Goodwill & capital structure : panel data evidence on the ability of companies to collaterize on goodwill in a sample of large listed corporates headquartered in the EEA

(1)

Abstract

Recent studies, including but not limited to (Menkhoff, L., et al., 2010) and (Rampini, A., Viswanathan, S., 2013), focused on the ability of companies to collateralize tangible assets which are easily re-deployable and in general an acceptable mean of securing debt, usually favored by the lender. This research paper evaluates the extent to which companies which are headquartered in the EEA are able to collateralize goodwill assets, in order to attract additional leverage. In line with the only available study which treated a South African sample, I find a 15% positive relationship. However, the relation is much stronger which I attribute to the maturity of the European Capital markets compared to the South African. The proposed model is robust and consistent in alternative model proposals. Coefficients on the control variables are in line with previous findings. Additionally, a 0.94 adjustment speed to target leverage, whereby the adjustment speed is higher than found in previous studies. I conclude that the statistically significant relationship has to be further investigated on reversed causality. Further room for improvement lies within the usage of two-stage GMM regression for effect estimations together with testing alternative definitions of leverage.

___________________________________________________________________________________ This document is written by student Philipp Garack, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is 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.

___________________________________________________________________________________

Introduction

Mergers and acquisition (“M&A”) are prevailing within Europe despite the financial crisis. In 2014 the

total volume of M&A activity of listed corporates in Europe amounted to €2.2trn1, picking up

year-on-year, on the back of a slight dip in deals in FY09, driven by the uncertainty stemming from the systematic shock of the failure of Lehman Brothers. In fact the number and volume of M&A deals remained unaffected by the financial turmoil. An acquisition is usually pursued, as the acquirer is projecting to achieve synergies from i) the integration of the business and ii) the consolidation of the market, i.e. lower competition. These estimated synergies are further quantified and evaluated to arrive at the final bid offer, which includes a premium (which in theory should roughly account for projected future synergies). This

1_{https://imaa-institute.org/statistics-mergers-acquisitions/}

GOODWILL & CAPITAL STRUCTURE: PANEL

DATA EVIDENCE ON THE ABILITY OF

COMPANIES TO COLLATERIZE ON GOODWILL

IN A SAMPLE OF LARGE LISTED CORPORATES

HEADQUARTERED IN THE EEA

PHILIPP GARACK, UNIVERSITY OF AMSTERDAM, JULY 2016

(2)

price premium has to be accounted for on the Balance Sheet. Along IFRS3, goodwill is an account included in the Balance Sheet, whereby it is measured as the difference between “the aggregate of (i) the value of the consideration transferred (generally at fair value), (ii) the amount of any non-controlling interest, and (iii) in a business combination achieved in stages, the acquisition-date fair value of the acquirer's previously-held equity interest in the acquiree, and the net of the acquisition-date amounts of the identifiable assets acquired and the liabilities assumed (measured in accordance with IFRS 3)” (IFRS 3.32). Pricing goodwill remains a difficult task for lenders in general. However, for a given company it indeed creates value, which is derived from the ability to achieve expected synergies, subject to annual impairment testing. In effect, goodwill is an artificial asset, which can very quickly disappear from the balance sheet. Lenders traditionally demand collateral in form of tangible security predominantly due to the lower cost of assets re-deploy ability and the implying higher bargaining position at default. Companies are holding more and more intangible assets on their Balance Sheet. Therefore it is academically relevant to investigate the ability of large listed companies, headquartered in the EEA, to collateralize on goodwill, i.e. take on additional leverage, secured with reference to this non-tangible account. Companies can collateralize on goodwill in several ways, including but not limited to direct collateralization, or indirectly through pledges on shares or other financial securities of companies which have this accounting position on their annual statements.

Research Question

My central research question oscillates around the extent to which “large European listed corporates

companies are able to collateralize on goodwill”. Prior studies generally find positive relationships

between asset tangibility and leverage, which implies a negative relationship between intangible assets and leverage. In my paper I am investigating the extent to which on Balance Sheet Goodwill can affect book leverage levels and therefore the financing choices between debt and equity and internal financing. An answer will be provided by means of an empirical study of a sample of large European listed companies over the period 2009-2014

Comprehensive previous research (Marsh (1982), Titman and Wessels (1988), Friend and Lang (1988), and Rajan and Zingales (1995), and Frank and Goyal (2004)) primarily focuses on the extent to which tangible assets enable companies to finance its operations with debt. Marsh (1982) sets leverage into perspective of asset compositions, Titman and Wessels (1988) confirm the significance of tangibility and Rajan and Zingales extend it for a set of control variables which became standard in these kind of studies. Clearly from other studies we know that tangible assets are a well preserved mean of collateralizing, which traditionally prevails. From a lenders’ perspective one reason for commonly accepting tangibles as collateral is that his bargaining power at default is larger (Hart and Moore, 1994). In addition, lenders can regain a higher salvage value at liquidation, as opposed to intangibles where the potential of a deadweight loss requires additional provisioning (Shleifer and Vishny, 1992). In effect a potential hold-up problem might arise in anticipation of the lenders position in default. Expecting such a hold up, lenders tend to finance tangible assets. On the contrary, firms with a high proportion of intangibles tend to move towards internal financing, whereby the capital structure of the big four accounting firms (primarily intangible assets, structured by means of partnership and highly capitalized) serves as a good example.

Subsequently, the available literature on the topic of intangible assets collateralization has been deemed limited. Therefore the paper will contribute to existing literature by verifying inconclusive existing findings on a very recent post-crisis sample, which has not yet been evaluated for this purpose. Rare recent studies on the topics will be presented accordingly. One relates to Loumitoti (2011) who considered a sample of US collateralized syndicated credit exposure. He concludes that driven by innovations or negative mutations 21% of these facilities were secured by means of intangible assets. This finding was subsequently confirmed in a study of South African firms by Matemilola and Ahmad (2013). They found a positive relationship between goodwill assets and leverage. My study will assess a similar methodology on a very recent European data set.In line with similar capital structure I will employ a study based on annual consolidated balance-sheet and income statement data. A dynamic panel regression

(3)

will be employed to capture the cross-section and time-series in years 2009-2014, whereby 31 countries are in scope. Below available literature is discussed, followed by a depiction of the sample and a description of the methodology applied. Lastly results are presented and discussed and further research in the topic is proposed.

Literature Review

The cornerstone of literature on capital structure was the paper of Modigliani and Miller (1958), who argue that under a set of main assumptions, which are absence of information asymmetry, no arbitrage and debt-related tax benefits and bankruptcy costs, the final form of capital structure is irrelevant. This implies from the fact, that under the above conditions, the value of a firm is independent of capital structure choices made by firms. As far as the above assumption have long proven to be unrealistic, initial claims of Modigliani and Miller (1958) formed the basis for capital structure theories, which are in essence building upon it. As assumptions were further loosened by them, capital structure turned out to be of relevance for shareholder value. First, modern capital structure theories will be confronted with empirical findings, followed by an elaboration of the potential causality of unidentifiable intangible assets on capital structure.

Theories of Capital Structure Static Trade-Off Theory

The trade-off theory (“TOT”) advocates that financing choices between debt & equity are a result of assessing the comparative proportion of costs and benefits for a given company. Bankruptcy & agency cost (potential costs associated with leverage and debt market financing in general, is “traded-off” against any potential benefits implying from the “tax shield” which is generated through the tax deductibility of interest payments (Keane, 1976). Proponents of the TOT, like Kraus and Litzenberger (1973), claim that there is an “optimal level” at the point at which the marginal costs and gains, attributable to this type of financing, balance. However, this view was challenged by Miller (1977) who described the proposals of the TOT as “a balance between horse and rabbit content in a stew of one horse and one rabbit” , implying that benefits of leverage can be enjoyed in almost every case, whereby the costs of debt are only witnessed in extreme cases. Reason for this is that the costs of leverage are witnessed by firms only in case of default, i.e. when the inability to finance current interest expenses results in bankruptcy, whereby an appointed bankruptcy agent is then utilizing available assets to extinguish as much of the indebtedness as possible. On the contrary, almost every company has to have an interest coverage above 1 (i.e. EBIT > Net interest expense), in order to be able to continue to operate. As a result companies tent to enjoy benefits of leverage to a far greater extent than its downturn. Therefore, the static trade-off was deemed arbitrary.

