Capital structure choice in the emerging markets

Размещено на http://www.allbest.ru/

Capital structure choice in the emerging markets

Introduction

capital diversified macroeconomic

What is the source of firm's funds? The answer to this question lies into the firm's capital structure. How do firms choose their capital structure and which factors affect their decision? The puzzle of capital structure has not an easy solution. Even now, many recent papers come to differing conclusions on this subject. Things get more complicated when we move setting from the developed countries, where markets work relatively efficient, to the countries with emerging economies, where known mechanics just may not work. Kumar et al. (2017) believe that one of the weakest points in research on the subject is a lack of works studying capital structure choice in emerging markets. This paper is intended to provide a quick dive into the scientific journey on important theoretical and empirical works, which try to answer the question of the firm's capital structure in developing economies, so the unfamiliar reader would have better perspective. After introducing you to the subject, I develop the model that tries to capture the effect of individual firm, institutional and macroeconomic variables on capital structure of the 6 representative developing economies. The countries I include in my analysis are Chile, Egypt, Malaysia, Poland, Thailand, and Turkey. The reason for the inclusion of these countries is that diversified macroeconomic conditions and institutional quality enable me to analyze their relevance for the capital structure.

Many research papers in this field try to analyze the relationship between macro institutional variables and adjustment costs and speeds. Conclusions and results in these papers are usually quite intuitive -- worse macroeconomic and institutional characteristics lead to higher adjustment costs and lower adjustment speeds. My work has different focus, which is to empirically look at the relationship between capital structure and firm factors with inclusion of macroeconomic and institutional variables.

The firm level data for analysis was obtained with Standard & Poor's Capital IQ platform. It consists of 1,763 firms from developing countries for the period from 2007 to 2014. The data on macroeconomic and institutional quality variables comes from different sources, including World Bank World Development Indicators, World Bank Doing Business, and Transparency International.

I do analysis for conventional firm factors, some macroeconomic, and institutional variables. The main contribution of my work is capturing some interesting interactive effects between some variables.

1. Literature Review

The theory of capital structure has not a long history relatively to some other topics in the field of economics and finance, as the first significant model that tries to capture the choice firm faces was developed by Modigliani and Miller (1958). Before this work nobody had not seriously addressed this question on the theoretical level. So, they laid the theoretical fundaments of the further works in the direction. The model itself features strong and strict assumptions, like existence of firms with homogenous cash-flow profiles, perfect information, perfect and frictionless markets, riskless debt, and absence of taxes. All these assumptions enable us to make a proposition about the choice capital structure. It tells us that firm's value only depends only on firm's cash flow profile, while the capital structure is irrelevant. Modigliani and Miller (1958) base this proposition on an argument that with the stated assumptions shareholders can construct mixed portfolio with any equity and leverage. Thus, arbitrage opportunities will exist if the presence of leverage in the capital structure changes firm's value. This result does not seem satisfying, as there are no implications and the assumptions are too unrealistic. But the subject gained an attention, what enabled further works in this field.

In the subsequent commentary on the work, Modigliani and Miller (1963) show that following same assumptions but with taxes included, the implications of model change. Optimal leverage in Modigliani-Miller world with taxes is not “does not matter” anymore but becomes the maximal possible amount. This happens because interest expenses are deducted from tax base, so to maximize its value firm should have more leverage in its capital structure. The implications of the second work were too extreme, as we rarely observe fully debt financed firms, but the argument for favoring debt in the capital structure started growing. This work is considered as a starting point of Trade-Off Theory, but not quite itself, as it has not featured any disadvantages of using debt to offset the benefits.

In the time between these important works, another influencing idea was born. Donaldson (1961) conducted a study of large firms, and the main result was that firms mostly use internally generated funds. This is a start of Pecking Order Theory, which was a hypothesis in the beginning, before the development of it by Myers (1984).

The first significant empirical works featuring conventional econometric models on capital structure appeared later than most of theoretical, as there was no convenient way to collect and analyze data using computers. Also, endogeneity had been a significant problem until the development of some estimation methods. Thus, the early empirical works featured looking for the patterns in data and surveys or using estimation methods that are capable with manual computation.

The pioneering work in empirical field of capital structure can be considered as Toy et al. (1974). Their model was relatively simple comparing to later researches. They used a sample of 816 manufacturing firms from France, Japan, Netherlands, Norway, and United States for the period 1966-1972. They separately estimated Ordinary Least Squares (OLS) model for each country using total debt as a dependent variable with growth, profits, and risk as explanatory variables. As a result, the model has good predictive power for every country, except France. The signs of explanatory variables are positive for growth and risk, negative for profits.

