Capital Structure Choice and Firm Value : New Empirical Evidence from Asymmetric Causality Test

This study aims to analyze the possible asymmetric causal relat ionship between capital structure and firm value by employing the asymmetric causality test of Hatemi-J (2012), on a time series data of Turkish manufacturing industry (consisting of Borsa Istanbul listed manufacturing firms) for the period of 1990.Q1-2015.Q4. Test results point out a unidirectional asymmetric causal relationship between capital structure and firm value, indicating that capital structure Granger-cause firm value when shocks are negative, but not when shocks are positive. More exp licit ly, a decrease in total debt ratio leads to a decrease in the market-to-book value rat io. Considering the effect of only negative shock, this empirical finding also supports partial evidence to the validity of trade-off theory which pred icts a positive relationship between debt level and firm value.


Introduction
Before the late 1950s, the conventional finance theory defended that a moderate debt financing increases firm value, as it is less costly compared to equity financing, imply ing U-shaped cost of capital function of leverage.In 1958, a new theory by Modigliani and M iller (hereafter MM), asserting that firm value is co mpletely independent of its capital structure (choice), opened a new era in corporate finance.Forming the basis for modern thinking on capital structure, MM (1958)'s capital structure irrelevance theory is clearly seen as one of the cornerstones in the modern finance theory.However, due to its set of some unrealistic and very restrict ive assumptions which do not hold in the real world and contradiction to main assumptions of mainstream academic finance, MM (1958)'s irrelevance theory has always been debatable, paving the way for the emergence of alt ernate capital structure theories.
The emergence of alternate capital structure theories is mostly related to researchers' attempts on introducing additional rat ionalizat ions for the irrelevance proposed by MM (1958) by relaxing its very restrictive assumptions, emphasizing that the capital structure choice may so mehow be relevant to (or may d irectly and/or indirectly affect) firm value.To date, many theories with their extensions (models) have been proposed on capital structure relevancy focusing on optimal use of debt, as point of origin.These widely accepted theories may broadly be categorized into trade-off [proposed by Modigliani and Miller, 1963;Kraus and Lit zenberger, 1973;Jensen and Meckling, 1976;Scott, 1976;Miller, 1977;Kim, 1978;DeAngelo and Masulis, 1980;Grossman and Hart , 1982;Bradley et al., 1984;Jensen, 1986;Diamond, 1989;Harris and Raviv, 1990;Stulz, 1990;Chang, 1999]; and the pecking order [Myers, 1984;Myers and Majluf, 1984] theories.Though their extensions are yet limited, signaling (Ross, 1977) and market t iming (Baker and Wurgler, 2002) theories may be considered among relat ively new capital structure theories.As more new theories and their extensions are also expected to be proposed as a consequence of the evolution in capital structure related studies, it can obviously be declared that neither the debate on the empirical relat ionship between capital structure choice and firm value is still truly clarified or conclusive nor is a consensus on it, even after about 50 years fro m MM (1958)'s pioneering theory.According to Tit man and Wessels (1988), one possible reason why empirical studies in this area have lagged behind the theories is because variables related to firm attributes embedded in research models are expressed by fuzzy and not directly observable concepts.Deesomsak et al. (2004), Beattie et al. (2006), andAl-Najjar andTaylor (2008) also provide supportive emp irical evidence on the incomp leteness and inconclusiveness of understanding capital structure theories.This paper that aims to analy ze the possible asymmetric causal relationship between capital structure and firm value,

Published by Sciedu Press
76 ISSN 1923-4023 E-ISSN 1923-4031 addresses two-fold problem.The first is to provide an insight into the relationship between capital structure choice and firm value in an emerging market.While literature o n the mentioned relat ionship is immense in developed markets such as United States (US) and Europe, the body of knowledge related to emerging markets (such as Turkey) is quite limited (Ebaid, 2009;Haron, 2014;Jaros and Bartosova, 2015).Studies of Chhibb er and Maju mdar (1999) on Indian firms; Chiang et al. (2002) on Hong Kong construction firms; Abor (2005;2007) on listed firms in Ghana and South Africa; and Zeitun and Tian (2007) on Jordan firms are the rare ones constituting this body.Compared to developed markets, emerging capital and stock markets are relat ively less efficient and incomplete with higher level of informat ion asymmetry and considerable degree of irregularity (Eldo miaty, 2007).These adverse conditions lead an unfavorable financial environ ment in emerging markets that necessitates reconsideration of capital structures theories which mostly focus on the effect of optimal use of debt on firm value fro m their perspectives.The second is about the complexity inherent in the optimization process of capital structure theories.As stated by Myers (2001) that there does not exist any universal capital structure theory and any reason to expect one, the outcomes of capital structure theories are valid only under certain conditions and with restrict ive assumptions.Therefore, these theories should be empirically readdressed by more co mp lex statistical methodologies.This study tries to fulfill this gap by emp irically testing the relat ionship between capital structure choice and firm value by Narayan and Popp (2010) unit root test with two structural breaks and the asymmetric causality test of Hatemi-J (2012).
