University of Amsterdam
Amsterdam Business School
Msc Finance: Corporate Finance Master Thesis
Capital Structure of European Firms: A Unified Model
Author: Pintilie Horia Toma June, 2021
Supervising Professor: Radomir Todorov
1
Statement of Originality
This document is written by Student Horia Toma Pintilie who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the
supervision of completion of the work, not for the contents.
2
Table of Contents
Abstract ...3
1. Introduction ...4
2. Literature Review ...6
2.1 Firm Specific Effects ...6
2.2 Monetary Policy Effects ...8
2.3 The Introduction of Negative Interest Rates by the ECB ...9
2.4 Macroeconomic and Fixed Effects ...10
3. Data & Methodology...12
4. Results & Discussion ...19
4.1 Policy Rates Correlations ...19
4.2 Summary Statistics ...20
4.3 Regression Analysis ...26
4.4 Limitations and Suggestions for Future Research ...31
5. Concluding Remarks ...32
References ...34
Appendix ...37
3
Abstract
The paper proposes a methodology for the analysis of corporate structure decisions of european firms. The model unifies previous findings in firm specific, macroeconomic and monetary policy predictors of market leverage ratios. As well as that, firms are separated into financially
constrained and unconstrained. The main findings of the analysis indicate that responses to firm specific variables are similar between both types of firms. Meanwhile, unconstrained firms are most responsive to changes in macroeconomic policy, due to its effects on costs of equity, while constrained firms show no significance in coefficients. Monetary policy is found to affect the constrained sample, mainly through borrowing responses, while the unconstrained sample is largely unaffected by policy rate changes.
4 1. Introduction
The field of capital structure has been under research since the introduction of the Modigliani Miller theorem (1958), which stated that in a perfect world where capital markets work flawlessly, capital structure does not matter, since the cost of debt equals the cost of equity.
Since then, the determinants of capital structure have been found to vary, due to the fact that the real capital markets have numerous imperfections. However, most of the studies are centered on firm specific characteristics, with few publications having been made on other determinants of leverage (Lemmon et al, 2003; Kumar et al, 2017).
Furthermore, most of the well established literature is based on firms from the United States, with Europe as well as the rest of the world being less researched, due to the extra complexity that investigating these areas brings (Kumar et al, 2017). Therefore, the aim of this study is to attempt a filling of the gap in literature by estimating a model for capital structure in which firm specific, industry specific and macroeconomic characteristics are considered. As well as that, the study attempts to provide a method that is able to bring european countries in the same sample, without biasing the results.
Conducting this research in an European setting is also crucial, since the area has gone through a unique monetary policy event, which was the introduction of negative policy rates by the European Central Bank (ECB). This event is shown to have had significant effects on the borrowing and lending behavior of economic agents.1 Therefore, this time frame might be revealing with regards to the relationship between policy rates and firms’ capital structures in the short run. As well as that, it could help in the creation of a comprehensive methodology that helps predict the borrowing behavior of firms once different policies are adopted by european institutions.
Another contribution of this paper is the separation of the sample into financially
constrained and unconstrained companies, as their reactions in borrowing behavior to changes in macroeconomic variables is shown to differ (Korajczyk & Levy, 2003). This will help advance
1 Heider et al. (2019), Bech & Malkhozov(2016)
5
the current debate in capital structure decisions of firms and well as provide european policy makers with a better understanding of the impact of their decisions on different types of firms.
Therefore, the research question of the paper addresses whether macroeconomic and monetary policy are useful in predicting capital structure decisions of european firms in a short time frame model. The model proposed will unify previous findings regarding determinants of capital structure, as well as make use of a shorter time frame than those used in studies which have attempted similar analyses in the United States (Graham et al, 2015).
The main findings of the paper suggest that in a unified model, which includes firm specific effects, macroeconomic and monetary policy effects, there exists a discrepancy in response to these effects based on financial constraints. The financially unconstrained sample returns significant coefficients for most firm specific variables (growth opportunities, firm size, tangibility, asset growth) and most macroeconomic variables (GDP growth, governement debt to GDP, inflation). On the other hand, the constrained sample’s response differs in that all firm specific variables are significant in predicting leverage ratios, while macroeconomic factors play no significant role. Policy rates are also able to influence leverage ratios in the short term for constrained companies, presenting a significant relationship to leverage. These findings are of importance to policy makers, since they show a localized impact of macroeconomic policy on unconstrained companies, since these changes are shown to have little effect on the capital structure of constrained firms. Meanwhile, the impact of monetary policy is localized on constrained companies.
Futhermore, the model is shown to be capable of high predictive power, explaining 82%
of the variation in leverage ratios for the full sample of firms. This is an important finding and an advancement in the field of corporate finance literature, since one of the main issues identified by literature was the lack of explainatory power of previous models (Kumar et al, 2017; Lemmon et al, 2003).
The next section of this paper will present the main previous findings from the existing literature and establish a theoretical foundation upon which this research is conducted. This section will analyse previous work on firm specific effects, macroeconomic and monetary policy effects on firm borrowing behavior. As well as that, the relevance of the study is discussed.
6
Section three describes the data sourcing and selection processes, as well as explain in detail the methodology of the empirical study. Section 4 is a discussion of the main results and implications of the paper’s findings, as well as the limitations and directions for future research which this paper identifies. Finally, section 5 presents the main conclusions of the study and the implications for policymakers as well as the academic field.
2. Literature Review
2.1 Firm Specific Effects
The vast majority of existing literature in the field of capital structure investigates firm specific variables that play an important role in the decision of leverage ratios. This is identified by the literature review of Kumar et al (2017), who provide a comprehensive overview of all the major determinants of capital structure which were found since the publication of Modigliani and Miller’s irrelevance theorem (1958). Their main findings are that the majority of well established literature focuses on firm specific factors and the United States market. The result of these findings is that the main existing models which analyze capital structures generally leave a large portion of variance in leverage unexplained or explained by onobservable factors. They sugest that future studies be aimed at finding the relationships between other broader factors that
determine capital structure decisions, as well as explore these relationships in emerging markets.
Regarding firm specific variables, a vast array of articles have been published since the 1990’s2, finding the main determinants of capital structure to be growth opportunities, denoted by Tobin’s Q or the Market to Book ratio, as well as firm size, tangibility of assets, profitability and growth of assets.