The complementary work to the previous paper was done by almost the same group of researchers. Stonehill et al. (1975) conduct the survey about financial goals and debt ratio determinants in the same list of countries. The main results are the following: only US firms appear to maximize its value plus dividends, while firms in other countries try to act as independent entities and maintain financial stability. As the determinants of leverage, the most important factors appear to be interest expenses coverage representing financial risk, availability of capital represented by current and historical capital markets conditions. Other factors, such as international factors, minimizing costs of capital, industry norm, technology, and liquidity appear to be partly significant or insignificant, depending on country.

The most famous theory of capital structure is the Trade-Off Theory, and the first work that has its main features was done by Kraus and Litzenberger (1973). The essence of this model is following: when firms decide on optimal leverage, they consider benefits and costs of it. The benefit is deduction of interest payments from taxes, or so-called tax shield. If there are no costs of leverage, then the results will be the same as in MM world with taxes. It is assumed that extra leverage brings additional costs of financial distress and increased risk of bankruptcy. As a result, firm should have an optimal mix of debt and equity, that maximizes the firm value. Haugen and Senbet (1978) state that costs of bankruptcy can be direct and indirect. Direct costs are the result of legal process associated with bankruptcy, for example legal charges and restructuring costs. The indirect costs are the result of discontinuing operational activity and reputation damage. They include worsening of vendor relationships, decrease in consumer base, and loss of workers.

Not everybody agreed with the significance of the bankruptcy costs. Miller (1977) argued that costs of bankruptcy are negligible in comparison with gains from tax shield. His solution is that other theorists ignored taxes on personal income. When there are different relationships between tax on personal interest and equity, the gain on leverage from the investor's point of view may not be as attractive as the gain in the value of firm. There can be even negative relationship between the gain of investor and firm's tax shield. Miller (1977) use his simple model to describe the equilibrium on a corporate bond market, which takes place due to progressive income taxes. This work was a critical argument against Trade-Off Theory.

To defend the grounds of Trade-Off Theory, there was a need in some other explanations for costs of increased leverage. Jensen and Meckling (1976) developed an agency cost approach to debt. They argue that moral hazard of managers can be a candidate for costs of financial distress. When firm has too much debt, it shifts to riskier behavior. Stockholders incentivize equity maximizing behavior, and with high leverage most value of non-risky projects go to debtholders. With riskier projects, the expected value of equity is higher because in case of success stockholders get more, and in case of unfavorable outcome, the losses are carried by bondholders. Myers (1977) by similar logic think that managers of high leveraged firms may pass the projects with positive NPV because they need to keep up with interest payments. This type of behavior is called underinvestment, which is along with risk-shifting considered as examples of managers' moral hazard problem and agency costs of debt. Debt, however, can have a positive effect on agency performance. Ross (1977) based on asymmetric information framework shows that issuing debt under certain assumptions is a good signal for the firm. Jensen (1986) argues that debt in the capital structure can have good impact on manager's behavior and efforts. His Free Cash Flow hypothesis states that additional debt has a disciplining effect on managers, as they need to consistently generate some amount of cash-flows to cover interest payments. These works theoretically reinforced Trade-Off Theory, adding arguments for balancing debt.

Myers (1984) with an influencing paper challenges Trade-Off Theory. He suggests that Pecking Order Theory is a good alternative as it better represent the firm behavior. The essence of this theory is that firms generally follow simple patterns in financing behavior. They prefer internally generated finances to issuing debt, which is preferred for issuing new equity. The theory can be based on the asymmetric information framework. Myers and Majluf (1984) in the complementary work define a more formal theoretical model. The intuition behind the model is simple. Managers have access to firm inside information, so from their observed behavior we can infer some conclusions. They do not want to issue equity when firm is undervalued because the share of existing shareholders is diluted. It is better to issue additional shares only when the firm is overvalued. Thus, the announcement of the future public offering can be interpreted as the signal that firm's stock price is too high. This is an example of adverse selection. Similar situation happens when firm takes more debt, as either it issues bonds or take a credit from a financial organization, it is required to reveal some information. So, firms should use internal finances to full extent, as their usage does not cause any asymmetric information consequences. Myers (1984) acknowledges that this theory fails to explain many individual cases, like the market timing behavior of firms when they issue equity on the rise of stock prices, but in his opinion, it represents aggregate behavior well and it covers some empirical facts, which static trade-off theory cannot. One of the main arguments in favor of Pecking Order Theory is that it covers an observed negative relationship between profitability and leverage, at which Trade-Off Theory fails.

The problem of profitability would be resolved in further works on Trade-Off Theory. The early theoretical works featured the static version of it, which considers capital structure optimization in one period. The main problem is that it does not allow incomplete adjustment. The firm is either at its optimum or it is not, and there are no other possibilities. This makes static trade-off theory hard to test empirically. The natural extension of static trade-off theory is the dynamic version of it, when many periods are considered, rebalancing is allowed, usually with some transaction costs.