The rest of the study is as follows: in section two, the theoretical background and empirical literature on capital structure theories are reviewed.Section three is about methodology and presents data, variables, the research model and empirical findings.Section four concludes and associates empirical findings with the finance theory.

Trade-off Theories
The capital structure of a firm is the mix of debt and equity used by the firm to finance its operations.Consequently, the capital structure choice is considered as one of the most important financial decisions confronting the firm Glen and Pinto (1994).Begun with the seminal paper of MM (1958), the debate on the relationship between capital structure choice and firm value has centered on whether there exists an optimal capital structure or at least a target capital structure or not.While optimal cap ital structure is the debt-equity mix that maximizes the firm value as defined by Weston and Brigham (1996), target capital structure is the financing mix toward which firms move their capital structure over time (Frydenberg, 2011).
MM (1958)'s theory of capital irrelevance has two propositions stating that (1) in (co mpetit ive, frictionless and complete) perfect capital markets with homogenous expectations, no taxes and no transaction costs, firm value is a constant, regardless of the proportions of market values of debt and equity; and (2) the weighted average cost of capital is also a constant, in other words, capital structure of a firm has no effect on its weighted average cost of capital [see Fama (1978) fo r the requirements of MM (1958)'s first proposition].However, the very restrictive assumptions of this orig inal theory that do not hold in the world led MM (1963) to ad just their earlier proposition by incorporating tax benefits as determinants of capital structure of firms.By this adjustment, MM (1963) do not only propose that firms should finance their operations by as much debt as possible to maximize their value, as interest paid on debt is a tax-deductible expense; but also, provide a basis for the emergence of trade-off theories of capital structure.
The idea behind trade-off theories of cap ital structure is that firm has options on choosing an optimal debt -equity mix to finance its investments by balancing related costs and benefits.These theories can be categorized into t wo groups as static and dynamic t rade-off theories.Static trade-off theories suggest that the benefits versus costs of debt such as tax savings versus direct and indirect bankruptcy costs are traded off against each other and the firm determines an optimal cap ital structure to a specified point, where marginal costs of debt and equity are equal.The notion static here imp lies focusing on only a single period decision.Static trade-off theories lack of target capital structure adjustments in the long run, because in these theories firm's debt rat io is determined by considering only a single period trade-off.
In the original version of the static trade-off theory by MM (1963) and its ancestors by Kraus and Lit zenberger (1973), Scott (1976) and Kim (1978), only the corporate tax benefits of debt has been in the foreground.According to these theories, the levered firm's value is h igher than the unlevered firm's value depending on the amount of debt tax savings.Miller (1977) extends these previous versions of the trade-off theory by considering both corporate and personal tax benefits together, and imp lies that the effect of debt on firm value might be much more than when only corporate tax benefits are considered.In another extension, DeAngelo and Masulis (1980)  credits and depletion allowances.They conclude the existence of negative relationship between leverage and non-debt tax shields, and positive relationship between leverage and corporate tax rates.
Apart from these tax-based trade-off theories, some other types of debt related costs such as bankruptcy and agency costs are also taken into account in empirical studies of Scott (1976), and Jensen and Meckling (1976).Scott (1976)'s study, though it resemb les Kraus andLitzenberger (1973)'s andKim (1978)'s studies in terms of considering only corporate taxes, also deals with bankruptcy costs.According to Scott (1976), a non -bankrupt firm's future expected earnings and assets' liqu idation value determine its market value.He also concludes the existence of a positive relationship between the optimal capital structure of the firm and its assets' liquidation value, the corporate tax rate and size.Weighting the agency costs against the tax benefits of debt, Jensen and Meckling (1976)'s agency theory argues the existence of an optimal capital structure with minimal agency costs.Their theory is based on two types of agency costs as agency costs of equity and those of debt, resulting fro m t wo types of conflicts between outside stockholders and managers, and those between stockholders and debtholders.Agency costs of equity are an inevitable consequence of delegation of financial and decision making rights of (mostly d ispersed) stockholders holding well-diversified portfolios to corporate managers, because this delegation eventually leads managers to act on their own interests conflict ing with those of outside stockholders.Though Rozeff (1982) and Easterbrook (1984) suggest to increase dividend payments as an alternate to reduce agency costs of equity, Jensen and Meckling (1976) propose two other ways: increasing managers' common stock ownership in the firm and boosting use of debt.However, the latter proposition contradicts with the former one giv ing rise to agency costs of debt, as higher debt ratio means lower equity financing.Besides, stockholders may transfer wealth fro m debtholders to themselves by bidding them less-risky projects but undertaking the risky ones (Jensen and Meckling, 1976: 335).As Jensen and Meckling (1976)'s agency theory argues the existence of an optimal capital structure with minimal agency costs without assumptions related to both taxes and bankruptcy costs, it can be regarded as comp lementary -rather than competing-to static trade-off theories as it considers debt as a control mechanism.
