt ratio. 3.3. Measurement of market development and transaction costs We classify 40 countries and regions into common-law countries and civil-law countries, according to LLSV. In addition, we collect country-level measures for market development, investor protection, and legal enforcement from previous studies as follows. 1) Capital market size. Two measures for capital market size, namely private credit relative to gross domestic product (GDP) (from Djankov et al., 2007) and external market capitalization to GDP (from La Porta et al., 2006) are adopted. Higher values for the two variables suggest more developed capital markets. 2) Investor protection. An index for anti-director rights, which takes a value from 0 to 6, comes from LLSV (1998). This index is computed by six indicators that aim to measure whether minority shareholders in a country gain fair access to shareholders meetings and the extent to which corporate charters protect their interests. A higher value denotes stronger protection. 3) Enforcement of securities law. Two measures for legal enforcement, namely private enforcement and public enforcement, are collected from La Porta et al. (2006). Each of the two variables takes a value between zero and one. Private enforcement measures whether corporate disclosure requirements on important transactions and irregularities are strict and if the burdens of proof on investors for the civil liabilities of directors, distributors, and accountants are low. A high value suggests strong private enforcement. Public enforcement indicates whether regulators in a country 13

act independently and have enough investigative power, and whether public enforcement covers non-criminal and criminal sanctions for securities law violations. As the above five variables are highly correlated with each other, to capture their commonality, we compute a composite market development index (MDI) by taking the first component from a principal component analysis of those variables. A higher value of MDI indicates a more developed capital market. To further show that transaction costs are associated with the strength of the markettiming effect on capital structure, we consider two specific rules that affect the cost of equity offerings or the cost of rebalancing after market timing. The first rule is existing shareholders pre-emptive rights. As discussed previously, a pre-emptive right will increase the cost and make the market-timing effort less effective because all the shares must be sold to existing shareholders first. Spamann (2006) offers a long discussion on the constructions of different governance variables in previous studies, including LLSV. Our pre-emptive right variable (Preemptive) is based on Spamann s preright variable. 4 It takes a value one if a pre-emptive right exists, and zero otherwise. The second rule relates to permission for stock repurchase. After timing the equity market, a firm can rebalance its debt ratio by raising debt financing or buying back its shares. If stock repurchase is not allowed, however, the firm will find it more difficult to adjust its debt ratio back to the target. Therefore, we expect the market-timing effect on the debt ratio is stronger in countries where stock repurchase is not allowed. From McLean et al. (2009), we collect the year in which stock repurchase was allowed in a country. However, as our markettiming measures capture the whole history of a firm s activities, during which a firm was not allowed to repurchase stock in earlier years and allowed in later years, we define a variable for 4 Spamann (2006) argues that the coding in LLSV (1998) is wrong for some countries and corrects some of the errors by re-visiting related legal documents. 14

permission for stock repurchase (Repurchase) that equals the percentage of years that a firm was allowed to buy back its stock up to time t. 5 4. Sample Selection and Sample Properties Our sample is composed of 16,857 firms (97,796 firm-years) in 40 countries and regions and covers the period from 1992 to 2006. We construct a list of all companies, active and dead, from Worldscope. For each firm, we collect annual financial variables from Worldscope and daily equity data from Datastream. We then construct variables for our univariate analysis and regression analysis. From the initial sample, we exclude firms with missing values on the key variables for the regressions in Table 4. We also exclude firms in the financial industry and firms with a book value of total assets of less than US$10 million. We exclude firms in the US and Japan from our sample because previous studies specifically examine the markettiming effect of capital structure in these two markets (see Baker and Wurgler, 2002 and Chang et al., 2009). In addition, the numbers of firms in these two countries are overwhelmingly large compared to those in other countries. Excluding them can avoid having our results be over-influenced by the effects of these two countries. Table 1 reports the distribution of observations across the 40 countries, as well as the values of country-level variables for market development, investor protection, and legal enforcement. Although just above one-third (14 out of 40) of countries are classified as common-law countries, about 47% of firm-years (45,832) come from common-law countries. The number of observations also varies largely across countries, from 98 of Sri Lanka to 13,776 of the United Kingdom. 5 If, for example, a firm has survived for ten years and was allowed to repurchase its stock in four of the ten years, then the value of Repurchase is 0.4. 15