Fischer, Heinkel, and Zechner (1989) made a paper on the dynamic trade-off theory. They find that with fixed costs of issuing equity, firms allow for capital structure to deviate from the target structure and adjust leverage only when it goes beyond extreme values. Whenever the firm is profitable it repays debt and allows leverage to fall. Firms adjust their leverage periodically to capture the benefits of debt tax shield. Thus, in dynamic setting Trade-Off with adjustment costs can explain the negative effect of profits, at which the Static Trade-Off theory fails. This paper alleviates one of the strongest disadvantages of Trade-Off Theory in comparison with Pecking Order Theory.

Leland (1994) made a very resembling Dynamic Trade-Off model with costly adjustments. In his work, the firm's leverage is a cumulative result of firms profits and losses. Firm allows this leverage to be off target if adjustment costs are higher than potential benefit of adjustment. When the leverage goes far away from the target and potential benefits are higher than adjustment costs, firm readjusts it, to capture all the benefits. The Dynamic Trade-Off Model was a breath of fresh air in the theoretical field of Capital Structure, but the empirical works incorporating its elements appeared later.

With time, empirical works with more sophisticated econometric models appeared on the subject. Marsh (1982) is one of the examples of it. He examines the debt and equity issues by UK firms for a period 1959-1972. He uses logit and probit models to estimate coefficients using data of 748 cases of security issue. By the structure of the model, the dependent variable is the probability of issuing equity, while the inverse probability is the issuing debt. As the main explanatory variable, they include the deviation from target structure in both long-term and short-term debt. First, they estimate the model with more variables, and then exclude some based on their performance. The proxies for determining leverage target in the final model are size, bankruptcy risk, and percentage of fixed assets. They also include in the model Market Timing conditions, such as forecast for debt and equity issues, and residual return from past year. On an estimated sample model showed probability of type I and II errors as 22% and 27% respectively, while for a hold-out sample of 110 issues it showed 25% and 37%. As a result, they conclude that market timing conditions are quite important for firms, and firms appear to have target capital structure determined by its variables.

The next significant work, which is mentioned in many papers on subject was done by Titman and Wessels (1988). They use factor analysis to test leverage theories prevailing at the time of paper, which were Trade-Off and Pecking Order among with some other ideas. They suggest that transaction costs may be important as size is negatively correlated with debt. Past profitability is also negatively related with leverage, what brings support to Pecking Order Theory of Myers (1984). Their results do not provide support for an effect of volatility, non-debt tax shields, future growth, collateral value on debt ratio.

Another paper that had impact on the further research was developed by Rajan and Zingales (1995). They analyze how conventional capital structure factors perform across industrialized G7 countries. First, they analyze the leverage statistics among the countries. Next, they look at the institutional differences that can affect the leverage. They include different tax systems, bankruptcy laws, market-based and bank-based financial systems, and ownership with control. Then they look at the cross-sectional evidence. The sample size is approximately 3500 firm observations from US, Japan, Germany, France, Italy, UK, and Canada. The market and book definitions of leverage are employed as dependent variables at year 1991. As the explaining factors they choose the market-to-book assets ratio, tangibility, profitability, and firm size. To overcome endogeneity problem, they use previous four-year averages as the explanatory variables (1987-1990). The model was estimated separately for each country. Overall, the evidence suggests that capital structure in industrialized G7 countries is not different from the US as it had been thought before. The factors overall perform relatively the same, with some exceptions among countries. However, the underpinnings are still unknown and further work linking theory with empirical findings needs to be done. They also suggest that further work explaining effect of institutional differences is required.

As the time passed, technology and knowledge developed enough for empirical works to be more available to the researchers from the technical point of view. The next wave of significant empirical works is dedicated to the problem of empirical testing of Pecking Order Theory.

Shyam-Sunder and Myers (1999) is one the first tests of pecking order theory against static trade-off theory. They find that both models perform very well independently with pecking order being better descriptor in their opinion, but when both models used jointly the coefficients of pecking-order theory almost do no change, while the power of target-adjustment model falls, but still statistically significant. There is an observation that firms use debt not to cover unexpected costs, but rather plan to finance with it anticipated projects. They also use simulations to see the power to reject both models. Target-adjustment model is not rejected when it is false in the setting of simulation, while pecking order model is rejected easily when it is false. This was the first paper provided support to pecking order theory. An answer to this work came 4 years after.