Another agency cost of debt related capital structure theory considering debt as a control mechanis m is the free cash flow theory of Jensen (1986).He defines benefits of using debt in his own words as "control hypothesis" (see Jensen, 1986: 324).As the free cash flows are controlled by firm managers, they may tend to waste firm's resources by allocating them to relat ively low-return p rojects instead of div idend payments, lead ing possible declines in the stock prices.To overcome this problem of agency costs of free cash flow, substitution of debt for equity is favored, as use of debt reduces the amount of cash controlled by firm managers.This argument on the effect of leverage as a control mechanis m is also supported by Grossman and Hart (1982), and Williams (1987) fro m a different perspective as through the threat of liquidation which eventually causes losses in salaries and repu tation of firm managers.
According to the proponents of trade-off theories, relatively profitable firms tend to borrow more to utilize tax advantages, as they have higher income to be shielded.Therefore, the relat ionship between debt level and such firms' financial performance (in terms of both profitability and market value) is expected to be positive.However, related emp irical literature yields mixed results.Though empirical studies of Taub (1975), Givo ly et al. (1992), Bos and Fetherston (1993), Petersen and Rajan (1994), Roden and Lewellen (1995), Champion (1999), Ghosh et al. (2000), Hadlock and James (2002), Abor (2005), Mo llik (2005), and Berger and Bonaccorsi di Patti (2006) support this expectation; several researchers testing the effects of debt on financial performance have reached contrary results.Fama and French (1998)'s study on the interrelationship among taxation, financing decisions and firm value concludes that excess use of debt does not create any tax benefits due to agency problems re sulting in negative relationship between debt level and financial performance.Studies of Kester (1986), Graham (2000), Booth et al. (2001), and Mesquita and Lara (2003) also emphasize the same negative relationship.Myers (1993) considers this inverse relationship between debt level and financial performance as the most devastating evidence contrary to the validity o f the trade-off theories.Another serious deficiency of these theories is the inconsistency between the firms' debt ratios predicted by the theories themselves and those of observed in the real world.According to Custodio and Ferreira (2013)'s study on the sample consisting of observations of Co mpustat firms fro m 1976 to 2008, the mean debt ratio of US firms is 0.27, a relatively very low ratio compared to what the trade-off theories predict.
Though limited in nu mber, related literature also includes empirical studies on agency and bankruptcy costs.Friend andLang (1988), andFirth (1995)'s studies on the agency problem proposed by Jensen and Meckling (1976) indicate the existence of negative and positive relationships between debt level and management ownership; and between debt level and external ownership, respectively.Empirical studies on the effect bankruptcy costs on capital structure choice mostly focus on the significance of this effect.Warner (1977)'s study on twenty railroad bankruptcies during the period of 1930-1935 reveals that bankruptcy costs are approximately 1% of a firm's market value long before bankruptcy.However, as time to bankruptcy approaches, these costs acceleratingly increase up to more than 5%.ISSN 1923-4023 E-ISSN 1923-4031 Studies of Miller (1977), Robertson and Tress (1985), and Pham and Chaw (1989) confirm this finding.
By the beginning o f the 1980s, static trade-off theories have begun to become discredited and fall out of fashion due to related incongruent empirical findings, and the era of dynamic trade -off theories have begun.Though the history of dynamic trade-off theories may be dated back to Stig lit z (1973)'s study on taxation, this study is not directly related to corporate finance theory, as the effects of taxation have been examined fro m the public finance perspective.Therefore, it is more appropriate to mention that the first dynamic trade -off theories of capital structure have been proposed by Brennan and Schwartz (1984), and Kane et al. (1984).These theories weight tax savings against bankruptcy costs under uncertainty, ignoring transaction costs.Though the rationale behind them is as the same as the static ones, they have some distinguishing features.One distinguishing feature is that they consider mu ltiple periods allowing dynamic adjustments in the firm's capital structure.In this regard, firstly Brennan and Schwartz (1984), Kane et al. (1984) and then Fischer et al. (1989) and Lewis (1990) argue the existence of a debt rate corridor within which firm's debt rat io is allowed to float over time.In cases that the debt ratio crosses the bounds of the corridor, firm reactively adjusts its capital structure to the optimal leve l.This ad justment, as the outstanding characteristic of the dynamic trade-off theories, allows these theories to be used in adoption of their static versions with the other capital structure theories that reject the existence of an optimal (or at least a target) capital structure, such as the pecking order and market timing theories.The adjustment may somehow be costly and its speed may differ.According to Leary and Roberts (2005), and Strebulaev (2007), the effect of adjustment costs is perceived mostly on leverage, rather than on capital structure changes.Therefore, the adjustment is required to be made gradually over t ime at a certain speed of adjustment, as supported by the extant literature (see, Welch, 2004;Kayhan and Titman, 2007;Bessler et al., 2008;and Huang and Ritter, 2009).Ho wever, there is no consensus on the magnitude of speed of adjustment.While related studies of DeMiguel and Pindado (2001), Ozkan (2001), Hussain (2005), and Flannery and Rangan (2006) indicate that the speed of adjust ment is relat ively high, some other studies of Fama and French (2002), and Huang and Ritter (2009) support evidence for relatively low speed of adjustment.
