Do valuation multiples differ between countries in the EU15?

Do valuation multiples differ between countries

in the EU15?

Master Thesis

MSc Business Administration – Finance

University of Groningen

Faculty of Economics and Business

Utrecht, April 2010

Author:

Tom Ros

Student number:

s1464701

Do valuation multiples differ between countries

in the EU15?

T.G. Ros*

ABSTRACT

Performing a multiple analysis for a European firm can be tough when the search for comparable companies is limited by the geographical boundaries of the target’s country. When searching for comparables abroad, country differences could affect valuation multiples. After controlling for firm fundamentals, I find that companies in 8 of the EU15 countries differ from other European firms when using EBITDA multiples. By using EBIT and sales multiples, I find respectively 3 and 5 significant country effects. The smaller the target’s country, the harder it is to find comparable companies within the same country. The three smallest countries within the EU15 in terms of available listed companies are Luxembourg, Denmark, and Ireland. My results indicate that EBITDA multiples differ from other European companies in Luxembourg and Denmark, but not in Ireland. EBIT and sales multiples from the three smallest European countries do not differ from other European firms.

Keywords: Valuation multiples, Country effects, Enterprise value JEL classifications: F30, G12

* Tom Ros

1

Introduction

In this study I find European valuation multiples to differ significantly between countries, and that these differences cannot be fully explained by fundamentals in all European countries. Practitioners who choose to value a European firm by using multiples have to take into account the geographical location of the firm under valuation, even when all comparable companies are located within the European Union. Furthermore, I find that the type of multiple influences whether or not countries differ from each other. This paper shows which European countries differ from others by looking at EBITDA, EBIT, and sales multiples.

When valuing a firm, one can employ a number of direct and indirect methods. Direct methods1 value a company based on the firm’s accounting figures and estimates of future performance, combined with the cost of capital the firm faces when driving business. To calculate the value, information about the future capital structure, sales realizations, costs, and profitability is needed. Indirect methods however, also called multiple analyses, value a company based on the valuation of comparable companies. Valuation multiples represent firm value measured against a (forecasted) accounting statistic. Examples of multiples are value-to-EBITDA (EV/EBITDA), price-to-earnings (P/E), and market-to-book (M/B). These multiples are calculated for a set of benchmark companies and serve as an estimate for the value of the firm under valuation (target). Multiples are used as follows: Target value = (Benchmark multiple) * (Target accounting statistic), wherein the benchmark multiple can be calculated from the set of comparable companies. To determine firm value, the combination of direct and indirect methods is often used in practice. Multiple analyses can be used for different purposes ranging from IPO valuations, leveraged buyout transactions, and as a check to examine whether a value calculated by direct valuation is in line with the market’s valuation of comparable companies.

When analysts want to perform a multiple analysis, comparable firms are used to serve as a benchmark for the valuation of the target. The underlying assumption in a multiple analysis is that firms with similar characteristics and performance forecasts will have equal firm values. Therefore, the best result of a multiple valuation is obtained when the comparable companies match the target as much as possible. Ideally, comparable companies are selected on the basis of variables that explain cross-sectional differences in multiples, such that the multiple constructed from comparable firms will be similar to the unknown multiple of the target. The chosen type of multiple also leads to different value estimates. Each type of multiple is driven by other factors. P/E multiples depend on both stock price and earnings, whereas EV/Sales multiples are driven by firm revenue and the value of equity and debt together.

Comparable companies can be selected on variables that explain differences in firm valuation: the value drivers. These value drivers are, amongst others; growth, profitability, and risk. A practitioner can compare firms on the basis of a value driver to chose the comparable companies, use industry classification as a proxy for firm growth and risk, or can combine firm specific factors in a so-called warranted multiple approach. Industry classification is often used in practice and research as a basis to construct peer groups. Industries are assumed to contain firms with similar growth prospects and business risk, and therefore similar valuations. Apart from fundamentals,

1 _{A widely used example of direct measures is the Discounted Cash Flow method (DCF) which values all future cash flows}

multiples can be systematically different among countries. These differences are caused by the legal system, the level of alignment of financial and tax accounting, shareholder protection, and the use of accrual accounting that differ in each country. Country effects can systematically affect valuation multiples, and therefore affect the accuracy of the valuation estimates when comparable companies are selected from outside the target’s country. Since European firms are located in relatively small countries, the chance of finding suitable comparable companies within the geographical boundaries of the target’s country is relatively small compared to larger nations like the U.S.. Therefore, corporate finance analysts are forced to search for comparable companies abroad when valuing a company in a small country. But do valuation multiples differ systematically between countries in the EU15?

