Is the extremely low-leverage policy a deliberate decision?

(1)

Is the extremely low-leverage policy a deliberate decision?

Master Thesis

Author:

Živilė Paulauskaitė 10604715

Supervisor: Dr. Liang Zou

University of Amsterdam

Faculty of Economics and Business (FEB)

Business Economics – Finance

(2)

2

Acknowledgements

I would like to thank dr. Liang Zou for the supervision and dhr. dr. J.E. Ligterink for the useful tips during the Thesis Seminar. I would like to thank family Paulauskai and family Commandeur for all the help and support. Special thanks to Niels Commandeur; I dedicate my thesis to you.

(3)

3

Abstract

This research is dedicated to answer the question “Is the extremely low-leverage policy a deliberate decision?” The extremely low-leveraged firms are analyzed in the market timing framework as well as the probability of a firm to be zero or almost zero-leveraged is tested. The results show that the decision to have zero or almost zero leverage is deliberate because the firms do not reach the extremely low leverage by market timing activities as well as the firms actively readjust their capital structure in the longer period. What is more, financial constraints are not the reason companies are zero or almost zero-leveraged. Further, the management’s ownership does not increase the probability of a firm to be extremely low-leveraged. Lastly, the analysis of the capital structure items revealed that the extremely low-leveraged firms are younger, smaller and rich with cash.

(4)

4

1. Introduction

Capital structure is a well-known topic in the academic literature. There are theories such as the trade-off theory or the pecking-order theory developed that address the choice between debt and equity. Both theories advocate taking up the debt: in the trade-off theory the advantage of the debt is explained by the tax-shield that firms get while the pecking-order theory proponents show that the advantage of the debt is lower agency costs. What is more, there is also academic literature on empirical evidence both on the theories (Fama, French (2002)) as well as on the evidence from surveying the CEO’s of firms about the capital structure (Graham, Harvey (2001)). Still, even though capital structure is discussed widely in the literature, it is not well-explained why companies choose a particular debt ratio. Among non-answered questions the puzzle why some firms choose not to have debt at all arises. Intuitively, one could argue that the unlevered firms are rare; however, there is evidence that the zero or almost zero companies are not an exception (Kortweg (2010)). It could be argued that being zero-leveraged is not beneficial because these companies leave the money on the table while not using the tax shield available. What is more, the choice of debt does not seem to fit to any of the predictions of theories on the capital structure. Therefore, it is interesting to research the reasons of such debt ratio as well as the capital structure patterns within the group of extremely low-leveraged companies.

In this paper the firms are researched in different aspects: the behaviour before becoming zero or almost zero-leveraged; the time when the firms keep the extremely low-leverage; and the probability of a company to be zero or almost zero-leveraged is tested. The question addressed in this paper is: “Is the extremely

low-leverage policy a deliberate decision?” To answer this question this paper integrates the components of

market timing theory and conventional capital structure elements into empirical research on zero or almost zero leverage companies. Firstly, it is tested whether the companies tend to time the market before they become zero or almost zero-leveraged. The hypothesis “ : Companies with almost zero leverage reach this level not by the issuing activities” is tested. The methodology used to test the first hypothesis is similar to

the methodology proposed by Baker, Wurgler (2002); in this paper adjusted with extra control variables. The book leverage is regressed against the weighted average market-to-book ratio, conventional market-to-book ratio and other control variables. The market timing activities are captured by the weighted average market-to-book ratio which corresponds to the historical valuation of the firms. The control variables which are included in this paper are tangibility, profitability, size, cash, capital expenditures, R&D expenditures and age. In one of the specifications the industry-fixed and time-fixed effects are included. Secondly, the proactive readjustment of the capital structure is tested to check whether the extremely low-leveraged firms readjust their market leverages to the target debt ratios. The hypothesis “ : Extremely low-leveraged companies readjust the capital structure” is tested. The adjusted methodology proposed by Welch (2004) is

used: the market leverage is regressed against the actual debt ratio and implied debt ratio. In this way it is tested to what extent the firms actively readjust to a target debt ratio and to what extent the market debt

(6)

6 ratio is allowed to float with the stock returns. Different readjustment periods are tested: 1 year, 2 years, 3 years and 5 years, to analyse the non-readjustment or readjustment behaviour within different time horizons. Lastly, besides researching the market timing activities of the zero or almost zero-leveraged firms, the capital structure items are discussed in order to see which of them increase the probability of companies to be extremely low-leveraged. In this part of the research two variables – equity ownership and financial constraints – are chosen as the variables of interest. Two hypotheses are tested: “ : The probability of a company to be extremely low-leveraged increases with the management’s equity ownership” and “ : The probability of a company to be extremely low-leveraged increases with the financial constraints”. To test the

hypotheses the probit model is used where the outcome is 1 if the firm is extremely low-leveraged and 0 if not. For the Hypothesis 3a the management ownership is defined as the percentage of the equity owned by top managers together to the total equity. For the Hypothesis 3b the SA Index (Hadlock, Pierce (2010)) is used as a proxy to determine the financial constraints of the firms. What is more, the control variables for the Hypothesis 3 are the same as the ones in the Hypothesis 1 which are tangibility, profitability, size, cash, capital expenditures, R&D expenditures and age. The industry-fixed and time-fixed effects are also included in one of the regression specifications.

The robustness checks encompasses the analysis of the firms that have zero short-term and long-term debt. It is checked whether the market timing behaviour also applies for these firms as well as the probability of a firm to have zero debt (short-term and long-term) is tested. What is more, additional results are given for the Hypothesis 3: the variables of interest are defined differently. Firstly, the financial constrains are measured by different proxy: the KZ Index (Lamont et al. (2001)). Secondly, the influence of the CEO ownership on the probability of a firm to be extremely low-leveraged is researched. Lastly, instead of the ownership, the dummy for a CEO turnover is taken up as a variable for further insight.

This paper contributes to the existing academic literature in several ways. First of all, the contribution of this paper is the focus on zero or almost zero-debt companies’ activities in the framework of the market timing. Baker, Wurgler (2002) find that the debt ratios depend on the historical market valuation of the firms as well as Welch (2004) state that leverage could be explained by the stock returns of the companies. However, the authors do not research explicitly the all-equity firms. What is more, another contribution of this paper is that the main research addresses the true zero or almost zero-debt companies. This comes from the two ways of determining book leverage. First one is a conservative way when the debt is determined as the total assets less the equity. Second way is when the debt is considered to be divided into borrowings and operational debt. In this way the leverage is the sum of the short-term debt and the long term debt (i.e. borrowings). The former way to describe the leverage is not used in the academic literature about the zero-leverage companies. Therefore, this paper addresses extremely low-zero-leverage firms where the zero-leverage is defined in the conservative way.