The other distinguishing feature is that wh ile the static trade-off theories restrictedly address benefits versus costs of debt such as tax savings versus direct and indirect bankruptcy costs; the dynamic trade -off theories broadly deal with the trade-off between the benefits of debt tax shields and a wide range of d iverse debt related costs, including n ot only bankruptcy costs, but also agency costs and other various costs of debt (Cheng and Tzeng, 2014).Modigliani (1982)'s study as an extension of Farrar and Selwyn (1967) on the interaction between marginal value of debt and inflation, Barnea et al. (1987)'s mu ltiperiod capital structure model on the differential costs of debt and equity financing, and firm growth possibility; and Goldstein (2001)'s study on the option values of shifting leverage related decisions to the next period as the other earlier extensions and studies of Sarkar and Zapatero (2003), Leary and Roberts (2005), Alti (2006), and Hovakimian (2006) as more recent extensions support emp irical evidence for the validity of the dynamic trade-off theories of capital structure.

The Pecking Order Theory
The orig in of the pecking order theory dates back to Donaldson (1961)'s book including a survey study on US firms.
Refuting the argu ment of existence of an optimal (or at least a target) cap ital structure, he suggests firstly internal funds to be used for financing, and follo wing this the preference of debt to equity.The empirical justification of his suggestion is that the insistence on debt financing for profitable firms should be unreasonable and become slighter due to incremental increase in profits, as such profits could be used for financing with min imu m cost.Unfortunately, his suggestions were not only paid so much attention, but also faced destructive criticis m by some scholars such as Myers (1984), Shyam-Sunder and Myers (1999), and Fama and French (2002).Later on, advancing Donaldson (1961)'s pioneering study, Myers (1984), and Myers and Majluf (1984) have increased the awareness of the so -called the pecking order theory.Though their theory is similar to its originated theory of Donaldson (1961) in terms of financing hierarchy, it d iffers fro m its ancestor by incorporating asymmetric info rmation and adverse selection, and signaling problems that have previously been discussed by Leland andPyle (1977), andRoss (1977).Myers (1984), and Myers and Majluf (1984) base their theory on the assumptions of perfect capital market, except that only owner(s) and/or manager(s) (insiders) of the firm have true information about the true value of firm's assets and growth opportunities, but the outside investors do not.Therefore, the precise valuation of securities to be issued to finance new investments is unfeasible for the outside investors.As a result of this information asymmetry between insiders and outsiders, market value of firm's new stocks may be undervalued relative to their fair value in non-existence of informat ion asymmetry.In such cases, as issuance of new stocks may destroy existing stockholders' value; financing of new investments will firstly be made by internal resources (i.e.retained earnings), then -if internal resources may remain insufficient-by debt financing over equity financing.Due to this financing hierarchy, the pecking order theory rejects the existence of an optimal (or at least a target) capital structure for firm to be ISSN 1923-4023 E-ISSN 1923-4031 pursued and the debt ratio of the firm reflects the cumulative requirement for external financing decisions, as contrary to trade-off theories.Consequently, the relat ionship between debt level and firms' financial performance (in terms of both profitability and market value) is expected to be negative for this theory.