3

Literature Review

The valuation of a company is in theory often described as a matter that can be accomplished via the DCF approach (Kaplan and Ruback, 1995). In a DCF analysis, future cash flows of the company are estimated and then discounted using a discount rate that is based on the risk associated with the cash flow. However, estimating the future cash flows accurately and choosing the appropriate discount rate is difficult. To overcome this difficulty, valuation multiples can be used.

How to use multiples for valuing a firm

Valuation multiples present firm value measured against an accounting statistic. These multiples are calculated for a set of benchmark companies and serve as an estimate for the value of the target. Multiples are used to estimate the target’s value as follows: Target value = (Benchmark multiple) * (Target accounting statistic), in which the benchmark multiple can be calculated from the set of comparable companies. The accuracy of the value estimate is measured by the variance of the estimation errors: actual (enterprise) value based on the book value of equity (plus net debt) – valuation estimate (Alford, 1992). Higher accuracy of the estimate results in a more reliable valuation range. Prior studies find that the most accurate estimate is obtained when both the benchmark multiple and the accounting statistic of the target are forecasts of future values (Liu, Nissim, and Thomas, 2002; Schreiner and Spremann, 2007). The multiples resulting from the calculation can be compared to the multiples of other firms and to past or expected multiples of the same firm. A properly executed multiple analysis can help test the plausibility of cash flow forecasts and explain mismatches between a company’s performance and that of its competitors (Koller et al., 2005). A practitioner uses multiples for different purposes such as IPO valuations, fairness opinions, leveraged buyout transactions, and as a check of calculated value under a different valuation method (Bhojraj and Lee, 2002; Kaplan and Ruback, 1995; Herrmann and Richter, 2003).Multiples are in general used to proxy firm value as if its shares were traded under market conditions.

Multiples can be calculated for either the equity value or entity value of a firm. Equity value multiples are based on the firm’s stock price or market capitalization, whereas entity value multiples are based on the enterprise value of a firm. Although equity multiples are widely used in practice, they come with two major drawbacks: equity multiples mix operating and non-operating items in their valuation (1), and equity multiples are systematically affected by the firm’s capital structure (2) (Koller et al., 2005). Entity value represents the enterprise value of the firm: equity + net debt. The usage of entity value is preferred in corporate valuation because the cash flows of the company are generated by the total funds available instead of only the equity-holders’ part of the funds. This concept refers to ‘matching’: the numerator (value) and denominator (accounting statistic) of the multiple should be constructed such that the numerator is directly linked to the denominator. Otherwise, the multiples are incomparable to each other from an economic perspective (Schreiner and Spremann, 2007). When choosing the entity value instead of equity value, the effect of the firm’s capital structure is not completely ignored. Since valuation depends on the Weighted Average Cost of Capital (WACC)2, a more efficient capital structure leads to a higher enterprise value (Grinblatt and Titman, 2002). As

interest payments for debt financing are tax deductable, the amount of debt can be enlarged to reduce the WACC as long as the tax benefits can be cashed by sufficient taxable income. However, the cost of this increased leverage is the larger risk for bankruptcy (Koller et al., 2005). Herrmann and Richter (2003) compare the accuracy of both entity and equity multiples in the U.S. and Europe in the period 1997 – 1999, and find entity multiples to outperform equity multiples. The EV/Sales multiple leads to more accurate results than M/B multiples, following the conclusions of Bhojraj and Lee (2003). Furthermore, prior research shows that multiples based on earnings lead to higher valuation accuracy than multiples constructed from book value of equity or sales do (Liu, Nissim, and Thomas, 2007; Schreiner and Spremann, 2007).

Factors that drive multiples

Valuation theory states that the firm’s enterprise value is driven by the fundamentals Growth and Return On

Invested Capital (ROIC) (Koller et al., 2005). Growth can be regarded as sales growth, and the ROIC is defined as being the return as a percentage of all funds invested in the firm. ROIC can be proxied by a firm’s

profitability: the return generated by the recorded sales. Prior studies argue in general that the best result from a multiple analysis is achieved when the peer group matches the target firm as much as possible in terms of fundamentals (Henschke and Homburg, 2009; Kim and Ritter, 1999; Richter, 2005).