Dynamic Trade-Off Theories

On the contrary, a different optimal capital structure, given perfect capital markets, is being hypothesized in the Pecking Order Theory (“POT”), as set out by Myers & Majluf (1984). Rather than having an optimal mix of debt and equity as outlined in the TOT, companies prefer certain types of financing over the other leading to a adjustable target capital structure. According to the POT, businesses in general prefer to finance investment opportunities through internally generated fund (operating cash flows), then through acquiring debt-related financing & lastly through floating equity. The reason for this is that external financing brings about two types of costs: i) transaction costs and ii) information asymmetry

costs.

The first relates to costs which imply from management knowing the exact quality and riskiness of issued securities and other financing participants’ need to invest material effort into assessing it (Fama E., & Fench, R., 2002). The latter stem from the variation in the cost of acquiring certain type of financing (due to exogenous risk levels), as along Fama and French debt-based financing is more expensive than equity

(4)

due to bankruptcy and agency costs. This differentiation of ownership and operations results in managers having superior information about the value of an entity. In effect, proponents claim that adjustments in a company debt-ratio is not a result of a benefit-cost trade-off, but dictated through a firms net cash flow, that is free cash flow (Fama E., & Fench, R., 2002).

Baker and Wurgler (2002), on the other hand approached capital structure, as being resultant of past attempts to time the market, i.e. use temporarily market equity mispricing, in order to maximize shareholder value. The authors agree with the proponents of the POT that optimal capital structure is not existing. Rather they state that managers who have asymmetric information about a given company, compared to other market participants, try to use this knowledge to float equity when the market valuation is in their opinion higher than fundamentals would suggest. In line, when market valuations are below fundamentals, share repurchases should take place. As such idiosyncratic optimal capital structures exist. Based on these outlays, the control factors will be applied (accounting for profitability, liquidity, size) and which should allow for the isolation of the effect of goodwill on leverage.

In addition, a target debt adjustment model was hypothesized and subsequently tested by De Miguel and Pindado (2001). They propose an estimation of the direction of the target capital structure adjustment pace based on the introduction of a lagged leverage variable, which coefficient should be an imperfect proxy for target debt adjustment speed. This is because passed leverage levels are deemed important and statistically significant predictors of current leverage levels in line with Leary and Roberts (2005) who claim that despite potential adjustment frictions capital structure is rebalanced. As my sample bears the systematic shock of 2008/2009, such rebalances are to be expected and to be taken account further in the study and hence a relatively high coefficient is to be expected on the lagged variable.

Goodwill & Intangibles Assets

From the above, I conclude that literature on the general determinants of capital structure is most definitely extensive. However, due to the fact that there is no single variable perfectly explaining target leverages, literature does not provide a simple answer. For the exact topic of the paper, the influence of Goodwill and Intangible assets on capital structure, available literature remains not only contradictive, but also largely non-existing. On one hand, material cumulative increases of intangible assets on the balance sheet could be witnessed globally, as firms were trying to enhance comparative advantages and service/product uniqueness (Nakamura, 2001). In the same time, driven by financial innovation, the disequilibrium in acceptable collateral for lender was addressed by increasing collateralization of intangibles (usually other than goodwill), as evidenced by a comprehensive study of Loimioti (2011), that concluded a positive relationship between leverage and various types of intangible collateral, except goodwill (used as control variable). One of the reasons for the aforementioned is the increasing liquidity of such assets and the presence of a market which attempts to value these with more and more ease (IRS, 1994-2005). Furthermore, given the large regulatory burden faced by regulated financial institutions, unregulated credit provider and other intermediaries are experimenting with unusual credit provision (Carey, Post and Sharpe, 1998). They are able to do as they are obliged to comply to the same regulatory capital rules as traditional banks. Furthermore, along Edwards (2001), banks did not manage to enter the market of “leveraged intangible residual collateralization”. The question remains, whether European companies were able to manage to enter (if existing) the market of goodwill collaterization.

Whether a company can collaterized Goodwill, in order to take on more leverage has been disputable. Another working paper, of Matemilola & Rubi Ahmad (2015), investigated the ability to collaterized on both fixed assets and goodwill. They find that both goodwill and tangible assets have positive and significant relations with leverage levels in a sample of South American firms. In this paper I would like to investigate whether these relationships also hold for European listed corporates. Reason for this remaining unclear, is the fact that no cash flows can be derived from this non-current accounting tool. Although expected synergies are often reported to banks in form of Investor Presentations, clearly asymmetric information between the lender and management/shareholders are present. As such one could expect a negative relationship between leverage and goodwill. However, as the sample is treating a set of

(5)

listed companies, which tend to have better credit metrics than non-listed companies, these asymmetric information might be of less relevance.

Hypothesis

The discussion above clearly shows that no clear picture on the relationship between goodwill and LT leverage ratio’s exists. In addition, previous research was pursued on an US and RSA and not EEA sample, wherewith materially different regulatory environment, different business culture, accounting standard, and other endogenous factors are present. My hypothesis will be in line with the aforementioned

study of Matemilola & Ahmad (2015). However, emphasis will be set primarily on the effect of goodwill

on leverage. The effect on assets tangibility has clearly been found to be positive in Europe in previous studies (refer to next section).Therefore the following two-sided main hypothesis was set out:

𝑯𝑯𝟎𝟎: European companies cannot collateralize on goodwill, that is goodwill does not affect the

capital structure of companies in the sample (statistically expressed as coefficient 𝛽𝛽₁ =0).

𝑯𝑯𝟏𝟏: European companies can indeed collateralize on goodwill. (statistically speaking a

two-sided test is applied: 𝛽𝛽₁ ≠0). Companies will be deemed able to collateralize on goodwill if

and only if the coefficient will be consistently above 0, i.e. the coefficient on goodwill will be positive.

The intuition behind the hypothesis is that, if companies are indeed able to collateralize on goodwill, then the positive coefficient would reveal a positive relation. As you take on more goodwill, leverage increases. The potential exist that this is also caused by reversed causality, which is further elaborated upon in the upcoming sections.

Data & Methodology

Data

The study will assess the influence of goodwill on leverage levels, using a dataset composed of large listed European corporates. A company will be deemed “European”, whenever their headquarters are located within the European Economic Area (“EEA”), as opposed to studies of Matemilola and Ahmad (2013) and Loumioti (2011), who treated a South African and US data set, respectively. The data input itself was obtained from the Compustata (Global Capital-IQ) for the period of 2009-2014, whereby annual audited fiscal year end (“FYE”) data is used. The exclusion from the sample of firms in

the financial services industry is a standard for capital structure research paper, given the heterogenic specificity in capital structure of these compared to non-financial corporates. We note, however, that real estate firms were left within the boundaries of the analyzed dataset. Further data restrictions include: (i) not fully consolidated companies were excluded, (ii) inactive companies were excluded. Given that generally accepted accounting standards regarding can vary within the EEA (European Commission, 2015), data represents standardized fiscal year end data (courtesy of COMPUSTAT). In effect, the sample constitutes 21,222 unique observations. Table 1 reveals that the average number of observations per year is 3,489. The data, despite the aforementioned adjustments, is unbalanced. Albeit no technical issues are expected in light of the substantial dataset. In addition, deviations are marginal (<5%). This translates into 31 EEA countries divided among 10 GIC sector. We refer to Graph 1 for a graphical depiction of the

Graph 2

Graphical representation data set

(6)

dataset distribution. The country classification is based on “ISO Headquarters per Incorporation”. As such the largest chunk of firms in the sample are British (~20%), French (~13%), German (~12%) and Polish (10%) firms. Data for Italian companies was not available. Excluding Italy, the skew towards these countries is in line with expectations and is representative. as its roughly corresponds to the market capitalizations of Stock Exchanges of given countries. From a sector perspective, Industrials (28%), Consumer

Discretionary (20%) and Information Technology (16%) firms conform the majority of the sample. This is again in line with expectations, as these sectors require continuous large investment outlays, whereby financing is often achieved through equity financing. We note that the division of sectors is constant over time as prescribed below. Please refer to the annex, for further depictions of data distributions.