Frank and Goyal (2003) try to empirically test whether the predictions of pecking order theory are consistent with the real-world data. They use a broad cross-section of US firms for the period from 1971 to 1998. The first observation from the data is that internally generated funds do not cover investment spending on average, so firms use external financing quite a lot. They find that net-equity issues are more in line with financing deficit than net-debt issues, and this contradicts pecking order theory. Financing deficit, when added to the model of capital structure choice, offers more explanatory power for a model, but does not influence the role of conventional factors. Another observation is that large firms in early times of data follow pecking order theory more reliably than small firms or firms at later times. This paper provides significant challenge for a pecking order theory, as many analysis results contradict it.

Another paper, trying to test pecking order theory, came 2 years after. Fama and French (2005) decided to test pecking order theory and its predictions about equity issuance and share repurchases. They argue that based on the original argument of asymmetric information, firms should avoid repurchases of equity too. They examine samples of non-financial firms from 1973 to 2002, with grouping by times 1973-1982, 1983-1992, and 1993-2002. The results of examination contradict the pecking order theory, as equity issues are quite common for the listed periods (54%, 62%, 72% with respective order for net equity issuers). Equity repurchase is not uncommon too, as on average 20% of firms retire equity each year. Overall, authors disagree with results of Shyam-Sunder and Myers (1999), but they do not show their support for Trade-Off Theory either.

At approximately same time frame, another idea was cultivated. Baker and Wurgler (2002) develop Market-Timing theory to explain changes in the capital structure. The main motivation is that neither trade-off nor pecking order theories explain the fact that weighted average market-to-book ratio is negatively related to leverage. They suggest that firms want to meet the market conditions and issue equity when stock prices are on the rice, without further correction of debt. So, the capital structure is the cumulative result of firm's attempts to issue equity at the good market conditions.

Some empirical studies find that Market-Timing Theory predicts well capital changes in the short-run (Leary and Roberts, 2005; Kayhan and Titman, 2007). In the long-run, however, these changes are reversed. This suggests that market-timing theory is not very reliable on its own, but it can be added to other capital structure models as an explanatory factor.

The paper done by La Porta et al. (1998), which looked at the relationship between law and finance does not directly contribute to the subject of capital structure. However, many other papers used it as reference for their research of international institutional factors of capital structure.

One of such works was written by Booth et al (2001), which is very relevant for my study. They examine capital structure in 10 developing countries. The main question of the study is whether the firms in developing countries exhibit the same financial patterns as in the developed countries? Countries included in the study are India, Pakistan, Thailand, Malaysia, Turkey, Zimbabwe, Mexico, Brazil, Jordan, and Korea. This study covers a period from 1980 to 1990 using IFC data with a sample number of firms for each country from 38 to 99. They employ three definitions of leverage, which are Total Debt Ratio, Book Long-Term Debt Ratio and Market Long-Term Debt Ratio. The data for some countries is not proper, so MLTDR is not available for Brazil, and Mexico, while both BLTDR and MLTDR are not available for Thailand. The first observation is that firms in developing countries use less long-term debt than in developed countries. They run a regression of average leverage against macroeconomic indicators, such as real GDP growth, inflation, ratio of stock market value to GDP, ratio of liquid liabilities to GDP, and Miller tax term. The regression itself has too few observations, but some conclusions can be drawn. The more developed equity markets the less debt is used, and the more ratio of liquid liabilities to GDP increases debts, except for market long-term. As individual firm factors for the model they employ tax rate, business risk, tangibility, size, profitability and growth. The model is estimated with pooled OLS and fixed effects methods separately for each country. The most consistent result is that profitable firms use less debt, which is consistent with pecking-order theory. The model also shows support for tangibility, which increases the long-term leverage in the capital mix. Overall, they conclude that conventional firm factors affect capital structure in the same way, but not so consistent and with lower impact. There are systematic differences in how macroeconomic and institutional factors affect the debt ratios. So, overall authors remain skeptic about making any strong conclusions from their results. This paper is significant, as it was one of the earliest works that concentrated on a broad set of developing countries.

At the same year, one of the most famous surveys on capital structure was published. Graham and Harvey (2001) surveyed CFOs of 392 US firms, and the questions contain the topic of capital structure. By the results, the most important factors affecting leverage are financial flexibility and credit ratings. For issuing additional equity CFOs consider EPS dilution and changes in the stock price value. They find moderate support for trade-off theory and target debt ratios. Many findings are consistent with pecking order theory; however, they are caused by factors other than asymmetric information considerations.