To find firms that match the target on fundamentals, a peer search is often started within the industry of the target. Industry classification is used in both practice and academic literature because firms within an industry are expected to have similar cost of capital and growth prospects (Koller et al., 2005). The cost of capital represents the risk for equity- and debt holders when investing in the company, and is expected to be similar for firms within the same industry. Growth prospects are also expected to be similar for companies within the same industry, because the firms drive business on the same market under the same market conditions. Henschke and Homburg (2009) search for differences between industry-based value estimates and actual stock prices in the U.S. in the period 1985 – 2004. Using equity valuation multiples, Henschke and Homburg find that these differences can be explained by firm specific differences in sales growth, risk and return. This means that industry multiples provide analysts a value estimate that has to be corrected for firm specific characteristics. It also implies that even within industries, company fundamentals differ from each other. Unfortunately, the authors do not inverse their research by asking whether industry effects add explanatory power in explaining differences between multiples next to the firm specific characteristics. Baker and Ruback (1999) also search for industry effects in S&P500 firms in the year 1995 and conclude industry classification to significantly affect multiples in all 22 industries. Alford (1992) examines the effect of the set of comparable firms on the accuracy of the P/E valuation method in the U.S. in the years 1978, 1982, and 1986. He finds no differences in the accuracy of valuation estimates between peer groups wherein firms are selected on a match between risk and growth, and peer groups based on industry classification. Moreover, Alford finds that combining industry classification with growth and risk variables does not increase valuation accuracy. This indicates that industry classification does not possess explanatory power when combined with firm risk and sales growth.

5 EV / Equity Multiples included in analysis Comparable

selection based on Outliers? S ample requirements

S ample

years S ample description

S ignificant country effects found? Results Alford (1992) Equity P/E Industry, firm size,

earnings growth

No information EBIT is available and positive, fiscal years ends in December

1978, 1982, 1986

1636 firms in 1978, 1591 in 1982 and 1471 in 1986

N/A Industry membership or a combination of risk and earnings growth are good proxies for comparable selection

Baker and Ruback (1999) unclear unclear M ultiple No information M V-equity, book value of debt, revenu, EBITDA and EBIT are available

1995 S&P500 firms as a basis, 225 left out after selection criteria

N/A EBITDA outperforms the EBIT and Sales multiple

Bhojraj and Lee (2002) Both EV/Sales + M /B ‘Warranted multiple’ Top and bottom 1% observations on each multiple and variable is deleted

U.S. firm with M arket Cap > $100 mln.

1982 - 1998 3,515 firms from the NYSE, NASDAQ, AM EX in 2000

N/A Warranted multiple outperforms straightforward methods based on industry classification

Bhojraj and Lee (2003) Both EV/Sales + P/E + forward P/E + M /B

‘Warranted multiple’ Top and bottom 3% observations on each multiple and variable is deleted

Earnings have to be available and positive

1990 - 2000 26,626 firm years from G7 country firms

No EV/S multiples outperform M /B multiples in terms of valuation accuracy. Growth and R&D information enhances explanatory power, but leverage does not

Ditmann and Weiner (2005) EV EV/EBIT Country, Region, OECD

Top and bottom 1% observations on each multiple and variable is deleted

Total assets, EBIT and total debt are positive and available

1993 - 2002 67,433 firm years over 10 years Yes U.K. and U.S. firms are comparable in firm valuation, and differ from the EU company valuations

Henschke and Homburg (2009) Equity P/E + M /B Industry Top and bottom 1% observations on each variable is deleted

Book value > $10 mln. and net sales are required to be positive

1985 - 2004 24,308 U.S. firm years N/A Valuation errors based on industry-based comparable company selection can be explained by differences in financial ratios

Herrmann and Richter (2003) Both EV/EBITDAAT + EV/EBIAT + EV/IC + EV/Sales + P/E + M /B

Expected growth rate, reinvestment rate, leverage

No information no information 1997 - 1999 524 U.S. and 830 large European firms in 1998. After criteria: 645 in 1997, 665 in 1998 and 664 in 1999

N/A Entity multiples lead to more accurate results than equity multiples. Entity multiples based on earnings outperform entity multiples based on invested capital