Related emp irical literature on the pecking order theory also yields mixed results.Studies of Kester (1986) on a comparison of debt policies of US and Japanese manufacturing firms; Krasker (1986), Narayanan (1988), Tit man and Wessels (1988), Pinegar and Wilbricht (1989) on US firms; Heinkel and Zechner (1990), Remo lona (1990) on US, United Kingdom, German and Japanese firms; Allen (1993) on Australian firms; Koh et al. (1993) on S ingapore firms; Griner and Go rdon (1995) on Un ited Kingdom SM Es; Kamath (1997) on New Yo rk Stock Exchange listed firms; Wald (1999) and Booth et al. (2001) on firms fro m different countries; Colo mbo (2001) on Hungarian firms; Fama and French (2002), and recent studies of Bharath et al. (2009), Autore and Kovacs (2010), and Bessler et al. (2011) all provide emp irical evidence supporting this theory.In accordance with these empirical find ings, Shyam-Sunder and Myers (1999) clearly evaluates the pecking order theory as an excellent first order descriptor of corporate finance behavior.Though it is obvious that the pecking theory is widely accepted, there exists an emp irical literature inconsistent with above mentioned empirical findings.For instance, studies of Mikkelson and Partch (1986), Brennan and Kraus (1987), Noe (1988), Constantinides and Grundy (1989), Ang and Jung (1993), and Bos and Fetherston (1993) contradict with assumptions of the pecking order theory.In another study of Frank and Goyal (2003), it is indicated that for US firms, the priority and do mination of debt financing over equity financing is not valid and deficits are mostly financed by equity, as contrary to financing hierarchy suggested by the pecking order theory.This finding may be clarified by Zhang and Kanazaki (2007: 34)'s comparison of "ideal vs. reality".The pecking order theory ideally assumes that while relatively small firms having more severe problems with asymmetric informat ion and adverse selection have tendency to finance their deficits by debt, the larger ones having insignificant problems with asymmetric informat ion and adverse selection mostly prefer equity financing.However, the realization is quite different that smaller firms face with more bankruptcy risks and agenc y problems leading difficulties in debt financing.Therefore, they rely heavily on equity issues rather than debt issues.Contrary to this, relatively large firms with lower bankruptcy risks, mo re reliab le management and easier access to debt financing mostly issue debt to finance their deficits.

Signaling Theory
In the trade-off and pecking order theories, cap ital structure serves as part of solution to problems of over-and underinvestment and changes in capital structures of firms cannot be regarded as a signal of quality of their financial position.This inefficacy has led researchers to propose alternate theories in which investment is fixed and capital structure serves as a signal of private insider information (Harris and Raviv, 1991).Ross (1977) 's theory is the ancestor of so-called signaling theory.His theory, assuming that the true distribution of firm returns is only known by managers -not by investors-, states that while firms with favorable prospects can raise additional capital by debt issuance, firms with unfavorable prospects can raise additional capital by equity issuance.Therefore, relat ively high debt levels are considered as a signal of higher quality for the firm.Managers of such highly leveraged firms are also considered as successful, as the market overvalues their securit ies.On the other hand, managers of lo w quality firms with unfavorable prospects cannot imitate higher quality firms by issuing more debt, because marginal expected bankruptcy costs for any debt level for those firms is higher.Hein kel (1982), Poitevin (1989) and Stein (1992) have extended the signaling theory by using models similar to Ross (1977)'s, but with some different assumptions, concluding that managers' choices of leverage are determined by subjective considerations.Supporting evidence for the signaling theory, Baker and Wurgler (2002), Welch (2004), and Kayhan and Titman (2007) ind icate that a firm's capital structure is significantly related to its historical stock prices.Masulis (1980), Baker et al. (2003), and Antweiler and Fran k ( 2006) also support the signaling theory assembling evidence on the positive market reaction on leverage increasing transactions.However, Agrawal and Mandelker (1987), Friend and Lang (1988) and Berger et al. (1997)'s empirical findings contradict with the signaling theory.

Market Timing Theory
Till the end of 1990s, so me firm-specific variables such as profitability, size, asset structure and tangibility were at the center of the debate on capital structure determinants .However, since the beginning of 2000s, Baker and Wurgler (2002), and Alti (2006)'s studies on the effects of capital market variables (such as issuance of stocks) on capital structure have led the emergence of a contemporary capital structure theory name d market timing.Until this theory, the idea that stock market performance is a determinant of issuance of new stocks as emp irically tested by Lucas andMcDonald (1990), andKorajczyk et al. (1992) has not gained dominance as a new strand of capital struct ure theories.Baker and Wurgler (2002) define the term "market timing" as the practice of issuing new stocks when the existing ones are perceived to be overvalued and buying them back when they are undervalued.The fluctuations in stock prices in such opportune times of under-and over-valuation affect capital structure of the firm as a determinant.In times of under-valuation, debt issuance becomes preferable.However, in t imes of over-valuation, the existing stockholders benefit fro m issuance of new and over-valued stocks.According to the market timing theory of Baker and Wurgler (2002), the effect of equity (stock) market timing on capital structure is persistent.They develop an external financing-weighted average market-to-book value rat io to analy ze market timing attempts and empirically demonstrate that their rat io is negatively related to leverage, indicating the effects of external financing on capital structure.
Co mpared to trade-off and pecking order theories, both theoretical and emp irical studies regarding market timing theory are yet underdeveloped and very rare.Leary and Roberts (2005), and Kayhan and Tit man ( 2007)'s studies on the role of market timings in capital structure related decisions, and Rossi and Marotta (2010)'s study on Brazilian IPOs in the period of 2004-2007 are among these rare studies supporting market timing theory.

Data, Variables and the Research Model
The data used in the analysis focuses on the period of 1990.Q1 -2015.Q4 for the Turkish manufacturing industry, consisting of Borsa İstanbul (BIST) listed manufacturing firms.Research model includes two variab les: market-to-book value rat io (denoted as M/B) as the dependent variable, and total debt ratio (denoted as TD) as the independent variable.