The models surveyed by Harris and Raviv (1991) have shown that leverage is positively associated with firm value. This comes as a result of larger firms being able to bear more debt, since they are less likely to default. This unlikelihood of default is mainly due to the fact that large firms have more liquid assets on their balance sheets, which can be sold in case of cash
2 Harris & Raviv (1991), Rajan & Zingales (1995), McConnell & Servaes (1995), Castro et al. (2016), Graham &
Harvey (2001)
7
flows’ insufficiency to cover debt installments. Smaller firms, however, only dispose of assets which are vital to their activity and less liquid in nature, therefore increasing the difficulty of selling such assets in order to repay debt. However, Rajan & Zingales (1995) argue that the effects of firm size may be ambiguous, due to the easier access to equity funding that large firms tend to have. Since equity is cheaper for them and attracts more demand, the argument is that larger firms find it easier to raise capital from equity than small firms do. However, most studies find a positive relationship between the two variables, perhaps indicating that the effect
described by Harris and Raviv (1991) is stronger.
The same studies also show a negative relationship between leverage and growth opportunities. The economic argument behind this finding is that firms are able to raise equity fairly easily when this equity is highly valued, due to the market believing in the future growth of the company. Since the cost of equity decreases as it becomes more demanded, companies should take advantage and raise more equity for new projects. As well as that, Myers (1977) argues that firms which already have a significant amount of leverage are more likely to not undertake projects which represent growth opportunities. This phenomenon is better known as precautionary savings, since companies pass up on potentially positive NPV projects when uncertainty about future income is high.3
As well as that, according to Rajan and Zingales (1995), the more tangible a firm’s assets are, the more debt that firm is capable of bearing, since tangible assets are able to serve as collateral. Therefore, that decreases the cost of debt for the lender, leading to an increased willingness to lend to the firm in question.
The relationship between actual growth and leverage is described best by Lang et al (1996), who find a negative relationship between the two variables irrespective of firm size. The reasoning behind this is that managers hold information about the growth opportunities of a company, and choose to reduce leverage if growth opportunities are high. As well as that, seeing actual growth numbers match their expectations, that enforces their views of being able to forecast firm growth. Therefore, growth and growth opportunities work in the same direction
3 Weil (1993), Friedman (1957)
8
when predicting leverage, actual growth being a confirmation of managers’ beliefs on growth opportunities.
As far as profitability is concerned, Castro et al (2016) argue that increased profitability should drive firms to borrow more, since it reduces bankruptcy costs as well as providing a more effective way to use tax shields. Due to higher profitability, lower information asymmetry serves as a reason for lower cost debt issuances. However, increased profitability should also increase a firm’s market value of equity, since investors can expect more return on their investment in the firm’s capital (Varaiya et al, 1987). Therefore, if market leverage is analysed, as opposed to book leverage, the relationship between the two variables becomes ambiguous.
However, the above mentioned methodologies do not help explain much of the variation in leverage ratios. Lemmon et al (2008) find that the majority of variation in leverage ratios is determined by an unobservable factor. These findings clearly point in the direction of a wider, more economy level based analysis of capital structure decisions.
2.2 Monetary Policy Effects
Researchers have started to slowly uncover the factors that are not observable at the firm level. The relationship between monetary policy and capital structure is still in its early stages.
Some attempts have been made at finding the exact link between the two variables, however little literature exists on the subject.
In order to understand the borrowing behavior of firms, the relationship between monetary policy and borrowing on a larger scale have to be understood. Gertler and Gilchrist (1993) analyse the role of market imperfections on monetary transmission channels. They theorise that the gap between the cost of funding with internal versus uncollateralized external sources is driven by the creditworthiness of a firm. However, this credit worthiness is impacted by the economic conditions of the time. Therefore, monetary policy disturbances are able to influence this gap and, consequently, leverage ratios of firms. Their main findings are that there exist large differences in borrowing responses between small and large borrowers after changes in monetary policy. As well as that, small borrowers are impacted most, with a decrease in
9
borrowing being observed after a period of tight money. Furthermore, they find the ratio of loans to sales rises sharply for large firms after a period of tight money, while staying constant for small firms. This happens due to large firms being able to borrow in order to smoothen out the impact of decreased sales in times of economic turmoil.
Adding to these findings, a study realised by Angeloni et al (2003) follows the meeting of the ECB and eurozone National Central Banks where findings of the 4 year long study on the monetary transmission mechanisms within the eurozone. The main findings of researchers from the european national banks reveal that in most european countries, long run effects of increases in interest rates reduce loan growth, mostly due to supply shifts.
2.3 The introduction of Negative Interest Rates by the ECB
The Eurozone countries have experienced a shock which is unique in every way. The introduction of negative interest rates by the European Central Bank has had effects which are still analysed today and are still under hot debate. The traditional monetary transmission channels dictate that with a lower interest rate, lending and borrowing have to be stimulated.
However, the impact of these policies is unclear.
Some academics argue that the impact of these interest rates were not at all as positive as previously thought. Heider et al (2019) find that these interest rates have produced a challenging environment for banks to create loans. Their argument is that due to less deposits being made, there is also less money to lend, and therefore banks are left with no choice but to seek higher risk investments instead.
This argument is reinforced by Bech and Malkhozov (2016), who show that the lack of transmission of these policy rates have created a disparity between banks’ assets and liabilities.
This has dropped volumes across all maturities since 2014, due to increasing excess of liquidity within the banking sector. However, they find an overall weak impact of the negative interest rates on money markets.
10 2.4 Macroeconomic and Fixed Effects
Macroeconomic conditions also play a crucial role in determining firms’ capital structures. Choe et al (1993) find that firms have the tendency to increase equity during expansionary cycles, while interest rate variables are generally insignificant. This effect was found due to more investment opportunities arising in times of economic upturn combined with higher returns on equity, leading to lower costs for raising capital through equity issues than debt. In addition to this finding, Korajczyk and Levy (2003) find that target leverage is counter- cyclical for financially unconstrained firms, while this relationship is pro-cyclical for constrained firms. This pro-cyclicality is driven by reduced costs of bankruptcy in times of economic boom, which lead to firms lending more in order to increase their tax shield. Furthermore, they find a lack of importance for macroeconomic conditions for constrained firms, while unconstrained firms seem to be more affected by these conditions. Already, a decision process separation can be observed between types of firms.
According to Friedman (1986), when governments have to borrow in order to finance large deficits, less is available for firms to borrow. Demirci et al (2019) also find a significant negative relationship between government debt and corporate leverage, indicating that
government deficits do ideed put pressure on the debt market, which becomes less accessible for companies. With the years following the 2009 financial crisis being characterised by large government deficits across the continent, leverage ratios could be affected by increased government borrowing, since that should increase the cost of debt due to the rise in demand.