Descriptive Statistics

Table 2 below depicts the summary statistics for each variable. We refer to the annex for an exact

definition of variables. The average level of leverage a company in the sample takes on amounts to 14.6%, whereby the 75th percentile is 22.1% and the 25th 0.9%. Average Leverage is roughly stable over time and ranges from 15.4% in 2008, through 14.7% in 2014. On first sight this leverage levels seems relatively low. Reason for this is, that the study only takes into account long term debt (maturity >1Y) and current debt, which is often a large chunk of total debt, is ignored for sake of drawing statistical inferences (see next section Dependent Variable). In line with reports of the European Central Bank (ECB, 2014), this evidences that the risk appetite of European financial institution was on average lower in the period 2009-2012, due to uncertainties caused by the systematic shock resultant of the 2008 financial crisis. Banks were initiating credit rationing. In addition, leverage was materially higher among entities active in the per Materials, Energy and Industrial sector, where material annual capital expenditure requirements prevail. Surprisingly, Utility companies in the sample have the lowest average leverage of all sectors. From a country perspective, Iceland, Portugal, Luxembourg, Norway, Spain and Ireland witness the highest leverage levels (all above 20%). Regarding goodwill, the highest levels can be observed for capital intensive sectors with the highest leverage levels, that is again the Energy, Materials and Industrial sectors reaching to 14.6% of all assets. Goodwill is least common among European, Healthcare, Utilities and Real estate companies, which on average have only 7.6% on the balance sheet, which again can be explained by the lower capital intensity of the latter industries. Dutch, British, Swedish and Irish corporate reveal the highest levels of on-balance sheet Goodwill,

Graph 1

Distribution per sector

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 2008 2009 2010 2011 2012 2013 2014

Ene rgy Material Industr ials

Consumer Discretionary Consumer staple s Health Ca re

Real Estate / Non exclu ded fin ancials Info rma tio n Technol ogy Tele-communication Utili ties

Table 1

Deviation per Panel year

FYE Obs Deviation

2009 3,312 - 177 2010 3,416 - 73 2011 3,552 63 2012 3,621 132 2013 3,667 178 2014 3,365 - 124 Average 3,489 -Grand Total 21,222 6

(7)

which is again intuitive, as most acquisitions in Europe were performed from firms located in the UK. Interestingly, the sub-sample of Netherlands, Portugal, Netherlands, Denmark Great Britain has highest levels of intangibles, excluding goodwill, which largely coincides with companies that also have material levels of average goodwill. Sector-wise, Utilities, Telecoms and Materials possess material average amounts of intangibles other than goodwill. For the data distribution, we refer to the annex, where a whole range of PIVOT tables depicting the sample is presented. Also find above a graph showing the development of averages of variables in the time period 2008- prevailing in FY08-FY09. This makes sense as the crisis resulted in material impairments of goodwill 2014. It can clearly be seen that in 2008 a material decrease of average EBITDA, Goodwill, and leverage was (ESMA, 2013) and EBITDA margins were decreasing due to general worse economic conditions. The increase in cash balances post-2008 is also intuitive, as companies faced with the systematic shock of 2008 halted investment and piled up cash balances for the expected economic downturn. Same logic also applied to developments in generated

Table 2

Summary Statistics

Summary Statistics LEVB GOOD INTEXG EBITDA LOGSAL ACID CASHEQ OPCF

Mean 0.1460 0.1460 0.0591 0.0899 2.4857 1.8815 0.1048 0.0689 Standard Error 0.0012 0.0012 0.0007 0.0008 0.0062 0.0210 0.0009 0.0009 Median 0.0962 0.0962 0.0176 0.0902 2.3611 1.4164 0.0690 0.0699 Mode 0.0000 0.0000 0.0000 0.0000 1.2145 2.0000 0.0000 0.0000 Standard Deviation 0.1765 0.1765 0.1007 0.1223 0.9101 3.0596 0.1289 0.1239 Sample Variance 0.0312 0.0312 0.0101 0.0150 0.8282 9.3614 0.0166 0.0154 Minimum 0.0000 0.0000 0.0000 -1.7168 1.0000 0.0143 -1.3817 -1.8890 25th Percentile 0.0092 0.0000 0.0030 0.0030 0.0481 1.7724 1.0253 0.0251 75th Percentile 0.2206 0.1844 0.0698 0.0698 0.1388 3.0796 2.0413 0.1446 Maximum 0.7648 0.5430 0.8692 1.4112 6.7421 215.3654 0.9710 5.7889 Count 21,222 21,222 21,222 21,222 21,222 21,222 21,222 21,222 Table 3

Average per sector

Average per sector LEVB GOOD INTEXG EBITDA LOGSAL ACID CASHEQ OPCF

Energy 0.1872 0.1455 0.0686 0.1191 4.6475 1.2712 0.0748 0.0935 Material 0.1948 0.1602 0.0700 0.1150 3.8335 1.4206 0.0848 0.0898 Industrials 0.1774 0.1250 0.0605 0.1108 3.0216 1.5890 0.0891 0.0872 Consumer Discretionary 0.1275 0.1087 0.0490 0.1029 2.3338 1.8023 0.1063 0.0780 Consumer staples 0.1197 0.1029 0.0500 0.0842 1.9820 2.1221 0.1141 0.0610 Health Care 0.1341 0.0949 0.0496 0.0738 1.7822 2.2164 0.1168 0.0515

Real Estate / Non excluded 0.1133 0.0764 0.0440 0.0554 1.6751 2.3019 0.1205 0.0524

Information Technology 0.1120 0.0949 0.0628 0.0502 1.4461 2.2815 0.1233 0.0378 Tele-communication 0.0965 0.1087 0.0744 0.0189 1.1672 2.7809 0.1506 0.0256 Utilities 0.1054 0.0957 0.0934 0.0106 1.0706 3.0515 0.1557 0.0000 Grand Total 0.1460 0.1152 0.0591 0.0899 2.4857 1.8815 0.1048 0.0689 Table 4 Correlation Matrix

Correlation Matrix LEVB lag1 GOOD INTEXG EBITDA LOGSAL ACID CASHEQ OPC

LEVB 1.0000 lag1 0.6915 1.0000 GOOD 0.1094 0.0927 1.0000 INTEXG 0.0720 0.0553 0.1361 1.0000 EBITDA -0.0031 0.0093 0.0456 0.0050 1.0000 LOGSAL 0.1696 0.1438 0.1193 0.0166 0.2021 1.0000 ACID -0.1052 -0.0934 -0.1168 -0.0455 0.0297 -0.1211 1.0000 CASHEQ -0.1914 -0.1410 -0.0982 -0.0280 0.0564 -0.1349 0.2370 1.0000 OPCFA -0.0095 0.0104 0.0267 0.0274 0.6450 0.1633 0.0090 0.0994 7

(8)

operating cash flows. Notably, the average amount of intangibles on the balance sheet, also increased. Average sales of the corporates were exactly flat at an average log(net sales) of ~2.49, which reflects the stagnant aggregate growth of the world-wide economy (CIA FACTBOOK). All in all, however, the sample averages are largely constant over time, implying that the systematic shock of 2009, should not have substantial influence on the ability to draw inferences from the data set, which seems to be roughly representative for the EEA.

