University of Amsterdam
Amsterdam Business School
BSc Economics and Business
Finance and Organization
A STUDY ON THE RELATIONSHIP BETWEEN LEVERAGE
AND FUTURE FIRM GROWTH
Author: L.V. Veenman
Student number: 5893674
Thesis supervisor: Dr. Jan Lemmen
Finish date: February 2016
ii
Acknowledgements
Foremost, I would like to show a deep sense of gratitude to my thesis supervisor dr. Jan Lemmen for his continuous guidance and support in the process of writing my bachelor thesis. Without his patience, motivation, preciseness and knowledge I could never have achieved completing this thesis.
Further, I would like to thank my thesis coordinator dr. Versijp for his help in the classes.
I would also like to thank my father, who was always willing to help and listen if I needed a sounding board.
iii
ABSTRACT
This study researches a firms’ capital structure decision and its effect on future firm growth.
Future firm growth is measured in employment and capital expenditures growth on a 1- and
3-year basis. Regressions on US manufacturing companies over the whole period 2004-2014 show
a negative relation between leverage and future firm growth. When making a distinction between
low and high q firms, there is a strong negative relation for low q firms and a weak/no relation
for high q firms. The sub periods mainly show the same results but the sub period 2010-2014
shows the possible influence of the global crisis.
Key words: Leverage, growth opportunities, Tobin’s q, Capital structure
JEL Classification: G32
NON-PLAGIARISM STATEMENT
By submitting this thesis the author declares to have written this thesis completely by himself/herself, and not to have used sources or resources other than the ones mentioned. All sources used, quotes and
citations that were literally taken from publications, or that were in close accordance with the meaning of those publications, are indicated as such.
COPYRIGHT STATEMENT
The author has copyright of this thesis, but also acknowledges the intellectual copyright of contributions made by the thesis supervisor, which may include important research ideas and data. Author and thesis supervisor will have made clear agreements about issues such as confidentiality.
iv
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ii ABSTRACT iii TABLE OF CONTENTS iv LIST OF TABLES v CHAPTER 1 Introduction 1CHAPTER 2 Literature review 3
2.1 Modigliani and Miller 3
2.2 Trade-off theory and agency costs 4
2.3 Pecking order theory and asymmetric information 5
2.4 Adjustment cost theories and other determinants of the leverage ratio 5 2.5 The capital structure decision’s influence on firm performance 7
CHAPTER 3 Data and methodology 10
3.1 Data 10
3.2 Methodology 11
3.2.1 Dependent variables 12
3.2.2 Explanatory variables 12
3.2.3 Regression and hypotheses 13
CHAPTER 4 Results 14 4.1 Whole period 2004-2014 14 4.2 Sub periods 18 4.2.1 2004-2006 18 4.2.2 2007-2009 19 4.2.3 2010-2014 20
CHAPTER 5 Discussion and concluding remarks 22
v
LIST OF TABLES
Table 2.1 Lang et al. (1986) study on relation between leverage and future firm growth 8 Table 2.2 Hurme (2010) study on relation between leverage and future firm growth 9
Table 3.1 General features of the data set 10
Table 3.2 Correlations between variables 11
Table 4.1 OLS regression over 2004-2014 15
Table 4.2 R-Square and Adj. R-square of previous studies 15 Table 4.3 Firms’ average debt and average debt ratios over 2004-2007 and 2010-2014 16 Table 4.4 Regression with anticipated growth opportunities over 2004-2014 17 Table 4.5 Regression with anticipated growth opportunities and R&D over 2004-2014 18 Table 4.6 Regression with anticipated growth opportunities and R&D over 2004-2006 19 Table 4.7 Regression with anticipated growth opportunities and R&D over 2007-2009 20 Table 4.8 Regression with anticipated growth opportunities and R&D over 2010-2014 21
1
CHAPTER 1 Introduction
Although capital structure is widely discussed in previous literature this thesis tries to give a contribution to the continuing discussion about the leverage in a company and its effect on firm growth. Decisions over and about the optimal debt level has been viewed in many literature, but modern capital structure was first discussed in Modigliani and Miller's (MM) work (Modigliani and Miller 1958; Modigliani and Miller 1963). In their work they state that under perfect market conditions capital structure does not have relevance within a firm and shareholders’ wealth does not depend on financing decisions. Later works have mitigated these findings of MM by viewing them in the light of real (imperfect) market conditions, which will be further discussed in the literature overview in Chapter 2.
One of the main questions within the capital structure debate is the relationship between the use of leverage and future firm growth. For example, Sharpe (1994) found that the impact of sales growth on firm growth is based on amount of leverage in firm. In recent literature Hurme (2010) had studied the relation between debt and future firm growth for 1990-2008, where the settings of Lang et al. (1996), which studied the relationship between 1970-1989, closely have been followed. In their studies future firm growth is measured in Investments, Employment growth and Capital expenditure growth. In both studies a negative relation was found between leverage and future firm growth. However, in 2011 Francis et al. (2011) found that this relation was significantly mitigated by controlling for remuneration of CEOs. This thesis will continue to examine this relation between debt and future firm growth but now also the global financial crisis will be taken into account and its effect on the relation of debt and future firm growth. Since Q1 of 2007 the Federal fund rate has declined from 5.25% to the, since 2009, set rate between 0-0.25%1 . As a result, the prime interest rate at which companies can borrow has also decreased, and thus makes it cheaper to borrow. The Securities Industry and Financial Markets Association (SIFMA) database shows that since 2008 the yearly corporate debt issuing volume has risen 200% to $1493,2 billion in 2015, and the debt market has been labeled as ”hot”. It is expected that this also has its influence on a company’s debt ratio and on the debt service cost of that chosen amount of debt.
Hence, this Thesis will research the relationship between debt and “future firm growth” for listed US manufacturing companies in the period 2004-2014. Besides I will also research the sub periods 2004-2006 (before financial crisis), 2007-2009 (financial crisis) and 2010-2014 (post financial crisis). Further the effect of the Tobin’s q (anticipated growth opportunities) on the relation between debt and future firm growth will be studied. To avoid problems with regulations issues within industries, the computation of data will be limited to all “manufacturing companies” (sic codes between 2000 and 3999) in the US for the period 2004-2014. Due to limited availability of data there is a minimum of $1 billion of
2 net sales for companies to be included in the regression. To investigate whether leverage is significant there will be controlled for variables that affect the growth measures. The setting described in Lang et al. (1996) and Hurme (2010) will be closely followed. However, some changes will be made. The main component that differs is that, that in both studies the effect of R&D expenditures on firm growth are not taken in account. The contribution of this factor to their model could be of important relevance to study the relationship between leverage and future firm growth. Firms can improve their competitive position by investing in R&D. Moreover, Bae and Kim (2003) find a positive relation between R&D expenditure and a firms’ value. This positive relation can be explained by the fact that new products and processes researched contribute to a firms’ tacit knowledge and hence it’s competitive advantage. In addition, Bracker and Ramaya’s work (2011) shows a positive relation between R&D investments and Tobin’s q. This relation becomes inverse when investing in R&D exceeds the optimal level. Other economic relevance of R&D investment is that, firms that will have a higher probability of survival when R&D expenditure increases (Hall 1987). Moreover, Doms et al. (1995) find that firms with high advanced manufacturing technologies have lower chances to default and have higher rates of growth. Hence, managers would be wise to adopt good levels of R&D innovation to maintain or increase the competitive position of their firm. The assumption of the study on hand is that; If managers see and take the opportunities of growth, they are thought to choose their optimal levels of capital structure so that they can achieve the foreseen growth. This suggests the negative relation between leverage and future firm growth found by Lang et al. (1996) and Hurme (2010) could have changed by adding R&D expenditure as a variable.
The results of this thesis mainly follow the findings previous literature (Lang et al. 1996 and Hurme 2010). Over the whole period 2004-2015 a negative relation is found between the debt ratio and future firm growth. When making a distinction between firms with low and high growth opportunities results show a negative relation between leverage and future firm growth for most of the growth measures of low q firms. Regarding high q firms mainly no relation is observed. The main assumption of this thesis about R&D expenditure isn’t observed and R&D expense seem to be a substitute for both capital and employment expenditure. However, an interesting fact was observed over the sub period 2010-2014 where although the coefficients were not significant, the sign of the relation between the debt ratio and the growth measures shifted to positive. This could implicate a positive relation between leverage and future firm growth for certain economic conditions. More results can be found in chapter 4.