M/B is a widely-used proxy for firm value in finance literature.It indicates both the value of common equity attributed by the market (Lee and Makhija, 2009) an d the ability of firm managers to use its assets effectively.Besides, M/B is also an indicator of riskiness of the firm (Griffin and Lemmon, 2002).In studies of Nayyar (1993), Becker and Gerhart (1996), Wiggins and Ruefli (2002), Chang (2003), Cho and Pu cik (2005) and Short et al. (2007), M/B is used as a market measure reflecting firm value.
Total debt ratio as a variable indicat ing the proportion of debt used to finance firm's assets is included in the research model regard ing the capital structure of the firm.Though there exists a very vast literature regard ing capital structure, the concept of leverage (debt used for investment purposes) does not have a clear-cut definit ion.The main problem here is whether or not to use book value or market value of debt in calculations.According to the defenders of book value, it is more reliab le than market value, as it does not fluctuate over time; and it is a mo re conservative approach to estimate debt ratios with their book values (Porras, 2011: 119-120).Besides, calculating the market value of debt is quite difficult, as the number of firms carrying their debt in bond form is relat ively rare.Therefo re, in this study, the book value of debt is used in calculations.
Definitions and calculations about the variables of the study are summarized in Table 3.
Table 1.Variables of the Research Model

Empirical Findings
This study aims to find out the possible effect of capital structure on firm value v ia various empirical analyses including the unit root test with two structural breaks developed by Narayan and Popp (2010) and the asymmetric causality test of Hatemi-J (2012).

Narayan and Popp (2010)'s Unit Root Test with Two Structural Breaks
In this study, one of the unit root tests with structural breaks which was recently introduced into the literature developed by Narayan and Popp (2010) Perron (1989), ignoring structural breaks as in Dickey Fuller test may lead to false acceptance of the unit root null hypothesis.Therefore, he develops the augmented Dickey Fuller test with one break.Later, Narayan and Popp (2010) propose a test allowing for structural breaks which resembles Lee and Strazicich (2004)'s test in the context of null hypothesis.Narayan and Popp (2010) consider some unobserved components model to represent the data generating process (DGP).The DGP of a time series   has two components, a deterministic co mponent (  ) and a stochastic component (  ) as given below: with   ∼  (0,   2 ).It is assumed that the roots of the lag polynomials  * () and (), which are of order  and , respectively, lie outside the unit circle.Narayan and Popp (2010) consider two d ifferent specifications which are both for trending data.One of the specifications allows for t wo breaks in levels (as denoted by 1), while the other allows for two breaks in levels as well as slope (as denoted by 2).Both 1 and 2 model specificat ions differ in terms of how the deterministic component,   , is defined: Here,  , ′ ,  = 1, 2, denote the true break dates.The parameters,   and   , indicate the magnitude of the level and slope breaks, respectively.The inclusion of Ψ * () in Equations ( 4) and ( 5) allows breaks to occur slowly over time.
The -statistics of  ̂, denoted   ̂, in Equations ( 7) and ( 8) is used to test the unit root null hypothesis of  = 1 against the alternative hypothesis of  < 1.The break dates are selected using the sequential procedure.
The empirical results of Narayan and Popp (2010)'s unit root test with two structural breaks g iven in Tab le 2 indicate that M/B is stationary at level; and TD is stationary at first difference and is integrated of order one, I(1) fo r both M1 and M2.Some of the structural break dates estimated such as 1998.Q2, 1999.Q3 and 2000.Q4 point out periods during which Turkish economy has faced with severe economic and financial crises, while the others, 2003.Q3 and 2004.Q1, refer the beginning years of economic recovery and growth following these crises.Narayan and Popp (2010).*, ** and *** denote significance levels of 1%, 5% and 10%, respectively.

The Asymmetric Causality Test of Hatemi-J (2012)
The starting point of analyzing the causality relationship between time series (i.e., determin ing whether one time series is useful in forecasting another one) can be attributed to Granger (1969) causality test which uses a vector autoregressive (VAR) model in analy zing the causality relationship.However, the F test procedure in this test loses its validity if the data are non-stationary, and it is known that time series data are mostly non -stationary leading the problem of spurious regression (Maddala, 2001).To overco me this proble m, Toda and Yamamoto (1995) propose a superior test which can be performed in case of existence of integrated variab les and even if there exists no cointegration between the variables.Unfortunately, Toda and Yamamoto (1995)' test on Granger non -causality cannot be performed to examine the possible asymmetric relationship between the variables.Later on, Hatemi-J (2012) suggests an asymmetric causality test by extending Granger and Yoon (2002)'s study to test only cointegration, -which is based on the idea of transforming data into both cumulative positive and negative changes -, to causality analysis.