[Insert Table 1 here] Table 2 reports the pair-wise correlations among country-level variables. Commonlaw countries have more developed capital markets, as measured by the composite market development index (MDI), than civil-law countries. MDI is positively and significantly related to each of its five components (correlation coefficients range from 0.49 to 0.83). This suggests that a capital market is more developed if the country has larger equity and debt markets relative to its GDP, and if it offers better investor protection and enforcement of laws. A preliminary analysis (unreported) also shows that more developed markets are less likely to have pre-emptive rights and more likely to allow stock repurchase, but none of the correlations is statistically significant at conventional levels. [Insert Table 2 here] Panel A of Table 3 reports the median values of debt ratios, market-timing measures, and control variables. All variables are winsorized at the 0.5 th and 99.5 th percentiles of their respective distributions for the sample firm-years. In particular, the median KTCOV is positive in 38 out of 40 countries (all except Argentina and Israel). This suggests that market timing is prevalent across countries with different types of market development and legal systems. Panel B reports comparisons between (1) civil-law countries and (2) common-law countries, and between (3) less developed markets and (4) more developed markets. Countries are classified into less developed and more developed according to the median value of MDI. Z-statistics for Wilcoxon rank-sum tests for the differences in medians between groups are reported. In general, two sets of comparisons offer the same patterns. 16

Firms in more developed markets (common-law countries, high MDI) have lower debt ratios than firms in less developed markets (civil-law countries, low MDI). This finding reflects that stock markets in common-law countries are generally more developed (LLSV, 1997). Indeed, both Giannetti (2003) and Fan et al. (2008) find that firms are less leveraged in common-law countries, countries that are less corrupt, and countries where stock markets are more developed. Firms in more developed markets also have higher market-to-book ratios and return-on-assets, and they invest more in fixed assets. This suggests that firms in more developed markets have more growth opportunities and those markets offer better corporate governance. Firms in more developed markets are smaller, however. This may reflect the fact that capital markets in civil-law countries are less developed and only larger firms can survive and tap the external equity market. In addition, since external capital markets are less developed in civil law countries, firms in those countries operate in diversified businesses so as to form an internal capital market. Diversification may also explain why firms in civil-law countries are larger. Finally, firms in more developed markets time markets more than firms in less developed markets, as reflected by their higher values of BWMB and KTCOV. This is consistent with our above argument that firms in developed markets face lower transaction costs of market timing, and therefore they time the market more easily and frequently. [Insert Table 3 here] 17

5. Results from Regression Analysis 5.1. Market-timing effect on capital structure: Regression analysis Table 4 reports the results from estimating the basic model for the market-timing effect on debt ratio. We adopt two measures, BWMB and KTCOV, to capture a firm s past markettiming activity and expect the two variables are negatively associated with a firm s current debt ratio. The panel regression model is based on the following specifications: TDB i, t = α i + β BWMB i, t 1 + γ X i, t 1 + ε i, t (5) TDB i, t = α 1 + β 1 KTCOVi, t 1 + β 2 KTMBi, t 1 + γ X i, t 1 + ε i, t 1 (6) TDB is the book leverage ratio. X is a vector of predetermined control variables, including firm size (SIZE), market-to-book ratio (MB), asset tangibility (TA G), and profitability (ROA). KTMB is the historical average of the market-to-book ratio up to t-1. All the explanatory values are lagged one period relative to the dependent value. Industry dummies are included but not reported in the table. The models are estimated with both OLS regressions and the Fama-Macbeth approach. For the OLS regressions, standard errors are calculated from the Huber/White/sandwich heteroskedastic consistent errors, which are also corrected for correlations across observations for a given firm. Columns (1a) and (1b) of Table 4 report the results using BWMB as the timing variable (equation (5)), and columns (2a) and (2b) report the results applying KTCOV (equation (6)). Columns (1a) and (2a) are estimated with OLS regressions, while columns (1b) and (2b) are estimated with Fama-Macbeth approach. The results suggest that large firms 18