There is another interesting survey done by Servaes and Tufano (2006). Its main advantage is that it uses sample of firms from different countries, including developing ones. They asked firms from different countries several questions regarding capital structure. The first question is what types of debt firms include in the term “debt”, and the most popular answers are long term debt maturing over year, long term debt shorter than year, and short-term debt (with percentage answered 96%, 90%, 86% respectively). The second question is does your firm have the target capital structure. The result is that 68% of firms do have target value and the proportion varied by region, with highest in Latin and North America (90% and 85% respectively), and lowest in Germany and Western Europe (67% and 56% respectively). The third question is which measures firm uses to define the target leverage. Most popular answers are EBITDA to interest (58%), debt relative to EBITDA (58%), debt to book value of equity (55%), and absolute level of debt (53%). The interesting part is that firms primarily look at their ability to keep up with interest, as the two most popular answers include EBITDA as a measure of firm's free cash flow, which can be used in debt repayment.

The very sophisticated analysis was done by Frank and Goyal (2009). They analyze which factors are most important in determining capital structure of US firms from 1950 to 2003. The work is not linked to testing theories but tries to empirically find reliable patterns. They start with the linear model that includes full list of candidate factors and exclude unreliable ones, using econometric information criterions (both Bayesian and Akaike) as a guide. Similar algorithm is applied to separated data, which were divided into two groups based on different characteristics of firms. Those pairs of factors are dividend paying and nonpaying, large and small, high and low growth. They even tried to account for missing data based on multiple imputation methods, which yield approximately the same results. Based on the analysis, the most reliable factors accounting for market leverage are median industry leverage, tangibility, logarithm of assets, inflation with positive effect, and market-to-book assets ratio, profitability with negative effect. The same set of factors are applied to the book definition of leverage, with logarithm of assets, the market-to-book ratio, and inflation. This work was so influencing, so it is already considered as classical. Still, the analysis was done for US firms, and the results cannot be universally applied to developing countries.

Work similar by scope was developed by Hang, M., et al. (2017). They conduct a Meta-regression analysis (MRA) based on randomly sampled results of 100 representative studies. Data is aggregated in a systematic way by the seven most common determinants of capital structure. These factors are tangible assets, non-debt tax shields, market-to-book ratio, firm growth, firm size, earnings volatility, and profitability. The MRA shows that the significant ones are tangibility with positive sign, market-to-book ratio with negative sign, and profits with negative sign. Variation in results are explained by publication selection bias when authors favor certain factors, which are more accepted by majority. Size of the selection bias depends on some publication characteristics. For example, this bias is less pronounced in less known articles and journals. Another factor is the definition of leverage adopted by authors. When market-based definition is used the publication selection bias is more pronounced. Same with the inclusion of short-term debt in the leverage. Overall, this paper provides useful insides looking at the situation on the higher level.

Another twist in the theory of capital structure was revival of Dynamic Trade-Off Model with adjustment costs. Strebulaev (2007) developed his version of it. This model allows him to generate data consistent with the most empirical findings, i.e. that resembles real world behavior. Using this data, he shows that cross-sectional tests can lead to conclusions that are not consistent with comparative statics of the core model. He brings attention to the problem in the validity for empirical testing of capital structures. The main argument is that in presence of small adjustment costs firms correct their leverage infrequently. Even with optimal debt ratio in mind, firms move slowly towards the target. Thus, we rarely observe firms at their target leverage levels. This happens because they adjust it at different moments and the magnitude of adjustment can be incomplete. He rejects the static trade-off model in favor of the dynamic trade-off model, which fits better with the simulated data. Another interesting contribution is that he finds another way to explain negative effect of profit on leverage in the context of dynamic trade-off model, which is considered one of the main drawbacks of the trade-off theory. The intuition behind the mechanism in model is that higher profitability affects future cash flows of the firm and increases firm value, so the leverage ratio decreases.

Danis et al. (2014) based on dynamic trade-off theory model with costly infrequent adjustment (dynamic inaction models) develop test methodology which enables them to test effect of profitability on capital structure. This paper can be considered as an answer to the point of Strebulaev (2007). As many previous dynamic inaction models noted firms' capital structure are often allowed to drift and be at suboptimal levels. To isolate cases when firms have optimal capital structure, they look at firms when they have large debt and equity readjustment. When they look at cross-section of firms when they do not make large readjustments to capital structure, the cross-sectional tests show the negative relationship between profit and leverage. This is consistent with most empirical findings, which find this fact as a support to pecking order theory. However, when we isolate the large refinancing points, the observed cross-sectional behavior shows the positive relationship between profitability and leverage choice. Then they show that in Time-Series dimension, firms which adjust their debt higher experience prior decrease in the leverage and increase in profitability. This test methodology imposes few restrictions on data and does not require exogenous variation at data. They look at the endogenous patterns emerging from dynamic inaction models. Thus, they conclude that there is no one trade-off model to explain everything, however the implications of dynamic inaction models are consistent with the observed data.

The econometric papers done by Arellano and Bond (1991) and Blundell and Bond (1998) and implementation of their methods in common statistical software opened a new field for a research in capital structure. Generalized Method of Moments (GMM) estimation methods, developed in those papers, allowed to overcome endogeneity problem in Short Dynamic Panels. This gave a boost to development of dynamic panel data models, including partial adjustment model. Most of the further works use those estimation techniques.