(7)

7 For the research I use the data of U.S. publicly traded firms obtained from the COMPUSTAT database as well as the data on the equity ownership and CEO’s obtained from the Execucomp database. After cleaning the data, the total sample consists of 193539 firm-year observations out of which 4571 is the firm-year observations of the extremely low-leveraged firms. The main results show that the companies do not time the market before they become zero or almost zero-leveraged. Secondly, the companies start timing the market after keeping the zero debt level for sufficiently long period of time (7 years). What is more, the zero debt firms tend to readjust their capital structure in the longer run. The readjustment is not observed within 1 year; however, in the longer periods firms tend to readjust their market leverage. Therefore, it is concluded that the market timing is not the reason firms reach and keep the extremely low leverage which suggest that such leverage policy is a deliberate decision. What is more, researching the probabilities of companies to be extremely low-leveraged it appears that the financial constraints are not the factor of the companies to be extremely leveraged. This contributes to the conclusion that the extremely low-leverage policy is a deliberate decision as the firms do not borrow even though they are able to. Furthermore, the management’s equity ownership and CEO equity ownership decrease the probability of a firm to be zero or almost zero-leveraged. This could be seen as counter intuitive results; however, the turnover of the CEO appears to be an important factor that increases the probability of a firm to have zero or almost zero leverage. Lastly, analyzing the specifics of the firms it could be seen that the extremely low-leveraged companies are rich with cash, relatively smaller and younger firms.

The rest of the paper is organized as follows. In Section 2 the literature review is presented. The literature review includes the discussion of the papers on general capital structure as well as the ones on zero-debt companies. The methodology used in the research is explained in Section 3. Section 4 consists of the description of the data and descriptive statistics. The results are discussed in Section 5 whereas the robustness checks of the results as well as the further insights are given in Section 6. Section 7 is dedicated to the conclusions.

2. Literature Review

Capital structure is discussed in the academic literature extensively for long time already; however, there are different ways for trying to find the answer why firms choose particular debt level. In the early and influential work, Modigliani, Miller (1958) argue that the capital structure does not make difference in the value of a firm because it just modifies the cash flow allocation between equity and debt. It is argued that the choice of leverage changes only the risk of equity. However, this argument holds only in the perfect capital markets which imply conditions such as: a) investors and firms trade securities that are the same as well as the value of the securities is equal to the present value of the generated cash flows; b) no taxes or transaction costs; c) capital structure decision does not affect the cash flows from the securities. If these

(8)

8 assumptions strictly hold, the capital structure research would not have any meaning; thus, the question of extremely low-leveraged firms would also not be interesting. In practice, these assumptions do not hold; therefore, even though the conclusions of Modigliani, Miller (1958) model are insightful, the capital structure topic is discussed way further in the academic literature. Even though the classification of the theories can differ, there are mainly two opposing streams to explain the choice of the capital. These are the trade-off theory and the pecking-order theory. In addition to these two theories, there is a relatively new approach which explains the capital structure by market timing. Further, the theories and the literature on zero-leveraged firms are discussed.

2.1. Trade-off Theory

The proponents of the trade-off theory try to find the balance between costs and benefits of the debt and finding the optimal level of debt. Summarizing the trade-off theory it could be said that the value of a firm is equal to the sum of the value of unlevered firm and the debt advantage less the costs of distress. The trade-off theory could be rooted to Kraus, Litzenberger (1973). With the trade-trade-off theory the authors try to solve the doubt of pros and cons of a debt. In their model they introduce the advantage of the tax shield and the cost of bankruptcy and conclude that the leverage is determined by the need of firms to maximize the market value. Tax shield is addressed in other academic articles too. For example, the model proposed by King (1974) is one of the classic works showing that the optimal financial policy is determinant for the cost of capital once the differences in taxes are allowed. Furthermore, Modigliani, Miller (1963) adjust their previous work and state that the tax advantage of debt is important for firm’s leverage. MacKie-Mason (1990) also studies the importance of the tax shield; he concludes that the preference of debt positively correlates with the effective marginal tax rate. On the disadvantage side of the debt there are costs of bankruptcy as well as the agency conflicts between debt providers and equity holders. The agency costs lower the debt ratio in companies because too much debt might lead to the disagreement between the management and shareholders (Jensen, Meckling (1976)). Taking all these factors into account, the trade-off theory is concerned about finding the optimal leverage which incorporates the advantages and disadvantages of the debt.

However, the trade-off theory is criticized because its inability to explain the behaviour of the firms of not rebalancing towards the optimal leverage. Modigliani, Miller (1963) suggest that even though the tax shield is an important factor of the capital structure, firms do not necessary keep the highest debt ratio possible. This might mean that companies leave money on the table; however, with static trade-off theory it is difficult to capture factors such as the possibility of other forms of financing or limitations to raise debt. Therefore, the early papers on capital structure fail to capture a lot of corporate financing factors. Also, empirical research concluded that companies are underleveraged: for example, Korteweg (2010) finds that

(9)

9 firms are on average underleveraged. The author also states that the reason of this conclusion is the companies which employ the conservative-debt (i.e. zero-debt) policy. These companies are particularly interesting because the management decides not only to have a company underleveraged but, in the extreme case, keep the debt close or equal to zero. What is more, Graham, Harvey (2001) conducted a survey and questioned 392 CFO’s about their firms’ budgeting. They find little evidence of the managers being concerned about optimal debt ratio. Managers tend to rebalance the debt ratio towards a range of leverage level instead. Denis, McKeon (2012) in general deny that firms manage the leverage towards the estimated optimal leverage. Rather, according to the authors, the leverage is a consequence of the capital needs for investments. The authors suggest considering low leverage as a financial flexibility which managers want to keep. Still, as zero or almost zero leverage companies are far away from the debt exhaust limit, this suggestion does not apply for these companies.