Hatemi-J (2012) examines the casual relationship between two integrated variables  1 and  2 defined as the following random walk processes: where  = 1,2, … , the constants  1,0 and  2 ,0 are the init ial values, and the variables  1 and  2 signify white noise disturbance terms.Positive and negative shocks are defined as given:  1 + = max ( 1 , 0) ,  2 + = max ( 2 , 0) ,  1 − = min ( 1 , 0) , and  2 − = min ( 2 , 0) , respectively.Therefore, it can be exp ressed as and It is possible to define the positive and negative shocks of each variable can be defined in a cumulat ive form as , and . Here, each positive and negative shock has a permanent impact on the underlying variable.In the next step, the causal relat ionship between these components is tested by focusing only on the case of testing for causal relationship between positive cumu lative shocks by using the vector   + = (  1 + ,  2 + ) .It is also possible to conduct tests for causality between negative cumu lative shocks by using the vector   − = (  1 − ,  2 − ) .The causality test is implemented by using vector autoregressive model of order , VAR () given as: where   + is the 2 × 1 vector of variab les,  is the 2 × 1 vector of intercepts, and   + is a 2 × 1 vector of erro r terms which corresponds to each of the variables representing the cumulat ive sum of positive shocks.The matrix   is a 2 × 2 matrix of parameters for lag order ( = 1, … , ).The optimal lag order () is selected by using the information criterion suggested by Hatemi-J ( 2003 ) ,  = 0, … , . ( where |Ω ̂| is the determinant of the estimated variance-covariance matrix of the error terms in the VA R model based on lag order ,  is the number of equations in the VAR model and  is the number of observations. After selecting the optimal lag order, the null hypothesis that th element of   + does not Granger-cause the th element of   + .The following denotations are used in order to define a Wald test in a compact form: The VAR () model can compactly be defined as given below: The null hypothesis of non-Granger causality,  0 ∶  = 0, is tested by the test method given below: where  16) has an asymptotic  2 distribution with the number of degrees of freedo m equal to the number of restrictions to be tested (in this case equal to ).However, as financial data do not usually fo llo w a normal d istribution, the existence of au toregressive conditional heteroscedasticity (ARCH) effects is mostly inevitable.Therefore, to overcome these problems, the bootstrapping simulation technique is used [see Hatemi-J (2012: 451) for details].
The null hypothesis of non-Granger causality is rejected at the  level of significance, if the Wald test generated in the final step is greater than the bootstrap critical value.The bootstrap critical values are produced for three different significant significance levels as 1%, 5% and 10%, respectively.
The emp irical results of Hatemi-J (2012) asymmetric causality between the market-to-book value ratio (M/B) and the total debt ratio (TD) are given in Tab le 3. Based on these results, only the 6 th null hypothesis that TD -does not Granger-cause M/B -can be rejected, while all other null hypotheses cannot be rejected.This result points out a unidirectional asymmetric causal relationship between TD and M/B, indicating that total debt ratio Granger-cause market-to-book value rat io when shocks are negative, but not when shocks are positive.More exp licitly, a decrease in total debt ratio leads to a decrease in the market-to-book value ratio.

Conclusion
Despite the existence of a vast theoretical and empirical literature on the relationship between capital structure and firm value, there is still no consensus on an answer to the question of what the optimal leverage that maximizes firm value is.The debate on this subject centers on whether there exists an optimal (or at least a target) capital structure or not.The idea behind trade-off theories of capital structure [proposed by Modiglian i and Miller, 1963;Kraus and Litzenberger, 1973;Jensen and Meckling, 1976;Scott, 1976;M ille r, 1977;Kim, 1978;DeAngelo and Masulis, 1980;Grossman and Hart, 1982;Bradley et al., 1984;Jensen, 1986;Diamond, 1989;Harris and Raviv, 1990;Stulz, 1990;Chang, 1999] is that firm has options on choosing an optimal (or at least a target) capital structure to finance its investments that maximizes firm value by balancing related costs and benefits of debt.However, the pecking order theory of Myers (1984), and Myers and Majlu f (1984); and market timing theory of Baker and Wurgler (2002) reject the existence of an optimal capital structure.
This apparent incompleteness and inconclusiveness of understanding capital structure theories has been subject to the studies of Tit man and Wessels (1988), Bevan and Danbolt (2002), Deesomsak et al. (2004), Beattie et al. (2006), Al-Najjar andTaylor (2008), andHaron (2014).According to these studies, the variety, ambiguousness and complexity of variab les related to firm attributes embedded in research models; both theoretically and empirically concentration on only a few of the developed theories, and using very different leverage definitions may be the major reasons.One solution to overcome the problem of inco mpleteness and inconclusiveness may be empirically readdressing capital structure theories by mo re co mp le x statistical methodologies.The other scantiness of studies on capital structure theories is the lack of diversity in samp le choice.While most of related studies focus on developed markets such as United States (US) and Europe; the body of knowledge rela ted to emerging markets have remained quite limited (Ebaid, 2009;Haron, 2014;Jaros and Bartosova, 2015).Th is study tries to fulfill these gaps by emp irically testing the relat ionship between capital structure choice and firm value by Narayan and Popp (2010) unit root test with two structural breaks and the asymmetric causality test of Hatemi-J (2012) by using an emerging market time series data.
