The effect of debt market conditions on capital structure,
how the level of interest rates affect financial leverage.
Tom Vegter1
Thesis MSc Finance
Faculty of Economics and Business, University of Groningen Supervisor: dr. P.P.M. Smid
June 8, 2017
Abstract
This paper studies the relationship between the level of interest rates and capital structure. Current interest rates are historically low compared to last decennia. This study uses existing capital structure theories and studies on corporate financial behavior to address the role of interest rates. It studies the determinants of capital structure of 2,166 publicly listed firms across 11 countries in the Eurozone from 2002 to 2016. Unbalanced panel data is used to perform regression analysis. I find no significant relationship between the level of interest rates and financial leverage. The results seems to be consistent with the trade-off theory that predicts no change in the attractiveness of debt over equity due to changes in interest rates.
Keywords: Interest rates, capital structure, debt market conditions JEL classification: G10, G30,G32
2
1. Introduction
There is growing attention for the role market conditions play in corporate financing decisions. Studies on market timing theory broadly discuss the effects of equity market conditions on financing decisions (Myers and Majluf, 1984). However the effect of debt market conditions on capital structure still lacks empirical result. This is remarkable since different scholars have addressed the importance of debt market conditions (Graham and Harvey, 2001; Marsh, 1982; Taggart, 1977). Especially the current period of historically low interest rates in the EU seems interesting to gain insight into the effect of debt market on financing decisions. The 10-year yield on German Government bonds in 1994 was 6.9% and decrease to 0.45% today (Spring, 2016). Moreover, considering the European Central Bank policy it is not unrealistic to expect these low interest rate, or even negative interest rates, for the upcoming decennia. This raises the questions on whether these changed conditions will affect corporations and their financing policy. This study focuses on how the level of interest rates affect financial leverage.
Marsh (1982) investigates how companies choose between financing instruments. He finds that historic and current market conditions influence the choice between debt and equity. A survey held under 392 CFO’s by Graham and Harvey (2001) indicates the importance of the level of interest rates. Forty-five percent of the respondents consider the level of interest rates as an important or very important factor affecting their firm’s debt policy. Based on the responses, they find that executives try to time the debt market and issue debt when interest rates are perceived to be low. More recently Baker and Wurgler (2011) found similar results in their survey on behavioral corporate finance. These findings demonstrate the importance of debt market conditions, and in particular the level of interest rates, for a firm’s trade-off between debt and equity. Especially the results of Graham and Harvey (2001) show that low interest rates increase the attractiveness for firms to issue debt. However current capital structure models do not include the level of interest rate as a factor influencing capital structure.
3
This paper starts with a discussion on theoretical frameworks and empirical evidence on the determinants of capital structure in section 2. Section 3 will discuss the hypothesis, method of analysis and data used to answer the research question. In section 4, results from the descriptive and correlation statistics as well as the results from the regression analysis will be discussed. Section 5 will conclude with a discussion on the implications of the results. Furthermore it will address the limitations of this study and ideas for future research.
2. Literature Review
The decision between debt and equity is extensively discussed and studied by different scholars. The predominant capital structure theories include the pecking order, trade-off and market timing theory (Frank and Goyal, 2007). These theories address different market imperfections to explain the attractiveness of debt and equity (e.g. taxes, bankruptcy costs, agency problems, adverse selection). The pecking-order theory suggests that firms prefer to use internal financing over external financing and, if needed, rather issue debt than equity. The trade-off theory argues that the choice between debt and equity is determined by the trade-off between the advantages of the tax-shield and the costs of financial distress. The market timing theory argues that share prices affect a firms decision between debt and equity. If managers believe the price of a company is overvalued, equity issuance is a relatively cheap source of capital compared to debt. Therefore, according the market timing theory, overvaluation of stocks can lead to lower leverage.
Many studies use the theories described above to investigate determinants of capital structure (De Jong, 2001; Fan et al., 2012; Frank and Goyal, 2007) As will be discussed later, the most commonly discussed and empirically proven determinants of capital structure are growth opportunities, firm size, profitability, asset tangibility, industry leverage and inflation. This paper will add the level of interest rates as a new variable to this existing set of factors influencing capital structure. The following section continues with a discussion on how interest rates fits into existing capital structure theories.
2.1. Interest Rates
2.1.1. Modigliani and Miller
4
firms can benefit from a tax shield because of tax deductible interest expenses. By increasing leverage, and thus interest expenses, a firm can create additional value compared to unlevered firms. This is illustrated by the proposition of the value of a levered firm by Modigliani and Miller (1963). The value of a levered firm is equal to the value of a unlevered firm plus the value of the tax shield. The value of the tax shield is determined by the amount of interest expenses and the level of corporate taxes.
The value of the tax shield is calculated by discounting the cash flows from tax deductible interest expenses. The capitalization rates plays an important role in determining the value of a within this model. Modigliani and Miller (1963) assume that the discount rate for income generated from the tax shield equals the average interest rate paid over total debt. This implies that the value of the tax shield is independent of the amount of interest expenses and the capitalization rate. Consequently the value of the tax shield only depends on the level of corporate taxes paid, Tc, and the total debt outstanding, D. The value of the tax shield, VTS, is
then calculated by the identity,
VTS= Tc * D. (1)
It is highly remarkable that this model does not take into account the effect of changing interest expenses. According to Modigliani and Miller (1963) the value of the tax shield stays constant even though interest expenses change. Debt is assumed to be fixed in perpetuity and tax savings are discounted by the effective interest rate. This implies that the value of the tax shield does not change when interest expenses change. Contrary to Modigliani and Miller (1963) one could argue that decreasing interest expenses reduces the cash flows from tax savings and thus the value of the tax shield.
Moreover, the model of Modigliani and Miller (1963) assumes firms to be in a steady state where there are no risks associated with the value of the tax shield. When valuing the benefits from tax deductible interest expenses the assumption is made these advantages last till infinity. However, these cash flows also carry risks associated with the firm. Therefore Myers (1974) proposed the adjusted present value method to separately value the financing benefits and risks of debt financing.