Further, a correlation matrix is presented, whereby all correlations are below |0.25|, meaning that no problems of multicollinearity should occur. The only variables with high correlation are CORREL(EBITDA|OPCFA)=0.65 and CORREL(LEVB|lag1)=0.69. However, given the substantial sample size, no problems of collinearity are expected to occur for (EBITDA|OPCFA), as the denominator of the variance of the error terms includes a multiplication with the number of observations. On the other hand, multicollinearity does not apply to the high correlation between LEVB and lag1, due to the fact that collinearity is a danger between independent variables.

The Model

In order to proceed with the analysis of the above hypothesis, the below model was drawn down: (𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿)𝑖𝑖,𝑡𝑡 = 𝜹𝜹(𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿)𝑖𝑖,𝑡𝑡−1+ 𝛽𝛽0+ 𝛽𝛽1(𝐺𝐺𝐺𝐺𝐺𝐺𝐺𝐺)𝑖𝑖,𝑡𝑡+ 𝛽𝛽2(𝐼𝐼𝐼𝐼𝐼𝐼𝐿𝐿𝐼𝐼𝐺𝐺𝐼𝐼𝐼𝐼𝐼𝐼)𝑖𝑖,𝑡𝑡+ 𝛽𝛽3(𝐿𝐿𝐿𝐿𝐼𝐼𝐼𝐼𝐺𝐺𝐼𝐼)𝑖𝑖,𝑡𝑡

+ 𝛽𝛽4(𝐿𝐿𝐺𝐺𝐺𝐺𝐿𝐿𝐼𝐼𝐿𝐿)𝑖𝑖,𝑡𝑡+ 𝛽𝛽5(𝐼𝐼𝐴𝐴𝐼𝐼𝐺𝐺)𝑖𝑖,𝑡𝑡+𝛽𝛽6(𝐴𝐴𝐼𝐼𝐿𝐿𝐶𝐶𝐿𝐿𝐶𝐶)𝑖𝑖,𝑡𝑡+𝛽𝛽7(𝐺𝐺𝑂𝑂𝐴𝐴𝑂𝑂𝐼𝐼) + 𝜀𝜀𝑖𝑖,𝑡𝑡

In the above model i and t relate to a given company and the fiscal year end at which the data was recorded. As data is both cross-sectional and in time-series, I will regress the above model using a panel regression with fixed effects.

Estimation Method

The proposed estimation methodology is a panel regression with a lagged variable and fixed effects which account for firm and country specificity. This estimation method allows for handling unobservable firm-specific characteristics both in the time series and cross section and also variables that change over time but not across entities, accounting for idiosyncratic heterogeneity. Section 3.0 Data a summarized analysis of the substantial data set is included and certain country-specific and industry-specific characteristics are listed. In order to avoid any disturbances from the heterogeneity, the fixed effects methodology (“FE”) is additionally applied. FE’s removes time-invariant characteristics, in order to assess the net effect of the predictors on leverage, as the error term is uncorrelated with the predictor, as evidenced by the Houseman test. Robust standard errors were used, in order to rule out the possibility of heteroscedasticity and serial autocorrelation.

Variables

Dependent Variable

The dependent variable is book leverage (“LEVB“), defined as long-term debt divided by total assets as proposed by Titman and Wessels (1988), per Fiscal Year End (“FYE”). Certain companies have broken book years (i.e. fiscal year varies from calendar year), however, this should not limit us from drawing statistical inferences, as data is standardized for constant one year long fiscal periods. It was decided to apply the regression based on book-values. Reason is that along Titman and Wessels (1988), statistical inferences may be drawn between market leverage and other predictors, even if firms would adjust their target leverage randomly, as with a higher market value both sides of the model increase. Long-term debt is used as a proxy, as current debt it very often used for working capital management, supply chain finance and payment cash management solutions, whereby current debt is used as a “Cash flow” optimization tool and this could lead to misleading levels of actual leverage aimed at required capital

(9)

outlays. Furthermore, the acid ratio (see liquidity) will in a sense serve as a proxy for the aforementioned type of short term uncommitted financing, as it is often derived from it. Another reason for choosing book leverage, is that previous studies, including but not limited to Matemilola and Ahmad (2013), which assess goodwill (which are defined in terms of net-book values) also primarily use this measure. Market-leverage could be affected by unintended consequences of the systematic shock of 2008 which lasted in 2009 and had a significant effect on market values of debt as found by Chen et al. (2015).

Explanatory Variable

The main explanatory variable is (“GOOD“)i,t, which represents the ratio of goodwill-to-total assets in period i,t. A two-sided test is applied and a relationship will not be hypothesized, as a two-sided test is applied

Asset Intangibility

An alternative test variable is “INTEXG”, which represents the ratio of a company’s intangible assets excluding goodwill to total assets. Non-current asset tangibility, i.e. PP&E to total assets has been found to be consistently positively related to leverage by among others Rajan and Zingales (1995) and Goyal (2004). However, as this study is concerned with the effect of asset intangibility on leverage, the variable has been adjusted in the study. It is not possible to account for goodwill, tangibility and intangibility, as the sum of these is unity and certain mechanical problems with the model could arise. On one hand, intangible assets are commonly believed to not preserve their value in default to the extent tangible assets do, resulting in a lower recovery rate for banks (Hart and Moore, 1994). In effect, the resultant hold-up problem should lead away banks from financing such entities. Furthermore, redeploy ability of tangible assets is materially more cost efficient (Harris and Raviv, 1991). Along this reason debt-financing should be in particular non-optimal for companies with a large proportion of intangible assets. On the other hand, recent studies like Loumioti (2011) find that in particular with syndicated loans (which are usually provided to large corporates as my sample), financial innovation allows for the collaterization of debt, albeit at increased costs. As such, the coefficient on this predictor remains disputable. Still, no doubt exist that this predictor will help to isolate the effect and strengthen the explanatory power of goodwill on leverage.

Lagged Leverage Variable

The lagged leverage variable (“lag1”) is introduced in order to account for the target leverage adjustment speed direction. The introduction of the lagged variable is proposed by Lemmon et al. (2008), in order to depict the materiality of fixed effects. It is expected to decrease the problem of endogeneity, whereby it is assumed the appropriate moment condition is satisfied. This will allow for the calculation of the target leverage adjustment pace. Lemmon et al. (2008) assess the explanatory power of leverage on the lagged variable and firm specifications (see below Control Variables). In theory, if the firm specific traits are materially variant in time and assuming non-substantial adjustment costs, these factors should be valid predictors. However, the introduction of a time lag, allows for additional explanatory power, when instruments are weak, target leverage adjustment costs are present or predictors are time-invariant. In this study a one year lag will be introduced along Matemilola and Ahmad (2013). The predictor is expected to yield significant explanatory power as the systematic shock of the 2008 financial crush should be contributing to higher adjustment costs due to ongoing credit rationing and investment irreversibility, as

found by Cao (2015). The direction of the adjustment speed is calculated as 1- δ and is expected to have

a positive relationship to LEVB, given market circumstances that prevail during the sampling period.

Control Variables

Below a set of control variables is presented, which should support the explanatory power of goodwill.