Drobetz and Wanzenreid (2006) investigate which factors affect target capital structure and adjustment speed towards it in Switzerland. The data consists of a sample of 90 Swiss firms over the 1991-2001 period. They build a partial adjustment model, where both target capital structure and speed of adjustment are endogenous. The following individual factors are employed: growth opportunities, size of firm, and distance from the target ratio. They also use a set of macroeconomic variables to catch time specific events, including the term spread, short-term interest rate, default spread, and TED spread. Capital structure is defined in two ways: the first is ratio of total non-equity liabilities to assets, the second is ratio of interest bearing debt to total capital. The second definition is especially useful to account for agency costs of debt. The main results from their analysis is that fast growing and firms further away from target capital structure have a higher speed of adjustment. The macroeconomic conditions that increase adjustment speed are the term spread and good economic prospects. The main limitation of this work is relatively small sample size.

The next papers I write about are more directly related to the problem of capital structure in developing countries. Fan et al. (2012) examine how institutional settings in different countries affect firm's capital structure and debt maturity choices. They use a sample 36,767 firms from 39 countries (both developed and developing) for a period 1991-2006. The dependent variables are Total Debt to Market Value of Firm and Long-Term Debt to Total Debt, which represent the capital structure and debt maturity choice respectively. First, authors analyze country financing patterns and link them to institutional settings. Then they employ regression analysis. The included firm specific factors are tangibility, profitability, size, and growth. The country specific factors are also included in the analysis, and they represent economic conditions with institutional settings. Generalized Method of Moments (GMM) technique is used to estimate the model. They include different specifications and find that institutional settings are better in explaining differences in capital structure and debt maturities than industry of the company. Another interesting finding is that in countries with higher gains from tax shields firms use more leverage. The poor institutional quality generally increases the debt-ratio and firms have shorter debt maturity. Peculiarities of economy also affect the capital structure. Bank oriented economies have shorter debt maturity; in countries with higher pension contribution more equity is used; in countries with higher pension benefits the debt has longer maturity; finally, in economies with high supply of government bonds there is less debt in capital structure and it has shorter maturity due to crowding out of corporate bonds.

Oztekin and Flannery (2012) explore how different institutional environment can affect the process of capital optimization. They use data from 37 different countries and build a partial adjustment model of leverage. The goal is to see how speed of adjustment is affected in different institutional environments. Based on indices found in the other literature they divide countries into groups sharing same institutional characteristics. These institutional factors are legal traditions, financial system structure and development, adjustment benefits and costs. All the factors they list show significant difference in speed of leverage adjustment between groups.

Oztekin (2015) follows in many ways his previous works. In this paper, he examines which factors are reliable in explaining capital structures decisions across 37 different countries. Then he looks how different institutional environment can affect the process of capital optimization. He uses a sample of 15,177 firms in total for period 1991-2006 with on average 7 years of observation per firm. Then he builds a partial adjustment capital structure model with the factors proposed by Frank and Goyal (2009) -- profitability, growth, size, tangibility, mean industry leverage, and inflation. He uses two methodologies for estimating the model. The first is separate one, where he estimates model separately for each country, which allows the coefficients to differ among countries. For a factor to be reliable, he requires it to be significant at 90% level and have a consistent sign across more than 50% countries as a rule of thumb. The second methodology is pooled, when the model is estimated for all the sampled data. He includes fixed countries effect to account for different institutional environments. The results on reliable factors on both methodologies almost coincide -- profitability is reliably negatively correlated with leverage, while size, tangibility, and mean industry level have positive reliable effect. The inflation is reliable only in pooled model with a negative sign. Then he divides countries into two groups: weak and strong institutional settings. By the separate methodology, mean industry level and size are not reliable in countries with weak institutions, while they are in strong institutional settings.

As the partial adjustment model includes costs of adjustment, Oztekin (2015) examines how different institutional features differ in costs of adjustments between strong and weak institutions. The general result is that countries with strong institutions generally have lower costs of debt and equity adjustments.

After that, the effects of institutional settings on capital structure are examined. Larger costs of debt have significant negative effect on leverage, while equity costs are insignificant. The following institutional factors have positive effect on leverage: better organized bankruptcy process, stronger creditor rights, better quality of government. The next listed institutional factors have negative effect on leverage: stronger shareholders rights, better contract enforcement, stronger law and order, greater financial transparency, better disclosure, higher level of standard enforcement, and laws preventing inside trading. Overall, this works shows how different institutional environments shape capital structure.