As described, in Chapter 2 there will be a review of existing and previous literature on capital structure. Chapter 3 will describe data and methodology used. AS mentioned chapter 4 will be used to present and analyze the results. Chapter 5 will be used to conclude and to give suggestions for future research.
3
CHAPTER 2 Literature review
As mentioned in Chapter one the study on capital structure has been subject to a broad and ongoing discussion. In the past century there have been many theories about the way companies choose their optimal debt level, however there isn't a mutual consent about a universal theory. But every study that focuses on the capital structure starts at Modigliani and Miller's work (Modigliani and Miller 1958; Modigliani and Miller 1963). Their theorem has been the reference point for many of the later published theories. To give a clear view of capital structure this chapter will give a review of previous literature on determinants of capital structure. Hereafter, a link will be placed with the subject this thesis covers, namely the relationship between leverage and future firm growth. As next item the empirical papers of Lang et al. (1996) and Hurme (2010) on the relationship between leverage and future firm growth will be discussed.
2.1 Modigliani and Miller
In their early work from 1958 Modigliani and Miller found proof that in perfect capital markets the way of financing doesn't matter when determining a firm’s market value or the weighted average cost of capital (WACC). From their first proposition follows that the market value (V) of a firm is totally independent of its capital structure (whether it is financed with equity (E) or debt (D)). Furthermore, the same holds for the firm's WACC, which depends on weighted levels of equity and debt and which is the required return on capital. (𝑟𝑟𝐸𝐸= 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑒𝑒 𝑜𝑜𝑒𝑒 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑒𝑒𝑒𝑒, 𝑟𝑟𝐷𝐷= 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑟𝑟𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑒𝑒 𝑜𝑜𝑒𝑒 𝑟𝑟𝑟𝑟𝑑𝑑𝑒𝑒)
𝑉𝑉 = 𝐸𝐸 + 𝐷𝐷 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 =𝐸𝐸𝑉𝑉 ∗ 𝑟𝑟𝐸𝐸𝐸𝐸 +𝐷𝐷
𝑉𝑉 ∗ 𝑟𝑟𝐷𝐷
From proposition 1 their second proposition can be derived and is written as:
𝑟𝑟𝐸𝐸= 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 + (𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 − 𝑟𝑟𝐷𝐷) ∗𝐷𝐷
𝐸𝐸
The above formula shows that the expected rate on return on equity increases when the debt-to-equity increases. Following from the MM-theorem the required return equity is automatically adjusted to the point that is doesn't payoff to borrow cheap debt and spend it on expensive equity. There isn't a free lunch. As mentioned, the MM-theorem assumes perfect capital markets but to what extent is it that the market can really be seen as “perfect”? For example, Myers states in his work (2001) that some investors are willingly to pay more for certain combinations of securities. Although the MM-theorem is still seen as
4 mother of all capital structure studies it has been confronted with criticism because of the stringent
conditions under which their theory holds.
The main critics aren’t focusing on the theory itself, but more on the fact the theory doesn’t take into account the realistic (imperfect) market conditions where capital structure does matter. For that reason, later studies focused on determinants of the optimal amount of debt in a firm when the market
assumptions are loosened. For example, Kraus and Litzenberger (1973) first introduced the so called trade-off theory in which a firm chooses its optimal level of debt based on the trade off the benefits of a tax shield and the cost of a possible bankruptcy. Later Jensen and Meckling (1976) introduced agency costs as a determinant for the optimal trade-off between debt and equity. In 1984 a different point of view was presented, the Pecking order theory (Donaldson, 1961, Myers and Majluf, 1984, Myers 1984). In their study they looked at perfect capital markets with the presence of asymmetric information, which led to the pecking order of financing. Furthermore, there have been mentioned many more determinants in literature. Harris and Raviv (1991) gave a comprehensive overview of studies on determinants for the optimal level of debt. In the next subsections I will give a brief description of the most important determinants and theories.
2.2 Trade-off theory and agency costs
As described above the Trade-off theory, first mentioned by Kraus and Litzenberger (1973), only took in account debt benefits and the possible costs of financial distress. In their model a company trades off the amount the marginal value added of additional debt to the present value of possible cost of financial distress (Myers 2001). Financial distress costs follow from a possible bankruptcy and include not only the direct financial distress costs like legal fees for attorneys but also indirect costs that occur during financial distress like for example losing customers, favorable credit terms or employees. In later work Jensen and Meckling (1986) argue that agency costs of debt, which arise due conflicts of interests, are need also be included to realize the optimal capital structure within a firm. According to Jensen and Meckling (1986) there are different types of conflicts of interest. First of all, there are conflicts between equity holders and managers. Second there are conflicts between equity holders and debtholders.
These conflicts between managers and equity holders arise when managers do not hold 100% of the firm. As a consequence, managers possibly do not capture the full 100% of the profits the firm and in this case a manager may possibly not choose for the maximization of shareholder value but rather choose a managing form where he/she maximizes his/her personal benefits (for example: expensive business trips or luxury transportation) where he/she does enjoy the full 100%. In this situation the possibility of taking out debt comes in handy. Jensen and Meckling (1986) suggest that by using debt instead of equity the fraction of ownership stays high and managers have less incentive to engage in non-favorable
5 shareholders’ activities. Further higher interest payments will reduce the availability of “free cash” to managers and according to Stulz (1990) and Jensen (1986) cash rich firms without good growth possibilities should have high debt levels to prevent overinvestment.
Conflicts between equity holders and debtholders originate as the result of the mismatch of risk and reward of equity holders. According to Jensen and Meckling (1976) equity holders may invest suboptimal. More specifically, in a situation that an investment results in a loss the costs are transferred to the debtholders. Hence, as a result equity holder might engage in value-decreasing investments. This situation where equity holders invest in riskier assets than debtholders is called the assets substitution problem, which is an agency cost of debt. However, if debtholders anticipate equity holders’ future behavior and pay less for the issued debt, than the cost of the assets substitution effect is incurred by equity holders.
2.3 Pecking order theory and asymmetric information
Another important determinant for explaining the optimal capital structure is private information. Private information occurs when insiders and managers of a firm have more knowledge about firm characteristics than investors. This problem of asymmetric information is according to Myers (1984) and Myers and Majluf (1984) the main determinant of the organization of capital. According to Myers and Majluf asymmetric information results in a hierarchy of using capital for new investments, called the “Pecking order theory of capital structure”. According to the pecking order the financing of new investments is done first with internal funds, e.g. profits, next with “safe” debt and ultimately as a last resort will be chosen for an issuance of equity. This follows from the fact that managers ought to be acting in the interest of existing shareholders and possible investors will value a company that issues equity as overvalued. Because of this equity underpricing by possible investors the net present value (NPV) of a new investment could be canceled out and result in a loss for existing shareholders, which result in underinvestment. As a consequence of this, managers tend to use first funds that are no or less subject to undervaluation, namely internal funds and less risky debt (Donaldson 1961, Myers 1984, Myers and Majluf 1984). Hovakimian et al. (2001) point out the major implication of the pecking order theory, namely that firms with a higher profitability finance themselves with retained earnings and decrease their leverage ratio. The opposite counts for firms with low profitability. This is also of important relevance for the connection between leverage and future firm growth discussed in the next section.
2.4 Adjustment cost theories and other determinants of the leverage ratio
Besides the traditional theories about capital structure like the trade-off theorem and the pecking order theorem there are many other factors that determine the optimal capital structure. One important factor which determines the target debt level in a firm are the adjustment costs of moving toward that level.