(10)

Profitability & Cash Flows

The first set of control variables to be employed relates to profitability, which has often been found to be negatively related to leverage and to be a significant determinant of target leverages (Baker and Wurgler, 2002). Considerable debate exists on whether profitability portraits a reason of variation in target capital structures or rather is causing the deviation. Furthermore, firms with higher operating profits (i.e. more profitable) tend to have larger tax benefits to be gained and should therefore accrue more leverage (Myers and Majluf , 1984). As a result this control variable also serves as a proxy for tax benefits from leverage as prescribed in initial capital structure literature (see above).

Cash flows also tend to be found to positively affect leverage. Intuitively this makes sense, as with increasing operational cash inflows, internally generated cash should be able to cover to a greater extent any required investment outlays. Along the aforementioned POT, firms prefer to finance themselves internally, then with debt and then by equity (Myers & Majluf, N. S., 1984). Baker and Wurgler (2002), use the ratio EBITDA (non-IFRS) to assets as a proxy, as does this study. Therefore also the control variable “OPCFA” is introduced. Given the relatively low correlation between EBITDA and operating cash flows (~0.65) no mechanical problems are expected. Reason for also introducing the cash flow variable is that in particular in times of financial systematic shocks, EBITDA may be a weak proxy for profitability. As mentioned above, such systematic shocks tend to come together with material non-cash book-losses like impairments and additional provisioning. Reported EBITDA, therefore might be magnified compared to the recurring EBITDA. As such, the cash flow predictor, is aimed at capturing the profitability and partially the liquidity factor which affects target leverage. In particular, I expect a negative relation of both predictors to leverage.

Liquidity

The last set of control drivers relates to pure liquidity measures, which are ACID and CASHEQ. The latter refers to the on balance sheet cash & cash equivalents to total assets ratio. Reason for the introduction of the cash control variable is that banks usually consider net-debt when assessing the creditworthiness of a potential obligor. As such, this variable allows for the extraction of the net-leverage effect and sets the target leverage in perspective. Furthermore, the first control variable is associated with the ratio of current assets to current liabilities. Again, this sets the dependent variable (long-term debt) into perspective, as we are controlling for the proportion of current assets, i.e. short terms means for credit repayment, with current liabilities, which serves as a proxy of current debt to be reimbursed within on year from the relevant reporting period. Given the above, a negative relationship between both control variables is expected.

Fixed Effects

As outlined before, the large data set consist of large listed corporates in a variation of industries and located throughout 31 EEA countries. Despite efforts within the European community to harmonize or integrate aspects which are of importance for the operations of capital market, bank regulation, standards and risk appetite of local credit-providing institutions remain largely diverse within the EEA (ADD SOURCE). In effect material unobserved country-specific effects are expected. One drawback of this method is that any determinants which do not change over time can no longer be identified, as any time-constant-firm-specific effects are swept away.

Results and discussion

The regression based on the above model, and taking into account the aforementioned variables, yields significant results, which will be discussed below. The table on the side depicts the optimal

model. All variables have the projected

Graph 5

Averages LOGSAL per sector

(11)

coefficient, which is in line with previous studies and the expectation set out in the previous section. The main explanatory variable (GOOD) has a positive coefficient to leverage (LEVB). This is in line with the results of previous studies, whereby Matemilola and Ahmad (2013) found a ~5% relationship between goodwill and leverage in their study of South African firms. A ~15% effect was estimated in this study. As far, as the modelled coefficient is far away from the RSA study, another working paper of Loimioti (2011) considers US firms and also finds a 15% relation in the one-stage regression (when performing 2-stage tests, the coefficient tends towards the ~5% found in the South African study). Nevertheless, my results suggest that companies can indeed collaterized on goodwill, or at least there is a persistent and statistically significant at the 1% level, relation between these variables. However, it can also be the cased that the draw I found is not causal, or in fact there is some reverse causality taking place. A reason for this might be that firms, when pursuing acquisitions, are very often finance through syndicated debt. So, the acquisition which results in additional goodwill (see accounting identity in section 1) actually also results in additional leverage. Furthermore, the lagged leverage variable (one period lag) is significant at the 1% level and reveals a positive relation with leverage at period i,t. The positive sign of this predictor is in line with the hypothesis of the dynamic trade-off theory. It can be interpreted as the average adjustment speed direction of companies in the sample to the a given target leverage level. The adjustment speed

determined in this sample, is notably very high at 1- 𝛿𝛿 = 1 − 0.0588 = 0.9412. Comparing to the

finding to previous studies, results are rather far away from the adjustment speed found by Fama and French (2002) and Ozkan (2001), who found an adjustment speed of approximately 0.38 and 0.55, respectively. However, when comparing it to the result of Flannery and Rangan (2006), who applied a methodology more similar to the one used in this paper, they observed an adjustment speed of 0.92. Also De Miguel and Piniado (2001) report a leverage adjustment speed of 0.79 on basis of a Spanish sample. Given the large variation in leverage adjustment pace and

the heterogeneity of the sample , the results obtained seem acceptable. However, in line with Ozkan, A. (2001), who evidences that certain statistical problems on the estimated coefficients, I conclude that a GMM regression would diminish any potential biases (see below Robustness of results, where this is further elaborated upon).

The coefficient of goodwill is almost double as high as for the intangibles excluding goodwill (0.0688). This suggests, that as much as companies are able to collaterized on goodwill they are much less so able to collaterized other intangibles. This opposes the findings of Loumiti (2011), who finds a materially stronger effect of other intangibles on capital structure. One explanation might be that American banks are more likely to pursue financial engineering (which collaterizing intangibles clearly represents), than European banks, which operate within boundaries of a more severe regulatory environment (ADD SOURCE).

Interestingly, only two of the estimated coefficients are not significantly different from zero. These are the logarithm of sales and EBITDA. Scale (LOGSAL) has indeed a positive relationship with leverage, as found by Titman and Wessels (1988), Kim and Sorensen (1986) and Mehran (1992). However, the significance is 0.9. This is rather counter-intuitive, however, might be explained by the usage of fixed effects in the regression. As mentioned above, this method has one drawback, namely that predictors which are time-invariant can no longer be estimated. In my opinion, this is the case with the LOGSAL variable, whereby the above diagram manages to display the average time in-variance of the variable relatively well, among all represented GIC sectors. A reasonable explanation for this is strikingly simple, namely average top-line growth of stock-listed European companies in the sample was on average 0%.

Table 5

Optimal Model results

LEVB Coeff. p.value Significance

lag1 0.0588 0.0000 *** GOOD 0.1546 0.0000 *** INTEXG 0.0688 0.0000 *** EBITDA -0.0182 0.0590 * LOGSAL 0.0003 0.9480 ACID 0.0009 0.0040 *** CASHEQ -0.0468 0.0000 *** OPCFA -0.0198 0.0100 *** _cons 0.1211 0.0000 *** 0.1391 7.26 Fixed Effects GIC

Country (EEA) * Significant at 10% ** Significant at 5% *** Significant at 1% R-sq: F(3964, 17248) 11

(12)

EBITDA, on the other hand, has a negative relationship with leverage. This finding is again in line with virtually all previous studies, including but not limited to Titman and Wessels (1988) and Rajan and Zingales (1995). However, the variable is only significant at the 10% level (p=0.059). A simple reason for this relatively low, compared to other variable, significance might be the high correlation between EBITDA and operating cash flows. As such, the effect of EBITDA on leverage might be partially covered through operating cash flow.

Further, the effect of the Acid ratio (ACID) is deemed statistically at 1%, however, the coefficient is relatively low. Reason for this is that the acid ratio is very often >1, as it is in the case of this sample, whereby 16,250 of the observation have such an acid ratio. Therefore, the coefficient is actually material, positive and statistically significant, suggesting that a higher proportion of current assets on the balance sheet of a given company, allows for taking on more long-term leverage. It simply ensures the repayment a short term liabilities towards the lender, who is less reluctant to lend to even a leveraged company. Reason is that most bank provided credit lines, which are uncommitted, implying a direct facility cancellation ability of the lender.