have more debt while more profitable firms have less debt in their capital structure. Both effects are stable across different model specifications and consistent with previous studies for the U.S. market. Inconsistent with the findings from previous studies, however, the marketto-book ratio s effect on the debt ratio is positive and the effect of asset tangibility is negative. 6 More importantly, columns (1a) and (1b) show that past market-timing activity, as measured by BWMB, has a negative and significant impact on the current debt ratio. This suggests that Baker and Wurgler s (2002) finding for the U.S. market is also applicable to emerging markets. In columns (2a) and (2b), the coefficient of KTMB, the measure of past investment opportunities, is negative and significant at the 1% level. This is consistent with Kayhan and Titman (2007) and Mahajan and Tartaroglu (2008) that KTMB may be a better proxy for a firm s investment opportunities than the lagged one-period market-to-book ratio. KTCOV, the preferred measure of market timing, also has a significantly negative effect on the debt ratio, and the finding is consistent across the two model specifications. This suggests that market-timing activity has a significant and persistent impact on capital structure in the world market. [Insert Table 4 here] 5.2. Transaction costs and the market-timing effect To examine if the market-timing effect on capital structure is stronger in countries where transaction costs are lower, we re-estimate the models with interaction terms between market- 6 In an alternative specification, we estimate the models with country dummies included. The coefficient of TA G becomes positive and the coefficient of MB is positive but statistically insignificant. The instability of the coefficients of TA G and MB is the result of our decision to use the book debt ratio rather than market debt ratio. 19

timing measurements and proxies for transaction costs (Proxy), namely a dummy for common-law countries (Origin), the market development index (MDI), a dummy for the existence of pre-emptive rights (Preemptive), and the permission for stock repurchase (Repurchase). In addition, as Fan et al. (2008) and Gonzalez and Gonzalez (2008) document that the debt ratio s sensitivity to control variables (X) differs across countries with different market development and investor protection, we also interact the control variable (X) with proxies for transaction costs. The model specification is as follows: TDB, i t = α i + β BWMB + φ Proxy X i, i, + θ Proxy BWMB + ε i, t i, + γ X i, (7) TDB, i t = α i + θ + β KTCOV 2 1 i, Proxy KTMB + β 2 i, KTMB + γ X i, i, + θ Proxy KTCOV 1 + φ Proxy X i, + ε i, t i, (8) Table 5 reports the regression results. Panel A reports the results from the estimation of equation (7) using the Fama-Macbeth approach. Consistent with H1a, the results in columns (1) and (2) indicate that firms in common-law countries and more developed markets have debt ratios that are more sensitive to market-timing measures than firms in civil-law countries and less developed markets, as shown by the negative coefficients (significant at the 5% level) of the interaction terms, BWMB Origin and BWMB MDI. The findings are consistent with McLean et al. (2009), who find that transaction costs of securities offerings are lower in more developed markets and that firms in those countries have more opportunities to time their equity offerings. We further show that more market-timing activity in those developed markets results in a cumulative negative effect on the capital structure of firms there. Consistent with H1b, pre-emptive rights (column (3)) and permission for stock repurchase (column (4)) both reduce the market-timing effect, as indicated by the positive coefficients (significant at 5% and 1% levels, respectively) of the interaction terms, BWMB 20

Preemptive and BWMB Repurchase. Therefore, besides market development, some specific securities rules may also affect the market-timing effect on the debt ratio, if those rules affect the costs of securities offerings or the costs of rebalancing. Panel B of Table 5 reports the results from a similar analysis, with KTCOV as the market-timing variable to capture past timing activity (equation (8)). The main results are qualitatively similar to those of Panel A, except that the interaction term KTCOV MDI is negative but insignificant at conventional levels. Other results are also worth of note. Firms in common-law countries and more developed markets have more debt in their capital structure than firms in civil-law countries and less developed markets, as indicated by the negative coefficients of Origin and MDI in columns (1) and (2). Pre-emptive right (Preemptive) is positively associated with the debt ratio, while permission for stock repurchase (Repurchase) is negatively associated with it. All the findings suggest that firms rely more on equity financing when they operate in countries where market institutions are more developed and liberal. In addition, the debt ratio is less negatively related to profitability in more developed and liberal markets, as indicated by the coefficients of ROA Proxy. The finding suggests that firms rely less on internal profits for financing when the cost of external financing is lower. The impacts of market development and transaction costs on the sensitivity of the debt ratio to other control variables, however, are less consistent. [Insert Table 5 here] In sum, our results in Table 5 are consistent with McLean et al s (2009) finding that capital-market development affects the transaction costs of equity offerings. We further show 21