6 Flannery and Ragan (2006) state regarding the trade-off theory that when adjustment costs are absent firms should always adjust to their target debt level. However, if adjustment costs are high, firms will be less likely to do so. Moreover, Leary and Roberts (2005) make assumptions that firms stay away from rebalancing to their target debt level until the adjustments cost are offset by the gains. This are adjustment cost theories and imply that firms rebalance to their optimal debt level rather partial than continuously. The speed at which adjustments are made isn’t a forgone conclusion either. For example, Flannery and Ragan (2006) find that the speed of adjustment (SOA), the rate at which firms reduce the gap with their target debt level, for a typical firm is 34.4% a year where Huang and Ritter (2009) find that firms only rebalance with 23.2% a year. These SOA are based on mean value of firms but firms’ specific SOA are depending on firms’ characteristics. Drobetz and Wanzenried (2006) for example find that firms will adjust quicker if they are further away from their target debt level. Mainly holds, the higher the adjustment costs the slower the speed of adjustment is.
To continue, country determinants for example could also be a determinant for choosing the optimal capital structure (Rajan and Zingales 1995). Further, many of determinants in practice are unknown while theories make assumptions about the debt target ratio. For example, according to Jensen and Meckling (1976) the agency problems between managers and shareholders could be decreased by the usage of debt. However, Brounen et al. (2006) find in their study that the debt decision isn’t based on the disciplining function of debt. Further, regarding the trade-off theory Brounen et al. (2006) find that tax-benefits of interest payments are the fourth most important factor for firms when targeting their optimal debt level. Next, according to Stulz (1990) firms that have a higher takeover threat are more likely to have higher levels of debt. A manager, which is facing a takeover threat, might be levering up a firm just to fence off such a takeover to maintain his personal benefits from being in control (Harris and Raviv 1988). Reputation is such a personal benefit for managers and can also influence the capital structure decision of a firm. To maintain their reputation managers could have an incentive to engage in relative safe investments (Harris and Raviv 1991). Because according to Hirschleifer and Thakor (1989) in a situation when a manager has two possible investment opportunities, the decision between those investment is based on the success rate rather than the expected return rate shareholders prefer. Hence, the manager will choose the investment A with the higher probability of success even if the expected return of Investment B (with the lower probability of success) is higher. Another related behavioral determinant that influences the managerial decision of capital structure is overconfidence. In his work Hackbarth (2007) shows that overconfident managers tend to choose higher level of debt then rational managers. Finally, to show the broadness of determinants in the capital structure choice, here is a summation of relation found by Murray and Vidhan (2009). They find that leverage tends to be higher when firms are larger, when the inflation is higher, when firms are less profitable and when firms have more tangible assets. As can be read in the previous chapter there are many theories on the decision of capital structure.
7 For the purpose of this study it isn’t necessary to elaborate on every possible determinant of leverage. For now, I will continue to focus on aspects that relate the subject (leverage) to the study on hand.
2.5 The capital structure decision’s influence on firm performance
Following the literature there are many factors and determinants that influence the choice of capital structure. Although it is very interesting to study for which reasons firms choose their aggregate levels of debt and equity. The real relevance of leverage doesn’t lie in the amount chosen but more in the fact, which economic consequences come with those aggregate chosen amounts of capital and debt. One of these consequences was found by Myers (1977). In his study he found that extreme high levels of debt lead to underinvestment. As a result, it is more likely that firms with high debt levels will pass up value creating opportunities. Following from this, firms with anticipated growth opportunities, in finance referred to as a Tobin’s q, should engage in lower levels of debt.
In literature a lot of models predict the relation between leverage and firm growth is to be negative (Jensen and Meckling 1986, Stulz 1990). This can be explained when firms have a “debt overhang”. In this situation managers find it difficult to raise new funds for new investment opportunities and they could possibly pass on a positive Net Present Value (NPV) project (Myers 1978). In addition, firms that have higher interest payments consequently have lower cash flow and thus fewer cash available to invest. Aivazian et al. (2005), Firth et al. (2008) and Dang (2011) find support for this negative relation. For example, Aivazian et al. (2005) find that there is a negative relation between debt and investment for Canadian firms in the period 1982-1999. Just like McConnell and Servaes (1995) they also find that there is a difference in this relation for “high growth firms” (measured in Tobin’s q) and “low growth firms”. Regarding high q firms a negative relation exist between firm value and leverage, on the other hand for low q firms this relationship is positive. These findings suggest that a firms’ anticipated growth opportunities determine their debt level. This can be explained by the fact that for firms with high growth opportunities the negative effect of debt, e.g. a lower cash flow available due to high interest payments dominate and for firms with low grow opportunities the positive effects, e.g. the disciplining effect of debt on CEOs, dominate.
Lang et al. (1996) studied the relationship between leverage and future firm growth between 1970-1989. In their study, which has almost the same methodology as the study on hand, they measure future firm growth by investment, which is defined as capital expenditures minus depreciation in year t+1 divided by fixed assets, employment growth and capital expenditure growth. In their study they find a negative relation between the book leverage and future firm value. However, the negative relation found was only significant for firms with low anticipated growth opportunities. Results of the relations found in their study are presented in Table 2.1 on the next page.
8 Table 2.1 Lang et al. (1996) study on relation between leverage and future firm growth
Data 1970-1989 (Net
investment-depreciation)/ Fixed assets 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Overall
Book leverage Negative Negative Negative Negative Negative
Industry-adjusted
leverage Negative Negative Negative Negative Negative
Low q firms
Book leverage Negative Negative Negative Negative Negative
Industry-adjusted
leverage Negative Negative Negative Negative Negative
High q firms
Book leverage No No No Negative No
Industry-adjusted
leverage No No No Weak negative No
Summary of the results by Lang et al. (1996). The sample period covers 1970-1989 and data is on US manufacturing companies with sales larger than $1 billion dollars. Relationships are Positive/Negative when significant at a 5% confidence level and Weak related when significant at a 10% level. If no significant relation exists in the regression, No is stated. “Low q” (“high q”) firms are firms with q<1 (q>1) or q < industry median (q > industry median) of the whole time period 1970-1989.
In the table 2.1 and 2.2 above and below book leverage is defined as the book value of short-term and long-term debt divided by the book value of total assets. The distinction between low q and high q is made based on a firms’ anticipated growth opportunities. Anticipated growth opportunities (measured in Tobin’s q), defined as the market value of a firm divided by its replacement cost of total assets, are high (low) when the calculated ratio is above 1 (below 1) or when the calculated q value is higher (lower) than the calculated industry median q value. A high (low) q implies overvaluation (undervaluation) of a firm. The industry-adjustment is done to control for industry effects. Industries are based on SIC codes, with a minimum of four companies in an industry. The calculation of the industry-adjusted variables is done by subtracting the median industry value from each firm specific value. In their study they mainly find a negative relation between book leverage and the growth measures. The results in the table for example show that in a low q firm there is a negative relation between book leverage and 1-year employment growth. Hence, if the book leverage goes up, ceteris paribus, the 1-year employment growth goes down. In 2010, Hurme restudied the relationship of leverage and future firm growth with more recent data (1990-2008), shown in table 2.2. She used the same settings as Lang et al. (1996) and for the new time period she studied the same strong negative relation was found between book leverage and future firm growth. However slight differences are seen when a distinction is made between low q and high q firms. When looking at the whole time-periods of both studies, we find more or less the same results.
9 Table 2.2 Hurme (2010) study on relation between leverage and future firm growth
Data 1990-2008 (Net
investment-depreciation)/Fixe d assets 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Overall
Book leverage Negative Negative Negative Negative Negative
Industry-adjusted
leverage Negative Negative No Negative Negative
Low q firms
Book leverage No Negative Negative Negative Negative
Industry-adjusted
leverage Weak negative Negative Negative Negative No
High q firms
Book leverage Negative No No Negative Negative
Industry-adjusted
leverage Negative No No Weak Negative Negative
Summary of the results by Hurme (2010). The sample period covers 1990-2008 and data is on US manufacturing companies with sales larger than $1 billion dollars. Relationships are Positive/Negative when significant at a 5% confidence level and Weak related when significant at a 10% level. If no significant relation exists in the regression, No is stated. “Low q” (“high q”) firms are firms with q<1 (q>1) or q < industry median (q > industry median) of the whole time period 1990-2008.