Robustness of results

Standard error robustness was avoided through the usage of robust standard errors. Further, the robustness of the model itself was tested, compared to alternative model set-ups. The results reveal, that the effect of goodwill on leverage is consistent regardless of the model choice. The relationship is positive and significant at the 1% level in every model except for the simple model 2. Furthermore, so is the adjustment pace which remains high throughout the applied models and remains significant at the 1% level. In addition, other variables are also consistent, regardless of model choice. When introducing the additional control variable, the explanatory power increases. The similar results suggest that goodwill indeed has a positive relationship with leverage, regardless of its reasons. The table on the previous page reveals the robustness of the results.

In line with Ozkan (2001), there biases on the OLS estimator might ooccur when us treating capital

structure panel data with fixed effect. The reason is the assumption of ἁ is not to be observed and

covariances are different to 0. In effect OLS brings about inaccurate parameter projection on the lagged

variable, due to the correlation with the constant ἁ. In line, an IV estimation, also might reveal statistical

problems, as it cannot always utilize existing moment conditions (Ozkan, 2001). Ozkan (2001), then shows that in order to address this statistical issue, a GMM regression should be applied, together with a 2-year lagged variable, that is uncorrelated with the error expression. In absence of serial correlation, this should result in efficiency gains related to the estimation process of coefficients. These claims were confirmed by Arellano and Bond (1991), who found the most negligent variance in the GMM conclusion.

(13)

Concluding, as much the first robustness check shows that the effect of goodwill on leverage is robust, it has to be noted that potential statistical problem could arise and these would better be addressed with the usage of a GMM regression. This however, remains outside the boundaries of this study.

Conclusion

The scope of the study included listed corporates headquartered in the EEA over the most recent period, where data is available and roughly balanced, of 2009-2014. In line with previous studies, a dynamic panel regression was applied, whereby time-invariant fixed characteristics, implying from an effectively still unharmonized European market were extracted using the fixed effect methodology. Results are in line with previous studies, however, do not yield a definitive answer to the research question.

The above regression yields significant and robust results. A positive effect of goodwill on leverage is found in line with the hypothesis, which is consistent over alternative model specification. The implied relationship is that a marginal change in goodwill results in an approximate adjustment of leverage by 15%. Even though the study confirmed this relationship statistically, the question arises, of whether this relationship is due to the positive influence of goodwill on leverage, or vice versa some kind of reversed causality is taking place. As such, the finding of Matemilola and Ahmad (2013) are confirmed, albeit the effect of goodwill is not the same it magnitude. A reason for this can be that European Capital markets are generally speaking more established and as such financial innovation is prevails to a greater extent. However, results on the main variable remain inconclusive, as it is also possible that it is actually leverage which results in higher levels of goodwill. Reason for this is that goodwill is derived from acquisitions only, along IFRS3. Therefore as company “take on” more goodwill, so might the leverage level behave, mechanically.

Furthermore, the study confirmed estimation of leverage adjustment directions, albeit a relatively quick and positive target leverage adjustment speed of 0.9412 is found. The positive sign of the lagged leverage variable confirms this level and is consistent regardless of model choice. No surprises are evident from

Table 6

Robustness of the model

LEVB Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8

lag1 0.0572 0.0575 0.0578 0.0580 0.0580 0.0581 0.0586 0.0588 *** *** *** *** *** *** *** *** GOOD 0.1629 0.1681 0.1680 0.1679 0.1691 0.1534 0.1546 ** *** *** *** *** *** INTEXG 0.0813 0.0801 0.0800 0.0809 0.0674 0.0688 *** *** *** *** *** *** EBITDA -0.0293 -0.0294 -0.0298 -0.0266 -0.0182 *** *** *** *** *** * LOGSAL *** 0.0002 0.0009 0.0003 0.0003 ** ACID 0.0006 0.0009 0.0009 ** *** *** CASHEQ -0.0517 -0.0468 *** OPCFA -0.0198 *** Cons 0.1376 0.1188 0.1134 0.1161 0.1155 0.1126 0.1213 0.1211 *** *** *** *** *** *** *** ***

Fixed Effects applies to each regression

R-sq 0.4782 0.1207 0.1134 0.1138 0.1145 0.1126 0.1413 0.1391

* Significant at 10% ** Significant at 5% *** Significant at 1%

(14)

the coefficients of control coefficients. All of them yield the same sign as found in previous studies. A positive effect of intangibles other than goodwill on leverage is found in line with Loumitio (2011), albeit it is noted that intangibles other than goodwill are treated as a whole, primarily in order to set goodwill in perspective. The introduction of this variable was a deviation of previous convention, whereby usually the asset tangibility is used. All in all, intangible assets are found to generally also serve as collateral of sufficient quality.

Profitability related control variables, namely EBITDA and operating cash flows have a negative relation to leverage, as predicted by the POT, which stresses that internal financing is in general the cheapest mean of financing. On the other hand, liquidity is positively related to leverage. A probable reason for this is that companies with high liquidity often being derive this liquidity from uncommitted current credit facilities (very often working capital and cash management related), which was excluded from the dependent variable. Interestingly, the logarithm of sales yields insignificant results, whereby this contradicts previous findings. One of the reason for this might be the application of fixed effects, which results in removing all time-invariant effects and due to the stagnant economic growth in the EEA in 2009-2014 this might result in an insignificant coefficient on this control variable.

Nevertheless a positive relation between goodwill and leverage can be concluded, subject to further investigative attempts. In order to improve the ability to draw causal inferences about the relation between these two accounting positions, I would propose further research to i) control for investment outlays, ii) investigate the relationship based on market leverage or to iii) perform a two-stage dynamic panel regression, in order to further reduce problems of endogeneity. In addition, robustness of the result could be improved, by also assessing the effect in light of a two-stage GMM regression, the addition of tangible assets as control variable, performing the same regression on market leverage and/or total leverage (as opposed to long term debt only).

(15)

Bibliography

Baker, M., & Wurgler, J., (2002). “Market timing and capital structure”. Journal of Finance, 57(1)

Bolaji Tunde Matemilola & Rubi Ahmad (2015) “Debt financing and importance of fixed assets and goodwill assets as collateral: dynamic panel evidence”, Journal of Business Economics and Management, 16:2, 407-421

Cao, V.N., (2015) . “What explains the value premium? The case of adjustment costs, operating leverage and financial leverage”. Journal of Banking and Finance, 59, 350-366

Carey, M.,. Post, M., & Sharpe, S.A., (1998) “Does Corporate Lending by Banks and Finance Companies Differ? Evidence on Specialization in Private Debt Contracting”. The Journal of Finance, 53(3)

Chen, R.W., Chidambaran, N.K., Imerman, M. B., Sopranzetti, B. J., (2013). “Liquidity, leverage, and Lehman: A structural analysis of financial institutions in crisis”, Journal of Banking and Finance, 45 Fama, E.F., & French, K.R. (2002). “Testing trade-off and pecking order predictions about dividends and debt”. Review of Financial Studies, 15, 1–33

Fama, F., & French, K.R., (1997). “Industry cost of Equity”. The Journal of Finance, 43, 153-193

Flannery, M.J., & Rangan, K.P., (2006). “Partial Adjustment towards target capital structure”. Journal of

Financial Economics, 79, 496-506

Goyal, K., & Frank, M.Z., (2009). “Capital Structure Decisions: Which Factors Are Reliably Important?”

Financial Management. 38(1).

Hart, O., & Moore, J., (1994). “A Theory of Debt Based on the Inalienability of Human Capital”.

Journal of Economics. 109(4).