that the transaction costs of equity offerings will affect firms ability to time their equity offerings, which in turns affects the market-timing effect on the capital structure. In particular, past market-timing activity s negative impact on the debt ratio is stronger for more developed capital markets than for less developed markets. In addition, by showing that securities rules, such as pre-emptive rights and permission for stock repurchase, also affect the market-timing effect, we confirm that transaction costs greatly affect market-timing activity in international capital markets. Our findings are also consistent with recent studies indicating that countryspecific factors have both direct effects (the level of debt ratio) and indirect effects (their influence on the roles of firm-specific factors) on capital-structure decisions. 7 In a robustness check, we examine market-timing activity s effect on contemporaneous change in the debt ratio, as in Kayhan and Titman (2007). The dependent variable is five-year change in the debt ratio and the explanatory variables include a five-year measure of market timing (KTCOV) and a set of control variables, including the change in the vector of variables X, market-adjusted return and change in target debt ratio from t-5 to t, and leverage deficit in t-5. As the dependent variable is measured based on overlapping windows, a bootstrapping technique is used to determine the coefficients statistical significance. See Appendix 1 for variable definitions, a description of the bootstrapping procedure, and the results. The results in Appendix 1 are largely consistent with those in Tables 5, except that while Preemptive KTCOV and Repurchase KTCOV are still both positive, they are statistically insignificant at conventional levels. 5.3. Analysis of debt-equity choice 7 See de Jong et al (2008), Fan et al (2008), and Gonzalez and Gonzalez (2008), for example. 22

Market timing implies that firms raise more external capital when market conditions are good. Previous studies generally show that firms issue equity when their firm values are high (see Loughran and Ritter, 1995, 1997) for firm-level evidence and Baker and Wurgler (2000) for aggregate-level evidence). A survey study by Graham and Harvey (2001) also documents that financial managers consider equity valuation to be a main consideration for offering equity. Having an ability to time the market, however, depends on both the existence of timing opportunities and the ability to take advantage of such opportunities. Following our arguments above, the existence of timing opportunities depends on the ability to take advantage of the opportunities depends on transaction costs. Therefore, firms in markets that are more developed and liberal are more likely to engage in market timing. To test the hypothesis, we estimate the following model that predicts a firm s choice between debt and equity issuance: Pr[ Eissue = 1] = F( α + β K + γ Proxy + ϕ Proxy Valuation ) (9) i, t i, i i i, Following Hovakimian et al. (2001), the dependent variable is a dummy variable that equals 1 if the net equity issued is more than 5% of total assets, and 0 if net debt issued is more than 5% of total assets in a year. The analysis includes pure equity issuers and pure debt issuers only. In other words, firms that issue both equity and debt and non-issuers are excluded. 8 Valuation is one of MB (Panel A of Table 6) and market-adjusted returns in year t (Panel B). 9 K is a vector of variables, including SIZE, MB (or market-adjusted return), TA G, and ROA used for previous tables. Following Hovakimian et al. (2001), and Chang et al. (2009), K also includes LDef, debt maturity, a dummy for MB > 1, and a dummy for ROA < 0 as additional control variables. LDef is the actual book-leverage ratio (TDB) minus the target-leverage ratio, 8 We set the cutoff point to remove extremely small issues that may be driven not by financing needs, but for some other reason, such as exercise of executive stock options. In an unreported robustness check, we set the cutoff point to 1% and the results are qualitatively unchanged. 9 Market-adjusted return is equal to cumulative stock return minus cumulative market return, from the end of year t-1 to the end of year t. 23