The next papers I am going to discuss mostly examine some particular developing regions and countries. Chakraborty (2010) considers capital structure in India. He uses sample of 1169 non-financial Indian firms from the Bombay Stock Exchange or the National Stock Exchange during the period from 1995 to 2008. The chosen factors are the following: profitability, tangibility, size, growth, non-debt tax shield, and uniqueness. He estimates models using fully modified OLS (FMOLS) and GMM techniques. Prior to FMOLS analysis he makes use of panel unit root and panel cointegration analyses and concludes that in long-run leverage is in equilibrium with its determinants. The estimation results are quite similar among both approaches, but the author prefers SMOLS as the tests show that its correct and they represent a long-run relationship. As a result, there is a negative effect on leverage with profitability, growth, size, uniqueness, and there is a positive relationship with tangibility and non-debt tax shields. He finds little evidence for agency costs of debt, and the explanation can be that in India most firm are family-owned. As a result, there is an evidence for both pecking-order and trade-off theory.

Espinosa et al. (2012) investigate the determinants of capital structure in Latin America. This paper includes sample of firms Argentina, Chile, Mexico, and Peru with sample sizes 23, 50, 41, and 19 respectively. At the same time, they have a sample of 466 US firms for comparison. The included factors are tangibility, growth, size, and profitability. Both book leverage and market leverage are used as dependent variables. They estimate a model using maximum likelihood and censored Tobit regression. While all the factors are significant in US with the usually expected signs, the results on Latin America are mixed. This might happen due to problem of endogeneity. To overcome it, they use GMM estimates. After re-estimating model, it is interesting that capital structure choice in Chile has same significant signs of factors as US for both definitions of leverage: negative profits and growth, positive tangibility and size. This can be explained by the fact that Chile has the most developed economy in the sample. Other countries show mixed result, with different significant factors for each leverage definition. Still, each remaining country has from two to three significant factors for each definition of leverage. The main limitation of this study is small sample size: for Peru there are only 19 firms and it is hard to make any accurate conclusions.

Xuan Vinh Vo (2016) did a paper on capital structure in Vietnam, one of the emerging markets. He develops a model that describes capital structure for firms listed Ho Chi Minh City stock exchange from 2006 to 2015. The used factors are growth, tangibility, profitability, firm size, and liquidity. The GMM dynamic estimates are used to control endogeneity. He estimates model for long-term and short-term leverages, and their ratio. The significant variables for long-term leverage are tangibility and size (significant at 1%), for the short-term leverage the profit is most significant with probability close to zero, while tangibility, size, and liquidity are significant at 10%. For the ratio of long to short term, all the variables are significant at 1% level, except profitability.

San Martin and Saona (2017) analyzed how different factors affect capital structure in Chile. They use some classical factors, such as firm size, tangibility, profitability, growth opportunities, as well as some unconventional, like non-debt tax shields, market timing, ownership structure, and dividend policy. Data includes 157 Chilean firms and GMM estimation is used. They find positive effects on leverage of firm's size and ownership concentration and negative effects of the pay-out policy, profitability, non-debt tax shields, and growth opportunities. Some findings are inconsistent with US market evidence, such as negative effect of tangibility.

2. Factors Choice

Profitability

The profitability is one the most debatable factors on the theoretical field. The pecking order theory implies that there is inverse relationship between profitability and leverage. Most empirical studies seem to be confirmed by most studies (Titman and Wessels, 1988; Rajan and Zingales,1995; Frank and Goyal 2009; Oztekin 2015). The effect of profits seems not change when we move to the developing countries (Booth et al. 2001; Chakraborty 2010; Espinosa et al. 2012; San Martin and Saona 2017). The underlying reason may be that debt has a higher cost of usage in the developed countries (La Porta et al., 1998).

Size

Most studies agree on the positive effect of size on the presence of leverage in the capital structure. Rajan and Zingales (1995) think this happens because larger firms have lower bankruptcy risks and there are lower information asymmetries due to more strict law requirements. Ghosh (2007) argues that larger firms have more diversified and less volatile cash-flows, so their debt capacity is higher. Bigger firms are also more financially sophisticated. This may reduce the transaction costs and lower the costs of debt. (Frank and Goyal, 2009; Booth et al., 2001). Overall, I expect the positive relationship between the leverage and firm size.

Tangibility

For developed countries, there is an established positive relationship between tangibility and leverage (Rajan and Zingales, 1995; Frank and Goyal, 2009). The reason for it is that tangible assets can be collateralized in debt contracts reducing adverse selection problems. This leads to lower costs of negotiating debt contracts and more optimal behavior of firms. However, this might not be the case for emerging economies. The traditional relationship can be inversed in settings of developing countries when tangible assets reduce asymmetries of information, so equity becomes less costly than debt (Booth et al., 2001). I expect positive significant effect of tangibility due to strong evidence of developed markets.