Harris, M. and Raviv, A. (1991), “The Theory of Capital Structure”. The Journal of Finance, 46: 297–355 Kraus, A., & Litzenberger, R.H., (1973). “A State-Preference Model of Optimal Financial Leverage”.

Journal of Finance, 28 (4), pp. 911-922

Leary, M., & Robert, M., (2005) “Do Firms Rebalance Their Capital Structures?”. Journal of Finance, 60(6)

Lemmon, M.L., Roberts, M.R., & Zender J.F., (2008) “Back to the Beginning: Persistence and the Cross-Section of Corporate Capital Structure”, Journal of Finance, 63 1575-1608.

Menkhoffa, L., Neubergerb, D., & Rungruxsirivorna,O., (2010) “Collateral and its substitutes in emerging

markets’ lending”. Journal of Banking & Finance”, Journal of Banking & Finance, 36(3)

Modigliani, F. & Miller, M.H., (1958). “The Cost of Capital, Corporate Finance, and the Theory of Investment”. American Economic Review, 48, p. 261

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

(16)

Nakamura, L. 2001. Investing in intangibles: Is a trillion dollars missing from GDP? Business Review, Federal Reserve Bank of Philadelphia, 27-37.

Ozkan, A. (2001), “Determinants of Capital Structure and Adjustment to Long Run Target: Evidence From UK Company Panel Data”. Journal of Business Finance & Accounting, 28: 175–19

Pindado, J., & de Miguel, A., ”Determinants of Capital Structure: New Evidence from Spanish Panel Data”. Journal of Corporate Finance, 7(1)

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

Rampini, A., & Viswanathan, S., (2013). “Collateral and capital structure”. Journal of Financial

Economics, 109(2)

Shleifer, A., & Vishny, R. (1992). “Liquidation Values and Debt Capacity: A Market Equilibrium Approach”. The Journal of Finance, 47(4), 1343-1366.

Keane, S. (1976). “The Tax-deductibility of Interest Payments and the Weighted Average Cost of Capital”. Journal of Business Finance and Accounting, 3(4)

(17)

Annex I - Additional Summary Statistics

Table 7

Averages per country

Averages by country LEVB GOOD INTEXG EBITDA LOGSAL ACID CASHEQ OPCFA

AUT 0.1621 0.0716 0.0366 0.1030 2.6930 1.6663 0.1068 0.0808 BEL 0.1586 0.1149 0.0541 0.1031 2.5997 1.8756 0.1008 0.0752 BGR 0.1077 0.0101 0.0168 0.1144 2.1253 1.9355 0.0742 0.0693 CHE _0.1285 _0.0884 _0.0579 _0.1054 _2.8436 _2.2504 _0.1535 _0.0912 CYP 0.1785 0.0388 0.0244 0.0727 2.0592 1.6073 0.0184 0.0636 CZE 0.1455 0.0386 0.0190 0.1273 3.7972 1.7410 0.0754 0.1124 DEU 0.1410 0.0959 0.0559 0.0953 2.4185 2.2189 0.1411 0.0723 DNK 0.1530 0.0906 0.0754 0.0750 2.9181 2.2749 0.0926 0.0600 ESP _0.2216 _0.1062 _0.0724 _0.0852 _2.7541 _1.3036 _0.0749 _0.0653 EST 0.1942 0.1066 0.0246 0.1061 2.2353 2.0126 0.1001 0.0944 FIN 0.1763 0.1569 0.0462 0.0937 2.4981 1.5343 0.1046 0.0760 FRA 0.1381 0.1481 0.0625 0.0858 2.4432 1.6045 0.1099 0.0649 GBR 0.1524 0.1759 0.0781 0.1145 2.2972 1.7909 0.1082 0.0861 GIB 0.0385 0.2843 0.2815 0.0387 2.7401 1.2756 0.2459 0.0842 GRC 0.1641 0.0190 0.0357 0.0419 1.9681 1.4730 0.0793 0.0379 HRV _0.1573 _0.0091 _0.0150 _0.0518 _2.7883 _2.4087 _0.0459 _0.0465 HUN 0.1373 0.0425 0.0660 0.1235 4.3504 1.8489 0.0848 0.1096 IRL 0.2224 0.1692 0.0644 0.0992 2.7225 2.0448 0.1307 0.0864 ISL 0.2692 0.2943 0.0663 0.1153 3.6217 1.4391 0.0691 0.1008 LTU 0.1150 0.0220 0.0207 0.1321 2.4432 1.6337 0.0494 0.1143 LUX 0.2502 0.1364 0.0694 0.1055 3.0260 1.4094 0.0731 0.0782 LVA 0.1469 0.0048 0.0187 0.0812 1.7070 2.0853 0.0452 0.0637 MCO _0.0354 _0.0000 _0.0060 _0.0417 _2.6010 _0.8852 _0.0786 _0.0577 MLT 0.1993 0.0066 0.0477 0.1056 1.7102 1.5296 0.0820 0.0870 NLD 0.1744 0.1893 0.0797 0.1131 2.9667 1.4190 0.0750 0.0834 NOR 0.2252 0.0767 0.0791 0.0672 2.8942 1.8474 0.1297 0.0526 POL 0.0859 0.0546 0.0375 0.0903 2.3081 2.0608 0.0764 0.0651 PRT _0.2554 _0.1266 _0.1064 _0.0735 _2.6006 _1.0530 _0.0613 _0.0581 ROU 0.0593 0.0022 0.0099 0.0637 2.1246 3.0195 0.0481 0.0422 SVK 0.1139 0.0505 0.0261 0.0803 2.1750 3.4793 0.0570 0.0498 SVN 0.1814 0.0152 0.0273 0.0816 2.3469 1.5758 0.0302 0.0723 SWE 0.1226 0.1639 0.0689 0.0579 2.7432 1.9581 0.1333 0.0425 Grand Total 0.1460 0.1152 0.0591 0.0899 2.4857 1.8815 0.1048 0.0689 17

(18)

Table 9

Sector per country

Sector by

country Energy Material Industrials

Consumer Discretion ary Consumer staples Health Care Real Estate Informati on Technolog Tele-communi cation

Utilities Grand _Total

AUT 6 BEL 12 BGR - CHE 43 CYP 9 3 CZE 18 DEU _{DNK 42 ESP 35 EST - 3 FIN 5 FRA _{GBR 111 277 GIB - GRC - HRV 8 HUN _{IRL 6 ISL 18 LTU - LUX 12 LVA - MCO - MLT - NLD 47 NOR 59 POL 36 PRT 6 ROU 4 SVK - SVN _{SWE 161 Grand Total Table 8 Observations per country AUT 290 ESP

25 133 61 31 13 - 17 2 2 290 36 153 142 24 19 3 46 5 5 445 - 35 52 22 5 2 39 2 3 160 145 326 242 45 34 4 34 10 12 895 39 35 16 21 6 90 3 6 228 14 19 0 - 4 - 1 - - 56}₁₃₃ ₁₉₀ ₆₀₀ 479 ₃₀₄ ₂₀₂ ₃₂ 429 ₆₁ ₇₄ _2,504 112 200 80 39 34 2 48 5 4 566 69 227 100 43 20 1 63 7 6 571 17 16 10 17 3 5 1 - 72 63 207 113 65 50 12 84 9 12 620}₁₅₅ ₂₄₈ ₆₀₂ 547 ₂₁₂ ₂₁₅ ₃₁ 446 ₅₇ ₁₁₆ _2,629 1,031 838 340 305 43 870 146 205 4,166 - 5 1 - - - 0 - - 6 27 114 185 105 94 16 250 32 47 870 19 200 93 18 7 - 11 - - 356}₃₄ ₁₃ ₂ 0 _- ₂ _- 8 _- _- ₅₉ 33 81 25 15 10 - 19 7 5 201 12 11 8 - 2 - 2 - - 53 - 64 53 12 16 1 14 - - 160 24 58 37 5 1 - 7 - - 144 - 2 11 10 17 3 28 - 6 77 - 2 4 - - - 0 - - 6 - - 7 8 9 1 9 2 2 38 73 203 61 31 22 3 32 4 11 487 134 251 119 47 35 3 66 4 15 733 109 515 550 266 178 28 332 36 52 2,102 22 91 58 25 15 - 28 - 1 246 18 65 138 77 51 16 91 9 15 484 5 1 11 4 8 3 5 - - 37}_- ₃ ₂₈ 33 ₇ ₉ _- 19 _- _- ₉₉ 263 556 330 133 109 21 211 25 53 1,862 960 1940 5838 4429 1914 1524 234 3304 427 652 21222 571 HUN 59 NLD 487