In the analysis, firstly, the stationarity of variables is tested by Narayan and Popp (2010) unit root test with two structural breaks.The estimated break dates refer to dramat ic turning points in Turkish economy.So me of these dates (1998.Q2, 1999.Q3 and 2000.Q4) point out periods during which Turkish economy has faced with severe economic and financial crises.Following the Asian and Russian crisis in 1997 and 1998, the attention of foreign investors to Turkey began to dimin ish as a result of loss of confidence.This led a sharp decrease in capital inflows into Turkey causing the economic growth rate decrease down fro m 7.5% in 1997 to 2.5% in 1998.A long with a very destructive earthquake in August 1999 hitting the major industrial zones of Turkey, the economy fell into a deep recession.On 22 December 1999, an exchange rate based stabilization program -also supported by the International Monetary Fund-was announced by the government in order to decrease the inflat ion rate to 10% by the end of 2001.Ho wever, this program fell short of solving problems.During 2000 and 2001, Turkish economy experienced very sev ere banking crises.Follo wing th is turmoil, a political crisis triggered by a dispute between the prime minister and the president about solving the problem of corruption in the banking sector emerged (Özatay and Sak, 2002).Trust in the sustainability of the program disappeared quickly and a currency crisis occurred.These adverse economic conditions also adversely affected market capitalizat ion and capital structure of Turkish firms.Market value of BIST (formerly known as Istanbul Stock Exchange) listed firms fell down to US$ 34.4 billion in 2002, co mpared to US$ 114.3 billion in 1999(CM B, 2002).After the 2000 and 2001 banking crises, the share of bank loans in total corporate sector liab ilit ies declined sharply.Firms either tended towards self-financing or began to use other forms of debt including trade credits.The increase in use of trade credits may also result fro m banks' panic and unwillingness on the supply of credit to corporate sector and from the absence of a well-developed bond markets.Then, to test the possible asymmetric causal relationship between capital structure and firm value, a recent approach of Granger causality developed by Hatemi-J ( 2012) is performed.The superiority of this test is that it considers the impacts of both positive and negative shocks of one variable on the other, remarkably differentiat ing fro m others allo wing no asymmetry in the causality testing.The empirical results support evidence on the existence of a unidirectional asymmetric causal relationship between capital structure and firm value, indicating that capital structure Granger-cause firm value when shocks are negative, but not when shocks are positive.More exp licit ly, a decrease in total debt ratio leads to a decrease in the market -to-book value rat io.Considering the effect of only negative shock, this empirical finding also supports partial evidence to the validity of trade-off theory.Because, the trade-off theory assumes that there are benefits to use debt within a capital structure, till the optimal cap ital stru cture is reached.Therefore, decrease in total debt within the capital structure will inevitably cause decrease in firm value.
=  () and vec indicates the column-stacking operator; ⨂ represents the Kronecker product, and  is a  × (1 +  ) indicator mat rix with elements ones for restricted parameters and zeros for the rest of the parameters.  is the variance-covariance mat rix of the unrestricted VAR model estimated as   =  ̂ ′  ̂ − , where  is the number of parameters in each equation of the VA R model.When the assumption of normality is fu lfilled, the Wald test statistics in Equation ( The other break dates(2003.Q3 and 2004.Q1)  refer the beginning years of economic recovery and growth following above mentioned crises.Between 2001 and 2006, Turkish economy grew appro ximately 7% in average terms, the inflation rates began to reduce and foreign d irect investments (FDIs) sky-rocketed especially after 2004.In 2006, the amount of FDIs to Turkey reached up to US$ 20 b illion, co mpared to only US$ 16 billion between 1980 and 2003.The trust on banking sector was restored.The ratio of non -performing loans to total loans was reduced to 0.4% in 2007, co mpared to 16.6% in 2001.The market value of BIST listed firms increased to appro ximately US$ 190 billion in 2007 (USAK, 2008).
analyze leverage-firm value relationship fro m the perspective of non-debt tax shields such as depreciation deductions, investment tax

Table 2 .
The Results of Narayan and Popp (2010)'s Unit Root Test Break dates are given in parenthesis.Crit ical values are obtained fro m

Table 3 .
The Results of the Asymmetric Causality Test of Hatemi-J (2012) Maximu m lag length is 4, and the HJC is used to determine the optimal lag length.a Critical values are obtained through 10,000 bootstrap replications.** denotes significance level of 5%.