2.2. Pecking-order Theory

Different from the trade-off theory, the pecking-order theory is not concerned with finding the optimal leverage; still, the advantage of debt is discussed. Myers, Majluf (1984) create the model of firm’s investing decisions based on asymmetric information and the managers acting in favour of the old shareholders. The investments can be financed by three means: retained cash, debt issuing or equity issuing. Assuming superior information held by the management, the authors suggest that the managers would sell equity in case it is overvalued. Still, the shareholders know it; therefore, the share price drops instantly. Then, for the signalling reason, the managers prefer investing by using own resources (cash) in the first place, issuing debt in the second and only in the last case managers issue equity. In other words, in the pecking-order theory the debt advantage is explained by the asymmetric information: the costs of the asymmetric information are lower with the debt; the agency costs of equity are reduced as the leverage is introduced (Myers (1977)). Still, even though the pecking-order theory has insightful conclusions as well as an intuitive application, the theory is also criticized. Fama, French (2002) test the predictions of the trade-off theory and the pecking-order theory. In the cases such as relationship between book leverage and profitability or finding desired debt leverage, the authors find that the pecking-order theory conclusions are more favourable compared to the trade-off theory. The pecking order theory predicts that more profitable firms have less book leverage and the leverage is not returning to its mean. However, in the cases when both theories share the predictions, it is not clear which of the theories is better in explaining the evidence. For example, the cases such as the negative relation between investment and leverage; the positive relation between firm size and leverage; and the relation between dividend payout and leverage are shared conclusions by both theories. Therefore, the explanation of the relationships cannot be attributed only to the pecking-order theory. What

(10)

10 is more, referring again to the companies which have zero or almost zero leverage, the puzzle of such low debt also does not fit into the predictions of the pecking-order theory.

2.3. Market Timing

The third approach to the capital structure is to consider the capital structure as a consequence of the ongoing activities of firms. Baker, Wurgler (2002) find that the managers time the market and issue equity when the value of the equity is high. Also, the authors claim that these issuing shocks to the leverage are persistent; therefore, it suggests that the leverage is the consequence of these issues. The idea of the mechanisms of the market timing is used in this paper to check if this particular market timing mechanism is working before companies become extremely low-leveraged (Hypothesis 1). Furthermore, interesting conclusions are drawn by Welch (2004) when the author tries to find the determinants of the capital structure in the stock returns. The idea of the research is that the author raises two hypotheses of extreme cases: perfect readjustment and perfect non-readjustment. He claims that the 40% of the debt ratio could be explained by stock return dynamics. Also, when taking into account the stock returns, a lot of variables used in the literature to explain the capital structure loses its significance or becomes less significant in explaining the debt- versus equity-financing decision in general. It is an interesting approach to test the proactive capital structure changes; therefore, this approach is used for the second hypothesis in this paper. The second hypothesis tests whether the companies readjust their capital structure (market leverage) when they are extremely low-leveraged.

Neither Baker, Wurgler (2002) nor Welch (2004) researches the companies that are only equity-financed. Previously it is discussed that the trade-off and pecking-order theories are not able to explain the zero leverage phenomena. Therefore, there is a gap in the academic literature regarding the market timing mechanism for zero-leveraged firms. I use the market timing approach in my own research and address specifically extremely low-leveraged companies. Intuitively, the extremely low-leveraged firms are different from the companies that have higher leverage. Also, the mechanics of the debt ratio should be different: it is unlikely to have the zero or almost zero debt only by passively pursuing the debt ratio. Therefore, the approaches used by Baker, Wurgler (2002) and Welch (2004) are particularly interesting as they are relevant for answering the research question as well as the analysis contributes to the existing literature on capital structure. The methodology taken from these two papers and the adjustments to the methodology is explained in detail in the methodology part.

(11)

11 2.4. Literature on Zero Leverage Firms and Debt Ratio

For more explicit analysis the decision to have zero leverage, other factors, besides market timing, are considered too. The management ownership and the financial constraints are the variables of interest in the specifications of the third hypotheses. Hypothesis 3a and 3b test the influence of these two variables on the probability of a company to be extremely low-leveraged. Further, the choice of the variables of interest as well as the choice of the control variables (for all the regressions) is explained.

First of all, an influential stream of literature proposes finding the answer of the zero-debt policy by exploring the management of the firms. Agrawal, Nagarajan (1990) report that the unlevered firms’ management is more likely to own more equity as well as the management is more likely to have family relationships. Similarly, Strebulaev, Yang (2013) conclude that companies with large CEO ownership and more CEO-friendly boards are more likely to employ the zero-debt policy. Schmid (2013) finds that in Germany family firms are less likely to take on debt. The author suggests that in this way families optimize their influence in the firms. However, Devos et al. (2012) conclude that the zero-debt policy is not driven by the entrenched managers. Following the idea of these findings in the academic literature I will check if the management ownership (all management’s ownership taken together) is an important factor for a company to be extremely low-leveraged. The CEO ownership influence is tested in the robustness checks.

Secondly, Besler et al. (2013) suggest that companies could be grouped into the deliberate zero-leverage companies and the financially constrained companies. What is more, Devos et al. (2012) find that the financial constraints have a big impact on the companies maintaining zero-debt. Also, the authors state that the debt-financing is chosen when the constraints are released. Therefore, the financial constraints are included as one of the variables in the regressions as a determinant which makes the influence on the companies to be extremely low-leveraged. One of the popular ways to measure financial constraints is the KZ Index (Lamont et al. (2001)). The calculation of the index requires financial attributes of the companies such as the value of property, plant and equipment, assets, cash flows, deferred taxes, long term debt and short-term debt, cash holdings and other. However, Hadlock, Pierce (2010) prove that KZ Index do not capture the financial constraints well and proposes size-age index (the SA Index). The index is sufficient for this research; therefore, it is included into the regressions as the variable that captures the financial constraints. Both of the indexes are probabilistic; therefore, even though the SA Index is chosen for the main research, the KZ Index is used in the robustness checks.

Next to the main variables of interest, there are control variables which are included in the analysis. The profitability is included in most of the papers that address the zero-leveraged companies (Devos er al. (2012); Besler et al. (2013)). It could be argued that the more profitable firms have mechanically lower leverage. Stabulev, Yang (2013) prove it in their empirical research and say that the zero-leveraged firms that pay dividends are more profitable and bigger. On the other hand, Kortweg (2010) find that the benefit

(12)

12 of the debt is higher for the firms which are more profitable. This implies different suggestion: the profitable firms should be keeping higher debt ratios. Next to it, tangibility and size in terms of sales are included as the control variables for the firm’s size and firm’s assets (Baker, Wurgler (2002)). Cash holdings are included to control for the ability to pay out the debt immediately; capital expenditures and R&D expenditures are included into the regressions as the controls for the growth opportunities (Besler et al. (2013); Stabulev, Yang (2013)).

What is more, Besler et al. (2013) find that in general the number of zero-leveraged firms is increasing. According to the authors, the reason for the increasing number of the zero debt companies is the IPO effect: there were more emerging companies during the period they studied. To capture this trend the time dimension is included in the empirical model as the time-fixed effects. The authors suggest a possibility that the competing firms lower their debt ratios too. Age is included because of the similar argument: firms after the IPO might hold lower debt ratio (Besler et al. (2013)). Also, the authors attribute the decision of zero-debt policy to the differences across industries. Because of the possible differences, the industry-fixed effects are included as controls in the regressions. Lastly, Berens, Cuny (1995) argue that the higher is the growth rate of firm’s future earnings the lower is the optimal capital structure. Therefore, the growth dimension is included in the descriptive analysis when the reference groups are constructed. The growth in this paper is measured by the market-to-book ratio which is suitable for capturing the internal capacity of the firms.

3. Methodology

In this section the hypotheses and the methodology used to test the hypotheses are presented. Next to the methodology, the sample selection procedure is introduced.

The main research focuses on the firms which keep the true zero or almost zero leverage. The book leverage is determined as the ratio of the value of the total assets less book equity to the total assets. The zero or almost zero-leveraged firms, then, are the ones that keep the true debt no higher than 5% of the total assets. This way of determining the debt ratio includes the account payables and other balance sheet items from the debt. In the robustness checks the book debt is described as the sum of only the short-term debt and long-term debt.

Through all the paper the whole dataset includes: a) the firm-year observations when companies are not yet zero-leveraged; b) the firm-year observations when companies keep the zero leverage; c) all other firm-year observations when companies are not zero or almost zero-leveraged. Furthermore, only the companies which have at least two years of zero leverage are considered as zero-leveraged firms, because one year of zero leverage might be reached by extra-ordinary events. When opposing the market timing as a way of

(13)

13 determining the capital structure to the deliberate decision of zero leverage, only one year of being zero-leveraged does not give any meaningful explanations for either option. Therefore, firms with only one year of zero leverage are treated as outliers.

What is more, the whole research is concerned about the capital structure choice; therefore, most of the control variables are expressed as the percentage of the total assets. In this way, the capital structure is addressed as well as the outlier problem is solved.

Hypothesis 1. : Companies with almost zero leverage reach this level not by the issuing activities.

Companies time the market and issue equity when the value is high. Therefore, if the debt ratio is driven by these issues, the weighted average market-to-book ratio is able to determine book leverage. Baker, Wurgler (2002) find that weighted average market-to-book ratio is better in explaining the book leverage than market-to-book ratio. This suggests that the historical valuation of a firm is important for the existing leverage. The authors confirm their hypothesis that the leverage is a consequence of firms’ issuing activities. However, it is not likely that the issuing activities would lead firms to have zero or almost zero leverage. Therefore, Hypothesis 1 is expected to be accepted which would strongly suggest that the companies with almost zero leverage have such a debt level deliberately. The methodology similar to the one proposed by Baker, Wurgler (2002) is employed. The specification of the regression is as follows:

The full explanation of the construction of the variables is given in the Appendix 1. The specification of the regression used by Baker, Wurgler (2002) includes only weighted average book ratio, market-to-book ratio, tangibility, profitability and size. Following the related literature, this paper, besides the mentioned variables, includes lagged one period cash holdings, lagged capital expenditures, lagged R&D expenditures and age. In the second specification of the regression the industry-fixed effects and the time-fixed effects are added.

For the sample I keep the companies which have at least 2 years (assuming that it is a relevant time period) of extremely low leverage. Hypnotizing the management deliberately reaching and keeping the zero or almost zero leverage, the time before them reaching the level is the most important. Then, intuitively, the market timing does not determine the book leverage. What is more, when the companies abandon

(14)

zero-14 leverage policy, the market timing mechanics might start working again. If the time after the zero-leverage period is not excluded, the results might be misleading.

Baker, Wurgler (2002) run the cross-section regressions; because the data used in this paper is a panel data, the robust standard errors are used. When including the industry-fixed effects and the time-fixed effects, the regressions are run without a constant.

Hypothesis 2. : Extremely low-leveraged companies readjust the capital structure.

The regression specification is as follows (Welch (2004)):

Table 1

The explanations of the variables of the Hypothesis 2 regressions

Variable Explanation

actual market debt ratio at time t+k constant

actual market debt ratio at time t implied debt ratio at time t+k:

, where is compounded return on stock

According to the methodology proposed by Welch (2004), the specifications are as follows when regressing without a constant:

Perfect readjustment: =0, =1. Perfect non-readjustment: =1, =0.

The idea of the regression is to analyze the implied debt ratio which floats with the market stock price and the actual ratio which evolves with the floating debt ratio plus net debt and/or net equity issues. Differently from Welch (2004) methodology, the linear regressions are run for the panel data (Welch (2004) uses the Fama-MacBeth regressions). If the companies readjust their capital structure, the actual debt ratio should be

(15)

15 more important than implied debt ratio when defining the leverage. I expect to have the situation of the (perfect) readjustment in the sample of extremely low-leveraged companies (opposite to what Welch (2004) finds in his sample). This would be in line with the Hypothesis 1 in the sense that extremely low-leveraged firms readjust to their target debt ratio and do not follow the debt ratio passively. Therefore, the results of the regression would also help answering the question if low leverage is a deliberate decision. Interestingly, if the constant is included, the could capture the target debt ratio.

For the second hypothesis only the years when the companies keep extremely low leverage are included. In this way it is captured whether the companies readjust their capital structure towards the zero or almost zero debt ratio when they reach the zero leverage.

Hypothesis 3a. : The probability of a company to be extremely low-leveraged increases with the management’s equity ownership.

Hypothesis 3b. : The probability of a company to be extremely low-leveraged increases with the financial constraints.

To test the third hypothesis (a and b) is tested using probit model:

The outcome variable is a categorical variable that takes up value 1 if the company is zero or almost zero-leveraged and value 0 when the company is not extremely low-zero-leveraged. To test the last hypotheses the whole sample of companies is taken.

The variables of interest are the Eq.Own (equity ownership) and the SA Index. Equity ownership is the total percentage of the equity that the top management owns together. The formula to calculate the SA Index (Hadlock, Pierce (2010)) is:

Asset size is the log of inflation adjusted book assets and age is the number of years since the IPO. Book assets are adjusted to inflation, setting the reference year to 2004; the inflation rates are taken from the

(16)

16 World Bank database1. Note that the size later in the regressions is defined differently. The higher is the SA Index, the more a firm is financially constrained.

Other variables included into the regression are lagged tangibility, lagged profitability, size, lagged cash holdings, lagged capital expenditures, lagged R&D expenditures and age. All the variables and their construction are given in the Appendix 1. What is more, following the consistency with other hypotheses and descriptive analysis, the regression is also run including the industry-fixed effects and time-fixed effects. Following the literature, the tangibility and age could have either positive or negative influence; I expect profitability and cash to have positive influence whereas size, capital expenditures and R&D expenditures to have negative influence. For the estimations, the clustered by firm standard errors are used.

In order to make the proper interpretation of the influence of the variables on the probability of a firm to be zero or almost zero-leveraged, the results are given as the marginal effects of variables on the outcome. The original probit coefficients and the statistics of significance are given in the Appendix 6.

4. Data and Descriptive Statistics

4.1. Data

The data used in the research is collected from the COMPUSTAT and Execucomp databases. To have as many observations as possible the longest time period 1961-2013 available is used. To create the dataset the data of the balance sheets items (COMPUSTAT) and the data of the equity ownership (Execucomp) are merged by the companies’ identifiers (gvkey) and fiscal years. The companies with SIC code between 6000-6999 are excluded from the sample because these are financial institutions which leverage levels are not comparable to leverage levels in any other industry. What is more, SIC code is recorded since 1987; therefore, the historical SIC codes are used for the earlier years. There were cases when companies had double entries of the stock prices per year. The adjustments were made manually; only the data which correspond to the fiscal year end information was left in the dataset. The firm-year observations with missing data on total assets are dropped because without the total assets neither book leverage nor market leverage can be determined.

To test the first hypothesis, the adjusted methodology by Baker, Wurgler (2002) for calculations of the variables is used. The observations with negative book leverage as well as the observations that book leverage is more than 1 (debt take up to more than 100% of total assets) are deleted as outliers. The book ratio is restricted to 10 as the upper bound. The formula for weighted average

1_{World DataBank. World Development Indicators.}

(17)

17 book ratio is given in the Appendix 1. For the formula of weighted average market-to-book ratio the net equity issues ( ) and the net debt issues ( ) are constrained to 0 as the lower bound to have a weighted average. In this way, in the years when or are zero or below zero, it only means that the variables do not contain any weight of the market valuation. For the second hypothesis the market debt ratio is also restricted to 0. For the third hypothesis the ownership is restricted to 100%. There were cases when the ownership exceeded 100%; however, these were deleted as outliers. According to the methodology of calculating the SA Index, the adjusted size is set to the $4.5bn if the total adjusted assets have higher value than $4.5bn and the age is set to the 37 years if a company is older than 37 years.

In total in the whole sample there are 193539 firm-year observations; zero or almost zero-leveraged firms’ firm-year observations come up to 4571. The longest period a company stays zero or almost zero-leveraged is 29 years. On average, the companies keep the extremely low leverage for 6 years.

4.2. Correlation

In order to have the information which is not redundant, the correlation between the variables is checked. The correlation matrix is given in the Appendix 3 for two samples: the whole sample and the zero or almost zero-leveraged firms’ sample. In the whole sample it could be seen that, not surprisingly, the book leverage is highly correlated to the market leverage (the correlation coefficient equals to 0.7811). Also, the book leverage and market leverage appear to be negatively correlated with the cash (the coefficients: -0.5137 and -0.4990 respectively) and the SA Index (the coefficients: -0.3924 and -0.3385 respectively). This could be interpreted as that the firms which have high cash holdings are more able to repay the debt; therefore, their debt ratios are lower, as well as the firms which are less financially constrained have lower debt ratios. The market-to-book ratio and the value weighted market-to-book ratio are correlated per definition; besides that, these two variables have the highest correlation coefficients with the cash holdings (the coefficients: -0.4990 and 0.4699 respectively). Tangibility and capital expenditures are correlated with each other with the correlation coefficient which equals to 0.6035. This relatively high coefficient could be explained by that the more firms spend on capital, the higher cumulated percentage of tangible assets in the total assets they could be attributed. Interestingly, the profitability is not highly correlated to the cash holdings (the correlation coefficient: 0.2330). One could expect that only the profitable firms are able to hold high cash percentage in the total assets. However, it does not seem to be the case. Profitability is negatively correlated (-0.4797) with the R&D expenditures instead, suggesting that the firms which invest more in R&D are more likely not to be or not yet to be profitable. Size (which is per definition related to the sales) is more correlated to the cash holdings (the coefficient: -0.4881) and relatively highly correlated to the age of the firms (the correlation coefficient: 0.5307) suggesting that the bigger firms tend to have more cash as well as they are older.

(18)

18 The correlation matrix in the sample of the zero or almost zero-leveraged firms is also presented in the Appendix 3. The tendencies of the correlation between the variables are similar; however, there are some notable differences. Firstly, the market and book leverages are relatively low correlated (the coefficient equals to 0.2495). This could be explained by that there is not a lot of variation in the book leverage for the extremely low-leveraged firms; however, the market leverage varies with the market valuation. This is related to the Hypothesis 2 which is dedicated to test whether the firms readjust their market leverage. What is more, the SA Index appears to be less correlated to the book leverage and market leverage (the coefficients equal to -0.0852 and 0.0446 respectively) in the sample of the zero or almost zero-leveraged firms. This could lead to the suggestion that among the zero or almost zero-leveraged firms (as the leverage is restricted when sampling) we could observe firms with various levels of constrains. Lastly, the cash holdings are less correlated to other included variables compared to the correlation seen in the sample of all firms. For example, the correlation coefficient to book leverage equals to as low as to -0.1581 and to market leverage equals to -0.0678.

4.3. Descriptive Analysis

In the following comparison (Table 2) the averages of the variables in the sample of the zero or almost zero-leveraged firms are referred to the reference groups. For every zero or almost zero-zero-leveraged firm-year observation a reference group is constructed. The groups are determined by three characteristics. Firstly, only the companies in the same industry according to the SIC codes are included into reference groups. Secondly, to capture similar companies, the value of total assets is taken. Only companies in the range of lower by 10% and higher by 10% total asset value is taken into reference groups. 10% is chosen arbitrary to get the most similar companies in the reference groups. If there are no companies in the reference group, the companies in the range of lower by 20% and higher by 20% total asset value is taken into reference groups. Third criterion to include a company into a reference group is market-to-book ratio. This ratio is taken because it captures the growth opportunities among the companies (the range is similar to the total assets: first stage of the reference group construction: 10%, second: 20%). Higher market-to-book ratio is attributed to companies which have bigger growth potential. Therefore, it is a good measure which includes market’s perception of the companies towards the investment policy, future profits as well as the cash flows. The proxy groups are as big as possible to have the least biased market proxy for every firm-year observation; therefore, it is important to highlight that one company might be included into more than one reference group. It is not an issue because the variables are averaged as well as the idea is to capture the most similar companies’ balance sheet items and have a proper comparison between zero-leveraged and non-zero-leveraged companies. What is more, there are no zero or almost zero-leveraged companies included in the proxy groups assuming that these companies are different from all the other companies. In this way the real difference of the balance sheet elements are captured. Furthermore, by company, the

(19)

19 variables and the proxies are averaged so every company has the same weight finding the averages no matter how long the company is zero or almost zero-leveraged.

The procedure to calculate and compare the averages of the variable between the zero-leveraged companies and the benchmark companies are similar to the one used in the paper of Stabulev, Yang (2013). However, the authors pay more attention whether companies are dividend payers or not. What is more, differently from the proxy groups used in this paper, Stabulev, Yang (2013) construct small proxy groups for the comparison (on average 3.4 companies in a proxy group).

Table 2

Descriptive statistics

The averages of the variables in the samples of zero-leveraged companies, in their reference group sample, in the initial extremely low-leveraged firms’ and in the whole sample. The construction of the variables is given in the Appendix 1. Book leverage, market leverage, profitability, tangibility, cash, capital expenditures (CapeEx), and research and development expenditures (R&D) are given as the percentage in the total assets. Ownership is given as percentage of total shareholders’ equity. In the last column the t-statistics are given for the significance of the differences in the sample averages of zero-leveraged companies and their reference groups averages (formula for the t-stat to compare two means:

).

The t-stat for the differences of book leverage and market leverage are not given because it does not give any insight as the differences are restricted when sampling.

Variable Zero leverage firms Reference firms Initial zero leverage firms Whole sample t-stat Book Leverage, % 2.93 29.51 2.67 44.89 NA Market Leverage, % 3.00 23.85 2.57 38.47 NA Tangibility, % 33.81 42.11 39.27 34.13 -7.96 Profitability, % -13.44 -10.82 -17.09 6.01 -3.47 Cash, % 43.80 24.29 42.41 15.39 21.91 CapEx, % 8.79 11.10 10.67 7.40 -4.22 R&D , % 9.07 8.62 9.45 7.56 0.74 Age 6.79 9.62 8.82 12.73 -14.75 Ownership, % 3.03 3.94 5.66 3.57 -1.38 SA Index -2.0438 -2.2613 -1.9127 -2.9791 8.08

From the Table 2 it could be seen that the differences in the averages of the variables are dramatic. In overall, almost all the recorded variables’ averages are statistically different in two samples: zero-leveraged companies and their reference companies.

Book Leverage and Market Leverage. The t-statistics are not presented for the differences of the book and

(20)

20 levels are restricted when sampling; therefore, the t-statistics do not give any insight. Interestingly, the book leverage and market leverage are lower in the reference groups than in the whole sample. This suggests that there is a pattern in such size and growth-opportunity companies as extremely low-leveraged companies. Following the explanation provided by Berens, Cuny (1995), the lower market and book debt ratios calculated in the sample of reference groups would lead to the conclusion that the extremely low-leveraged companies and their leveraged pairs are on average higher growth companies. What is more, the unreported analysis of the market and book leverages’ dynamics show that companies start decreasing the leverage level approximately 5 years before becoming extremely low-leveraged. When period t is the first year of a company being zero or almost zero-leveraged: at period t-5 the decrease in the book leverage is slight: 5%; later (periods: t-4 and t-3) the decrease is more significant: approximately 14%; and lastly the decrease is dramatic: on average 25% (period: t-2) and 82% (period: t-1). The changes for the market leverage during the years are slightly different, still qualitatively similar.

Equity Ownership. Interesting enough are the statistics about the ownership which is one of the variables of

interest for the Hypothesis 3. It is one of the two variables that the average in the sample of extremely lowleveraged firms is not statistically different from the average in the reference groups (tstatistics equals to -1.38). The management’s ownership is relatively low: on average 3.03% of total equity of zero or almost zero-leveraged firms. The average of management’s ownership in the reference groups’ sample is 3.94% and 3.57% in the whole sample.

SA Index. What is more, on average the zero or almost zero-leveraged companies are the most financially

constrained compared to the whole sample and their reference groups. Still, firms in the the reference groups are more financial constrained compared to the whole sample. The dynamics of the SA Index show that firms that have zero or almost zero debt level have a slight increase in the SA Index in the year that they become extremely low-leveraged. However, later they become less financially constrained. What is more, the SA Index varies less in the sample of zero or almost zero-leveraged firms: minimum value that the SA Index obtains is -4.4843 and the maximum value 2.2271 (the values in the whole sample ranges between -4.6380 and 4.0674).

Tangibility and CapEx. Looking at the tangibility, it could be seen that the zero or almost zero-leveraged

firms have the least tangible assets as the percentage of the total assets (33.81%) compared to both: reference groups (42.11%) as well as the whole sample (34.13%). Still, looking at the dynamics of the tangible assets it could be seen that it increases gradually during the time. The average increase of 3.5% could be observed during the time period 5 years before and 5 years after the company becomes zero or almost zero-leveraged. What is more, the average of the capital expenditures is in the middle of the average capital expenditures in the whole sample and in the reference groups suggesting that the companies are growing. However, when not using the debt, the extremely low-leveraged companies are somewhat limited

(21)

21 to reach the level of the capital structure items applying for their industry pairs. These observations lead to the suggestion that extremely low-leveraged firms are relatively smaller.

R&D expenditures. The growth factor could be seen looking at the R&D expenditures too: even though the

R&D expenditures decrease sharply (more than 40%) at the time company becomes extremely low-leveraged, the average R&D expenditures are the highest in the sample of the zero or almost zero-leveraged firms (9.07% of the total assets) compared to other samples. However, the difference appears to be not statistically significant; therefore, the statistics on the R&D expenditures should be taken with the caution.

Profitability and Cash. The profitability variable follows similar pattern as the book and market leverages.

The zero or almost leveraged firms exhibit the lowest profitability compared to other samples. The average profitability is even negative (-13.44% of the total assets) in the zero-leveraged companies’ sample. This negative value leads to two suggestions. Firstly, the companies which are not profitable might be financially constrained; therefore they are extremely low-leveraged. Secondly, this correspond to the logic that for the companies which have the negative taxable earnings it is not optimal to raise debt because then firms are not able to use the tax shield provided by debt. These suggestions are tested with the third hypothesis. However, the zero-leveraged companies appear to have very high levels of cash and short-term investments. On average zero-leveraged companies have the cash holdings of 43.80% of the total assets. It is almost twice as much as their reference group companies hold (24.29%) and almost three times more compared to the total sample in the analyzed dataset (15.39%). It is an interesting and puzzling insight because the zero-leveraged companies on average are not more profitable; still, they have very high cash holdings. Looking at the profitability and cash holdings dynamics it could be seen that, during the period of 5 years before companies become zero or almost zero-leveraged, the profitability decreases sharply. The cash holdings on average is in its peak at the year companies become extremely low-leveraged; after that the cash holding percent in the total assets decreases gradually.

Age. The zero or almost zero-leveraged companies tend to be younger than their reference groups’ firms

and the firms in the whole sample. The average age of the zero or almost zero-leveraged firms is 6.79 years while the reference firms are on average 9.62 years old. Further insight about differences across various age groups of the firms is given in the Appendix 4. The book and market debt levels are restricted when sampling the zero and almost zero-leveraged companies; therefore, there is no big difference in the averages of different-age companies. However, in the whole sample we see that both: the book leverage and the market leverage increases as the companies become older. In both samples (i.e. zero and almost zero-leveraged companies and whole dataset) the profitability increases as the companies age. The different patterns we see in the tangibility and cash. The older zero and almost zero-leveraged firms have fewer tangible assets and hold more cash while in the whole sample we see that older companies get more assets and hold less cash and short-term investments. The capital expenditures and R&D expenditures expressed

(22)

22 as the percentage of the total assets become smaller as the companies become older. This pattern could be seen in both samples. Lastly, the ownership percentage does not follow a clear pattern in the extremely low-leveraged firm’s sample but the percentage decreases for the whole sample of companies.

Industry dimension. In the academic literature industry dimension is one of the reasons of companies having

different capital structures (Besler et al. (2013)). The averages across the industries are reported in the Appendix 5. It could be seen that most of the variables differs a lot across the industries even though the averages of variables are given as the percentages from the total assets. The lowest average book leverage and market leverage are in the mining industry: 32.32% and 27.28% accordingly. The highest ratios of book debt and market debt are in the construction industry: 58.37% and 56.62% accordingly. Profitability also ranges significantly across the industries: mining industry and public sector show negative profitability; in the other industries the profitability is positive. All the other variables range significantly too. The differences seen in the descriptive statistics is a good proof to include the industry-fixed effects in the regressions as control variables. What is more, looking at the zero or almost zero-leveraged firms by industries it appears that the most of these firms are either in the mining industry (2740 firm-year observations) or in the manufacturing industry (1148 firm-year observations). The lowest numbers of extremely low-leveraged firms are in the construction; agriculture, forestry and fishing; and retail trade industries.

The companies which are extremely low-leveraged from the first year when they are recorded in the database have somewhat different averages of the capital structure. Still, the averages are more similar to the whole sample of the extremely low-leveraged companies. Also, later sampling differently (excluding the initial extremely low-leveraged firms) for the regressions of the Hypothesis 3 does not change the results. Therefore, the differences in the capital structure items are not discussed explicitly. Only one thing to note: the companies which have initial zero or almost zero leverage are less financially constrained compared to other firms.

5. Results

5.1. Results Hypothesis 1

In the Table 3 the results of the regressions are reported. Two specifications are presented in the table: the regression without the industry-fixed and the time-fixed effects and the regression with the industry-fixed and the time-fixed effects. Hypothesis 1 is accepted which states that the companies do not reach the zero or almost zero level of debt by timing the market. It could be seen that in both of the specifications the value weighted market-to-book ratio and market-to-book ratio are not statistically significant (the t-statistics are 0.00 and -0.43 accordingly in the Panel B). The conclusion that the market timing is not the reason for

(23)

23 companies to become extremely low-leveraged as well as the evidence that other companies’ (i.e. leveraged ones’) capital structure follow market timing strongly suggests that the zero or almost zero leverage is a deliberate decision. Further, the results are discussed in detail.

Table 3

The results of the regressions run for the Hypothesis 1

The outcome variable is the book leverage in percentages. Panel A represents the results on the regression without the industry-fixed and time-fixed effects. Panel B represents the results on the regression with the industry-fixed and time-fixed effects. M/Befwa,t-1 is the weighted average market-to-book ratio calculated with formula

and lagged one

period. M/Bt-1 is the lagged value of the market-to-book ratio. Tangibility, profitability, cash, capital expenditures (CapEx), and

R&D expenditures are expressed as percentage of the total assets and lagged one period. Size is the natural logarithm of the sales; lagged one period. Next to the coefficients the t-statistics are presented; the number of the observations and the R2 are given below. The coefficients are marked as significant at the *5% significance level and the **1% significance level.

Panel A Panel B

Variable Coefficient t-stat Coefficient t-stat

M/Befwa,t-1 -0.0068 -1.18 0.0000 0.00 M/Bt-1 -0.0010 -0.20 -0.0022 -0.43 Tangibility (PPE/At-1), % 0.1166* 2.41 0.1069* 2.15 Profitability (EBITDA/At-1), % -0.0983* -2.23 -0.1100* -2.13 Size (Log(S)t-1) 0.9000* 2.12 1.1886* 2.58 Casht-1, % -0.2629** -9.64 -0.2438** -8.83 CapExt-1, % 0.0285 0.27 0.0115 0.10 R&D expenditurest-1, % 0.0718 1.44 0.1010 1.81 Age 0.0561 0.47 0.1813 1.48

Industry-fixed effects No Yes

Time-fixed effects No Yes

Constant 28.0672 11.18 - -

Number of observations 1063 1063

R2 0.1948 0.6758

In the first specification of the regression only the variables for the capital structure are included. The results show that the weighted average market-to-book ratio is not statistically significant factors (the t-statistics is -1.18). The same could be seen in the regression specification with the industry-fixed effects and the time-fixed effects. The t-statistics for the coefficients of the weighted average market-to-book ratio is 0.00. These results show that the historical valuation of the equity does not influence the capital structure. As already stated, this strongly suggests that the firms that decide to become zero or almost zero-leveraged do not time the market which is the sign of the deliberate seeking for the zero leverage. What is more, the results

(24)

24 of the market-to-book ratio are striking. The coefficients next to the market-to-book ratio are not statistically significant in none of the specifications (the t-statistics are -0.20 and -0.43 respectively). This suggest that even the growth opportunities which are commonly considered to have negative influence on the capital structure (Berens, Cuny (1995); Bakker, Wurgler (2002)) do not have impact on the capital structure during the period when companies are seeking to reach the zero or almost zero leverage. This further suggests that the conventional explanation of the capital structure does not hold here.

Only cash variable is statistically significant at the 1% significance level. The coefficient of the cash and short-term investments enters the model with the negative sign meaning that the higher are the cash holdings the lower is the book leverage during the time before companies become zero or almost zero-leveraged. The increase of one percent of the cash in the total assets decreases the book leverage by 0.2438%. It is not a surprising observation since the companies which have enough cash are able to repay the debt. Three of the variables are statistically significant at the 5% significance level. Tangibility enters the regression with a positive sign meaning that the more the company has tangible assets, the higher is the debt ratio in the period before companies become zero or almost zero-leveraged. According to the second specification of the regression, everything else kept equal, the increase of one percent of the tangible assets increases the book leverage by 0.1069%. Another variable that enters the regression with the positive sign and is significant at the 5% significance level is the size variable. According to the result of the Panel B, everything else kept the same, the increase of the sales by one percent increases the book leverage by 1.1886%. Tangibility and size could be seen as interrelated variables because the interpretation of the results is similar: the bigger is the company, the more difficult it is to get zero or almost zero debt level. The coefficient of profitability is also significant at the 5% significance level in both of the specifications of the regression. All things equal, the increase of the ratio of the profitability to the total assets by one decreases the book leveraged by 0.11% during the time when companies are not yet zero or almost zero-leveraged. This suggests that only profitable firms tend to decrease the leverage. Capital expenditures, R&D expenditures and age appear to be insignificant when determining the book leverage before the firms become extremely low-leveraged. The further discussion of the capital structure variables is left for the Hypothesis 3 results.

(25)

25 Table 4

The regression results for the Hypothesis 1 in the different samples

The outcome variable is the book leverage in percentages. In all the panels the results of the regressions without and with the industry-fixed and time-fixed effects are given. Panel A sample includes the firm-year observations when the companies are zero or almost zero-leveraged. Panel B sample includes the firm-year observations when the companies are zero or almost zero-leveraged and keep the extreme low leverage for at least 5 years. Panel C sample includes the firm-year observations when the companies are zero or almost zero-leveraged and keep the extreme low leverage for at least 10 years. M/Befwa,t-1 is the weighted average market-to-book ratio calculated with formula

and lagged one period. M/Bt-1 is

the lagged value of the market-to-book ratio. Tangibility, profitability, cash, capital expenditures (CapEx), and R&D expenditures are expressed as percentage of the total assets and lagged one period. Size is the natural logarithm of the sales; lagged one period. Next to the coefficients the t-statistics are presented; the number of the observations and the R2 are given below. The coefficients are marked as significant at the *5% significance level and the **1% significance level.

Panel A Panel B Panel C

Variable Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat

M/Befwa,t-1 -0.0001 -0.17 -0.0002 -0.51 -0.0004 -0.88 -0.0004 -0.99 -0.0046** -3.07 -0.0053* -2.17 M/Bt-1 0.0006* 2.10 0.0006* 1.97 0.0011** 3.06 0.0010* 2.42 0.0014 1.72 0.0014 1.11 Tangibility (PPE/At-1), % -0.0019 -0.50 0.0017 0.51 0.0013 0.28 0.0046 0.98 0.0636** 3.04 0.0927* 2.56 Profitability (EBITDA/At-1), % -0.0002 -0.09 -0.0003 -0.09 -0.0108** -2.66 -0.0081 -1.91 -0.0054 -0.70 0.0046 0.31 Size (Log(S)t-1) 0.1989** 8.52 0.2076** 7.82 0.2504** 8.28 0.2621** 7.46 0.1748 1.96 0.1077 0.71 Cash, % 0.0005 0.27 -0.0005 -0.29 0.0012 0.53 -0.0017 -0.67 -0.0049 -0.83 0.0016 0.16 Capital expenditures, % 0.0107 1.16 0.0085 0.91 0.0095 0.77 -0.0079 -0.55 0.0212 0.38 -0.0217 -0.33 R&D expenditures, % 0.0169** 3.55 0.0147** 3.16 0.0126* 2.01 0.0135* 2.01 -0.0272 -1.14 -0.0053 -0.15 Age 0.0150** 2.60 0.0140* 2.12 0.0292** 4.01 0.0237* 2.56 0.0010 0.08 0.0152 -0.54

Industry-fixed effects No Yes No Yes No Yes

Time-fixed effects No Yes No Yes No Yes

Constant 2.4780 15.32 - - 1.8963 8.57 - - 3.1578 6.22 - -

Num. of observations 756 756 406 406 90 90

(26)

26 The results in the Table 4 are given for the further insights of the Hypothesis 1. The regressions with the conventional capital structure variables (from Baker, Wurgler (2002)) as well as with additional variables including the time-fixed effects and the industry-fixed effects are run. What is more, the regressions are run on three different samples: a) the sample of extremely low-leveraged companies during the years when they are zero-leveraged (Panel A); b) the sample of extremely low-leveraged companies which hold the zero or almost zero leverage for 5 or more years (Panel B); c) the sample of extremely low-leveraged companies which hold the zero or almost zero leverage for 10 or more years. The sample size decreases dramatically with the increase of the years of extremely low leverage kept. Furthermore, the R2 differs: in the Panel A the regression variables can explain approximately 19% (R2=0.1899)of the variation of the book leverage while in the Panel C the variables can explain more that 45% (R2=0.4519) of the variation. Still, the main coefficients of interest keep similar values across the panels. Even though, the book leverage in the sample varies from 0 to 5%, the capital structure variables appear to be significant.

In general, when the companies reach the zero or almost zero leverage level, the market-to-book ratio becomes significant at the 5% significance level even though the book leverage do not vary that much anymore. This is different from the period before companies become zero or almost zero-leveraged when the market-to-book ratio does not influence the book debt ratio. However, when the companies actually are zero or almost zero-leveraged, the historical market valuation of the firms’ assets are still not important for the book debt ratio. Same results could be seen in the regression which was run on the sample of the companies which keep the zero or almost zero leverage for 5 or more years. Only the companies which keep the zero or almost zero leverage for more than 10 years tend to time the market2_{. According to the}

Panel C, the market-to-book ratio is not significant anymore. Instead, the weighted average market-to-book ratio becomes significant (the t-statistics in the regression specification which includes the time-fixed and the industry-fixed effects equals to -2.17). It could be seen as a puzzling result; however, looking more closely, the results could be explained. When the companies reach the extremely low leverage, they might start looking for the opportunities to time the market to readjust to the zero leverage. However, it might not be possible in the short period of time. Therefore, only the companies which are longer zero-leveraged show the market-timing activities. What is more, unreported results of the analysis of the companies which keep the zero or almost zero leverage long enough show that these companies also do not time the market before they become zero or almost zero-leveraged. Therefore, even though the results in Table 4 could look like contradictory ones, they show similar results that were found in the Hypothesis 1 testing.

2_{Coefficients on M/B}

efwa,t-1 appears to be significant when the companies keep the zero or almost zero leverage at

Is the extremely low-leverage policy a deliberate decision?