BEL 445 EST 72 IRL 201 NOR 733

BGR 160 FIN 620 ISL 53 POL 2,102 CHE 895 FRA 2,629 LTU 160 PRT 246

CYP 228 GBR 4,166 LUX 144 ROU 484

CZE 56 GIB 6 LVA 77 SVK 37

DEU 2,504 GRC 870 MCO 6 SVN 99 DNK 566 HRV 356 MLT 38 SWE 1,862 Grand Total 21,222

Observation per country

(19)

Annex II – Variable definitions

LEVB – book leverage, defined as LTD Debt to Total Assets

Lag1 -- LEVB lagged by 1 period (1FY)

GOOD – ratio of foodwill to total assets

INTEXGTAN – ratio of intangibles asset corrected for goodwill to total assets

INTEXG – ratio of Intangible assets corrected for goodwill to total assets

EBITDA – ratio of EBITDA to Assets

LOGSAL – logarithm of sales

ACID – ratio of current assets to current liabilities

CASHEQ – ratio of cash & cash equivalents to total assets

OPCFA -- ratio of operating cash flows to total assets

(20)

Annex III – Data Distribution

(21)

Annex III – Regression output

Fixed-effects (within) regression Number of obs = 21,221

Group variable: ISIN1 Number of groups = 3,965

within 0.0064 min 1

between 0.8012 avg 5.4

overall 0.4782 max 7

F(1,17255) 111.93 corr(u_i,Xb) 0.7413

LEVB Coef. Std. Error t P>|t| [95% Conf. Interval]

lag1 0.0572 0.0054 10.5800 0.0000 0.0466 0.0678 _cons 0.1376 0.0010 139.7800 0.0000 0.1357 0.1396 sigma_u 0.1565 sigma_e 0.0858 rho 0.7690 F(3964, 17255) 7.4800

R-sq: Obs per group

within 0.013 min 1

F(2,17254) 113.47

corr(u_i,Xb) 0.2205 Prob>F 0

lag1 0.0575 0.0054 10.6700 0.0000 0.0469 0.0681 GOOD 0.1629 0.0152 10.6900 0.0000 0.1331 0.1928 _cons 0.1188 0.0020 58.9800 0.0000 0.1149 0.1228 sigma_u 0.1555 sigma_e 0.0855 rho 0.7679 F(3964, 17255) 7.5100

within 0.0144 min 1

F(3,17253) 83.79

lag1 0.0578 0.0054 10.7400 0.0000 0.0473 0.0684 GOOD 0.1681 0.0153 11.0100 0.0000 0.1382 0.1980 INTEXG 0.0813 0.0166 4.9100 0.0000 0.0489 0.1138 _cons 0.1134 0.0023 49.3400 0.0000 0.1089 0.1179 sigma_u 0.1554 sigma_e 0.0854 rho 0.7679 F(3964, 17255) 7.5100

(22)

within 0.015 min 1

F(4,17252) 65.62

lag1 0.0580 0.0054 10.7700 0.0000 0.0474 0.0685 GOOD 0.1680 0.0153 11.0000 0.0000 0.1380 0.1979 INTEXG 0.0801 0.0166 4.8300 0.0000 0.0476 0.1125 EBITDA -0.0293 0.0088 -3.3100 0.0010 -0.0466 -0.0119 _cons 0.1161 0.0024 47.6200 0.0000 0.1113 0.1209 sigma_u 0.1554 sigma_e 0.0854 rho 0.7681 F(3964, 17255) 7.5100

within 0.015 min 1

F(5,17251) 52.49

lag1 0.0580 0.0054 10.7700 0.0000 0.0474 0.0685 GOOD 0.1679 0.0154 10.9300 0.0000 0.1378 0.1980 INTEXG 0.0800 0.0166 4.8100 0.0000 0.0474 0.1126 EBITDA -0.0294 0.0091 -3.2400 0.0010 -0.0471 -0.0116 LOGSAL 0.0002 0.0047 0.0500 0.9570 -0.0089 0.0094 _cons 0.1155 0.0114 10.1600 0.0000 0.0932 0.1378 sigma_u 0.1554 sigma_e 0.0854 rho 0.7680 F(3964, 17251) 7.4000

(23)

within 0.0152 min 1

F(6,17250) 44.48

lag1 0.0581 0.0054 10.7900 0.0000 0.0475 0.0686 GOOD 0.1691 0.0154 11.0100 0.0000 0.1390 0.1992 INTEXG 0.0809 0.0166 4.8600 0.0000 0.0483 0.1135 EBITDA -0.0298 0.0091 -3.2800 0.0010 -0.0476 -0.0120 LOGSAL 0.0009 0.0047 0.1900 0.8510 -0.0083 0.0100 ACID 0.0006 0.0003 2.0900 0.0370 0.0000 0.0012 _cons 0.1126 0.0114 9.8400 0.0000 0.0902 0.1351 sigma_u 0.1555 sigma_e 0.0854 rho 0.7683 F(3964,17250) 7.3800

within 0.0167 min 1

F(7,17249) 41.75

lag1 0.0586 0.0054 10.8900 0.0000 0.0481 0.0692 GOOD 0.1534 0.0157 9.7900 0.0000 0.1227 0.1841 INTEXG 0.0674 0.0168 4.0000 0.0000 0.0344 0.1004 EBITDA -0.0266 0.0091 -2.9300 0.0030 -0.0444 -0.0088 LOGSAL 0.0003 0.0047 0.0500 0.9570 -0.0089 0.0094 ACID 0.0009 0.0003 2.9400 0.0030 0.0003 0.0015 CASHEQ -0.0517 0.0103 -5.0000 0.0000 -0.0719 -0.0314 _cons 0.1213 0.0116 10.4900 0.0000 0.0986 0.1440 sigma_u 0.1543 sigma_e 0.0854 rho 0.7657 F(3964, 17249) 7.2600

(24)

within 0.017 min 1

F(8,17248) 37.38

lag1 0.0588 0.0054 10.9300 0.0000 0.0483 0.0694 GOOD 0.1546 0.0157 9.8600 0.0000 0.1239 0.1853 INTEXG 0.0688 0.0168 4.0900 0.0000 0.0358 0.1018 EBITDA -0.0182 0.0096 -1.8900 0.0590 -0.0372 0.0007 LOGSAL 0.0003 0.0047 0.0700 0.9480 -0.0088 0.0094 ACID 0.0009 0.0003 2.8800 0.0040 0.0003 0.0014 CASHEQ -0.0468 0.0105 -4.4600 0.0000 -0.0673 -0.0262 OPCFA -0.0198 0.0077 -2.5800 0.0100 -0.0348 -0.0048 _cons 0.1211 0.0116 10.4600 0.0000 0.0984 0.1437 sigma_u 0.1544 sigma_e 0.0853 rho 0.7660 F(3964, 17248) 7.2600