Furthermore, Demirgüç-Kunt and Maksimovic (1999) were able to find that inflation has a significant negative effect on the use of long term debt. This paves the path to understanding the relationship between the stability of the macroeconomic environment and firm leverage.
Booth et al (2001) further enforce this view by showing that firms are more likely to borrow when expecting real growth, but not when growth is due to inflationary periods. However, the effects of inflation on both total debt and long term debt were insignificant in the regressions, due to high standard errors, which were a result of small sample size. Nevertheless, the literature on the effects of inflation on capital structure is lacking and inconclusive. Further investigating
11
the potential effects of inflation could pave the way to a better understanding of firms’ borrowing behavior.
As well as that, Giannetti (2003) provides a clear image of why country factors matter. In their study, it is found that firms from countries with good creditor protection are more likely to take on debt, due to the ease that these protections bring to the lending system. These effects are especially relevant for firms in sectors with highly volatile returns. Similar findings are also found by Alves and Ferreira (2011), with the addition that these effects are similar no matter the country.
Fan et al. (2011) add to these findings by showing the importance of the quality of tax system and the presence of corruption for firm’s debt decisions. Unsurprisingly, they find that countries with a stable taxation system and low corruption level tend to have a more active debt market. These findings are supported by earlier empirical works of Graham (1996), which also link tax shields to capital structure choice. They find a positive relationship between the marginal tax rate on corporate profits and debt issuance. The reasoning behind this finding is that a
company will use nondebt tax shields as a way to reduce its marginal tax rate. Therefore, higher corporate tax rates should lead to more borrowing.
Adding to this research, literature already exists on the effects of industry fixed effects on firm borrowing behavior. However, this part of the literature is also in its incipient stages, as found by Kumar et al (2017). To exemplify the early stages of this research, Koralun-Bereźnicka (2018) find significant industry effects on debt maturity choice, as well as country effects for firms in 11 European Union countries.
All these aspects create an opportunity to test and rectify the findings of previous research regarding capital structure decisions of firms. Putting together a comprehensive model that uses all previous findings could be a step in the right direction by filling the gap in the literature with regards to the effects of macroeconomic, industry and monetary policy effects on firm leverage. Furthermore, there is no alternative to finding how much of the variation in leverage ratios can be explained by macroeconomic and monetary policy variables. A model that unifies all previous findings will likely be able to find the predictive power of the variables at
12
hand. Similarly, a study of this nature could help to create an economic model that aids the prediction of the borrowing behavior of firms once differing policies are adopted.
One additional expansion that this paper brings to the literature is the analysis of macroeconomic variables while separating firms based on financially constrained and
unconstrained status. It is expected that unconstrained firms are more responsive to changes in the economic environment, as that attracts or deflects investment in their equity4. However, constrained companies are not likely to attract as much investment in expansionary cycles, due to their general unattractiveness to investors. Therefore, it can be expected that macroeconomic variables will have different impacts on financially constrained and unconstrained firms.
Such a study has already been attempted by Graham et al (2015), who also identified this gap. However, this study is based on public US companies, which leaves a gap in the literature for European firms, as well as other geographical areas. The author of this study also takes a wide time frame in order to analyse these factors (1920-2010). However, technological advances alone could create a discrepancy in the found effects for different time periods. Therefore, the present study will try to analyse a shorter time frame which is centered around the introduction of negative interest rates. As well as that, the main finding of Graham et al (2015) is that in the long run, firm specific variables do not help predict leverage ratios, while macroeconomic wide variables do. The present study will attempt to show that macroeconomic variables as well as firm specific variables are able to predict leverage ratios in the short run, therefore completing another gap which exists in the literature.
3. Data & Methodology
The data is comprised of yearly accounts from european firms’ annual reports, gathered from Compustat’s Global Annual Database. Firms were omitted from the sample if they were from the following industries, due to lack of available data: Transportation, Communications, Electric, Gas and Sanitary service (SIC codes 4000-4999), Finance, Insurance and Real Estate (SIC codes 6000-6999) and Non Classifiable companies (SIC codes 9900-9999). Similarly, firms
4 Lucas (1978), Baker et al (2005), Chang & Dasgupta (2002)
13
with missing total assets, sales, intangible assets, book equity or common share counts were dropped from the sample. Furthermore, Compustat Daily was used in order to source stock prices for each company. The data includes 1800 companies from 25 european countries over 10 years (2009-2019).5 Only firms whose data was available for the entire time frame, containing no gaps in years, were kept in the sample.
Macroeconomic data was obtained from two main sources, namely Eurostat and the World Bank Development Index. This data includes total GDP for each year, debt to GDP ratios, inflation, as well as corporate tax rates. In order to obtain data on policy rates for all countries, manual collection is required from each individual Central Bank’s website. Such a selection has already been compiled by the Bank of International Settlements.6 In the case of corporate tax rates, Eurostat holds data for all the countries in the sample, excluding Romania and Cyprus.
This data has been obtained individually from the Romanian National Agency for Fiscal Administration (ANAF) and the Inland Revenue Department of the Republic of Cyprus.
The time frame mentioned was chosen based on two main considerations. The first is a need of as large a time frame as possible in order to properly capture time fixed effects and the impact of monetary policy. Secondly, the results would be affected if the impact of the Covid-19 pandemic was present in the sample. Therefore, the time frame is ended in 2019, when the virus first entered Europe and the governments imposed restrictions on business practices. Including the year 2020 in the sample would lead to an exogenous shock other than the sovereign debt crisis having an impact on the data, hence the results and both the internal and external validity of the paper.
The methodology used for this study is largely similar to that of Graham et al (2015).
However, their study benefited from the advantage that the American economic system brings to analysing firms from across the country. On the other hand, Europe is a continent made up of sovereign states. These states are free to take different decisions regarding monetary and fiscal
5 Austria, Belgium, Switzerland, Cyprus, Germany, Denmark, Spain, Estonia, Finland, France, Great Britain, Greece, Hungary, Ireland, Italy, Lithuania, Luxembourg, Latvia, The Netherlands, Norway, Poland, Portugal, Romania, Slovenia, Sweden.
6 https://www.bis.org/statistics/cbpol.htm
14
policy and hence, a method has to be developed for finding which states are similar enough to be part of the same sample.
In the case of the euro zone, the question is easily answered by the governance of all encompassed states by the ECB, who set monetary policy. However, central banks of states outside the euro zone have been shown to import monetary policy directly from the ECB. This is primarily driven by the increased euroization of loans in countries adjacent to the euro zone (Brown & Stix, 2015). They find that in the neighbouring areas of the euro zone, increasingly more loans are written in euro, with countries seeing as much as half of all loans in euro. As well as that, Brzoza-Brzezina et al (2010) find a strong relationship between policy changes by the ECB and euro borrowing in economies close to the euro zone. This indicates a direct impact of the monetary policy set by the ECB on other countries.
Hence, this paper assumes that the ECB influences monetary policy of foreign central banks if a very high correlation between policy rates can be observed between itself and aforementioned central banks. In order to ensure that the sample consists only in the strongest matches, correlations of 0.8 or more will be considered strong. This constraint is added in order to minimise any potential effects that might bias the regression results.
Similarly, studies conducted in the context of the United States economy benefit from the lack of need to compensate for country effects. Hence, wide time frames can be used, such as in the paper of Graham et al (2015). In a european setting, this becomes more difficult, as a
sufficiently large number of firm-year observations need to be collected for each individual country in order to be able to estimate such effects. In consequence, this study is limited to using a more constrained data frame, in order to try and maximise the number of firms with available data for each country in the sample. However, this imposes limitations on the significance of effects of variables such as the corporate tax rate, due to its small and rare movements in developed economies. This limitation is further discussed in the hypotheses formation.
In the next section, data panel regressions will be conducted. The majority of literature on the subject of capital structure uses data panel regressions, as does the study upon which this methodology is based (Graham et al, 2015). This study will follow a similar approach and include firm and year fixed effects. Country fixed effects are not needed, as they would cancel
15
out the effect that the macroeconomic variables are producing. However, it is important to state that simple OLS regressions cannot be used, as the standard errors they produce would be biased.
Therefore, standard errors will be clustered at the industry and firm level in order to account for different borrowing practices of separate industries and differing lending environments of different countries.7 Furthermore, Petersen (2009) finds that data panel standard errors will only be unbiased if both firm and year fixed effects are used. The way these have to be implemented is by creating dummy variables for each individual year, as well as clustering at both the industry and firm levels. These additions will ensure that OLS coefficients are not biased, as Petersen (2009) explains would happen in the absence of these clustered standard errors and fixed effects.
In order to account for financial constraints, the study will split the sample into
financially constrained and financially unconstrained firms. This dummy variable is calculated according to the theory of Fazzarri et al (1988). They argue that financially constrained firms have a payout ratio of less than 10% of Net Income for any of the past 10 years. By this
definition, approximately 67% of firms in this study’s sample would be financially constrained.
This includes companies which are in high growth stages and therefore do not pay dividends.
Hence, this study adds an additional constraint that ensures fast growing companies do not get classified as financially constrained. The additional constraint is a maximum asset growth of less than 3% for any of the years in the sample. This results in a 32,2% share of firms that are
financially constrained.
The basic hypotheses for these regressions are that all factors, be they firm specific or economy-wide are significant predictors of leverage ratios. A simple regression equation looks as follows:
𝐿𝑒𝑣𝑖𝑡 = 𝛽1∗ 𝑀𝑇𝐵𝑖𝑡+ 𝛽2∗ 𝐿𝑛(𝐴𝑇)𝑖𝑡+ 𝛽3∗ 𝑇𝑎𝑛𝑖𝑡+ 𝛽4∗ ∆𝐴𝑇𝑖𝑡 + 𝛽5∗ 𝑃𝑅𝑂𝐹𝑖𝑡+ 𝛽6∗ ∆𝐺𝐷𝑃𝑖𝑡+ 𝛽7∗ 𝐺𝑉𝑇𝑑𝑒𝑏𝑡𝑖𝑡+ 𝛽8∗ 𝜋𝑖𝑡+ 𝛽9∗ 𝑃𝑜𝑙𝑖𝑡+ 𝛽10∗ 𝐶𝑇𝑅𝑖𝑡+ 𝛽11∗ 𝐹𝑖𝑛𝐶𝑜𝑛𝑖𝑡
Formalised, the hypotheses are as follows:
𝐻0: 𝛽𝑖 = 0 ; 𝐻1: 𝛽𝑖 ≠ 0 ∀ 𝑖 ∈ [1; 11]
7Koralun-Bereźnicka, J. (2018), Kumar et al (2017)
16
The regression firm specific variables’ definitions are as follows: Market Leverage is used as a definition for Leverage. This is calculated as Total Debt (Compustat item 9 + Compustat item 34) divided by Total Debt plus Market Value of Equity (Compustat item 24 times Compustat item 25). The Market to Book ratio (MTB) is calculated as the Market Value of Equity divided by the Book Value of Equity (Compustat item 60). Company size is denoted by the natural logarithm of Total Assets (Compustat item 6). Tangibility of assets is calculated as the difference between Total Assets and Intangible Assets (Compustat item 33) divided by total assets. Profitability is defined as the ratio of EBITDA to Total Assets. All firm specific variables are cleaned of outliers, with the largest (or lowest) observations as well as their corresponding firms being excluded from the sample. As well as that, the data is winsorized using cuts of 0.5%
on either side, in order to further remove outliers.
As well as that, economic and monetary policy variables are pre-defined by the institutions which created the datasets. GDP growth is given by the difference in yearly GDP divided by the previous year’s GDP (World Bank National Accounts Data). The GDP accounts data is denominated in constant 2010 US dollars. Government debt is defined as the percentage of GDP that is held in public debt (World Bank Development Index). Inflation (𝜋) is based on the Consumer Price Index and is calculated and provided by Eurostat. Policy Rates are sourced from each country’s Central Bank by the Bank of International Settlements. Finally, Corporate Tax Rate data is compiled by the OECD and is defined as the nominal tax rate that a company has to pay on each dollar of earnings before taxes each year.
Based on the theories provided in the theoretical framework of the paper, the expected effects of these variables are as follows: the Market to Book ratio is expected to have signficantly negative correlation with leverage ratios. Hence, a firm with more growth opportunities is
expected to make less use of debt in comparison to equity, since its’ equity is highly valued.
However, the effect should be smaller in size for financially constrained firms, due to lower access to debt issuance.
Firm size is expected to be positively correlated with leverage ratios, due to lower cost of bankruptcy that results from the decreased likelihood of default. Asset growth should negatively affect leverage ratios, similarly to growth opportunities, since actual growth will attract equity
17
demand and therefore ease the access to equity capital. There is no reason to presume that financially constrained firms will behave differently to unconstrained companies regarding this relationship, therefore the hypothesis being the same for both groups.
Tangibility will likely positively correlate with leverage ratios for unconstrained companies, while constrained companies are hypothesised see a negative relationship. This is due to financially constrained companies likely having already liquidized tangible assets in order to ease funding constraints. Meanwhile, unconstrained companies are able to keep tangible assets as collateral. However, it is unclear what proportion of these assets are indeed used as collateral.
Therefore, the relationship could likely be insignificant for unconstrained firms.
The main variables of interest will be the macroeconomic and monetary policy variables, since this is one of the main areas of novelty that the paper brings to the existing literature. This is due to their aggregation in one unified model, as opposed to investigating their impact on capital structure independently.
Expansionary cycles have previously been found to negatively coincide with leverage8. This is also expected to occur in the sample at hand. However, there might be a difference in significance between responses of financially constrained and unconstrained companies. This is due to the general inability of financially constrained firms to receive funding for new projects.
Hence, the state of the economy should not play a significant role in the easing of funding constraints.
Government debt is also expected to be negatively correlated with leverage ratios for unconstrained companies, due to increased demand for debt in times of increased budget deficits.
Due to the same reasons mentioned previously, it is expected that there might be a difference in significance in results between constrained and unconstrained companies.
Inflation has previously been shown to have negative but insignificant effects on
borrowing, therefore leading to these effects being this paper’s prediction as well. However, no separation was previously made between unconstrained and constrained companies. Hence, the
8 Choe et al (1993), Lemmon et al (2008)
18
present study attempts to explore these potential differences, with inflation being expected to be significant only for the unconstrained sample.
Under the assumption that monetary transmission channels work normally, policy rates should be negatively correlated with leverage ratios, due to the increased borrowing response created by lowering policy rates and hence decreasing overall costs of debt. However, previous studies have found insignificant regression coefficients for interest rates due to small sample sizes (Demirgüç-Kunt & Maksimovic, 1999). This paper increases firm sample size to 1800 firms, in an attempt to find whether these results change in significance. However, it is hypothesized that the results will be similar to those of previous literature.
On the other hand, corporate tax rates are expected to have an insignificant effect, due to small movements present in the sample. The maximum yearly increase is 2.5%, while the maximum yearly decrease is -0.1%. In the case of a long term study, the corporate tax rate could potentially generate significant coefficients. However, this data set is limited by a relatively short time frame9 and developed economies which do not typically experience large shifts in tax rates.
It is important to note that macroeconomic variables can influence each other. In order to make sure that these effects are not present, and therefore do not bias the results, the paper presents evidence from Barro (1995). The author finds that effects of inflation on GDP growth are only significant in periods of high inflation, which are due to monetary policy which
stimulates inflation. The average yearly inflation that would lead to such effects is approximately 10% yearly. These inflation numbers are not present in the current sample, the maximum yearly inflation being 6.09%.
Another concern is the effect of the Corporate Tax Rate on GDP growth. Existing literature indicates that the effects of the corporate tax rate on GDP growth are negative and significant (Lee & Gordon, 2005). However, the shifts necessary to distort economic growth are large. The authors report that a 10% drop in corporate tax rates is necessary in order to stimulate economic growth by 1.1%. In the present sample, the maximum yearly decrease in corporate tax rates is -0.1%, equating to a potential increase in GDP of 0.011%. The maximum increase in
9 Papers such as Graham et al (2015) expand time frames to as much as 90 years (1920-2010)
19
corporate tax rates is 2.5%, equating to a potential decrease in GDP of 0.275%. These effects are largely negligible, since the macroeconomic variables do not present large movements in mature or advanced developing, such as the european ones in the present sample.
4. Results & Discussion
4.1 Policy Rate Correlations
This section analyses the results derived from the pairwise correlation analysis of policy rates from the Euro area and its neighbouring economies. These results are available in Appendix 1. The main takeaway from this analysis is that it confirms the findings of Brown and Stix (2015) and Brzoza-Brzezina et al (2010), according to which neighbouring economies import monetary policy directly from the Euro Area. Between 2006 and 2020, all countries in the sample present highly significant correlations in policy rates. However, Croatia’s policy rates only have a correlation of 0.495 with the Euro area’s, which is the lowest rate obtained. All the other correlation rates are higher than 0.8, showing very strong correlations. The methodology proposed attempts to use countries which directly absorb policy from the ECB. In the event of including countries for which correlations are not as high as possible, there is a possibility that other effects may be captured as well. Any regression coefficients that represent the impact of policy rates on leverage may then be biased, since they include other unobserved effects.
Therefore, Croatia will be omitted from the sample since the intention is to apply as strict a requirement as possible when confronted with the question of inclusion in the sample.
Similarly, even though the Czech Republic’s policy rates are highly correlated with the ECB rate, the sample size of Czech firms is too small to be included, with only two firms satisfying the conditions mentioned in section 3 of the paper. The same has to be applied to Slovakia, where only one firm was able to meet the requirements. This step has to be taken due to the impossibility of calculating macroeconomic effects when such a small sample is in play.
The Czech and Slovak observations are automatically dropped when the regressions are ran, therefore there is no point in holding on to these observations.
20
A question that this paper is unable to answer is whether the monetary transmission channels work the same way in all the countries mentioned in this study. Even though policy rates are highly correlated, there still is room to question the similarity in their effects on borrowing at large and on firms’ leverage ratios. Hopefully, the macroeconomic variables will capture some of these effects, and account for the lack of data on monetary transmission channels.
Another important note refers to what the reported correlations between countries interest rates actually capture. Intuitively, they mainly capture transmissions of monetary policy from the central bank of the most influential economy of the area to central banks of the surrounding economies. However, there might be a number of other factors which are included in these correlations and are worth further exploring. Unfortunately, that is outside the scope of this paper and constitutes a limitation of the model presented. Hence, it serves as a direction for future research to further expand on this issue and perhaps find more robust ways to analyse the transmission of policy from the ECB to other european central banks.
4.2 Summary Statistics
This next section will look into summary statistics of firms from the entire sample, as well as the financially constrained and unconstrained samples. The main points of interest are potential differences which are observable from such early analysis between the groups of firms described.
Firstly, large movements within the variables signal that some effects are occurring in the short run of the study’s time frame. Alternatively, these movements show a different borrowing behavior for firms of different types. For instance, Market to Book ratios for the entire sample average at 2.536, with a standard deviation of 3.505 (Table 1). Similarly, leverage ratios average 0.243, the standard deviation being 0.228. Such variability in these variables signal a non
stagnant environment where policy could have the same effect over all types of firms. In order to understand the impact of policy change on firm’s borrowing behavior, it is helpful to observe non symmetric responses and behaviors. It is also important to note that out of 18,000
21
observations, 1662 contain market leverage ratios equal to zero. Had this number been too great (50% or more of observations), there would likely not be enough non zero observations to produce unbiased results, especially for the country specific variables such as GDP growth or Policy rates.
Table 1. Descriptive Statistics Entire Sample
The table includes summary statistics of company specific variables for the entire sample, representing 1800 firms over 10 years (2010-2019). The variables presented follow the definitions established in Section 3 of the paper. The variables are also cleared of outliers and further winsorized with cuts of 0.5%.
Variable Obs. Mean Std. Dev. Min Max
Market to Book Ratio 18000 2.536 3.505 0.104 33.02
Log(Assets) 18000 6.062 2.211 1.102 12.018
Market Leverage 18000 0.243 0.228 0 0.927
Tangibility 18000 0.488 0.222 0.001 0.99
Asset Growth 18000 0.039 0.162 -0.843 0.603
Log(Sales) 17999 5.951 2.286 -3.124 15.524
Profitability 17998 0.096 0.126 -2.167 3.177
A table of correlations is also presented in Appendix 3. The table correlates firms specific variables for the entire sample and shows significance levels for the correlations between
brackets. The most important aspect when considering such analyses is to ensure there are no correlations close to 1 or -1 present in variables which will be used as part of a regression model.
Such a correlation between variables would lead to high R squared values being most likely attributed to multicollinearity. However, this is not the case, with the highest correlation being - 0.324 between the Market to Book ratio and Market Leverage. A high correlation is observed between Book and Market Leverage, however that is due to the numerator of both variables being Total Debt. Hence, such a result is expectable.
Furthermore, two variables of interest and their correlation are the natural logarithm of total assets and asset growth. A correlation of 0.097 is observed between the two variables, with
22
a significance level of under 1%. A positive but low correlation is indicative of faster growth in smaller companies, while larger companies experience lower growth rates. This is in line with expectations, since companies that are biggest in size are expected to have reached maturity or near maturity.
Another possible concern is the similarity between growth opportunities and actual growth of assets. If the theories mentioned in section 2 of the paper are correct, managers’ beliefs about growth opportunities should also match the actual growth numbers. However, there is a correlation of only 0.068 between the two variables, indicating that perhaps these beliefs are not as consistent as theory might suggest. Hence, any concern that the two variables might produce multiple collinearity or capture the same effects should be neglected.
Following the analysis of the entire sample, the differences between financially
unconstrained and constrained companies are investigated. The tables in Appendix 2 and 3 are revealing to some of the differences between the two types of companies. Firstly, unconstrained companies show a lower market leverage ratio than constrained companies. This might be indicative of financially constrained firms having difficulty in clearing debt from their balance sheets, while unconstrained firms can set leverage ratios only slightly lower and still be able to pay out dividends. Otherwise, this difference might be due to the difference in market evaluation of constrained and unconstrained companies, since there is reason to believe that the market prefers investing in unconstrained companies, hence driving their market value of equity up.
As well as that, the largest differences between the two groups can be observed in
profitability and the ratio of sales to turnover. Unsurprisingly, unconstrained firms are more than double as profitable as their constrained counterparts, as well as seeing a large gap between the logarithm of sales. These factors are expected to be main determinants of leverage, alongside growth opportunities, since they determine how much debt a firm is able to repay at the end of the year. Furthermore, these factors are expected to predict market values of equity, since more profitable firms as well as firms that have the biggest sales are expected to be preferred by the market.
The difference between profitability in constrained and unconstrained companies is also
23
visible when plotting yearly average profitability ratios. While constrained firms keep a somewhat constant profitability ratio during the time frame, with the exception of 2012, unconstrained firms experience a drop of approximately 1% which was not recovered from. In spite of this, movements seem to be similar in terms of peaks and lows for both types of firms.
This indicates a systemic impact, rather than firm specific characteristics being the main drivers of these movements.
Figure 1. Average Profitability Constrained Figure 2. Average Profitability Unconstrained
However, some similarities in capital structure behavior can be observed when plotting Leverage ratios against time for unconstrained and constrained firms. Both types of firms experienced a linear and somewhat uninterrupted reduction in leverage ratios between 2011 and 2017, only to see a sharp increase in the two following years, leading up to the Covid-19
pandemic. This type of sudden shift in borrowing behavior is indicative of a potential exogenous shock which affected the economic environment. Alternatively, these movements could be the result of increasing market values of equity, while debt numbers were kept largely unchanged.
The possibility of a shock in 2017 is enforced by the sudden drop in growth opportunities for both unconstrained and constrained firms. As well as that, a negative relationship between growth opportunities and leverage can be observed, the graphs showing a near mirror image of each other when visually compared. When leverage increases, the Market to Book ratio
decreases and vice versa, a trend which is unchanged for nearly every year in the sample.
24
Figure 3. Average MTB ratio Constrained Figure 4. Average MTB ratio Unconstrained
Figure 5. Average Leverage Ratios Constrained Figure 6. Average Leverage Ratios Unconstrained
Further investigating total debt and market value of equity graphs tells a clearer story.
While leverage ratios are similar, their components’ progressions differ radically. Constrained firms’ total debt decreases rapidly from 2011 until 2015, only starting to rise again from 2017 onwards. In contrast, Unconstrained firms present a much smoother borrowing trajectory, with a stabilisation of total debt between 2015 and 2017, after which the upwards trend resumes.
Differences of the same magnitude are also present when investigating market equity values. Constrained firms saw a halving in market equity from 2010 until 2012, with a slow recovery until 2014. Afterwards, another decrease ensued until 2018, with a rapid recovery afterwards. These trends could be due to the market responding to the firms’ financial constraints, as all the firms in the sample were active for the 10 years. When looking at
25
unconstrained companies, the years following the 2008 crisis are stagnant, with a fivefold increase of market equity between 2014 and 2016, after which a rapid decrease is again noted.
Such different responses in terms of market equity and debt issuance for both types of companies indeed indicates that the separation between financially constrained and unconstrained
companies is crucial to obtaining meaningful results. As well as that, it indicates that firms’
responses to economic environment changes differ significantly, even if leverage ratios may follow similar trajectories.
Figure 7. Average Total Debt Constrained Figure 8. Average Total Debt Unconstrained
Figure 9. Average Market Equity Constrained Figure 10. Average Market Equity Unconstrained
26 4.3 Regression Analysis
The main subsection of the results and discussion section analyses the main findings gathered from regression analyses. This section will reveal the main qualities and drawbacks of the empirical model devised, as well as potentially shedding some light on the effects of
economy wide variables’ effects on firms’ capital structures.
The first set of regressions, present in Table 3, are ran following the methodology described in section 3. The main conclusions which can be derived from the results are that the impact of firm specific characteristics is significant regardless of firm type. Both financially constrained and financially unconstrained companies’ characteristics are significant predictors of leverage ratios. However, as predicted, financially constrained companies present insignificant coefficients for most non firm specific variables. This confirms the predictions made on the basis of previous literature’s findings.
When analysing the Market to Book ratio, it is clear that the results are in line with previous literature’s findings, as well as with the present paper’s hypotheses. Both constrained and unconstrained firms present a significant negative relationship between growth opportunities and leverage. This indicates that for both types of firms, when the market believes that these firms have good opportunities to launch profitable projects, the cost of equity decreases, hence encouraging firms to take on less debt for these projects. One additional step in investigating this relationship could be the separation of the sample into high MTB and low MTB ratio firms, however that is beyond the scope of this paper.
Similarly, company size, denoted by the logarithm of total assets, is significantly and positively correlated with leverage ratios. This is also in line with previous findings and supports the main hypotheses regarding both unconstrained and constrained firms. Larger firms are therefore able to use more leverage in order to fund new projects, since they are also less likely to default on their debt.
Profitability follows the same pattern, confirming the hypotheses made and the findings of previous literature. A negative and significant relationship is observed for all firms. This
27
indicates that companies which are able to generate more profit per sales are also able to attract more investment and therefore need less debt in order to fund new projects.
The main similarity which can be observed between the three above mentioned variables and their effects on leverage is that unconstrained firms see larger effects than constrained firms for all of the variables. This might be due to the larger variability observed in the leverage ratios of unconstrained versus constrained firms (Appendix 2). Alternatively, financially constrained firms may give more importance to other markers. This is the more likely answer, since fast growing firms are excluded from the financially constrained group. Since these firms are expected to have higher growth opportunities, as well as increased profitability compared to constrained companies, their leverage ratios are expected to be more reliant upon their performance in these metrics.
Similarly to growth opportunities, actual growth was expected to be negatively correlated with leverage ratios. The results confirm this hypothesis, as the effects are negative and similar in size between all three samples. This is due to equity investment being attracted by both firm types when the market sees actual growth occur. However, there is a slight difference in significance level, since the coefficient is only significant at the 10% level for unconstrained companies, while it is significant at the 5% level for constrained companies. This might simply be due to unconstrained firms attracting more investment regardless of their asset growth, since higher payout ratios are intended for that specific purpose. Hence, asset growth holds less significance for unconstrained companies than it does for constrained companies.
Tangibility is the first variable which sees a large difference in prediction of leverage ratios between unconstrained and constrained firms. While financially constrained firms see a negative and significant relationship, the financially unconstrained sample returns a coefficient which is insignificant. This confirms the hypotheses made in section 3, further enforcing the possibility of firms with more tangible assets not using all of these assets as collateral.
Meanwhile, constrained companies are likely making use of their tangibility in order to gain access to capital.
28
Table 2. Regression Analysis of Market Leverage
The table presents regression results for the full sample containing 18000 firm-year observations, as well as the financially constrained and financially unconstrained samples. The coefficients are reported under each column according to their variables.
Significance levels are denoted by stars, the significance levels being explained below the table. R-squared values are calculated after the incorporation of firm and year fixed effects, as well as clustering of standard errors at the level of firms and industries.
Market Leverage Entire Sample Financially Constrained Financially Unconstrained Market to Book -0.0066*** -0.0034*** -0.0079***
Log(Total Assets) 0.0681*** 0.0621*** 0.0738***
Profitability -0.2653*** -0.1605*** -0.3792***
Asset Growth -0.0305* -0.0395** -0.0313*
Tangibility -0.0460** -0.0916*** 0.0018 GDP growth -0.3807*** -0.0088 -0.5366***
GVT debt -0.0003*** -0.0004 -0.0003**
Inflation -0.0057*** -0.3777 -0.5258***
Policy Rate -0.0105 -0.0190** -0.0045 Corporate Tax -0.0134 0.0335 -0.0223 Intercept -0.0218 0.0906 -0.1079
𝑅2 0.8205 0.8365 0.8112
Number of Obs. 18,000 5,800 12,200
Firm & Year FE Yes Yes Yes
Statistical significance levels: *** p<0.01, ** p<0.05, * p<0.1
Turning to economy wide variables, the difference in response between types of firms becomes apparent. When considering the results of the entire sample, GDP growth is a
significant and negative predictor of leverage ratios. However, unconstrained firms are the cause of that relationship. On the other hand, constrained companies’ insignificant coefficient indicates that regardless of a recession or a boom, their cost of equity in relationship to the cost of debt does not differ significantly. This might be due to the market preferring to invest in
unconstrained companies in times of economic boom, since the payouts are larger, as well as actual growth numbers, hence expecting higher returns on investment. Another possibility is that regardless of the economy’s state, financially constrained companies do not have the funding
29
resources to undertake new projects. Hence, their leverage ratios are not dependent on the economic environment as much as they are on firm characteristics.
Similarly, government debt returns negative coefficients for all samples, confirming the hypothesis established in section 3. However, the main point of interest is the difference in significance between samples, given that coefficients only differ by 0.0001. In spite of a smaller coefficient for the full and unconstrained samples, the constrained sample coefficient is the only insignificant one. However, the results confirm the theory according to which constrained firms do not respond as significantly to changes in economy-wide variables. Regardless, the full sample confirms previous findings according to which most firms negatively adjust leverage ratios to changes in government debt levels. This is, as shown, mainly driven by the
unconstrained sample of firms.
Inflation follows the same pattern as the previous two variables, with negative coefficients confirming the hypotheses established. The results also confirm that previous insignificance found when regressing borrowing on inflation was due to non separation of samples into financially constrained and unconstrained firms. This conclusion stems from the highly significant coefficient of inflation when analysing the unconstrained sample, indicating a negative borrowing response when inflation rises. Hence, unconstrained companies prefer to increase equity capital in relation to debt capital in periods of higher inflation, possibly due to the market’s desire for increased returns in order to compensate for inflation.
Building on the previous findings of Demirgüç-Kunt and Maksimovic (1999),
significance levels for policy rates do not seem to become significant for the entire sample, even after increasing sample size to 18000 observations belonging to 1800 firms. However, for
constrained companies, the predicted negative relationship does hold. This possibly signifies that a drop in policy rates does coincide with loan take up for these firms, due to easing of costs for capital raising. Constrained firms may see such movements in policy rates as an opportunity to attempt climbing out of the financially constrained status. Meanwhile, the results keep
insignificant for unconstrained firms. This does not come as a surprise, since unconstrained companies do not primarily rely on interest rates’ movements in order to raise capital. This finding has important consequences for policy makers as well. The separation between types of
30
firms reveals a localized impact of changes in policy rates in developed economies such as european ones, and contributes to better understanding the effect of the ECB’s policy rate impact on borrowing behavior of companies.
The last variable of interest is the corporate tax rate. The hypothesis of the paper regarding the reduced time frame and its’ consequence on the significance of tax rates’
coefficients on leverage ratios is confirmed. Although previous long term studies show significant effects of this variable (Graham et al, 2015), the predictive power of tax rates is insignificant when only 10 years are present in the sample. Therefore, in order to better estimate such effects, longer term studies of capital structure are necessary. These findings are,
nevertheless, of importance to policy makers, since they show that the impact of corporate tax rates on capital structure decisions is not immediate, rather taking effect in the long run. In the event of more abrupt shifts in tax rates, however, the effects might differ.
One of the main advantages of these models is also the explanatory power that they were able to produce. The lowest R squared value is 0.8112, meaning that approximately 81% of variance in Market Leverage ratios was explained by the regressors and the fixed effects used.
To illustrate the predictive power of the fixed effects, a regression was ran without firm fixed effects, while using a Financially Constrained variable interacted with all firm specific variables (Appendix 5). Immediately, the discrepancies between the two models become clear. While the R squared value of the model is much smaller, at 0.2425, the significance levels have also changed. All variables used in the previous models are now significant, with the exception of profitability. This confirms the findings of Petersen (2009), who argues that in the event of misuse of either fixed effects and clustering of standard errors, the coefficients’ significance levels will be incorrect, as long as the researcher knows that there exist both a time and firm effect.
Furthermore, this paper uses a set of regressions similar to the main set as a robustness check. The only difference between the two sets of regressions is the use of Book Leverage as the dependent variable, as opposed to Market Leverage. Book leverage is simply defined as the total value of debt divided by the book value of equity (Compustat item 60). If the results of the main regression set are robust, the macroeconomic variables should not play a significant role in
31
the determination of Book Leverage. This is mainly due to their previously discussed impact on equity markets, which are the investment channel for public firms. Furthermore, public firms such as those present in the sample are mainly interested in the market value of their equity in relation to the debt they hold. Book equity is not expected to play a major role in their decision of capital structure, as it would for private firms. Hence, insignificant results for macroeconomic variables would be indicative of a localized effect of these variables on public firms’ capital structure decisions.
The regression results are present in Appendix 4, and indeed confirm the robustness of results obtained in the main regression analysis. As predicted, macroeconomic variables are insignificant when attempting to predict book leverage ratios. The only exception are policy rates, which are negatively correlated with book leverage and significant. This is likely due to policy rates having a direct effect on the debt that firms opt to use for their projects, therefore implying that monetary transmission channels might have an effect in the short term. Firm specific variables also see a shift in the estimated effects, with asset growth and profitability lacking significance in the results, due to the absence on market equity in the dependent variable.
4.4 Limitations and Suggestions for Future Research
This section will briefly cover the main limitations of this paper and some directions for future research that could be undertaken. Firstly, availability of data is always a concern for such studies. In the context of european firms, Compustat unfortunately does not provide as large a database as its United States Annual Fundamentals database. Hence, future studies into capital structures of European companies are advised to attempt finding a broader data source, as it could potentially increase sample size. This is especially relevant when attempting to estimate country effects for smaller countries, from which fewer public companies are expected to exist.
This study has already been confronted with this issue, since the Czech Republic and Slovakia had to be excluded from the sample due to insufficient observations.
32
Another issue that this study has confronted is the inability to separate firms into a sample from the southern european region10 from the rest of the sample. The impact of the sovereign debt crisis was more severe in these states than elsewhere in Europe, therefore possibly impacting firms’ borrowing behavior differently. The simple use of a dummy variable for southern european countries could be used, however separation into differing samples would be more useful in order not to affect standard errors of other variables. This study was unable to perform such a separation due to lacking sample size for financially constrained firms from that region. Hence, this could be a direction for future research into the subject of capital structure.
Secondly, there could be some use in analyzing the effects of lagged variables on leverage ratios. This paper only takes into account the short run effects of policy changes on firms’ capital structures. However, monetary policy transmission channels do not achieve their full effect in the short run (Egert & MacDonald, 2008). Hence, it might be useful to include lagged policy rates into such a model in order to account for the slow transmission of monetary policy into the economy, and this transition’s effects on firms’ capital structures.
Furthermore, the analysis of correlations between policy rates is perhaps incomplete, since there might be more effects at play than the direct import of monetary policy from the ECB. Further investigation into these effects is well mandated, and could serve towards a better understanding of the transmission of monetary policy from the ECB to other european central banks. As well as that, this could lead to an augmentation of the methodology used to determine predictors of capital structure for european firms, and therefore an improvement in policy makers’ expectations regarding the impact of their policy on debt markets.
5. Concluding Remarks
There are multiple reasons for which the model presented in this paper brings an improvement for the capital structure literature. Firstly, the use of both firm specific and macroeconomic variables in a short run analysis of leverage ratios proves to be successful in explaining a large portion of variation in the dependent variable. This was one of the main issues
10 Greece, Italy, Spain, Portugal
33
identified by the literature review of Kumar et al (2017), as well as by the previous work of Lemmon et al (2003), who found that the majority of variation within leverage ratios was unobservable and external to firm specific effects.
Secondly, the separation of the sample into financially constrained and unconstrained companies is also crucial to understanding the effects of economy wide variables on leverage ratios. As predicted by previous literature, the borrowing behavior of these groups of companies differs significantly, since financially constrained companies’ purpose is to become
unconstrained, in order to attract investment. This process is largely unaffected by the state of the economy, be it in a boom or a recession. The exception to this rule is the policy rate, which in the short term seems to affect borrowing practices of constrained firms, while unconstrained firms are unaffected. Such findings are of great use to policy makers, since the impact of their policy is localized. Furthermore, these findings could lead to a better application of policy since the impact is known to be focalized, as opposed to economy-wide.