where target leverage is the predicted value of the leverage ratio obtained by regressing TDB on a vector of explanatory variables (X) used by Rajan and Zingales (1995) together with country and industry dummies. If a firm has a target debt ratio, it will issue equity when its current debt ratio is higher than the target. Therefore, we predict a positive sign for this variable. Debt maturity is defined as long-term debt / (short-term debt + long-term debt). Hovakimian et al. (2001) argue that wealth transfer from shareholders to debtholders is more likely when a financially distressed firm has more long-term debt. Therefore, we predict that a firm is less likely to issue equity if it has a negative ROA and more long-term debt. Finally, a dummy for MB > 1 is included because a firm will be reluctant to issue equity if its stock price is low relative to its book value, to avoid the threat of takeover. Proxy is one of Origin, MDI, Preemptive, and Repurchase. All explanatory values are lagged one period relative to the dependent value. Consistent with other empirical studies, Panel A of Table 6 shows that the lag marketto-book ratio (MB) has a positive effect on the probability of equity issuance. More important, firms are more likely to time their equity issuance when transaction costs are lower, as indicated by the positive coefficients of MB Origin and MB MDI. The negative coefficient of MB Preemptive also suggests that the existence of pre-emptive rights discourages firms from timing the market. The coefficient of MB Repurchase is negative rather than positive, however, which is inconsistent with our prior belief that permission for repurchase may encourage firms to time their equity issuance more aggressively. Table 6 offers other important findings as well. We do not find strong evidence to support our prediction that firms issue equity to rebalance their capital structure when their current leverage is higher than the target. The coefficients of LDef are positive but all are statistically insignificant. The result is surprising because Appendix 1 clearly shows that 24

long-term change in the debt ratio is negatively related to deviation from the target leverage. Therefore, the failure of finding target behavior from Table 6 is unlikely to be caused by measurement problems of target leverage. Notice that there are two major differences between Table 6 and Appendix 1. First, Appendix 1 examines the change in the debt ratio, which is driven by both external financing decisions and the decision to retain earnings for investments, but Table 6 considers external financing decisions only. In an unreported test, we run an OLS regression for change in retained earnings on LDef with other control variables and find that the coefficient of LDef is statistically significant at the 1% level. In other words, firms do retain more earnings when their debt ratios are above their targets. Second, Appendix 1 examines long-term change in the debt ratio after the debt ratio deviates from the target, while Table 6 examines the external financing decision after a short-term deviation of the debt ratio from the target. It is possible that firms take a longer time to rebalance their capital structure or that firms rebalance their capital structure only when the deviation is not temporary. To test the latter possibility, we calculate average LDef by taking the average of LDef from t-3 (or t-5) to t-1, and the coefficient of the variable is positive and significant at the 5% level. The finding confirms that firms do not rebalance their debt ratios immediately after a deviation from the target. 10 Other variables have signs mostly consistent with our expectations and with previous studies. The coefficient of ROA is negative, but the coefficient of the dummy for ROA < 0 is positive. Therefore, more profitable firms are less likely to issue equity, and loss-making firms are more likely to issue equity. The findings are consistent with the pecking-order theory that when firms are profitable, they will first issue debt if they need external financing, but when firms are losing money, they will seek equity financing after their debt capacity has been exhausted. The interaction term, debt maturity the dummy for ROA < 0 is negative. 10 The above two unreported results are available upon request. 25

This suggests that loss-making firms are less likely to issue equity when they have more longterm debt in their capital structure, to avoid value transfer from shareholders to debt-holders. Panel B replaces MB with market-adjusted return in year t. The result indicates that firms are more likely to issue equity than to issue debt if their share prices go up more in the period. Consistent with Panel A, the impact of share price run-up on the probability of equity issuance is stronger in more developed markets, and weaker in countries with pre-emptive rights and permission for repurchase. The coefficients of other explanatory variables are qualitatively unchanged. [Insert Table 6 here] 5.4. Test on speed of adjustment Transaction costs affect not only market timing but also the speed of adjustment to the target leverage ratio. Huang and Ritter (2009) demonstrate that if firms rebalance slowly to the target leverage ratio, market timing could have a more persistent effect on the current leverage ratio. In general, firms facing a lower cost of market timing also face a lower cost of adjustment towards the target debt ratio. Therefore, the existence of a market-timing effect on the debt ratio does not necessarily mean that firms are unconcerned about their debt ratios. To estimate the adjustment speed towards the target debt ratio, we run the following model that is constituted by two structural equations: TDB = α + λ ( µ TDB ) + η + v + ε, (10) * i, t i, t i, ti, i t i, t 26