5
savings does not take into account the costs associated with default (Koziol, 2014). However, the probability of bankruptcy can highly affect the value of the tax shield. Default is a major issue for companies since the this heavily affects the profitability of a firm. Lower profits could reduce or even let the value of the tax shield disappear (Koziol, 2014). Therefore several scholars have proposed to use an alternative discounting rate which take into account the risks associated with the probability of default. This replaces the assumption that firms are in a steady state lasting till infinity (Krause et al., 2016; Molnár and Nyborg, 2013).
Existing scholars clearly have indicated that the estimated risk of the tax shield is too low in the original Modigliani and Miller (1963) theorem. The assumption that the tax shield generates cash streams to infinity leads to an unrealistic overvaluation of the tax savings. Moreover, current economic conditions present the situation where negative interest rates are part of a realistic scenario. The effect of non-positive interest rate on the value of the tax shield is even more interesting. If non-positive interest rates reduces interest expenses to zero this will completely wipe out the value of the tax shield. This reduces the corporate finance theory to a situation of Modigliani and Miller (1958) where the value of a firm is independent of its capital structure. Thus, according the Modigliani and Miller theorem non-positive interest rates could reduce the value of a levered firm to being equal to the value of an unlevered firm.
Furthermore, there can be discussion about the effect of negative interest expenses on firm value. Complex taxation issues might occur if a firm receive interest on their outstanding debt. It is unclear how possible interest revenues on outstanding debt are taxed. Moreover, creating return on outstanding debt might affect other determinants of leverage and risk associated with the firm, i.e. risk of financial distress and bankruptcy. However, it is difficult to estimate the consequences of negative interest rates.
2.1.2. Trade-off theory
6 2.1.3. Market timing and firm behavior
Several studies have examined the effect of debt market conditions on the issuance of debt. Graham and Harvey (2001) conduct a survey to study corporate financing practices within firms. Based on the trade-off model they investigate the importance of the interest rate level on decisions to issue new debt. Forty-five percent of the respondents consider the level of interest rates as an important or very important factor affecting their firm’s debt policy. Based on these responses they state that executives are affected by the level of interest rates when making a trade-off between debt and equity. They find that low interest rates increase the attractiveness of debt. These findings are confirmed by a survey from Bancel and Mittoo (2004) on the issuance of debt and equity held under 87 managers in 16 European countries.
Bosworth, Smith and Brill (1971) have investigated how interest rates affect fluctuations in debt issuance. Their hypothesis is that firms will issue long-term debt if they feel that the current level of interest rates is relatively low compared to their expectations of future interest rates. By issuing debt when interest rates are relatively low firms can speculate to take advantage of this. (Baker et al., 2003) perform an analysis on the relation between maturity of debt issued and the shape of the yield curve. They find that large spreads between short- and long-term interest rates are negatively related to the issuance of long-term debt2. This indicates that firms tend to stay
away from long-term debt if long term interest rates are relatively high compared to short term rates.
The latter result still does not actually proof that interest rates influence the total leverage of a company. However, the surveys of Graham and Harvey (2001) and Bancel and Mittoo (2004) indicated that firms tend to issue debt when interest rates are low. This would imply that the issuance of debt increases in periods of relatively low interest rates what will cause higher financial leverage.
2.2. Extensions of the model
As discussed earlier, various studies have investigated determinants of financial leverage. Following from the theories mentioned I will discuss most widely accepted and empirically proven determinants of financial structure.
7 2.2.1. Growth opportunities
According to the trade-off theory, firms with high growth potential face higher costs of debt and therefore lower leverage. This mainly relates to the fact that the cost of financial distress is higher for growth firms. High growth firms can for example face underinvestment and asset substitution problems (Myers, 1977). This increases the cost of debt. Jensen and Meckling (1976) argue that free cash flow problems are less prevalent in firms with high growth opportunities. This reduces the agency costs of equity and implies lower leverage. In contrast, the pecking order theory suggest that high growth firms have a lack of internal financing and therefore need to attract debt for investments.
Many studies have investigated the role of growth opportunities in capital structure (Barclay, Smith and Watts, 1992; Bradley, Jarrell and Han Kim, 1984; Frank and Goyal, 2009; Rajan and Zingales, 1995) Most of these studies find that there is a negative relationship between growth opportunities and leverage (see for example Rajan and Zingales (1995)).
2.2.2. Firm size
The trade-off theory proposes a positive relationship between firm size and leverage. Typically, larger firms are more diversified and therefore face a lower risk of default (Chen, 2003). Besides this, larger firms mostly have a stronger reputation at debt holders (Frank and Goyal, 2004). This results in lower debt related agency and bankruptcy costs which reduces the cost of debt. Lower costs of debt implies an increase in leverage. The pecking order theory claims a negative relationship between firm size and leverage. The argument suggest that a strong reputation reduces equity related asymmetry costs. This reduces adverse selection and thus increases the attractiveness to issue equity instead of debt (Chen, 2003).
Different studies address the relationship between firm size and leverage (Alves and Francisco, 2014; de Jong, Kabir and Nguyen, 2008; Rajan and Zingales, 1995). Most studies find a positive relationship between the size of firms and leverage.
2.2.3. Asset tangibility
8
reduces agency costs. The reduction of these bankruptcy and debt related agency costs increases the attractiveness of debt. Following the pecking order theory asset tangibility reduces information asymmetry which lowers the cost of equity. This implies a negative relation between asset tangibility and leverage (Harris and Raviv, 1991).
Empirical studies predominantly point in the direction of a positive relation between asset tangibility and financial leverage (e.g. Frank and Goyal, 2009; Rajan and Zingales, 1995; Titman and Wessels, 1988).
2.2.4. Profitability
As discussed before it is important for firms to be profitable and have positive earning to take advantage of the tax shield. Moreover profitability is associated with higher credit ratings which implies a lower cost of debt. According to the trade-off theory higher profitability therefore implies higher leverage. According to the pecking order theory firms prefer to use internal financing over external financing. This implies that profitable firms therefore use less debt and thus a negative relationship between profitability and leverage.
Empirical evidence indicates a negative relationship between profitability and leverage (De Jong et al., 2008; Rajan and Zingales, 1995; Zarebski, 2012)
2.2.5. Industry
Leverage ratios can vary widely across industries. This variation can be explained in different ways. One explanation is that the industry leverages ratios are used by firms to benchmark their own leverage. Consequently managers will adjust their leverage ratios towards the industry median. Different studies find empirical proof that firms use the median industry leverage as a proxy for their target capital structure (Gilson, 1997; Flannery and Rangan, 2006; Hovakimian, Opler and Timan, 2001). Besides this some scholars have argued that different omitted variables are captured by an industry measure, this can also explains the explanatory power of this variable (Frank and Goyal, 2009).
2.2.6. Inflation
9
lower the relative price of debt and therefore increase leverage. Frank and Goyal (2009) find a positive relationship between expected inflation and leverage.
3. Methods & Data
3.1. Hypothesis
As discussed in section 2.1 interest rates play an interesting role in existing capital structure theories. From the Modigliani and Miller (1963) theory we see that their assumptions on interest expenses and the value of the tax shield are unrealistic and overvalue the true value of tax savings. Scholars that have addressed this issue propose an alternative discount rate to value cash flows from the tax shield. This would indicate a negative relationship between interest rates and the value of the tax savings. According to the trade-off theory on the cost of debt and equity there is little effect on leverage. The change in interest rates will also call a change in required return on equity which probably will not affect the relative attractiveness of debt over equity. When looking at debt market timing behavior of firms there is proof that lower interest rates triggers the issuance of debt. In their surveys Graham and Harvey (2001), Bancel and Mittoo (2004) and Baker and Wurgler (2011) find proof for this.
The arguments following from the different theories are not in line with each other. Therefore the predictions from existing theories are ambiguous and do no lead to a clear hypothesis. However, considering that leverage is a result of managerial decisions the hypothesis point in the direction that firms try to time the market as is discussed an proven by the work of Graham and Harvey (2001) and Bancel and Mittoo (2004). As a result the hypothesis is:
Hypothesis: The level of interest rates is negatively related with firm leverage.
3.2. Methodology & Model
10
perform a Hausman test. This analysis tests whether the unique errors, Ui, are correlated to the
dependent variable, leverage. The null hypothesis that Ui is uncorrelated with leverage is
rejected and therefore this study uses fixed effects in the regression analysis. To minimize heteroscedasticity problems cross-section white standard errors are used in the regression estimation. To control for outliers in the dataset all firm-specific variables are adjusted by winsorization. This includes the variables leverage, growth opportunities, size, tangibility of assets and profitability, respectively LEV, GROW, SIZE, TANG and PROF. Winsorizing is done at the 1st and 99th percentile of the observations. A correlation matrix is inspected to check
for multicollinearity problems among the independent variables. Table 4 reports the correlation statistics which are discussed in section 4.2.
To control for the effect of the global financial crisis a dummy variable is included which equals one in 2008 and 2009 and zero in all other periods. All independent variables are included with a lag of one year considering the fact that firms take time to adjust their leverage based on the determinants of capital structure.
The complete model is represented by Eq. 2.
LEVit = α + β1CRIDt + β2INTit-1 + β3GROWit-1 + β4SIZEit-1 + β5TANGit-1 + β6PROFit-1 +
β7INDUit-1 + β8INFLit-1 + Xi + Uit (2)
With:
α the intercept,
LEVit the leverage of company i in period 1,
INTt-1 the level of interest rates in period t-1,
GROWit-1 the growth opportunities of company i in period t-1,
SIZEit-1 the size of company i in period t-1,
TANGit-1 the asset tangibility of company i in period t-1,
PROFit-1 the profitability of company i in period t-1,
INDUit-1 the related industry median leverage of company i in period t-1,
INFLit-1 the inflation rates in the country of company i in period t-1,
CRIDt the crisis dummy which equals 1 in 2008 and 2009 and zero in all
other years,
Xi the unobserved firm-specific and time invariant variable of
company i,
11
To investigate whether the relationship between the level of interest rates and financial leverage varies between periods an additional analysis is performed. Therefore the sample period is split into three categories. Firstly, years characterized by relatively high interest rates. Secondly, years characterized by relatively low interest rates. Thirdly, years during the global financial crises as captured in the dummy variable CRID. As discussed in section 4.1 the level of interest rates varies from 4.8% in 2002 to 0.5% in 2015. Furthermore the years 2002-2007 are characterized by interest rates greater than 3.3% and the years 2010-2015 by interest rates smaller than 2.8%. Therefore an additional period dummy, DLOW, is created which equals one in the years 2010-2015 and zero in all other years. The period dummy is used to control for changes in the intercept as well as for changes in the regression coefficient of the variable INT. The interaction variable INT*CRID was added to control for changes in the relationship between INT and LEV during the crisis. This results in an additional model presented by Eq. 3.
LEVit = α + β1CRIDt + β2INTit-1 + β3GROWit-1 + β4SIZEit-1 + β5TANGit-1 + β6PROFit-1 +
β7INDUit-1 + β8INFLit-1 + β9DLOW + β10INTit-1*CRIDt + β11INTit-1*DLOWt + Xi +
Uit (3)
With:
DLOWt the period dummy which equals 1 in the years 2010-2015 and zero
in all other years. 3.3. Variables
This section describes the proxies used for the variables in the model. All values from financial statements and share prices are measured at the 31st of December of the calendar year.
3.3.1. Leverage
12
examine the impact of using book values rather than market values of debt. By using the changes in bond-market yield and detailed information on corporate debt they construct a measure for the market value of debt. They find that using book values rather than market values of debt sometimes leads to serious differences when measuring the financial leverage. Especially in the context of this study it is important to consider that the market value of debt can be highly affected by changes in the interest rate. Unfortunately the data does not provide detailed information on the corporate debt. Therefore we use the book value of total debt and should consider this issue mismeasurement when doing the analysis.
Furthermore the value of a company can be estimated by the total asset value on the balance sheet as well as the market value of the assets. Frank and Goyal (2009) state that leverage based on book values of assets are backward looking where leverage based on marked value of assets is forward looking. Therefore I use the market value of assets to measure the total value of the assets. The leverage of a company as captured in the variable LEV. LEV is calculated as the total debt over the market value of the assets (TDM). The total debt includes all debt items on the balance sheet. The market value of the assets is calculated as the total book value of assets −
book value of equity + number of shares outstanding x share price. For a robustness check I also
consider the leverage based on the book value of assets, TDB. TDB is calculated as the total debt over the book value of assets at the end of the calendar year.
3.3.2. Interest rates
13 3.3.3. Growth opportunities
To capture the growth opportunities of a company this study uses the market-to-book ratio. This measure is used as a proxy for earning opportunities of the assets by many other scholars. (Barclay et al., 1992; Frank and Goyal, 2007; Rajan and Zingales, 1995). The market to book ratio is calculated as (total book value of assets − book value of equity + number of shares
outstanding x share price) divided by total book value of the assets.
3.3.4. Size
This paper uses the market value of assets as a measure for firm size. To adjust for normality I transform the market value of assets by taking the natural logarithm of it. Therefore, the variable SIZE is represented by the natural logarithm of market value of the assets. Other studies often use this measure to estimate the size of a firm (Fan et al., 2012; Iqbal and Kume, 2014; Korteweg, 2010).
3.3.5. Asset tangibility
The asset tangibility of a firm is represents how capable a firm is to use assets as collateral value in case of distress or bankruptcy. Therefore this study looks at the net value of property, plant and equipment (PPE) as part of the total assets. The asset tangibility is therefore calculated as net PPE / book value of total assets.
3.3.6. Profitability
14 3.3.7. Industry median
The industry median leverage is calculated by categorizing firms based on the Standard Industry Classification(SIC). This classification resulted in 11 industries. For each of these industries the yearly industry median was calculated. The leverage median is derived from the leverage variable as discussed in section 3.3.1.
3.3.8. Inflation
The level of inflation is measured by the yearly consumer price index inflation rate per country. Historical inflation rates are obtained per country. Inflation rates were assigned based on the country where the firm is headquartered according the Standard & Poor’s CapitalIQ database.
3.4. Data description
The sample used in this study contains unbalanced panel data from the period 2002-2016. Since this study makes use of one lag in the regression model, independent and dependent variables are taken from the period 2002-2015 and 2003-2016 respectively. Firm-specific data is obtained from Standard & Poor’s Capital IQ database. Economic variables are retrieved from the Oxford Economics database. The data includes all publicly listed firms headquartered in the countries which were part of the Eurozone from 20023. Companies located in Greece are
excluded from analysis due to the complex situation regarding the national debt crisis. Table 2 gives an overview of the countries included in the sample and the distribution per country. Firms active in financial services and public utilities are excluded from analysis since their capital structure heavily depends on the regulatory environment they operate in (Schaeffler and Weber, 2012; Schober et al., 2014). Therefore firms with SIC codes starting with 46, 49, 60-64 and 67 are excluded from the sample. Based on these selection criteria the sample consist out of 2,305 individual firms and 26,473 unique firm year observations. Observations with missing data for the firm-specific variables are excluded from analysis. This result in an adjusted sample of 2,166 individual firms and 21,907 unique firm year observations.
Table 1 and 2 present the sample distribution by industry and country respectively. The Full Sample columns show the absolute number of firm year observations as well as the percentage
3The Capital IQ database is used to determine in which countries the firms are headquartered. Listing of these mid-
15
of total observations. The Adjusted Sample columns show the distribution in absolute number and percentages after the adjustment for missing data. For both, the distribution by industry and by country, the adjustment did not incur any major changes in the relative composition of the sample. Table 1 shows that approximately half of the firm year observations are coming from firms active in manufacturing. The service industry accounts for almost 23 percent of the observations, other industries each account for approximately 0.5 to 7 percent of the total observations. Table 2 shows that over 65 percent of the observations are from countries headquartered in France, Germany or Italy. Furthermore other countries each account for approximately 2 to 7 percent of the total firm year observations.
Table 1
Sample distribution by industry
This table presents the distribution of the firm year observations by industry. The data contains firm year observations from all publicly listed firms headquartered in countries which are part of the Eurozone since 2002. The sample period is 2002-2016. Greek and financial service firms are excluded from the sample. The columns ‘Full Sample’ show the distribution after these firms are excluded. The columns ‘Adjusted Sample’ show the distribution after observations with missing data are excluded. The columns ‘Obs.’ Show the absolute number of firm year observations for the industry. The columns ‘%’ show the number of firm year observation as a percentage of the total number of observations.
Full Sample Adjusted Sample
Industry Obs. .% Obs. .%
Agriculture, Forestry and Fishing 164 0.62 135 0.62
Mining 577 2.18 410 1.87
Construction 855 3.23 719 3.28
Manufacturing 13,031 49.22 11,101 50.67
Transport, Communication, Electric, Gas and Sanitary service 1,787 6.75 1,528 6.97
Wholesale Trade 916 3.46 771 3.52
Retail Trade 1,056 3.99 918 4.19
Finance, Insurance and Real Estate 1,303 4.92 1,061 4.84
Services 6,332 23.92 4,998 22.81
Public Administration 0 0.00 0 0.00
Non Classifiable 452 1.71 266 1.21
16
4. Results
4.1. Descriptive statistics
Table 3 present the descriptive statistics for the variables in the sample. The median of the dependent variable TDM is 0.199. Similar to what is found in other studies, the median for the leverage ratio TDB, 0.232, is slightly higher than TDM (Frank and Goyal, 2009). Besides this, the standard deviations of TDM and TDB are very similar, 0.205 and 0.204 respectively. Overall the leverage ratio in this data shows similar results to what is found in other studies on capital structure (Frank and Goyal, 2009; Iqbal and Kume, 2014; Korteweg, 2010). Fig. 1 shows the development of the yearly median leverage of the complete sample for both TDM and TDB. The yearly median of TDM shows a sharp increase from 0.168 in 2007 to 0.259 in 2008. This is explained by the drop in market value of the assets during the global financial crisis causing TDM to increase. Furthermore the yearly median of TDB is more constant during the sample period and fluctuates around 0.23.
Table 2
Sample distribution by country
This table presents the distribution of the firm year observations by country. The data contains firm year observations from all publicly listed firms headquartered in countries which are part of the Eurozone since 2002. The sample period is 2002-2016. Greek and financial service firms are excluded from the sample. The columns ‘Full Sample’ show the distribution after these firms are excluded. The columns ‘Adjusted Sample’ show the distribution after observations with missing data are excluded. The columns ‘Obs.’ Show the absolute number of firm year observations for the country. The columns ‘%’ show the number of firm year observation as a percentage of the total number of observations.
Full Sample Adjusted Sample
Country Obs. .% Obs. .%
17
The independent variable INT has a mean of 3.006 and a standard deviation of 1.262. The maximum value for INT is reported in 2002 as 4.824. From 2010 on, the value of INT stays below 3.0. During the sample period INT decreased to a minimum of 0.541 in 2015. EURIBOR has a mean of 2.172 which is lower than INT. This is as expected since EURIBOR has a shorter maturity than the 10-year German bond yield which is represented by INT. Except for the crisis period, EURIBOR show a similar development over time as INT.
Figure 1
Development of interest rates and financial leverage
This graph shows the development of the variables INT, EURIBOR and the median values of TDM and TDB in the period 2002-2016. INT is the yearly average of the yield on 10-year German government bonds. EURIBOR is the yearly average of the 12-month Euribor rate. TDM is the total debt divided by the market value of the assets. TDB is the total debt divided by the book value of the assets. The values for INT and EURIBOR are indicated by the primary axis on the left. The values for TDM and TDB are indicated by the secondary axis on the right of the graph.
The variable GROW has a median of 1.103. This values is lower compared to other studies of Frank and Goyal (2009) and Alves and Francisco (2013) who reports a mean value of 1.76 and a median value of 1,34 respectively. The size of the companies, measured in the variable SIZE, is slightly higher than what is found by Frank and Goyal (2009). They report a mean value of SIZE of 4.58 where this study shows a value of 5.335. The variables TANG and PROF have median values of 0.158 and 0.086 respectively. This is comparable to what is found in other studies (Frank and Goyal, 2009; Iqbal and Kume, 2014; Korteweg, 2010).
The variable INDU represents the median leverage per industry. Based on the two different leverage measures, TDM and TDB, the INDU variables have a mean of 0.238 and 0.250 respectively. Again, the leverage ratio based on the market value of the assets is lower than the
18
leverage based on the book value of the assets. During the sample period 2002-2015, the retail trade industry reports the highest average leverage of 0.503, the real estate industry reports the lowest average leverage of 0.146 based on TDM.
The variable INFL reports the yearly inflation for all eleven countries in the sample and has a mean of 1.844. In general, the values for INFL are relatively high in 2007 and 2008 and drop to even negative values, implying deflation, in 2009. In Ireland the reported inflation took the maximum value of 4.905 in 2007 and the minimum value of -4.459 in 2009. Typically the variable INFL has values greater than 2.0 before 2007 and below 1.5 after 2012. During the period 2007-2012 there are high fluctuations in the variable INFL.
Table 3
Descriptive statistics
This table reports the descriptive statistics for the variables used in this study. Given the use of lags the sample of the independent and dependent variables covers the period 2002-2015 and 2003-2016 respectively. TDM is the total debt divided by the market value of the assets. TDB is the total debt divided by the book value of the assets. INT is the yearly average of the yield on 10-year German government bonds. EURIBOR is the yearly average of the 12-month Euribor rate. GROW is the market-to-book ratio value is calculated as (total book value of assets − book value of equity + number of shares outstanding x share price) divided by total book value of the assets. MVA represent the market value of the assets in million euros calculated by the total book value of assets − book value of equity + number of shares outstanding x share price. The variable SIZE is the natural logarithm of MVA. TANG is calculated asnet PPE divided by the book value of total assets. The variable PROF is calculated as EBITDA divided by the book value of total assets. INDU TDM is the yearly industry median of the variable TDM based on SIC classification. INDU TDB is the yearly industry median of the variable TDB based on SIC classification. INFL is the yearly consumer price index inflation rate per country. All firm-specific variables are winsorized at the 1st and
99th percentile. The economic variables INT, EUR12M and INFL are acquired from the Oxford Economics database,
other variables are acquired from Standard & Poor’s Capital IQ database. N presents the number of observations.
Variable N Mean Median Max Min
19 4.2. Correlation statistics
Table 4 shows the correlation statistics for the variables used in this study. The correlation between the two measures of leverage is 0.825 and highly significant. However, the correlation coefficients between the interest rates and leverage ratios have opposite directions with the two leverage measures. The level of interest rates has a small but positive correlation to leverage based on market values and a negative correlation to leverage calculated by book values. This is the case for both variables, INT and EURIBOR, measuring the level of interest rates. The variable measuring the growth opportunities is negatively correlated to both measures of leverage. However, the correlation between growth opportunities and leverage based on market values is much stronger. Hence, both variables are a function of the market value of the assets. The size of a company is positively correlated to leverage but is only significant with the leverage based on book values. The correlation coefficient between asset tangibility and leverage is positive and significant. The correlation between profitability and leverage is negative and highly significant. Both variables measuring the industry median are positively correlated to leverage. The level of inflation has a small but positive and significant correlation coefficient with leverage.
20
Table 4
Correlation statistics
This table report the correlation coefficients for all variables used in this study. Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively. Given the use of lags the sample of the independent and dependent variables covers the period 2002-2015 and 2003-2016 respectively. All firm-specific variables are winsorized at the 1st and 99th percentile. TDM is the total debt divided by the market value of the assets. TDB is the total debt divided by the book value of the assets. INT is the yearly average of
the yield on 10-year German government bonds. EURIBOR is the yearly average of the 12-month Euribor rate. GROW is the market-to-book ratio value is calculated as (total book value of assets − book value of equity + number of shares outstanding x share price) divided by total book value of the assets. MVA represent the market value of the assets in million euros calculated by the total book value of assets − book value of equity + number of shares outstanding x share price. The variable SIZE is the natural logarithm of the variable MVA. TANG is calculated asnet PPE divided by the book value of total assets. The variable PROF is calculated as EBITDA divided by the book value of total assets. INDU TDM is the yearly industry median of the variable TDM based on SIC classification. INDU TDB is the yearly industry median of the variable TDB based on SIC classification. INFL is the yearly consumer price index inflation rate per country.
Variable TDM TDB INT EURIBOR GROW SIZE TANG PROF INDU TDM INDU TDB INFL
21 4.3. Regression results
Table 5 present the regression result from the OLS analysis with the independent variable TDM. Model 1 includes a one variable regression including the crisis dummy. Model 2 presents the full model with all independent variables included. In model 3, insignificant variables are excluded from the regression. As discussed in section 3.2, an additional analysis is performed to investigate a possible change in the effect across periods characterized by either relatively low and high interest rates. Model 4 is the full model including the second period dummy what represents time periods characterized by low interest rates. In this model the dummy as well as the interaction term with the variable INT are included. In model 4, time fixed effects are left out due to the fact that the inclusion of a second period dummy yet increase the control for time fixed effects.
As discussed in section 2.2 this study includes the most commonly accepted and empirically proven determinants of leverage. My results confirm most of these expectations and the findings of other studies. The results of model 2, 3 and 4 indicate that direction of the relationship between leverage and growth opportunities, size, asset tangibility and profitability are consistent with earlier findings. The variable GROW has a significant negative relation to TDM. However, the effect of the relationships is relatively small. The variable SIZE has a significant positive relation to TDM. The relationship between the variable TANG and TDM is proven to be significantly positive. Furthermore the results show that the relationship between the variable PROF and TDM is significantly negative. The results from model 2 and 3 show that INDU has significant positive relationship with TDM. However, model 4 does not support this relationship. The results from model 2 do not support the expectation that the variable INFL has a significant positive relationship with the dependent variable. Therefore the variable was excluded in model 3. Leaving out the variable INFL in model 3 does not affect the coefficients of the other variables greatly. Only the value of the intercept and the crisis dummy change from 0.177 to 0.151 and 0.041 to 0.058 respectively. However, the adjusted r squared does not change between model 2 and 3. Overall, the relationship between leverage and growth opportunities, firm size, asset tangibility and profitability appear to be relatively constant across the models 2 and 3. Furthermore, we see minor changes in the estimated coefficients of the these independent variables in model 4.
22
fixed effects are used4. Analysis of the full model as presented in model 2 give similar results for
the relationship between INT and TDM. Moreover, analysis of model 3 gives the same results.
4 Running the regression with fixed effects only resulted in an R2 and adjusted R2 of 0.676 and 0.643 respectively.
Results are available upon request.
Table 5.
Regression results
This table reports the regression statistics from the Ordinary Least Square regression. The use of fixed effects is indicated at the bottom of the table. Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively. All independent variables, except for the dummy variables, are lagged by one year. Given the use of lags, the sample of the independent and dependent variables covers the period 2002-2015 and 2003-2016
respectively. All firm-specific variables are winsorized at the 1st and 99th percentile. To minimize
heteroscedasticity problems cross-section white standard errors are used. The independent variable, TDM, is the total debt divided by the market value of the assets. CRID is the crisis dummy which equals one in 2008 and 2009 and zero in all other years. DLOW is a dummy variable which equals 1 in the period 2010-2016 and zero in all other years. GROW is the market-to-book ratio value is calculated as (total book value of assets − book value of equity + number of shares outstanding x share price) divided by total book value of the assets. MVA represent the market value of the assets in million euros calculated by the total book value of assets − book value of equity + number of shares outstanding x share price. The variable SIZE is the natural logarithm of the variable MVA. TANG is calculated asnet PPE divided by the book value of total assets. The variable PROF is calculated as EBITDA divided by the book value of total assets. INDU TDM is the yearly industry median of the variable TDM based on SIC classification. INFL is the yearly consumer price index inflation rate per country. INT is the yearly average of the yield on 10-year German government bonds. N is the number of firm year observations.
TDM (1) (2) (3) (4) Intercept 0.256*** 0.177*** 0.151*** 0.086* CRID (dummy) 0.046*** 0.041*** 0.058*** 0.042** DLOW (dummy) 0.039 GROW -0.040*** -0.039*** -0.040*** SIZE 0.011** 0.011** 0.017*** TANG 0.083*** 0.083*** 0.065*** PROF -0.160*** -0.160*** -0.154*** INDU 0.337*** 0.337*** 0.105 INFL 0.001 0.004** INT -0.007 -0.011 -0.011 0.019 INT*CRID 0.003 INT*DLOW -0.006 R2 0.702 0.722 0.722 0.715 Adjusted R2 0.669 0.691 0.691 0.683 F-statistic 21.323*** 23.413*** 23.425*** 22.723***
Entity fixed effects Yes Yes Yes Yes
Time fixed effects Yes Yes Yes No
23
Model 4 allows for changes in the relationship between the variable INT and TDM. By using the dummy variables DLOW and CRID the sample is split into three periods. Firstly, the period 2003-2007 represent years characterized by relatively high interest rates. Secondly, the years 2008-2009 represent the period of the global financial crisis. Thirdly, the period 2010-2016 represent years characterized by relatively high interest rates. Model four includes two interaction terms to see whether the relationship between TDM and INT differs between the three time periods. The results show that the variables INT*CRID and INT*DLOW are not significant. This gives proof that the relationship between TDM and the variable INT, does no significantly differ between the three time periods.
Furthermore, the results from model 4 show that there is no significant coefficient for the dummy variable DLOW. These results provide no proof that leverage changes during periods characterized by relatively high interest rates. Similar to model 1, 2 and 3 the variable CRID has a significant positive value, implying that leverage increased during the period 2008-2009.
Based on the models tested I cannot conclude the variable INT has a significant effect on the variable TDM. The relationship between these two variables is consistently insignificant in all models. Moreover, including the dummy DLOW does not change these conclusions. According model 4 the level of leverage does not significantly change between periods characterized by relatively low or high interest rates. The results found in my analysis provide no evidence to support my hypothesis. Therefore I fail to reject the null hypothesis that the level of interest rates has a positive or no relationship with financial leverage. Furthermore, I found no evidence to assume that there is a positive relationship between the level of interest rates and financial leverage.
4.4. Robustness checks
To check for robustness of the results I perform additional analysis by replacing the variables TDM and INT by other measures. The results of this are presented in table 6. Model 5, 6 and 7 show the estimated regression equation with the leverage based on book values, TDB, as a measure for financial leverage. Model 8 shows a model where the level of interest is measured by the variable EURIBOR representing the yearly average 12-month Euribor rates. In model 8 TDM is used as a measure for leverage.
24
model 5, 6 and 7. Furthermore, the results for the dummy variable DLOW and the interaction variables INT*CRID and INT*DLOW show no significant relationships when using TDB as a measure for leverage. These results are consistent with the results found in table 5.
In model 8, the yearly average of the 12-month Euribor rate is used as a measure for the level of interest rates. The results show that there is no significant relation between the variable EURIBOR and TDM. These results are similar to the results from table 5.
Based on the robustness tests I can conclude that the relations leverage and growth opportunities, firm size, asset tangibility, profitability and industry median leverage are consistently significant in most of the models tested in table 5 and 6. The level of inflation, INFL, shows a significant but very weak relation with financial leverage in model 7.
25
Table 6.
Regression results robustness checks
This table reports the regression statistics from the Ordinary Least Square regression used for robustness checks. The use of fixed effects is indicated at the bottom of the table. Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively. All independent variables, except for the dummy variables, are lagged by one year. Given the use of lags, the sample of the independent and dependent variables covers the period 2002-2015 and 2003-2016 respectively. All firm-specific variables are winsorized at the 1st and 99th percentile. To
minimize heteroscedasticity problems cross-section white standard errors are used. The independent variable is indicate at the top of the table. TDB is the total debt divided by the book value of the assets. TDM is the total debt divided by the market value of the assets. CRID is the crisis dummy which equals one in 2008 and 2009 and zero in all other years. DLOW is a dummy variable which equals 1 in the period 2010-2016 and zero in all other years. GROW is the market-to-book ratio value is calculated as (total book value of assets − book value of equity + number of shares outstanding x share price) divided by total book value of the assets. MVA represent the market value of the assets in million euros calculated by the total book value of assets − book value of equity + number of shares outstanding x share price. The variable SIZE is the natural logarithm of the variable MVA. TANG is calculated asnet PPE divided by the book value of total assets. The variable PROF is calculated as EBITDA divided by the book value of total assets. INDU TDM is the yearly industry median of the variable TDM based on SIC classification. INFL is the yearly consumer price index inflation rate per country. INT is the yearly average of the yield on 10-year German government bonds. EURIBOR is the yearly average of the 12-month Euribor rate. N is the number of firm year observations.
TDB TDB TDB TDM
Model 5 Model 6 Model 7 Model 8
Intercept 0.293*** 0.045 0.031* 0.027 Crisis dummy 0.037*** 0.034*** 0.037*** 0.034*** DLOW -0.015 GROW -0.009*** -0.010*** -0.009*** SIZE 0.021*** 0.024*** 0.021*** TANG 0.104*** 0.097*** 0.104*** PROF -0.236*** -0.232*** -0.236*** INDU 0.446*** 0.400*** 0.467*** INFL -0.002 0.004*** 0.000 INT -0.011 0.002 -0.001 EURIBOR 0.000 INT*CRID 0.000 INT*DLOW 0.003 R2 0.693 0.709 0.709 0.710 Δ R2 0.659 0.676 0.676 0.678 F-statistic 20.395*** 22.080*** 22.050*** 22.088***
Entity fixed effects Yes Yes Yes Yes
Time fixed effects Yes Yes No Yes
26
5. Conclusion
5.1. Main conclusions
This paper studies the effect of interest rates on financial leverage. Based on existing corporate finance studies I hypothesize that interest rates are negatively related to leverage. In the model, the level of interest rates is added to a set of empirically proven determinants of leverage. To test the hypothesis I perform an OLS regression on unbalanced panel data including 21,907 unique firm year observations. The data contains 2,166 publicly listed firms headquartered in the Eurozone.
The results show that there is no significant relationship between the level of interest rates and financial leverage. Based on the findings of Graham and Harvey (2001), Bancel and Mittoo (2004) and Baker and Wurgler (2011) I hypothesize a negative relationship between interest rates and financial leverage. The predictions of existing capital structure theories on the relation between interest rates and financial leverage are ambiguous. According Modigliani and Miller (1963) there is no effect of interest rates on the optimal capital structure. Other studies argue that lower interest rates could decrease interest expenses and so the value of the tax shield (Fernandez, 2004). From our results I cannot conclude whether a change of interest rates directly affects the value of the tax shield. According to the trade-off theory a change in interest rates will not change the attractiveness of debt relative to equity. Under the assumption that firms will adjust leverage to the optimal capital structure, our findings provide proof for this statement.
In contrast to my results, the surveys discussed find that firms increase debt issuance in periods of relatively low interest rates. The contradiction between these findings might be explained by the fact that managers issue debt when they perceive interest rates to be low. The perception of low interest rates can be very abstract and might not be captured by just the absolute level of interest rates. Furthermore it could well be that changes in capital structure due to the level of interest rates are very small or nihil. This makes it difficult to observe adjustments in capital structure because of changes in the level of interest rates. Unfortunately the data does not provide details on quantity and size of debt issuances.
27 5.2. Limitations and future research
28
6. References
Alves, P., Francisco, P., 2014. The impact of institutional environment on the capital structure of firms during recent financial crises. The Quarterly Review of Economics and Finance 57, 129–146.
Baker, M., Greenwood, R., Wurgler, J., 2003. The maturity of debt issues and predictable variation in bond returns. Journal of Financial Economics 70, 261–291.
Bancel, F., Mittoo, U.R., 2004. Cross-country determinants of capital structure choice: a survey of European firms. Financial Management 33, 103–132.
Barclay, M.J., Smith, C.W., Watts, R.L., 1992. The determinants of corporate leverage and dividend policies. Journal of Financial Economics 32, 263–292.
Bosworth, B., Smith, W., Brill, D., 1971. Patterns of corporate external financing. Brookings Papers on Economic Activity 253–284.
Bradley, M., Jarrell, G.A., Han Kim, E., 1984. On the existence of an optimal capital structure: theory and evidence. The Journal of Finance 39, 857–878.
Chen, J., 2003. Determinants of capital structure of Chinese-listed companies. Journal of Business Reserach 57, 1341–1351.
Cooper, I.A., Nyborg, K.G., 2006. The value of tax shields is equal to the present value of tax shields. Journal of Financial Economics 81, 215–225.
De Jong, A., 2001. The disciplining role of leverage in dutch firms. European Finance Review 48, 31–62.
De Jong, A., Kabir, R., Nguyen, T.T., 2008. Capital structure around the world: The roles of firm- and country-specific determinants. Journal of Banking and Finance.
Deesomsak, R., Paudyal, K., Pescetto, G., 2004. The determinants of capital structure: Evidence from the Asia pacific region. Journal of multinational financial management 14, 387–405. Dempsey, M., 2013. Consistent cash flow valuation with tax-deductible debt: A clarification.
European Financial Management 19, 830–836.
Doğan, M., 2013. Does firm size affect the firm profitability? Evidence from Turkey. Research Journal of Finance and Accounting 4, 2222–2847.
Fan, J.P.H., Titman, S., Twite, G., 2012. An international comparison of capital structure and debt maturity choice. Journal of Financial and Quantitative Analysis 47, 23–56.
Fernandez, P., 2004. The value of tax shields is not equal to the present value of tax shields. Journal of Financial Economics 73, 145–165.
29 of Financial Economics 79, 469–506.
Frank, M.Z., Goyal, V.K., 2009. Capital structure decisions: Which factors are reliably important? Financial Management 38, 1–37.
Frank, M.Z., Goyal, V.K., 2007. Trade-off and pecking order theories of debt, Handbook of Corporate Finance. Elsevier, North Holland.
Frank, M.Z., Goyal, V.K., 2004. The effect of market conditions on capital structure adjustment. Finance Research Letters 1, 47–55.
Gilson, S.C., 1997. Transactions costs and capital structure choice: Evidence from financially distressed firms. The Journal of Finance 52, 161–196.
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.
Harris, M., Raviv, A., 1991. The theory of capital structure. The Journal of Finance 46, 297– 355.
Hovakimian, A., Opler, T., Titman, S., 2001. The debt-equity choice. The Journal of Financial and Quantitative Analysis 36, 1–24.
Iqbal, A., Kume, O., 2014. Impact of Financial Crisis on Firms’ Capital Structure in UK, France, and Germany. Multinational Finance Journal 18, 249–280.
Jensen, M.C., Meckling, W.H., 1976. Theory of the firm: Managerial behaviour, agency costs and ownership structure. Journal of Financial Economics 3, 305–360.
Korteweg, A., 2010. The net benefits to leverage. The Journal of Finance 65, 2137–2170. Koziol, C., 2014. A simple correction of the WACC discount rate for default risk and
bankruptcy costs. Review of Quantitative Financial Accounting 42, 653–666.
Krause, M.V., Lahmann, A., De, A.L., 2016. Reconsidering the appropriate discount rate for tax shield valuation. Journal of Business Economics 86, 477–512.
Lee, K., Lee, C.F., 2009. Cash holdings, corporate governance structure and firm valuation. Review of Pacific Basin Financial Markets and Policies 12, 475–508.
Marsh, P., 1982. The choice between equity and debt: an empirical study. The Journal of Finance 37, 121–144.
Massari, M., Roncaglio, F., Zanetti, L., 2007. On the equivalence between the APV and the wacc approach in a growing leveraged firm. European Financial Management 14, 152–162. Modigliani, F., Miller, M.H., 1963. Corporate income taxes and the cost of capital: A correction.
The American Economic Review 53, 433–443.
30
Molnár, P., Nyborg, K.G., 2013. Tax-adjusted discount rates: A general formula under constant leverage ratios. European Financial Management 19, 419–428.
Myers, S.C., 2001. Capital structure. Journal of Economic Perspectives 15, 81–102.
Myers, S.C., 1977. Determinants of corporate borrowing. Journal of Financial Economics 147– 175.
Myers, S.C., 1974. Interactions of corporate financing and investment decisions-implications for capital budgeting. The Journal of Finance 29, 1–25.
Myers, S.C., Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that investors do not have. Journal of Financial Economics 13, 187–221. Rajan, R.G., Zingales, L., 1995. what do we know about capital structure? Some evidence from
international data. The Journal of Finance 50, 1421–1460.
Saliha, T., Abdessatar, A., 2011. The determinants of financial performance: an empirical test using the simultaneous equations method. Economics and Finance Review 1, 1–19. Schaeffler, S., Weber, C., 2012. An Evaluation of Different Approaches to Capital Structure
Regulation Chair for Management Sciences and Energy Economics An Evaluation of Different Approaches to Capital Structure Regulation.
Schober, D., Schaeffler, S., Weber, C., Schober, D., Schaeffler, S., Weber, C., 2014. Idiosyncratic risk and the cost of capital: the case of electricity networks. Journal of Regulation Econonmics 46, 123–151.
Sweeny, R.J., Warga, A.D., Winters, D., 1997. The Market value of debt, market versus book value of debt, and returns to assets. Journal of Financial Management 26–21.
Taggart, R.A., 1985. Secular patterns in the financing of U.S. corporations, in: Corporate Capital Structure in the United States. University of Chicago Press, Chigaco, pp. 13–80. Taggart, R.A., 1977. A model of corporate financing decisions. The Journal of Finance 32,
1467–1484.
Titman, S., Wessels, R., 1988. The determinants of capital structure choice. The Journal of Finance 43, 1–19.