Growth

Empirically, higher market-to-book ratio is mostly negatively related to leverage (Rajan and Zingales, 1995; Frank and Goyal, 2009). This seems to be the same case for developing countries (Espinosa et al., 2012; Chakraborty, 2010). High growth opportunities serve as an alternative sign of high quality firm which needs less leverage (Titman and Wessels, 1988). Myers (1977) also argue that high leverage results in passing positive NPV projects, which consolidates the negative relationship between growth opportunities and leverage. Thus, I expect the negative effect of growth opportunities.

GDP Growth

GDP is the most common macroeconomic variable, yet still there is no consensus about its relationship with the capital structure. Bokpin (2009) finds that per capita GDP growth is significantly negatively associated with leverage in capital structure based on analysis of 34 emerging markets.

Inflation

Mokhova and Zinecker (2013) argue that Inflation leads to higher expected rate of return, which is negatively affects the stock markets. So, inflation should positively affect leverage, as the debt becomes relatively more attractive than equity. Frank and Goyal (2009) finds that inflation is the only relevant macroeconomic factor in explaining market leverage for US firms. I expect positive effect of inflation on the leverage.

Institutional Credit Quality

The better quality of credit institutions lowers the transaction costs and should increase the leverage of the firms. World bank Doing business methodology includes legal, private, and public aspects of the credit system based on cross-country methodology developed by Dejankov et al. (2006). I expect that the institutional quality of credits affect leverage in a positive way.

Institutional Tax Quality

The family of trade-off theories see taxes as the main reason of having debt in capital structure. Empirical works that are focused on the international capital structure often include tax system as the factors (Fan et al., 2012; Booth et al., 2001). World bank Doing business include rating index on the effective corporate rate and other aspects of paying taxes, methodology for which was developed by Djankov et al. (2010). I expect the institutional tax quality to have negative impact on leverage.

Corruption Perception

Corruption is one the significant market frictions in the developing countries. It reflects the lack of effective channel communications and increases the transaction costs. Fan et al. (2012) included corruption index in their model and it is significant for every group of countries. I use the Corruption Perception Index by Transparency International. The index increases with better anticorruption environment. I expect more leverage with higher Corruption Perception Index.

3. Model and variables

In this section, I will introduce you to the structure of my model. It includes the dependent and independent variables, and estimation methods.

The initial model can be written in a following short way:

(1)

Where is the vector of firm factors, is the vector of macroeconomic factors, and is a vector of institutional factors.

If we expand each vector, we will get the following:

The dependent variable is calculated in four ways: market value of total leverage (MTLev), which is equal to Total Debt to Enterprise Value; book value of total leverage (BTLev), which is equal to Total Debt to Total Debt; Book and Market Long-Term leverages (BLLev; MLLev) are calculated in the same manner, but with Long-Term Debt instead of Total Debt. Frank and Goyal (2009) argue that common leverage definitions yield approximately the same results in the US settings. In case of developing countries this might be not true.

All the firm variables are lagged to overcome the problem of endogeneity. The definitions of the firm variables are the following: is the measure of profitability and calculated as the return on assets; is the measure of size and calculated as the logarithm of total assets; is the measure of firm's growth potential and calculated as the Enterprise Value divided by Total Assets; represents tangibility of assets and calculated as Book Tangible Value divided by Total Assets.

The definitions of the country macroeconomic variables are the following: is the growth of GDP; is the annual inflation measured by GDP deflator.

The definitions of the country institutional variables are the following: is the rating of institutional quality of credit system estimated by World Bank Doing Business methodology; is the rating of institutional quality of tax system estimated by World Bank Doing Business methodology; is the corruption perception index estimated by Transparency International methodology, with lower values for higher level of corruption.

I do not differentiate between countries because I expect macroeconomic and institutional variables to capture the cross-country differences. So, the data is pooled for all the countries.

My goal is to test the following hypotheses:

H1: The effect of conventional firm factors has a same effect on leverage as in Developed Countries

H2.1: There is a negative association between GDP growth on capital structure

H2.2: Inflation is associated with higher leverage

H3.1: The better institutional quality of credit system is associated with higher leverage

H3.2: The better institutional quality of tax system is associated with lower leverage

H3.3: The better perception of corruption is associated with higher leverage

The objectives of first four hypotheses is to study the effect of firm's conventional individual factors on capital structure in the setting of developing country. The hypotheses 2, 3, and 4 asses the impact of macroeconomic variables. The hypothesis 5 assess the impact of important qualities of financial institutions on capital structure.

To explore some additional capital structural interactions in the developing markets, I propose to add to the model the following interactive variables: , ,. So, the model can be rewritten in the following way: