The effect of market timing on capital structure: a cross-industry analysis

(1)

The effect of market timing on capital

structure: a cross-industry analysis

Bob van Alphen, 11334193 June 2020

Bachelor Thesis

BSc Economics and Business: Economics & Finance Supervisor: E. Seregina

(2)

Statement of Originality

This document is written by Bob van Alphen who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

Prior research has shown that firms tend to issue equity when their market value is high. The recently proposed market timing theory of capital structure suggests that these market timing actions determine a firm’s capital structure. This thesis examines the explanatory power of this market timing theory of capital structure. Additionally, a cross-industry analysis is conducted to study whether the explanatory power of the market timing theory differs across different industry sectors. I first test whether firms engage in market timing and subsequently research whether these market timing actions have a persistent effect on capital structure. It can be concluded that market timing, in line with the market timing theory, has a persistently negative effect on leverage among firms in the full sample as well as firms in the chemical industry. However, the results on the remaining industry sectors do not support the market timing theory. For the oil and gas production industry and the IT industry, the reported evidence for market timing does not seem to explain any persistent deviations in the capital structure. For the food and construction industry, there is no reported evidence of market timing actions.

(4)

1 Introduction

Deciding on a firm’s capital structure is an important topic in corporate finance. When making external financing decisions, a firm can choose between either equity or debt. In order to maximize firm value, the right amount of debt and equity must be chosen.

There exist many theories about how a firm’s capital structure is determined. The recent theory proposed by Baker and Wurgler (2002) suggests that a firm’s current capital structure is a result of previous attempts to time the equity market.

When a firm’s management issues equity when the share price is high and repurchases shares when the share price is low, this is referred to as ‘equity market timing’ (Baker & Wurgler, 2002). The existence of market timing is studied multiple times. Loughran and Ritter (1995) find evidence of market timing as well as Lucas and McDonald (1990) and Graham and Harvey (2001).

Researching the effect of market timing on capital structure has been done several times. Baker and Wurgler (2002) first opened the discussion by claiming that the market timing theory has ‘substantial explanatory power’ on a firm’s capital structure. Following Baker and Wurgler (2002), there have been many other studies on the relation between market timing and capital structure but there doesn’t seem to be a consensus on which theory best explains a firm’s capital structure. A remarkable conclusion from the study of Bruinshoofd & de Haan (2012) was that, despite the absence of evidence for market timing determining capital structure on their entire sample, they did find evidence of market timing determining the capital structure among firms in the ICT sector. This raises the question of how the explanatory power of the market timing theory of capital structure differs across industry sectors.

To answer this question, I will use the method proposed by Baker and Wurgler (2002). In order to examine whether firms engage in market timing, I will regress the market-to-book ratio on net equity issues as well as the other channels of the change in leverage, to see if misvaluation affects leverage through net equity issues, as proposed by the market timing theory. Subsequently, I will regress the EFWAMB, which is a variable that captures all relevant market timing actions, on leverage, to measure the persistent effect of market timing on capital structure. The sample used in this research consists of firms in the US from 2003 until 2018. In order to perform the cross-industry analysis, I will use the set of industry sectors introduced by Talberg et al. (2008) to construct different subsamples. The construction-, food-, oil and gas production-, chemical- and IT industry are used in the analysis. These industry sectors all experience different economic conditions which affect their issuing behavior and capital structure.

Although the existing research on the market timing theory is quite extensive, apart from the research conducted by Bruinshoofd and de Haan (2012), there has (to my knowledge) not been a sectoral analysis of the explanatory power of the market timing theory. This thesis examines further this difference in evidence of the market timing theory on capital structure across different industry sectors.

In the remainder of this thesis, I will first discuss the prevailing theories on capital structure, followed by a summary of existing literature on the market timing theory and on capital structure across different industries. In chapter 3, the research methodology is explained followed by a description of the data and the different variables. In chapter 4, the results are presented and discussed. Finally, I will conclude the findings and provide recommendations for further research in chapter 5.

(6)

2 Literature review

2.1 Prevailing capital structure theories

The oldest theory on capital structure is the one proposed by Modigliani and Miller (1958). In their study, they suggest 2 propositions on the cost of capital faced by firms. The first proposition states that ‘the market value of any firm is independent of its capital structure’ (M&M, 1958, p.268). Their second proposition states that the cost of equity is related to the level of leverage as a higher amount of debt increases the probability of default. These propositions are based on the assumption of perfect capital markets. This implies that firms pay no taxes, there are no transaction costs, borrowing costs, and bankruptcy costs and that there exists perfect symmetric information. This is obviously an unrealistic assumption. In their second version, Modigliani and Miller (1963) adjust their theorem by including the assumption that firms face taxes. They then conclude that the firm value is affected by the amount of debt used in the capital structure as the firm’s interest expense is deductible from the taxes a firm has to pay.

Kraus and Litzenberger (1973) added the bankruptcy costs into the equation, which lays the foundation of the trade-off theory. The argued that a firm’s value is affected by costs and benefits of debt through bankruptcy costs and the interest tax shield respectively and that firms choose their optimal capital structure by weighing the two against each other.

Jensen and Meckling (1976) criticize the findings of Modigliani and Miller. According to M&M, debt should never be used in the absence of tax benefits if bankruptcy costs are taken into account. Which would imply that prior to any tax benefit, no debt would be used. Since debt was used even in the absence of tax benefits, the theory proposed by M&M is incomplete. Jensen and Meckling (1976) argue that an important additional cost to consider is the agency cost. Agency costs are the costs associated with a conflict of interest between the managers and the shareholders. Jensen and Meckling (1976) argue that debt could be a way to restrict managers from taking actions that are not in line with the interest of the shareholders.

Another well-known theory on capital structure, which builds upon the concept of agency costs, is the pecking order theory proposed by Myers and Majluf (1984). The idea behind the pecking order theory is that firms prioritize between different sources of financing. First internal funds are preferred to debt issues and debt issues are preferred to equity issues. Managers are perceived to know more about the condition of the company than investors, so all actions management take, are interpreted by the outside investors. Assuming managers care more about the interest of existing shareholders compared to the interest of ‘new’ shareholders, equity is least preferred as this signals that the current share price is overvalued which would drive down the share price (the value invested in the firm by existing shareholders). Whereas a debt issue would signal that the current share price is undervalued.

2.2 Market timing

Market timing by a firm’s management is the practice of strategically planning equity issues and repurchases. The existence of market timing is studied multiple times.

According to Lucas and McDonald (1990), issuing behavior is affected by the adverse selection problem. This is a problem that occurs when there is asymmetric information. In this case, the firm’s management has more information than the investors do. Lucas and McDonald (1990) find that firms that are undervalued will wait to issue equity until this undervaluation is corrected, whereas firms that are overvalued will immediately issue equity as waiting may

(7)

cause a devaluation of the firm. Graham and Harvey (2001) try to describe the current practice of corporate finance by surveying roughly 400 CFO’s about various finance topics. They find that that ‘recent stock price performance is the third most popular factor affecting equity-issuance decisions’ (Graham & Harvey, 2001, p.222) which could be interpreted as evidence of market timing. In their study on the relation between overvaluation and issuing behavior, Dong et al. (2012) found strong evidence that overvaluation relates to greater equity issuance. For their research, they used residual income valuation to price (V/P) as a measure of misvaluation to test the correlation between misvaluation and new issues. They found that especially equity issues are sensitive to misvalution. Loughran and Ritter (1995) research the underperformance of firms following an equity issue. They differentiate between firms issuing an IPO and firms issuing an SEO. In both cases, firms seem to be underperforming significantly following an equity issue. This thus represents an overvaluation on which management acts by issuing equity. Jensen (2005) argues that financial management feels great amounts of pressure to act during times of overvaluation. Among others, this is due to the target-based corporate budgeting system in which managers are expected to reach a certain target and because of the management’s personal compensation. All these studies provide quite some evidence of the existence of market timing.

2.3 Previous studies on market timing and capital structure

The first to conduct research on the market timing theory of capital structure are Baker and Wurgler (2002). In their study, they argue that a firm’s capital structure is the cumulative outcome of previous attempts to time the equity market. According to B&W (2002), the effect of market misvaluation on capital structure is also quite persistent. They regress leverage on the ‘external finance weighted average market to book ratio’ (EFWAMB1_{). This variable}

summarizes the historical variation in market valuations which they use as a measure of market timing. Baker and Wurgler (2002) find a significantly negative relation between the EFWAMB and leverage which they see as evidence in favor of the market timing theory of capital structure. Other research on this topic found the same results as Baker and Wurgler (2002). Huang and Ritter (2009) find evidence for the market timing theory, they claim that ‘historical equity risk premia have long-lasting effects on leverage through their influence on securities issuance decisions’ (Huang & Ritter, 2009, p.36)

There are however many studies that also test the market timing theory but conclude otherwise. Bruinshoofd and de Haan (2012) conducted research on the market timing theory but their findings did not corroborate with the market timing theory but rather backed the pecking order theory as firms in their study preferred debt over external equity. Although Kayhan and Titman (2007) found evidence of market timing, they argue that this effect is short-lived as firms rebalance towards a target debt ratio. They suggest that market timing only has a short-term effect on a firm’s capital structure. In line with these findings, Leary and Roberts (2005) argue that the effect of market timing on capital structure disappears within 2 to 4 years. Similarly, Alti (2006) finds evidence for market timing but claims this effect on a firm’s capital structure is “completely vanished” in 2 years following the IPO. In his study, Alti (2006) compares the equity issuing behavior at IPO of ‘hot’ markets and ‘cold’ markets. He finds that firms in hot markets indeed do issue more equity at IPO but in the years following the IPO, issue relatively more debt to rebalance the firm’s capital structure. Hovakimian (2006) also tests the market timing theory by using the same method proposed by Baker and Wurgler. Hovakimian (2006) finds no significant evidence of market timing and thus also no persistent

(8)

effect on leverage. De Bie and de Haan (2007) used the same method proposed by Kayman and Titman (2007) for analyzing the effect of market timing on the capital structure of Dutch firms. Contrary to Kayhan and Titman, they found no (short-term) market timing effects but find evidence for capital structure targeting and the pecking order theory.

Altogether, there seems to be no real consensus on whether the market timing theory properly explains capital structure.

2.4 Sectoral differences

In their study on the relation between market timing and capital structure, Bruinshoofd and de Haan (2012) use the same method Baker and Wurgler (2002) proposed, by regressing leverage on EFWAMB and several control variables. They conducted their research on multiple countries, both in the US and in several countries in Europe. Bruinshoofd and de Haan (2012) concluded that the empirical evidence of the negative correlation between market timing and leverage found in the US did not extend to countries in Europe. Apart from a difference between countries, they also found this correlation to be different among different industry sectors. In Europe where Bruinshoofd and de Haan (2012) found no significant evidence of correlation between market timing and capital structure on the entire sample, they did find evidence of market timing effects on capital structure among ICT firms.

In their paper in which they study capital structure across industries, Talberg et al. (2008) differentiate between 5 different industries that all experience different economic conditions. The first industry used by Talberg et al. (2008) is the construction industry. According to Talberg et al. (2008), this industry sector is characterized as being sensitive to the general state of the economy. The second industry used is the food industry, which is known as a more stable industry as food is an essential product. At the same time, the food industry is highly competitive (Talberg et al., 2008). The third industry used in their analysis is the oil and gas production industry. This industry is a representative for a capital-intensive industry with high operating margins. The fourth industry sector used is the chemical industry which is known as being a mature industry. Finally, the IT (information technology) sector is used in the analysis to represent a relatively new industry. Talberg et al. (2008) find that various factors like profitability, growth, age, asset structure, and size characterize different industries and these factors seem to affect a firm’s capital structure.

(9)

3 Research methodology

For testing the theory of market timing, I will use the method introduced by Baker and Wurgler (2002) and later also used by Bruinshoofd and de Haan (2012) and many others. First, I will test whether firms engage in short-term market timing. Subsequently, in order to test the persistency of market timing, I will regress leverage on the EFWAMB, which represents a measure of historical market timing actions, and a set of control variables.

3.1 Current market timing

In order to test the market timing theory of capital structure, I first need to test whether firms engage in short-term market timing. To test whether firms participate in short-term market timing, Baker and Wurgler (2002) regress leverage on the market-to-book ratio and several control variables. According to the market timing theory, firms issue equity when the valuation of the firm is high. In order to test this relation between the valuation of a firm and equity issues, the dependent variable, leverage, is decomposed into three parts to see whether the market-to-book ratio affects leverage through net equity issues. The annual change in leverage can be decomposed into the change in equity issues, the change in retained earnings, and the residual change in leverage (Baker & Wurgler, 2002).

The annual change in leverage is defined as the annual change in book debt divided by total assets. As book leverage should always be between 0 and 1, values higher than 1 are excluded from the sample (Baker & Wurgler, 2002). Book debt is calculated by subtracting the book value of equity from total assets [Assets total – book equity]. Book equity, in turn, is calculated as follows: Assets total – Liabilities total – preferred stock + deferred taxes + convertible debt. Net equity issues !_"!" are defined as the sale of common and preferred stock minus the purchase of common and preferred stock, divided by total assets. Newly retained earnings !∆$%_" " is defined as the change in retained earnings divided by total assets. Finally, the residual change in leverage #$_&'(!_"(

!−

(

"!"#"& is defined as lagged book equity divided by total assets minus

lagged book equity divided by lagged total assets (Baker & Wurgler, 2002).

These components of the change in leverage are regressed on the market-to-book ratio and several control variables used by Rajan and Zinghales (1995). the market-to-book ratio is calculated by dividing the market value of equity by the book value of equity [market eq / book eq]. To mitigate any extreme values, market-to-book values larger than 10 are excluded from the sample (Baker & Wurgler, 2002). Rajan and Zinghales (1995) analyzed the capital structure of firms in industrialized countries across the world and found 3 factors to be of large explanatory power of capital structure.

The first important determinant they find is asset tangibility which is the ratio of fixed to total assets. For representing fixed assets, the value of property, plant, and equipment is used [PPENT / Assets total]. Tangible assets reduce the agency cost of debt because this class of assets is easily collateralized. A second important determinant of capital structure found by

!" #$!− ! " #$!"# = − ( ) #*!− ! ∆,-# $!− .-!"#! 1 #_!− 1 #_!"#$0 (1)

(10)

Rajan and Zinghales (1995) is profitability. According to Rajan and Zinghales (1995), Profitability and leverage are negatively correlated. Profitability is calculated by dividing the earnings before interest, taxes, depreciation, and amortization by total assets [EBITDA / Assets total]. Finally, an important determinant of capital structure found is size. Rajan and Zinghales (1995) argue that size could be related to the probability of default. Size is calculated by taking the logarithm of net sales (Baker & Wurgler, 2002).

As leverage is always between 0 and 1, when this variable reaches one boundary it can only go in one direction regardless of the effect from other variables. For this reason, lagged leverage is added to the set of control variables (Baker & Wurgler, 2002).

Summing up, this results in the following three models:

!_#"$ $= ' + ) * + ,-$%&+ . * //0 # - $%&+ 1 * 0,234# # -$%&+ "5678(:)< $%&+ 1 * 4 #-$%&+ =$ (2) >0$%&* 1 #$− 1 #$%&-A_$= ' + ) * + ,-$%&+ . * //0 # - $%&+ 1 * 0,234# # -$%&+ "5678(:)< $%&+ 1 * 4 #-$%&+ =$ (4)

The empirical evidence of market timing is tested by determining the correlation between the market-to-book ratio and equity issues. For this analysis, I am interested in knowing the ‘within-firm’ effect of this regression as I want to analyze how the market-to-book ratio determines a firm’s capital structure and how it influences the issuance of equity. For this reason, I will perform a regression analysis using the fixed effects model. This method looks at how the market-to-book ratio affects the dependent variable over time, within the same firm.

3.2 Persistency of market timing

Evidence of short-term market timing on itself does not necessarily imply that a firm’s capital structure is determined by these market timing actions. As found by Kayhan and Titman (2007), Leary and Roberts (2005), and Alti (2006), the effect of market timing on capital structure is only short-lived as these effects are eventually balanced away. To test the persistence of the effect of market timing on capital structure, Baker and Wurgler (2002) regress leverage on a newly proposed variable: the external finance weighted average market-to-book ratio (EFWAMB). This variable summarizes the historical variations in the firm’s market-to-book ratio by summing market-to-book ratios in each year, applying more weight to market-to-book ratios in times where significant amounts of external financing are issued.

*∆C0 # -$= ' + ) * + ,-$%&+ . * //0 # - $%&+ 1 * 0,234# # -$%&+ "5678(:)< $%&+ 1 * 4 #-$%&+ =$ (3) $'()*+_& = - .)+ 0) ∑&–(*-.(.*+ 0*) ∙ 4* +5₎ & )-( (5)

(11)

Net equity issues are defined as the sale of common and preferred stock minus the purchase of common and preferred stock. Net debt issues are defined as long-term debt issuance minus long-term debt reduction plus changes in current debt. Finally, the market-to-book ratio is calculated by dividing the market value of equity by the book value of equity.

The EFWAMB thus captures relevant previous market timing efforts. When, for example, the market-to-book ratio is relatively high in a certain year in combination with large amounts of external equity and debt issued that year, this will give high values of EFWAMB. The EFWAMB is essentially the sum of weighted averaged market-to-book ratios. Calculating the EFWAMB for 2005 would thus be done as follows:

This variable summarizes the historical variations in the firm’s market-to-book ratio by summing market-to-book ratios in each year, applying more weight to market-to-book ratios in times where significant amounts of external financing are issued.

Similar to the method used by Baker and Wurgler, any negative values for the sum of total financing are excluded from the sample as well as any values of EFWAMB higher than 10. (Baker & Wurgler, 2002).

Eventually, leverage will be regressed on EFWAMB and the 3 control variables found by Rajan and Zingales (1995) discussed above. The current market-to-book ratio is added as a control variable because it may be related to investment opportunities. The current market-to-book ratio controls for the cross-sectional variation in the market-to-market-to-book ratio and leaves the EFWAMB to explain the within-firm variation in the market-to-book ratio (Baker & Wurgler, 2002).

This results in the following regression model:

As opposed to the regression model used to test for short-term market timing, when regressing leverage on the EFWAMB, more interesting is the result on the cross-section. To make sure this model measures the effect of the EFWAMB on leverage between panels, I will use the random effects regression model.

3.3 Samples and summary statistics

The research sample will consist of firms in the US as more empirical evidence for market timing is found in the US than for example in Europe (Bruinshoofd & de Haan, 2012). The period on which the research is conducted is from 2003 until 2018 to ensure enough observations are included but to minimize the chance of obtaining ‘noisy’ data from the financial crisis in 2001. The data is obtained from the merged COMPUSTAT CRSP set $'()*+/..0 = ._/..1+ 0_/..1 ∑(._/..1'/..0+ 0_/..1'/..0)∙ * +/..1+ ._/..2+ 0_/..2 ∑(._/..1'/..0+ 0_/..1'/..0) ∙* +/..2+ ./..0+ 0/..0 ∑(./..1'/..0+ 0/..1'/..0)∙ * +/..0 (6) "# $%!= ' + )(+,-$./)!"#+ 1 " . /%!"#+ 2 " 33+ $ % !"#+ 4 " +/56#$ $ %!"#+ 789:;(<)= !"#+ >! (7)

(12)

provided by WRDS. In accordance with Baker and Wurgler (2002), firms within the financial sector (with SIC (Standard Industrial Classification) codes 6000-6999) are excluded from the sample as this sector is heavily regulated. Take for example the Basel 2 agreement which regulates banking firms’ capital structure. (Talberg et al., 2008). Furthermore, firms with a book value of assets below $10 million and firms with missing data on total assets are excluded from the sample. With these restrictions taken into account, the entire dataset consists of 7659 firms with a total of 54656 firm-year observations.

To test the market timing theory between different industries, I will use the set of fundamentally different industries introduced by Talberg et al. (2008). In their paper in which they study capital structure across industries, Talberg et al. (2008) differentiate between 5 different industries as discussed in chapter 2.

Talberg et al (2008) used the ICB classification codes to differentiate between these industries. Unfortunately, this classification system is not available on the merged COMPUSTAT CRSP database I used. Therefore, I used the Standard Industrial Classification system to differentiate between the different industry sectors. For the construction industry, I used SIC codes 1500-1799. For the food industry, I used SIC codes 2000-2099. For the oil and gas production industry sector, I used SIC codes 1311-1389. For the chemical industry, I used the SIC codes 2800-2899. Finally, for the IT sector, I used SIC codes representing the software and computer service industry: 7371-7373.

Table 1 gives the descriptive statistics of the capital structure as well as the financing decisions per sector. Extreme values at the 1st_{and 99}th_{percentile have been eliminated for the net debt}

issues, net equity issues and newly retained earnings to correct for skewness. Book leverage is shown in the first panel. The average level of book leverage for the full sample is 42%. The construction industry seems to be the most highly levered industry with a debt level of 53,2%, whereas the chemical sector on average holds the lowest level of debt with a reported mean value of book leverage of 32,9%. Panel B shows the summary statistics of the debt issuing activities per sector. The full sample issues 1% of total assets in debt on average. The oil and gas production industry issue the highest amount of debt with an average of 3,7%, whereas the IT sector seems to issue the lowest amount of debt with a mean value of 0,5%. Panel C gives the equity issuing activity per sector. On average, the full sample issues equity worth of 4,2% of total assets. The chemical industry reports the highest issuance of equity with a mean value of 19%, whereas the construction industry reports the lowest equity issuance with an average of 0,5%. Finally, panel D shows the newly retained earnings relative to total assets. This panel shows similar activities across all subsamples with a reported mean of 0,1% of the full sample and 0,3% for the chemical industry which is the highest value of newly retained earnings observed across the different industries.

Table 2 shows the summary statistics of the variables used in the regression models. As previously mentioned, for the purpose of calculating EFWAMB, any negative values for the total sum of external financing are excluded from the sample. This is done to ensure that the weights are increasing in the total amount of debt and equity issues (Baker & Wurgler, 2002). This restriction significantly decreases the sample size as can be seen in table 2. The average market-to-book ratio of the full sample is 2,37 which indicates that on average, the firms observed in this research are overvalued. The highest average values of the market-to-book ratio are found for firms operating in the chemical industry with a reported mean value of 3,28. The lowest mean value of market-to-book is found in the construction industry with an average of 1,37. The average value of EFWAMB for the full sample is 2,73. The highest mean value of EFWAMB is found among firms operative in the chemical industry with an average of 3,56, whereas the lowest value of EFWAMB is found among firms in the construction industry with

(13)

a mean value of 1,8. The summary statistics of the control variables used in the regression models are also shown in table 2.

The statistics of the net equity issuing activity in table 1 and the market-to-book and EFWAMB in table 2 are of particular interest to this research as certain expectations for market timing can be inferred from these reported values. The market timing theory predicts that overvalued firms issue relatively large amounts of equity. High values of EFWAMB would then imply that high values for market-to-book ratio are relevant for the issuance of external equity and debt. For the full sample, this relation could be inferred from the reported values. A combination of relatively high market-to-book ratio, high equity issuance, and relatively high values for EFWAMB could imply that market timing has the expected persistent effect on leverage. This relation is also found in the oil and gas production industry, the chemical industry, and in the IT sector. Especially the chemical sector reports the potential of this persistent effect of market timing on capital structure with the highest mean values of market-to-book, equity issuance, and EFWAMB. In addition, the chemical sector reports the lowest level of leverage across all subsamples which supports the market timing theory of capital structure even more.

(14)

Table 1: Summary statistics of capital structure and financing decisions

Table 1 gives the number of observations, mean, standard deviation, minimum value, maximum value, 25% quantile, median and 75% quantile of the capital structure and components of the change in the capital structure per sector. Book leverage is defined as book debt divided by assets. Net equity issues are the sale of common and preferred stock minus the purchase of common and preferred stock, divided by total assets. Net debt issues are long-term debt issuance minus long-term debt reduction plus changes in current debt, divided by total assets. Newly retained earnings are defined as the change in retained earnings divided by total assets. Outliers at the 1st_{and 99}th_{percentile in net}

debt issues, net equity issues and newly retained earnings are eliminated.

Panel A: Book leverage !!

""

Sector N Mean S.D. Min Max q25 Median q75

full sample 54656 42,0% 21,6% -21,6% 99,7% 24,6% 41,6% 58,0% Construction 753 53,2% 16,8% 3,7% 95,9% 44,8% 53,4% 63,6% Food 1400 46,6% 18,5% 5,2% 93,9% 32,3% 47,6% 59,7% oil and gas production 2756 42,2% 20,5% 0,0% 99,0% 27,7% 41,9% 55,4% Chemical 6654 32,9% 21,5% -8,1% 98,2% 14,2% 29,0% 48,6% IT 2814 36,5% 17,3% 2,7% 95,2% 23,4% 34,2% 47,9%

Panel B: Net debt issues !#_""

full sample 28028 1,0% 7,2% -21,1% 33,6% -0,5% 0,0% 1,3% Construction 233 0,9% 7,5% -21,1% 33,6% -1,7% 0,0% 4,5%

Food 733 1,2% 7,9% -21,1% 33,6% -1,9% 0,0% 3,5%

oil and gas production 1298 3,7% 10,4% -21,1% 33,6% -0,5% 0,3% 8,3% Chemical 3701 1,4% 7,7% -21,1% 33,6% 0,0% 0,0% 0,0%

(15)

Panel C: Net equity issues !_"$"

full sample 49925 4,2% 17,4% -20,5% 97,9% -0,5% 0,0% 1,1% construction 697 0,4% 6,4% -20,5% 46,4% -0,6% 0,0% 0,3%

food 1320 0,5% 9,0% -20,5% 97,9% -1,2% 0,0% 0,3%

oil and gas production 2545 5,5% 13,9% -20,5% 97,9% 0,0% 0,1% 5,5% chemical 5574 19,0% 32,3% -20,5% 97,9% 0,0% 0,7% 35,5%

IT 2624 2,9% 13,9% -20,5% 97,9% -0,8% 0,3% 2,1%

Panel D: Change in retained earnings !∆&'_" "

full sample 46746 0,1% 0,8% -3,2% 4,6% 0,0% 0,0% 0,1%

construction 678 0,0% 0,7% -3,2% 4,6% 0,0% 0,0% 0,0%

food 1258 0,1% 0,7% -3,2% 4,6% 0,0% 0,0% 0,0%

oil and gas production 2177 0,1% 0,8% -3,2% 4,6% 0,0% 0,0% 0,1%

chemical 5571 0,3% 0,8% -3,2% 4,6% 0,0% 0,1% 0,3%

(16)

Table 2: Summary statistics of the variables used in the regression models

Table 2 gives the number of observations, mean, standard deviation, minimum value, maximum value, 25% quantile, median and 75% quantile of the independent variables used in the regression models, per sector. The variable EFWAMB is defined as the sum of market-to-book ratios weighted by the relative amount of financing in every respective year. To eliminate too extreme values, values of EFWAMB higher than 10 are excluded from

the sample as well as outliers at the 1st_{and 99}th_{percentile. The market-to-book ratio is defined as market equity}

divided by book equity. Outliers of the market-to-book ratio are excluded by dropping firm observations with a market-to-book ratio larger than 10. Asset tangibility is calculated by dividing the value of property, plant, and equipment by the total assets. Profitability is defined as the EBITDA divided by total assets. Size is defined as

the log of net sales.

EFWAMB

full sample 21536 2,73 1,98 -1,64 8,97 1,34 2,21 3,69

construction 212 1,80 1,18 -1,64 7,69 0,99 1,70 2,38

food 498 2,47 2,08 -1,64 8,97 1,32 1,75 3,09

oil and gas production 1423 2,22 1,61 -1,64 8,97 1,17 1,78 2,83

chemical 3189 3,56 1,94 -1,64 8,97 2,14 3,30 4,78

IT 1337 3,50 2,01 -1,64 8,97 1,96 3,01 4,83

Marketbook

full sample 21536 2,37 1,86 0,01 9,99 1,10 1,79 3,09

construction 212 1,37 0,88 0,10 5,68 0,82 1,26 1,65

food 498 2,26 1,87 0,06 9,87 1,15 1,70 2,74

oil and gas production 1423 1,83 1,53 0,04 9,86 0,83 1,39 2,29

chemical 3189 3,28 2,13 0,06 9,97 1,68 2,79 4,41

(17)

Tangibility

full sample 21521 27,5% 28,2% 0,0% 99,6% 4,7% 14,9% 46,1% construction 203 8,3% 12,9% 0,0% 55,6% 0,8% 2,2% 9,9%

food 498 30,0% 16,8% 0,3% 90,2% 18,3% 27,7% 39,2%

oil and gas production 1423 74,2% 20,4% 0,0% 99,6% 65,2% 81,1% 89,1%

chemical 3189 11,0% 15,0% 0,0% 93,4% 1,1% 4,2% 15,1%

IT 1336 5,9% 6,1% 0,0% 49,8% 2,2% 3,9% 7,3%

Profitability

full sample 21503 0,00 0,26 -4,74 10,70 -0,03 0,08 0,13

construction 212 0,05 0,11 -0,43 0,36 0,03 0,07 0,10

food 498 0,10 0,11 -0,54 0,45 -0,43 0,10 0,15

oil and gas production 1422 0,06 0,23 -2,81 0,86 0,02 0,11 0,17 chemical 3182 -0,24 0,43 -3,57 10,70 -0,46 -0,22 0,08

IT 1334 0,02 0,18 -1,67 0,63 -0,02 0,06 0,12

Size

full sample 20256 5,58 2,42 -6,91 12,74 4,08 5,63 7,21

construction 212 7,31 1,77 2,20 10,01 6,37 7,66 8,69

food 498 6,58 1,96 1,87 11,41 5,29 6,47 7,86

oil and gas production 1384 5,87 2,27 -2,22 10,78 4,73 6,32 7,38

chemical 2603 3,65 3,05 -6,91 11,36 1,80 3,53 5,57

(18)

4 Results

As discussed in the previous chapter, in order to measure the effect of market timing on the capital structure, I will perform several regression analyses. Regression models 2 to 4 are performed to measure whether firms engage in short-term market timing and regression model 7 is performed to see whether these potential short-term market timing actions have a persistent effect on a firm’s capital structure. Finally, I check the robustness of the results of regression model 7 by using a different definition of leverage: market leverage.

4.1 Short-term market timing regression results

The total change in leverage can be decomposed into 3 parts as described by model 1. Each component individually affects the change in leverage. Models 2 to 4 measure the effect of the market-to-book ratio on these components of the change in leverage to see through which channels the market-to-book ratio affects the change in leverage.

Table 3 gives the regression output of models 2 to 4. In panel A, the effect of the book ratio on net equity issues is shown. According to the market timing theory, the market-to-book ratio is positively correlated with net equity issues. This would imply that firms with a high market-to-book ratio issue new equity. As shown in panel A, the regression on the full sample reports a significantly positive coefficient of 0,0069 with a t-value of 18,58. This positive correlation indicates that a higher market-to-book ratio would result in more net equity issues, which is what the market timing theory proposes.

Comparing the results across different industries, this positive correlation is reported among firms operating in the oil and gas production industry, the chemical industry, and the IT industry with very significant coefficients of 0,0059, 0,0158 and 0,0071 respectively. Among firms in the food industry, the market-to-book ratio doesn’t seem to be correlated with net equity issues with a reported insignificant coefficient of -0,0003 with a corresponding t-value of -0,19. Firms operative in the construction industry show the opposite of what the market timing theory predicts. With a reported significant coefficient of -0,006 with a t-value of -2,34, the market-to-book ratio is negatively correlated with net equity issues which implies that a higher market value would decrease the net equity issues.

The impact of the market-to-book ratio on the newly retained earnings is shown in panel B. For the full sample and all individual industries, the reported correlations seem to be small and insignificant. Both the market-to-book ratio and the other variables don’t seem to be of large impact on the newly retained earnings with all relatively small and insignificant reported coefficients.

Finally, the regression output on the residual growth in assets is shown in panel C. These results show negative and significant coefficients across all industries with a coefficient of -0,0329 and a t-value of -30,76 for the full sample. This negative correlation implies that a higher market-to-book ratio would increase leverage through the residual growth in assets.

These results indicate that the market-to-book ratio seems to have the expected correlation with leverage. Except for the food and construction industry, all subsamples show a positive and

(19)

significant correlation between the market-to-book ratio and net equity issues. These findings can be interpreted as evidence of market timing.

Table 3: short-term market timing regression output

Table 3 gives the fixed regression output of models 2, 3, and 4. Coefficients are reported as well as t-statistics reported in parentheses. The coefficients are labeled with * if significant at the 10% level of alpha, ** if significant at the 5% level of alpha and *** if significant at the 1% level of alpha. Net equity issues are the sale

of common and preferred stock minus the purchase of common and preferred stock, divided by total assets. Newly retained earnings are defined as the change in retained earnings divided by total assets. The residual change in leverage is defined as lagged book equity divided by total assets minus lagged book equity divided by

lagged total assets. The market-to-book ratio is defined as market equity divided by book equity. Outliers of the market-to-book ratio are excluded by dropping firm observations with a market-to-book ratio larger than 10.

Asset tangibility is calculated by dividing the value of property, plant, and equipment by the total assets. Profitability is defined as the EBITDA divided by total assets. Finally, Size is defined as the log of net sales. Lagged book leverage is not reported. The number of observations differs slightly in each panel due to some

missing values of the dependent variable.

Panel A: regression on net equity issues

full sample construction food oil and gas prod chemical IT

N 42012 609 1180 2158 4076 2188 MB 0,0069*** -0,006** -0,0003 0,0059*** 0,0158*** 0,0071*** (18,58) (-2,34) (-0,19) (3,54) (8,57) (4,98) Tangibility 0,0447*** -0,005 -0,0288 0,0772*** 0,1055** 0,1133** (6,74) (-0,09) (-0,92) (3,91) (2,42) (1,96) Profitability -0,1261*** -0,0276 -0,1209*** -0,0189 -0,2314*** 0,0079 (-27,35) (-1,11) (-3,91) (-1,53) (-12,9) (0,36) Size -0,0212*** -0,0157*** -0,0206*** -0,0272*** -0,0133*** -0,0212*** (-26,11) (-3,97) (-4,67) (-9,97) (-3,99) (-5,56) Cons. 0,0882*** 0,0967*** 0,1218*** 0,1305*** 0,0364* 0,0493** (16,69) (3,21) (3,65) (6,51) (1,92) (2,47)

Panel B: regression on newly retained earnings

N 44655 653 1245 2111 4780 2313 MB 0,0003 0,0006 -0,0009 0,0001 -0,0007 0,0045 (0,34) (0,36) (-1,15) (0,1) (-1,2) (0,34) Tangibility -0,0014 -0,0254 -0,0051 0,0038 -0,0042 0,1776 (-0,1) (-0,67) (-0,4) (0,29) (-0,3) (0,34) Profitability -0,0132 -0,0117 0,0125 0,0003 0,0037 -0,1828 (-1,33) (-0,7) (0,97) (0,04) (0,66) (-0,91) Size 0,0007 0,0033 -0,0029 0,0012 0,0001 -0,0063 (0,38) (1,24) (-1,55) (0,65) (0,1) (-0,18) Cons. 0,0043 -0,0138 0,0185 -0,0071 0,0037 0,0957 (0,38) (-0,7) (1,32) (-0,52) (0,65) (0,51)

(20)

Panel C: regression on residual growth in assets

N 45674 654 1245 2335 4806 2319 MB -0,0329*** -0,0204*** -0,0091** -0,0477*** -0,0404*** -0,0326*** (-30,76) (-3,49) (-2,32) (-12,65) (-15,67) (-9,51) Tangibility -0,057*** -0,0245 -0,1655** -0,0084 -0,14** -0,4204*** (-2,95) (-0,19) (-2,5) (-0,19) (-2,21) (-3,12) Profitability -0,0363*** -0,3374*** -0,4443*** -0,1011*** 0,1338*** -0,2262*** (-2,77) (-5,97) (-6,59) (-3,56) (5,39) (-4,35) Size 0,0399*** 0,0555*** 0,0299*** 0,0593*** 0,0008 0,0418*** (17,32) (6,2) (3,09) (9,76) (0,17) (4,58) Cons. -0,2257*** -0,4211*** -0,2154*** -0,3925*** 0,1283*** -0,1089** (-15,1) (-6,29) (-2,95) (-8,85) (5,12) (-2,28)

4.2 Results on the persistency of market timing

As mentioned in the previous subsection, there seems to be reported evidence of short-term market timing. Only firms in the food and construction industry report an (insignificantly) negative correlation between the market-to-book ratio and net equity issues which is the opposite of what the market timing theory predicts. In line with these findings, I expect to find potential evidence for any long-term effect of market timing on capital structure among firms operating in the oil and gas production industry, the chemical industry, the IT industry, as well as for the full sample.

In table 4, the regression results of model 7 are shown. As mentioned previously, the number of observations in model 7 decreases significantly due to the additional restrictions for the purpose of computing the EFWAMB. The results on the full sample show a significantly negative correlation between the EFWAMB and leverage with a reported coefficient of -0,0084 with a corresponding Z-value of -7,94. This significantly negative correlation is also found among firms in the chemical industry with a reported coefficient of 0,0108 with a Zvalue of -3,3. The residual industry sectors all report relatively small and insignificant correlations between the EFWAMB and leverage. These results imply that market timing actions have a persistently negative effect on leverage on the full sample as well as among firms operating in the chemical industry. Evidence of market timing found among firms in the oil and gas production industry and firms in the IT industry, does not seem to induce a persistent effect on leverage. In line with the absence of evidence of market timing for firms operating in the food and construction industry, there is also no reported persistent effect of market timing on leverage.

(21)

Table 4: EFWAMB regression

Table 4 gives the random effects- regression output of model 7. Coefficients are reported as well as Z-statistics reported in parentheses. The coefficients are labeled with * if significant at the 10% level of alpha, ** if significant at the 5% level of alpha and *** if significant at the 1% level of alpha. The dependent variable book leverage is defined as book debt divided by assets. The independent variable EFWAMB is defined as the sum of

market-to-book ratios weighted by the relative amount of financing in every respective year. To eliminate too extreme values, values of EFWAMB higher than 10 are excluded from the sample as well as outliers at the 1st

and 99th_{percentile. The market-to-book ratio is defined as market equity divided by book equity. Outliers of the}

market-to-book ratio are excluded by dropping firm observations with a market-to-book ratio larger than 10. Asset tangibility is calculated by dividing the value of property, plant, and equipment by the total assets. Profitability is defined as the EBITDA divided by total assets. Finally, Size is defined as the log of net sales.

N 16430 170 414 1144 2018 1044 EFWAMB -0,0084*** 0,0065 -0,0035 0,0009 -0,0108*** 0,0021 (-7,94) (0,74) (-0,67) (0,17) (-3,3) (0,56) MB 0,0026*** 0,0146 0,0076* 0,0014 0,0029 0,0006 (3,36) (1,39) (1,76) (0,37) (1,4) (0,22) Tangibility 0,139*** -0,134 0,3622*** 0,1366*** 0,2271*** 0,1911* (14,76) (-0,83) (5,62) (3,57) (5,71) (1,89) Profitablility -0,1557*** -0,2688*** -0,3045*** -0,1292*** -0,0752*** -0,3035*** (-22,66) (-3,71) (-4,29) (-5,44) (-6,05) (-9,44) Size 0,0322*** 0,0267** 0,0474*** 0,0264*** 0,0189*** 0,0252*** (29,59) (2,46) (5,83) (6,37) (8,13) (5,03) Cons. 0,2249*** 0,344*** 0,0807 0,2058*** 0,2536*** 0,2129*** (29,52) (4,16) (1,28) (5,31) (15,59) (7,57)

(22)

4.3 Robustness check

To see whether the results differ significantly when leverage is defined differently, I also ran the EFWAMB regression on market leverage as opposed to book leverage. Ferris et al. (2018) studied the relation between book leverage and market leverage and found that although these two measures move closely together, market leverage is more volatile. Market leverage is calculated by dividing the book value of debt by total assets corrected for the market equity [book debt/ (Assets total – book equity + market equity)]. Book debt is calculated by subtracting the book value of equity from total assets [Assets total – book equity]. Book equity, in turn, is calculated as follows: Assets total – Liabilities total – preferred stock + deferred taxes + convertible debt. Finally, market equity is calculated by multiplying the total shares outstanding by the share price [shares outstanding * share price].

From table 5 it can be concluded that the results did not change significantly overall. Only the coefficients for the market to book ratio turned significantly negative in the above regression results whereas, in the regression on book leverage, the market to book ratio did barely show any significant coefficients. This can be explained through the correlation between market value which is quantified by the market to book ratio and its effect on market equity. A higher market to book ratio increases the market equity which in turn decreases the market leverage, hence the significantly negative correlation.

Table 5: EFWAMB regression (market leverage)

Table 4 gives the random effects- regression output of model 7. Coefficients are reported as well as Z-statistics reported in parentheses. The coefficients are labeled with * if significant at the 10% level of alpha, ** if significant at the 5% level of alpha and *** if significant at the 1% level of alpha. The dependent variable market leverage is calculated by dividing the book value of debt by total assets corrected for the market equity.

The independent variable EFWAMB is defined as the sum of market-to-book ratios weighted by the relative amount of financing in every respective year. To eliminate too extreme values, values of EFWAMB higher than

10 are excluded from the sample as well as outliers at the 1st_{and 99}th_{percentile. The market-to-book ratio is}

defined as market equity divided by book equity. Outliers of the market-to-book ratio are excluded by dropping firm observations with a market-to-book ratio larger than 10. Asset tangibility is calculated by dividing the value

of property, plant, and equipment by the total assets. Profitability is defined as the EBITDA divided by total assets. Finally, Size is defined as the log of net sales.

N 16430 170 414 1144 2018 1044 EFWAMB -0,0085*** 0,0102 -0,009 -0,0021 -0,0106*** -0,0025 (-7,47) (0,90) (-1,40) (-0,35) (-3,86) (-0,74) MB -0,024*** -0,0458*** -0,020*** -0,035*** -0,0134*** -0,0241*** (-27,89) (-3,37) (-3,55) (-7,49) (-7,43) (-8,45) Tangibility 0,1779*** -0,2960 0,2834*** 0,1959*** 0,2123*** 0,2836*** (18,57) (-1,46) (3,81) (4,44) (6,37) (3,05) Profitablility -0,1219*** -0,3427*** -0,4485*** -0,0787*** -0,0532*** -0,1423*** (-16,38) (-3,65) (-5,06) (-2,68) (-5,10) (-4,60) Size 0,0318*** 0,0265* 0,0298*** 0,0236*** 0,0182*** 0,0138*** (28,45) (1,93) (3,44) (5,14) (9,37) (3,13) Cons. 0,1956*** 0,4248*** 0,2243*** 0,2087*** 0,1924*** 0,2204*** (25,11) (4,08) (3,27) (4,67) (14,28) (9,03)

(23)

5 Conclusion

The purpose of this research is to test the explanatory power of the market timing theory on a firm’s capital structure across different industry sectors. The results of the short-term market timing regression analysis show that firms in the full sample engage in short-term market timing. Evidence for market timing is also seen in the oil and gas production industry, chemical industry, and the IT industry.

Results of the EFWAMB regression show that market timing has the expected persistently negative effect on leverage in the full sample as well as among firms in the chemical industry. For the oil and gas production industry and the IT industry, the reported evidence for market timing does not seem to explain any persistent deviations in the capital structure.

The results for the full sample and the chemical industry are in line with the findings of Baker and Wurgler (2002). The results for the oil and gas production industry and IT industry, however, coincide with the findings of Kayhan and Titman (2007), Leary and Roberts (2005), and Alti (2006), in the sense that the effect of market timing on capital structure is only short-lived. The results for the food and construction industry, on the other hand, show no evidence of (short-term) market timing and thus no reported effect on capital structure, in line with the findings of Hovakimian (2006) and de Bie & de Haan (2004). It can be concluded that although market timing actions reduce leverage in the long run in some subsamples, this trend is not observed in all subsamples.

This research merely exhibits the differences in market timing behavior and its effect on the capital structure across different industries. Future research could look more into the fundamentals of these different industries to determine what specific economic conditions could explain this difference in equity issuance and its effect on the capital structure.

(24)

6 References

Alti, A. (2006). How Persistent Is the Impact of Market Timing on Capital Structure? The

Journal of Finance, 61(4), 1681-1710.

Baker, M., & Wurgler, J. (2002). Market Timing and Capital Structure. The Journal of

Finance, 57(1), 1-32.

Bruinshoofd, A. & De Haan, L. (2012). Market timing and corporate capital structure – A transatlantic comparison. Applied economics, 44(28), 3691-3703

De Bie, T. & De Haan, L. (2007). Market Timing and Capital Structure: Evidence for Dutch Firms. Quarterly Review of The Royal Netherlands Economic Association, 155, 183-206. Dong, M., Loncarski, I., Horst, J., & Veld, C. (2012). What Drives Security Issuance

Decisions: Market Timing, Pecking Order, or Both? Financial Management, 41(3), 637-663. Ferris, S. P., Hanousek, J. & Shamshur, A. (2018). Asymmetries in the Firm's use of debt to changing market values. Journal of corporate finance, 48, 542-555.

Graham, J. R. & Harvey, C. R. 2001. The theory and practice of corporate finance: evidence from the field. Journal of Financial Economics, 60, 187-243.

Hovakimian, A. (2006). Are Observed Capital Structures Determined by Equity Market Timing? The Journal of Financial and Quantitative Analysis, 41(1), 221-243.

Huang, R., & Ritter, J. (2009). Testing Theories of Capital Structure and Estimating the Speed of Adjustment. The Journal of Financial and Quantitative Analysis, 44(2), 237-271.

Jensen, M. (2005). Agency Costs of Overvalued Equity. Financial Management, 34(1), 5-19. Jensen, M. C. & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3, 305-360.

Kayhan, A. & Titman, S. (2007). Firms’ histories and their capital structures. Journal of

Financial Economics, 83(1), 1-32.

Kraus, A., & Litzenberger, R. (1973). A State-Preference Model of Optimal Financial Leverage. The Journal of Finance, 28(4), 911-922.

Leary, M. T. & Roberts, M. R. (2005). Do Firms Rebalance Their Capital Structures? The

Journal of Finance, 60(6), 2575-2619.

Loughran, T., & Ritter, J. (1995). The New Issues Puzzle. The Journal of Finance, 50(1), 23-51.

Lucas, D., & McDonald, R. (1990). Equity Issues and Stock Price Dynamics. The Journal of

(25)

Majluf, N. S. & Myers, S. C. (1984). Corporate Financing and Investment Decisions When Firms Have InformationThat Investors Do Not Have. Journal of Financial Economics, 13(2), 187-221.

Modigliani, F., & Miller, M. (1959). The Cost of Capital, Corporation Finance, and the Theory of Investment: Reply. The American Economic Review, 49(4), 655-669.

Modigliani, F., & Miller, M. (1963). Corporate Income Taxes and the Cost of Capital: A Correction. The American Economic Review, 53(3), 433-443.

Rajan, R., & Zingales, L. (1995). What Do We Know about Capital Structure? Some Evidence from International Data. The Journal of Finance, 50(5), 1421-1460.

Talberg, M., Winge, C., Frydenberg, S. & Westgaard, S. (2008). Capital Structure Across Industries. International Journal of the Economics of Business, 15(2), 181-200.

The effect of market timing on capital structure: a cross-industry analysis