The studies of Lang et al. (1996) and Hurme (2010) have shown a negative relation between leverage and future firm growth. Although from previous literature (Lang et al. 1996, Hurme 2010) follows that leverage is negatively correlated with future firm growth, it is interesting to see if their findings still hold when controlling for additional factors that could influence future firm growth. As described above Bradley (1984) et al. found a negative relation between R&D expenditure and leverage. Further, Hall (1987) and Greenhalgh et al. (2001) found a positive relation between R&D investments and 1-year employment growth. The adding of R&D expense to respectively Lang et al. (1996) and Hurme (2010) estimated equations could have major implications for their findings (negative relation between leverage and future firm growth). Further, I will just as Hurme (2010) look if the relations differ when looking at sub periods. In the next chapter I will elaborate on the data and methodology.
10
CHAPTER 3 Data and methodology
3.1 Data
The data of this sample is based on “large” manufacturing firms in the United States in the years 2004-2014. To qualify for the selection, firms need to have a minimum of 1 billion dollar of sales in each base year. This thesis follows previous literature, Lang et al. (1996) and Hurme (2010). In their work Lang et al. (1996) have given three reasons for choosing large firms. At first, due the fact that data needed for our regression is more accessible for large firms then for small firms the possible selection bias is lower. Next, because large firms generally are more settled compared to small firms, the possible relation between leverage and future firm growth is thought to be smaller for large firms. Hence, when finding this relation in our “large” firm sample it is more convincing to say such a relation between leverage and future firm growth exists. Another important factor is that to compare the study on hand with the named previous literature (Lang et al. 1996 and Hurme 2010) it is better to use the same data settings.
As said the sample consists of large manufacturing firms, SIC codes 2000-3999, this is because to steer clear of regulations issues. Further for firms to be included in the sample of firms need to have information in the base year, from which the growth is calculated, on both depended variables and explanatory variables. The data needed on depended variables are: capital expenditures and employees. The data needed to calculate the explanatory variables are: assets, debt, market value, cash flow, sales, capital expenditures and R&D expense. Final data exist of 528 firms and 5369 firm years. All the data is found on COMPUSTAT. Table 3.1 gives an overview of the final sample set. One outlier due to wrong data (1 and 3-year capital expenditure growth of respectively 9200% and 19000%) was removed from the sample. Other outliers, although large, are based on correct information and have been treated as normal values and are included in the sample.2 In table 3.2 the correlation matrix is presented.
Table 3.1 General features of the data set
Mean 25th percentile 50th percentile 75th percentile Standard deviation Number of firm-years Min Max 1-year employment growth .0329 -.0335 .0138 .0671 .2008 4580 -1 3.8512 3-year employment growth .0956 -.0768 .0476 .1884 .3787 3359 -1 6.0404 1-year capital expenditure growth .1106 -.1290 .0563 .2676 0.4295 4634 -1 7.1118 3-year capital expenditure growth .3200 -.1753 .1545 .5646 .9010 3409 -1 15.8022 Debt ratio .2297 .1274 .2101 .3065 .1653 5368 0 2.1263 Cash flow(0)/TA(-1) .0898 .0276 .0655 .1190 .1436 5368 -.7829 3.0724 Capital expenditures/FA(-1 .1033 .0588 .0831 .1219 .0830 5368 0 1.8956 Tobin’s q 1.466 .8403 1.2123 1.7972 .9907 5368 .0006 11.5787 Sales growth .0738 .0162 .0618 .1419 .1996 5368 -.8073 5.1786 R&D expense/TA-1 .0416 .0094 .0233 .0535 .0531 5368 0 .8218
Features of the data set. Data is collected at COMPUSTAT. Sample is collected over the period 2004-2014. Debt ratio is defined as book debt divided by book total assets. Cash flow and R&D expense are divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
2
11 Table 3.2 Correlations between variables
1-Year employment growth 3-Year employment growth 1-Year capital expenditures growth 3-Year capital expenditures growth Debt ratio 1-Year employment growth 1.00 3-Year employment growth 0.6561 1.00 1-Year capital expenditures growth 0.2554 0.1752 1.00 3-Year capital expenditures growth 0.2194 0.3601 0.4632 1.00 Debt ratio -0.0833 -0.0899 -0.0516 -0.0185 1.00 Cash flow(0)/TA(-1) 0.0956 0.1165 0.0791 0.0401 -0.0020 Capital expenditures/FA(-1) 0.1458 0.2347 -0.0983 -0.0891 -0.1556 Tobin’s q 0.1629 0.2077 0.1292 0.0804 -0.1274 Sales growth 0.1107 0.1132 0.1455 0.0465 -0.0699 R&D expense/TA(-1) 0.0520 0.0820 -0.0078 -0.0194 -0.1886
Cash flow Capital
expenditures/FA(-1)
Tobin’s q Sales growth R&D
expense/TA-1 Cash flow/TA(-1) 1.00 Capital expenditures/FA(-1) 0.3237 1.00 Tobins q 0.1478 0.2587 1.00 Sales growth 0.2940 0.3212 0.1836 1.00 R&D expense/TA-1 0.2877 0.2367 0.3474 0.1191 1.00
Features of the data set. Data is collected at COMPUSTAT. Sample is collected over the period 2004-2014. Debt ratio is defined as book debt divided by book total assets. Cash flow and R&D expense are divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
3.2 Methodology
This section of the chapter discusses the methodology used. First, dependent variables are explained in chapter 3.2.1. Next, in chapter 3.2.2 the explanatory variables used, are discussed. Finally, the regression model and hypothesis are shown in chapter 3.2.3. All data has been found at COMPUSTAT and has been processed by Stata/SE 13.1 to give the regression results. To calculate the dependent variables and to perform a regression the setting of Lang et al. (1996) mainly has been followed, but in the study on hand there has also been controlled for R&D expense.
12
3.2.1 Dependent variables
To measure the discussed relation between leverage and future firm growth I have chosen for similar measures as in Lang et al. (1996), namely employment growth and capital expenditure growth. These growth rates are on a 1- and 3-year basis. These growth measures are shown below.
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑜𝑜𝑒𝑒𝐸𝐸𝑟𝑟𝑒𝑒𝑒𝑒 𝑔𝑔𝑟𝑟𝑜𝑜𝑔𝑔𝑒𝑒ℎ (1 𝑜𝑜𝑟𝑟 3 𝑒𝑒𝑟𝑟𝑦𝑦𝑟𝑟) =𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑜𝑜𝑒𝑒𝐸𝐸𝑟𝑟𝑒𝑒𝑒𝑒 (𝑒𝑒𝑟𝑟𝑦𝑦𝑟𝑟 1 𝑜𝑜𝑟𝑟 3) 𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝑜𝑜𝑒𝑒𝐸𝐸𝑟𝑟𝑒𝑒𝑒𝑒 (𝑒𝑒𝑟𝑟𝑦𝑦𝑟𝑟 0) − 1
𝑊𝑊𝑦𝑦𝐸𝐸𝑟𝑟𝑒𝑒𝑦𝑦𝐸𝐸 𝑟𝑟𝑒𝑒𝐸𝐸𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑟𝑟𝑒𝑒 𝑔𝑔𝑟𝑟𝑜𝑜𝑔𝑔𝑒𝑒ℎ (1 𝑜𝑜𝑟𝑟 3 𝑒𝑒𝑟𝑟𝑦𝑦𝑟𝑟) =𝑊𝑊𝑦𝑦𝐸𝐸𝑟𝑟𝑒𝑒𝑦𝑦𝐸𝐸 𝑟𝑟𝑒𝑒𝐸𝐸𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑟𝑟𝑒𝑒 (𝑒𝑒𝑟𝑟𝑦𝑦𝑟𝑟 1 𝑜𝑜𝑟𝑟 3 )𝑊𝑊𝑦𝑦𝐸𝐸𝑟𝑟𝑒𝑒𝑦𝑦𝐸𝐸 𝑟𝑟𝑒𝑒𝐸𝐸𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑒𝑒𝑟𝑟𝑟𝑟𝑟𝑟𝑒𝑒 (𝑒𝑒𝑟𝑟𝑦𝑦𝑟𝑟 0) − 1
In the following text of the study on hand the emp1, emp3, capx1 and capx3 will be substitutes for respectively 1-year employment growth, 3-year employment growth, 1-year capital expenditure growth and 3-year capital expenditure growth.
3.2.2 Independent variables
First, the main explanatory variable is the debt ratio, which is the ratio of the book value of debt to the book value of total assets. The decision to choose the book value of debt rather than the market value of debt is fueled by the fact that market leverage valuation would give too much importance to equity value fluctuations (Lang et al. 1996).
𝐷𝐷𝑟𝑟𝑑𝑑𝑒𝑒 𝑟𝑟𝑦𝑦𝑒𝑒𝑟𝑟𝑜𝑜 = 𝐵𝐵𝑜𝑜𝑜𝑜𝐵𝐵 𝑣𝑣𝑦𝑦𝐸𝐸𝑟𝑟𝑟𝑟 𝑜𝑜𝑜𝑜 𝑒𝑒𝑜𝑜𝑒𝑒𝑦𝑦𝐸𝐸 𝑦𝑦𝑒𝑒𝑒𝑒𝑟𝑟𝑒𝑒𝑒𝑒𝐵𝐵𝑜𝑜𝑜𝑜𝐵𝐵 𝑣𝑣𝑦𝑦𝐸𝐸𝑟𝑟𝑟𝑟 𝑜𝑜𝑜𝑜 𝑒𝑒𝑜𝑜𝑒𝑒𝑦𝑦𝐸𝐸 𝑟𝑟𝑟𝑟𝑑𝑑𝑒𝑒
Next, there is controlled for variables that also could influence the growth measures. These variables are those suggested by Lang et al. (1996), namely Tobin’s q, cash flow divided by total assets, capital expenditures divided by fixed assets and sales growth from year -1 to year 0 and an added variable, R&D expense divided by total assets.
Tobin’s q is the market value of a firm divided by its replacement cost. It is important to control for this variable because a higher “q” implies more valuable growth opportunities for a firm. The common way to calculate the Tobin’s q is by the Lindenberg and Ross algorithm (Lindenberg and Ross, 1981). However, due to the comprehensiveness of these calculations a simpler, less data intensive calculation, the approximation given by Chung and Pruitt (1994), is followed.
𝑊𝑊𝐸𝐸𝐸𝐸𝑟𝑟𝑜𝑜𝑒𝑒𝑟𝑟𝐸𝐸𝑦𝑦𝑒𝑒𝑟𝑟 𝑇𝑇𝑜𝑜𝑑𝑑𝑟𝑟𝑒𝑒′𝑒𝑒 𝑟𝑟 =𝑀𝑀𝑦𝑦𝑟𝑟𝐵𝐵𝑟𝑟𝑒𝑒 𝑣𝑣𝑦𝑦𝐸𝐸𝑟𝑟𝑟𝑟 𝑒𝑒𝑒𝑒𝑜𝑜𝑠𝑠𝐵𝐵 + 𝑂𝑂𝑟𝑟𝑒𝑒𝑒𝑒𝑒𝑒𝑦𝑦𝑒𝑒𝑟𝑟𝑟𝑟𝑒𝑒𝑔𝑔 𝐸𝐸𝑟𝑟𝑟𝑟𝑜𝑜𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑒𝑒𝑒𝑒𝑜𝑜𝑠𝑠𝐵𝐵 + 𝐷𝐷𝑟𝑟𝑑𝑑𝑒𝑒
𝑇𝑇𝑜𝑜𝑒𝑒𝑦𝑦𝐸𝐸 𝑦𝑦𝑒𝑒𝑒𝑒𝑟𝑟𝑒𝑒𝑒𝑒
≈
𝑊𝑊𝐸𝐸𝐸𝐸𝑟𝑟𝑜𝑜𝑒𝑒𝑟𝑟𝐸𝐸𝑦𝑦𝑒𝑒𝑟𝑟 𝑇𝑇𝑜𝑜𝑑𝑑𝑟𝑟𝑒𝑒′𝑒𝑒 𝑟𝑟 = 𝑀𝑀𝑦𝑦𝑟𝑟𝐵𝐵𝑟𝑟𝑒𝑒 𝑣𝑣𝑦𝑦𝐸𝐸𝑟𝑟𝑟𝑟 𝑜𝑜𝑟𝑟𝑟𝑟𝐸𝐸
13
3.2.3 Regression and hypotheses.
The variables described above will be used in an OLS regression with STATA 13.1. To correct for heteroscedasticity a robust regression will be done. First this study will perform the same regression as done in the studies of Lang et al. (1996) and Hurme (2010):
� 𝐸𝐸𝐸𝐸𝐸𝐸1, 𝐸𝐸𝐸𝐸𝐸𝐸3, 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒1, 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒3� = 𝛽𝛽0 + 𝛽𝛽1𝐷𝐷𝑟𝑟(0) + 𝛽𝛽2𝑇𝑇𝑊𝑊(−1) + 𝛽𝛽3𝑊𝑊𝑜𝑜(0) 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒(0)𝐹𝐹𝑊𝑊(−1) + 𝛽𝛽4 𝑇𝑇𝑟𝑟(0) + 𝛽𝛽5𝑆𝑆𝑦𝑦(−1)𝑆𝑆𝑦𝑦(0)
where Dr stands for Debt ratio, Cf for Cash flow, Capx for Capital expenditures, Tq for Tobin’s q, Sa for sales, TA for total assets and FA for fixed assets. Numbers in parenthesis relate to the year of the data.
After the first regression a distinction will be made between high q and low q firms.
�𝐸𝐸𝐸𝐸𝐸𝐸1, 𝐸𝐸𝐸𝐸𝐸𝐸3,𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒1, 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒3� = 𝛽𝛽0 + 𝛽𝛽1𝐷𝐷𝑟𝑟𝐷𝐷𝑟𝑟 + 𝛽𝛽2𝐷𝐷𝑟𝑟𝐷𝐷𝑟𝑟 + 𝛽𝛽3𝑇𝑇𝑊𝑊(−1) + 𝛽𝛽4𝑊𝑊𝑜𝑜 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒(0)𝐹𝐹𝑊𝑊(−1) + 𝛽𝛽5 𝑇𝑇𝑇𝑇 + 𝛽𝛽6𝑆𝑆𝑦𝑦
where DrHq stands for Debt ratio if firm has q>1, DrLq for Debt ratio if firm has q<1, Cf for Cash flow, Capx for Capital expenditures, Tq for Tobin’s q, Sa for sales, TA for total assets and FA for fixed assets. Numbers in parenthesis relate to the year of the data.
Next, the control variable R&D expense will be added to research whether a difference can be found between previous studies and the study on hand.
�𝐸𝐸𝐸𝐸𝐸𝐸1, 𝐸𝐸𝐸𝐸𝐸𝐸3,𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒1, 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒3� = 𝛽𝛽0 + 𝛽𝛽1𝐷𝐷𝑟𝑟𝐷𝐷𝑇𝑇 + 𝛽𝛽2𝐷𝐷𝑟𝑟𝐷𝐷𝑇𝑇 + 𝛽𝛽3𝑇𝑇𝑊𝑊(−1) + 𝛽𝛽4𝑊𝑊𝑜𝑜 𝑊𝑊𝑦𝑦𝐸𝐸𝑒𝑒(0)𝐹𝐹𝑊𝑊(−1) + 𝛽𝛽5 𝑇𝑇𝑇𝑇 + 𝛽𝛽6𝑆𝑆𝑦𝑦(−1) + 𝛽𝛽7𝑆𝑆𝑦𝑦(0) 𝑇𝑇𝑊𝑊(−1)𝑅𝑅&𝐷𝐷
where DrHq stands for Debt ratio if firm has q>1, DrLq for Debt ratio if firm has q<1, Cf for Cash flow, Capx for Capital expenditures, Tq for Tobin’s q, Sa for sales, TA for total assets and FA for fixed assets, R&D for R&D expense. Numbers in parenthesis relate to the year of the data.
Finally, these regressions will also be done for the different time periods: 2004-2006, 2007-2009 and 2010-2014. These above mentioned three regressions are going to be tested against a null hypothesis to see if a relation exist between leverage and future firm growth. These are the main hypothesis:
For regression 1 For regressions 2,3,4,5 and 6
𝐷𝐷0: 𝛽𝛽1(𝐷𝐷𝑟𝑟) = 0 𝐷𝐷0: 𝛽𝛽1(𝐷𝐷𝑟𝑟𝐷𝐷𝑟𝑟) = 0 𝑦𝑦𝑒𝑒𝑟𝑟/𝑜𝑜𝑟𝑟 𝛽𝛽2(𝐷𝐷𝑟𝑟𝐷𝐷𝑟𝑟) = 0
𝐷𝐷1: 𝛽𝛽1(𝐷𝐷𝑟𝑟) ≠ 0 𝐷𝐷1: 𝛽𝛽1(𝐷𝐷𝑟𝑟𝐷𝐷𝑟𝑟) ≠ 0 𝑦𝑦𝑒𝑒𝑟𝑟/𝑜𝑜𝑟𝑟 𝛽𝛽2(𝐷𝐷𝑟𝑟𝐷𝐷𝑟𝑟) ≠ 0
The null hypothesis regarding regression states that there is no significant relation between the debt ratio and the growth measures separately. The null hypothesis is rejected if a relation is found. For regressions 2,3,4,5 and 6 the debt ratio is also tested but now there is difference between low and high q firms. The null hypothesis is rejected if a relation is found between low q firms (high q firms) and the growth measures.
14
CHAPTER 4 Results
This chapter presents the results of the regression analysis. The empirical results found are compared to the finding in similar previous literature (Lang et al. 1996 and Hurme 2010). The last decade has economically been turbulent. After the dot com bubble in 2000, the next crisis, the global financial crisis of 2007, introduced itself. Since its introduction interest rates have been under pressure, for example the prime lending rate, which banks charge to their most creditworthy clients, declined from 8.1% in 2007 to 3.3% in 2009. Hence, in this light it is interesting whether old assumptions about the relation between leverage and future firm growth still hold.
Results of the regression analysis in this chapter are divided in two sections. In chapter 4.1 the results of the period 2004-2014 as a whole is presented. After this in chapter 4.2 the period 2004-2014 will be split up in the following three sub periods: 1. 2004-2006 2. 2007-2009 and 3. 2010-2014.
4.1 Whole period 2004-2014
Table 4.1 shows the results of the regression over the period 2004-2014. The coefficients in table 4.1 depict the change of the growth measures in the 2nd row caused by a change in the independent variables on the left side of the table. For example, if the Tobin’s q increases from 1 to 2, the employment growth in column 1 is expected to grow with 2.78%. The R-square value, which is roughly speaking the explanatory power of the model, is for each of the growth measures comparable to both the studies of Lang et al. (1996) and Hurme (2010). R-Square values are shown in table 4.2.
Further, to correct for heteroskedasticity robust standard errors have been used. Results from the regression over the whole period 2004-2014 show a negative relation between most of the growth measures and the debt ratio. Only for 3-year capital expenditures growth the relation is not statistical significant. Further capital expenditure is significant for all growth measures. The negative relation found, with both 1- and 3-year capital expenditures growth, is also found by Lang et al. (1996) and Hurme (2010) between respectively 1970-1989 and 1990-2008. The same holds for the anticipated growth opportunities, the Tobin’s q. The other control variables Sales growth and Cash flow show significant relevance for all measures except for 3-year employment growth. However, significance is wishful, it is also important to address the economic theory behind these coefficients. First, regarding the positive coefficient of cash flow divided by total assets, it is expected that cash flow has a positive effect because if firms have more internal cash available they are likely to invest these available funds into Net Present Value projects. The findings, a positive relation between cash flow and investment, of Fazzari et al. (1988) support this. Next, regarding capital expenditures divided by fixed assets, the amount of capital expenditures in the base year (year 0) determines the eventual growth of capital expenditures. For example, the percentage growth is much higher when capital expenditures grow from 200 in year 0 to 400 in year 3(100%) relatively to when it grows from 2000 in year 0 to 2200 in year 3 (10%). Hence, a negative relation between capital expenditures and the 1- and 3-year capital expenditures growth is expected. Further, firms with a high Tobin’s q are more likely to invest. Blundell et al. (1992) also find that the level of investment depends on the anticipated growth level. To elaborate, in case of overvaluation (q>1), where the market captures some unrecorded assets of the firm, firms can issue new shares and investment this in new projects. It follows that this becomes more interesting when the Tobin’s q is higher. Finally, there is controlled for the multiplier effect of the sales growth between year -1 and year 0. When a firm sees its sales grow, it will need new fixed assets and employees to cover that demand. It’s expected that for all growth measures the relation coefficient is positive, hence it isn’t surprising that table 4.1 shows an overall positive relation between sales growth and all growth measures. Now, I will focus on the relationship studied (the relation.
15 between the debt ratio and the growth measures). As can be read in chapter 2.5, Myers (1977) for example found that firms with extreme high leverage will underinvest. Moreover, if firms have high debt their “free cash” is reduced by high interest payments.
Table 4.1 OLS regression over 2004-2014
1 2 3 4 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Intercept -.0830 (0.00) *** -.1061 (0.016) ** -.2438 (0.00) *** .0582 (0.0572) Debt ratio -.0507 (0.013) ** -.1035 (0.008) *** -.1445 (0.001) *** -.1442 (0.105) Cash flow/TA(-1) .0493 (0.019) ** .0899 (0.102) .2253 (0.00) *** .4140 (0.003) *** Capital expenditures/FA(-1) .1790 (0.00) *** .8030 (0.00) *** -1.0310 (0.00) *** -1.8750 (0.00) *** Tobin’s q .0287 (0.00) *** .0581 (0.00) *** .0672 (0.00) *** .0978 (0.00) *** Sales growth .05816 (0.00) *** .0461 (0.256) .3472 (0.00) *** .2824 (0.003) *** R-square 0.0453 0.0787 0.0736 0.0304 Adj. R-square 0.0443 0.0774 0.0726 0.0290 N 4580 3359 4634 3409
OLS regression over the period 2004-2014 on US Large Manufacturing firms with a minimum of 1 billion dollars in sales for each base year. Regression results are corrected for heteroscedasticity. Significance levels are shown for a 10% level (*), 5% level (**) and a 1% level (***) and p-values are in parenthesis. Employment (Capital expenditures) growth is defined as the percentage change between year 1 (year 3) and the base year (year 0). Debt ratio is defined as book debt divided by book total assets. Cash flow is divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
Table 4.2 R-Square and Adj. R-square of previous studies (Lang et al. 1996 and Hurme 2010)
Lang et al. (1996) 1-year Employment
growth 3-year Employment growth 1-year Capital expenditure growth 3-year Employment growth R-Square 0.060 0.094 0.087 0.120 Adj. R-Square 0.05931 0.0932 0.0863 0.1192
Hurme (2010) 1-year Employment
growth 3-year Employment growth 1-year Capital expenditure growth 3-year Employment growth R-Square 0,054 0,0628 0,0894 0,085 Adj. R-Square 0.053 0.0616 0.0885 0.0838
R-square and adj. R-Square of previous studies (Lang et al.1996 and Hurme 2010) with the same settings which are depicted in table 2.1 and 2.2.
16 Hence, these firms have less cash available for investment. Regarding the study on hand it is expected that negative relations are found between the debt ratio and the growth measures. The economic implications of the results found are that, keeping everything else constant, firms can steer their annual employment growth by 0.0507%, their 3-year employment growth by 0.1035% their annual capital expenditures growth by 0.1445% and their 3-year employment growth 0.1442% for each percentage less in debt ratio. This implicates that firms can increase their future firm growth by setting a lower debt ratio. For example, by decreasing their debt ratio by 10%, firms are, ceteris paribus, expected to have 1.442% point more capital expenditures growth. However, such reductions in debt aren’t observed, as can be seen in table 4.3. On the contrary, both average debt ratio and total debt have risen when comparing the sub periods 2004-2006 and 2010-2014.
Table 4.3 Firms’ average debt and average debt ratios over 2004-2007 and 2010-2014
Firms’ average debt ratio Firms’ average total debt
2004-2006 .2066 3644.406
2010-2014 .2440 5263.676
Firms’ average total debt is in millions.
Following the results (read: the significance of the Tobin’s q on all growth measures), the next step is to examine the differences between the relation of leverage and future firm growth at firms with a “high” Tobin’s q (q>1) and a “low” Tobin’s q (q<1). Table 4.4, where the debt ratio has been split up for high q and low q firms, shows mainly the same empirical result for the control variables and still support the economic theory. However, when looking at the relation between the debt ratio of both low q and high q firms and the growth measures clear differences are seen. Regarding high q firms there is now no significant relation to be found between the debt ratio of high q firms and all the growth measures. For low q firms the coefficients of the debt ratio found in table 4.4, roughly speaking don’t differ from those found in table 4.1. The results of this study are comparable to those of Lang et al. (1996); Lang et al. (1996) found a negative relation with all growth measures for low q firms and there was no significant relation for high q firms, besides the 1-year capital expenditure growth. Hurme’s results (2010) however indicated a less convincing distinction between low q and high q firms. As in Lang et al. (1996) their study, apart from the 3-year capital expenditure growth, all growth measures show a significant relation with the debt ratio for firms with a low Tobin’s q. For high q firms only a significant negative relation can be found with 1-year capital expenditure growth at a 5% confidence level. These results are in line with the findings of Servaes and McConnell (1995) in which the relation between leverage and future firm growth depends on a firms’ anticipated growth opportunities.
17 Table 4.4 Regression with anticipated growth opportunities over 2004-2014
1 2 3 4 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Intercept -.0718 (0.00) *** -.0831 (0.061) * -.2284 (0.00) *** .0791 (0.451) Debt Ratio if q>1 -.0260 (0.255) -.0581 (0.165) -.1102 (0.019) ** -.1002 (0.296) Debt Ratio if q<1 -.1238 (0.00) *** -.2223 (0.00) *** -.2462 (0.00) *** -.2585 (0.04) ** Cash flow/TA(-1) .0479 (0.023) ** .0909 (0.099) ** .2239 (0.00) *** .4154 (0.003) *** Capital expenditures/FA(-1) .1878 (0.00) *** .8245 (0.00) *** -1.0191 (0.00) *** -1.8555 (0.00) *** Tobin’s q .0224 (0.00) *** .0476 (0.00) *** .0584 (0.00) *** .0877 (0.00) *** Sales growth .0581 (0.00) *** .0409 (0.314) .3474 (0.00) *** .2787 (0.004) *** R-Square 0.0486 0.0812 0.0750 0.0308 Adj. R-Square 0.0473 0.0796 0.0738 0.0291 N 4580 3359 4634 3409
OLS regression over the period 2004-2014 on US Large Manufacturing firms with a minimum of 1 billion dollars in sales for each base year. Regression results are corrected for heteroskedasticity. Significance levels are shown for a 10% level (*), 5% level (**) and a 1% level (***). Employment (Capital expenditures) growth is defined as the percentage change between year 1 (year 3)and the base year (year 0). Debt ratio if q>1 (q<1) is defined as book debt divided by book total assets when Tobin’s q >1 (q<1). Cash flow is divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
Where the above regressions are mainly used to compare results between previous literature and this study, the following regression is mainly used to compare results within the study. In the next regression R&D expense divided by total assets is included to control for its effect on the growth measures. For example, Hall (1987) and Greenhalgh et al. (2001) found a positive relation between R&D investments and 1-year employment growth. However, table 4.5 doesn’t show major differences regarding the relation between the debt ratio and the growth measures for both low and high q firms, compared to the results in table 4.4. However, in contrast with what I expected, R&D expense divided by total assets has a significant negative relation with all growth measures. This negative relation can possibly be explained by the fact that investing in R&D is a substitute for both investing in new employees and investing in new fixed assets. Another explanation is perhaps to be found in the fact that R&D investments are mostly for the long-term and these effects are not captured by the 1- and 3-year growth measures used in the study on hand. Further, although a significant relation with all growth measures exists, by adding R&D expense as a control variable, the Adjusted R-Square only slightly improved. Hence, by adding the R&D model no major improvement on the explaining power of the model is seen.
18 Table 4.5 Regression with anticipated growth opportunities and R&D over 2004-2014
1 2 3 4 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Intercept -.0674 (0.00) *** -.0753 (0.089) * -.2167 (0.00) *** .1021 (0.330) Debt Ratio if q>1 -.0336 (0.139) -.0718 (0.083) * -.1312 (0.006) *** -.1429 (0.148) Debt Ratio if q<1 -.1317 (0.00) *** -.2354 (0.00) *** -.2673 (0.00) *** -.2981 (0.019) ** Cash flow/TA(-1) .0605 (0.005) *** .1166 (0.032) ** .2573 (0.00) *** .4910 (0.00) *** Capital expenditures/FA(-1) .1941 (0.062) * .8392 (0.00) *** -1.0014 (0.00) *** -1.8093 (0.00) *** Tobin’s q .0250 (0.00) *** .0522 (0.00) *** .0656 (0.00) *** .1016 (0.00) *** Sales growth .0568 (0.00) *** .0378 (0.352) .3437 (0.00) *** .2692 (0.005) *** R&D expense/TA(-1) -.1642 (0.001) *** -.2796 (0.019) ** -.4355 (0.006) *** -0.8292 (0.019) ** R-Square 0.0501 0.0825 0.0774 .0328 Adj. R-Square 0.0486 0.0806 0.0760 .0308 N 4580 3359 4634 3409
OLS regression over the period 2004-2014 on US large manufacturing firms with a minimum of 1 billion dollars in sales for each base year. Regression results are corrected for heteroskedasticity. Significance levels are shown for a 10% level (*), 5% level (**) and a 1% level (***). Employment (Capital expenditures) growth is defined as the percentage change between year 1 (year 3)and the base year (year 0). Debt ratio if q>1 (q<1) is defined as book debt divided by book total assets when Tobin’s q >1 (q<1). Cash flow and R&D expense are divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0. R&D expense is divided by total assets in the previous year.
4.2 Sub periods
In order to see differences arisen due the global financial crisis the next section will discuss three sub period separately. First, the sub period 2004-2006, then the sub period 2007-2009 and finally 2010-2014. For each growth measure two regressions are done. One regression is done without (A) and the other with (B) R&D expense divided by total assets.
4.2.1 2004-2006
The results of the regression in the sub period 2004-2006 of this study don’t correspond with the results found by Hurme (2010), which she observed over the sub periods of 1991-1996 and 2002-2005. Where in her work all control variables show a significant relation with 1- and 3-year capital growth rates, the opposite is seen in this study. For the purpose of this study it isn’t interesting to further discuss this difference. The regression results can be found in table 4.6.
19 Table 4.6 Regression with anticipated growth opportunities and R&D expense over 2004-2006
N 1A 1B 2A 2B 3A 3B 4A 4B 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Intercept -.0996 (0.036) ** -.0940 (0.052) * -.1688 (0.141) -.1584 (0.174) .3072 (0.000) .2929 (0.001) .7781 (0.002) .7532 (0.003) Debt Ratio if q>1 -.0804 (0.038) ** -.0890 (0.026) ** -.1229 (0.127) -.1372 (0.095) * -.1089 (0.332) -.1321 (0.259) -.2384 (0.127) -.2771 (0.072) * Debt Ratio if q<1 -.1920 (0.00) *** -.1984 (0.00) *** -.3750 (0.00) *** -.3846 (0.00) *** -.0873 (0.456) -.1037 (0.384) -2618 (0.317) -.2864 (0.277) Cash flow/TA(-1) .0548 (0.347) .0781 (0.202) .0926 (0.381) -.0468 (0.672) .0737 (0.0481) .1364 (0.222) ** .1448 (0.474) -.0303 (0.895) Capital expenditures/FA(-1) .3083 (0.004)** * .3185 (0.003) *** 1.377 (0.00) *** 1.3944 (0.00) *** -.9450 (0.00) *** -.9186 (0.00) *** -2.0928 (0.00) *** -2.0506 (0.00) *** Tobins q .0196 (0.00) *** .0219 (0.00) *** .0387 (0.004) *** .0427 (0.001) *** .0407 (0.002) *** .0470 (0.00) *** -.0554 (0.068) .0662 (0.047) * Sales growth .08775 (0.016) ** .0839 (0.024) ** .0951 (0.314) .0872 (0.363) .4491 (0.00) *** .4388 (0.00) *** 1.1811 (0.00) *** 1.1620 (0.00) *** R&D expense/TA -.1420 (0.150) -.2489 (0.218) -.3751 (0.087) * -6385 (0.253) R-Square 0.0666 0.0675 0.1141 0.1149 0.0662 0.0687 0.04 0.0409 Adj. R-Square 0.0621 0.0624 0.1095 0.1095 0.0618 0.0637 0.0351 0.0352 N 1275 1164 1293 1182
OLS regression over the sub period 2004-2006 on US large manufacturing firms with a minimum of 1 billion dollars in sales for each base year. Regression A (B) is done without (with) controlling for R&D expense divided by total assets. Regression results are corrected for heteroskedasticity. Significance levels are shown for a 10% level (*), 5% level (**) and a 1% level (***). P-values are shown in parentheses. Employment (Capital expenditures) growth is defined as the percentage change between year 1 (year 3)and the base year (year 0). Debt ratio if q>1 (q<1) is defined as book debt divided by book total assets when Tobin’s q >1 (q<1). Cash flow and R&D expense are divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm dividend by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
4.2.2. 2007-2009
The results found in table 4.7 in the sub period 2007-2009 are comparable with those of the whole period 2004-2014. In addition to the whole period, there is also a significant negative relation between 3-year capital expenditure growth and the debt ratio for low q firms. By contrast, the debt ratio for high q firms only shows a weakly negative relation at a 10% significance level for both the 1-year employment growth as well as for the 1-year capital expenditure growth. Again there are no major effects observed by adding R&D expense as a control variable.
20 Table 4.7 Regression with anticipated growth opportunities and R&D over 2007-2009
Regression 1A 1B 2A 2B 3A 3B 4A 4B 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Intercept .0138 (0.597) -.0180 (0.495) .0881 (0.130) .0973 (0.098) * .0926 (0.239) .1043 (0.182) .7243 (0.00) *** .7350 (0.00) *** Debt Ratio if q>1 -.0379 (0.141) -.0476 (0.067) * -.0626 (0.218) -.0805 (0.108) -.1288 (0.134) -.1565 (0.074) * -.1225 (0.397) -.1443 (0.343) Debt Ratio if q<1 -.1111 (0.011) ** -.1199 (0.006) *** -.2130 (0.002) *** -.2290 (0.001) *** -.3857 (0.00) *** -.4105 (0.00) *** -.3627 (0.037) ** -.3817 (0.028) ** Cash flow/TA(-1) .0660 (0.028) ** .0780 (0.016) ** .2479 (0.002) *** .2703 (0.002) *** .3474 (0.00) *** .3780 (0.00) *** .8292 (0.00) *** .8532 (0.00) *** Capital expenditures/FA(-1) .2134 (0.020) ** .2185 (0.022) ** .8013 (0.00) *** .8235 (0.00) *** -.6754 (0.002) *** -.6584 (0.002) *** -1.5493 (0.00) *** -1.5205 (0.00) *** Tobin’s q .0321 (0.001) *** .0354 (0.00) *** .0381 (0.001) *** .0442 (0.00) *** .0829 (0.00) *** .0920 (0.00) *** .0951 (0.004) *** .1024 (0.004) *** Sales growth -.0528 (0.051) * -.0528 (0.052) * -.1232 (0.071) * -.1259 (0.066) * -.0681 (0.360) -.0687 (0.348) -.4391 (0.00) *** -.4107 (0.00) *** R&D expense/TA -.1906 (0.021) ** -.3585 (0.066) * .5094 (0.094) * -2.955 (0.475) R-Square 0.0759 0.0790 0.0917 0.0947 0.0743 0.0777 0.0558 0.0565 Adj. R-Square 0.0717 0.0741 0.0873 0.896 0.0701 0.0729 0.0513 0.0512
OLS regression over the sub period 2007-2009 on US large manufacturing firms with a minimum of 1 billion dollar in sales for each base year. Regression A (B) is done without (with) controlling for R&D expense divided by total assets. Regression results are corrected for heteroskedasticity, Significance levels are shown for a 10% level (*), 5% level (**) and a 1% level (***). P-values are shown in parentheses. Employment (Capital expenditures) growth is defined as the percentage change between year 1 (year 3) and the base year (year 0). Debt ratio if q>1 (q<1) is defined as book debt divided by book total assets when Tobin’s q >1 (q<1). Cash flow and R&D expense are divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
4.2.3 2010-2014
If differences with the sample as a whole would have to be found, it should be in the regression of the years 2010-2014. The lending rate, at which companies can borrow for the short- and medium-term, has since 2009 been around 3.3% (source: The World Bank). Hence, debt service costs are, ceteris paribus, low compared to other periods and as a result the capital structure decision becomes different. In table 4.8 we find that for high q firms there is, besides a weak negative relation with 1-year capital expenditures growth, no significant relation between the debt ratio and the growth measures.
Although, the relations aren’t significant, a point of interest is the sign of the relation coefficient found. If these positive relations are the result of lower interest rates, it suggests that in certain times the debt ratio could have a positive effect on future firm growth. When that is the case, it suggest that the positive effects of debt predominate the negative effects of debt.
21 Table 4.8 Regression with anticipated growth opportunities and R&D over 2010-2014
Regression 1A 1B 2A 2B 3A 3B 4A 4B 1-year employment growth 3-year employment growth 1-year capital expenditures growth 3-year capital expenditures growth Intercept -.0950 (0.00) *** -.0898 (0.00) *** -.2708 (0.001) *** -.2661 (0.001) *** -.3375 (0.002) *** -.3301 (0.003) *** -.2050 (0.314) -.1739 (0.399) Debt Ratio if q>1 .0094 (0.810) .0023 (0.952) .0311 (0.719) .0222 (0.788) -.1266 (0.057) * -.1365 (0.039) * .0880 (0.666) .0323 (0.877) Debt Ratio if q<1 -.0911 (0.009) *** -.1015 (0.003) *** -.1231 (0.300) -.1354 (0.233) .2293 (0.023) ** -.2155 (0.012) ** -.0422 (0.864) -.1201 (0.630) Cash flow/TA(-1) .0250 (0.358) .1226 (0.242) .0858 (0.110) .0900 (0.099) * -1.259 (0.040) ** .2379 (0.034) ** .4780 (0.068) * .5054 (0.062) * Capital expenditures/FA(-1) .1183 (0.056) * .2185 (0.049) ** .3581 (0.036) ** .3615 (0.034) ** -1.2591 (0.00) *** -1.2534 (0.00) *** -1.7411 (0.00) *** -1.7208 (0.00) *** Tobins q .0153 (0.00) *** .0182 (0.00) *** .0641 (0.001) *** .0671 (0.002) *** .0628 (0.001) *** .0668 (0.001) *** .1442 (0.00) *** .1627 (0.00) *** Sales growth .0960 (0.00) *** .0957 (0.00) *** .2281 (0.00) *** .2288 (0.00) *** .4831 (0.00) *** .4826 (0.00) *** .4318 (0.012) ** .4359 (0.011) ** R&D expense/TA -.2026 (0.025) ** -.2016 (0.473) -.2868 (0.321) -1.2870 (0.047) ** R-Square 0.0318 0.0336 0.0670 0.0675 0.0967 0.0975 0.0500 0.0537 Adj. R-Square 0.0288 0.0302 0.0610 0.0605 0.094 0.0943 0.044 0.0467 N 1981 943 2002 956
OLS regression over the sub period 2010-2014 on US large manufacturing firms with a minimum of 1 billion dollars in sales for each base year. Regression A (B) is done without (with) controlling for R&D expense divided by total assets. Regression results are corrected for heteroskedasticity. Significance levels are shown for a 10% level (*), 5% level (**) and a 1% level (***). P-values are shown in parentheses. Employment (Capital expenditures) growth is defined as the percentage change between year 1 (year 3) and the base year (year 0). Debt ratio if q>1 (q<1) is defined as book debt divided by book total assets when Tobin’s q >1 (q<1). Cash flow and R&D expense are divided by total assets in the previous year. Capital expenditures is divided by fixed assets in the previous year. Tobin’s q is defined as market value of the firm divided by the book value of total assets. Sales growth is defined as the growth of sales between year -1 and 0.