Kaplan and Ruback (1995) EV EV/EBITDA Industry No information EV > $40m. and industry classification must be available

1980 - 1989 51 highly levered transactions N/A Comparable company multiple valuation performs worst, followed by comparable transaction valuation, whereas DCF performs best

Kim and Ritter (1999) Both P/E + M /B + P/Sales Recent IPO's in a similar industry

P/E ratios above 100 are maximized at 100, M /B ratios above 10 are maximized at 10

no information 1992 - 1993 190 IPO's N/A Valuation accuracy is higher for old firms than for young ones. EV/Sales accuracy is improved when multiples are corrected for profitability and leverage

Lie and Lie (2002) Both EV/EBITDA + EV/EBIT + EV/Sales + EV/Book Value + P/E + forward P/E

Industry No information no information 1998 8,621 firms from the worldwide Compustat Database

N/A The estimation errors are negative in mean and EBITDA multiples outperform EBIT multiples in terms of accuracy

Liu, Nissim and Thomas (2002) Equity E/P + CF/P + Dividends/P + Sales/P

Industry The interquartile range of the valuation errors is used

M ultiples and value drivers have to be positive

1987 - 2001 Firms from Australia, Canada, France, Germany, Hong Kong, Japan, South Africa, Taiwan, U.K., and the U.S.

Yes M ultiples based on forward looking estimates outperform trailing multiples. There are significant differences among countries in terms of valuations and pricing errors

Liu, Nissim and Thomas (2007) Both EBITDA/P + M /B + E/P + Sales/P

Industry The interquartile range of the valuation errors is used

M ultiples have to be positive 1982 - 1999 19,879 U.S. firm years N/A M ultiples based on forward looking estimates outperform EBITDA and M /B valuations. EBITDA and M /B valuations dominate estimates based on cash flow

Richter (2005) Both P/E + M /B Expected growth rate, reinvestment rate, leverage

No information no information 1996 - 1998 332 U.S. firms, 210 German firms and 455 other European firms (997 total)

N/A Valuations based on value-driver-comparables outperform simple SIC-code classified peer group valuations

Schreiner and Spremann (2007) Both P/E + M /B + P/Sales Industry No information M arket Cap > $200m and positive net debt are required

1996 - 2005 Dow Jones STOXX 600 + S&P500 N/A Equity value multiples outperform EV multiples, knowledge related multiples outperform traditional ones, forward looking multiples outperform trailing ones

TABLE I

Overview of prior research

Instead of using industry classification to proxy fundamentals, several studies focus on actual fundamentals. Growth is included in the papers of Herrmann and Richter (2003), and Alford (1992) to explain differences in multiples. Herrmann and Richter find that growth explains valuation differences for both entity and equity multiples. Alford only finds growth to explain differences in P/E multiples when the information concerning the company’s growth is combined with a measure of risk such as firm size.

Risk is another important aspect in firm valuation, because the risk that debt- and equity holders face by investing in the firm will affect firm value as well. Henschke and Homburg (2009) chose firm size to be a measure for risk by using the log of total assets. Although Henschke and Homburg hypothesize that smaller firms are more vulnerable to bankruptcy and will therefore be traded at lower multiples, no empirical evidence for their statement is found. Lie and Lie (2002) split their sample of 8,621 firms worldwide in three size classes based on the book value of total assets and conclude that valuation errors are positive for small firms and negative for large firms. This implies that small firms are overvalued whereas large firms are typically undervalued.

Bhojraj and Lee (2003) aim at constructing an alternative company selection method to compete with the often used industry classification. The study of Bhojraj and Lee indicates that the most accurate value estimate is obtained when the selection process is based on a number of firm specific factors instead of only one. One of the significant factors in explaining EV/Sales multiples is profitability. Profitability is defined as EBIT/Sales and turns out to be positively and significantly correlated to the valuation multiple.

Country differences when using multiples

7

global scale, by selecting 10 large countries including France, Germany and the U.K.. The researchers find estimation errors to differ significantly between these 10 large countries in the period 1987 – 2001, implying country effects are relevant in worldwide valuation practices. Both Dittmann and Weiner (2005) and Liu, Nissim, and Thomas (2002) construct their peer groups on the basis of one variable (respectively return on assets (ROA) and industry), and therefore only explain differences in valuation errors from one perspective. The explanatory effect of the country variable could disappear or diminish when the errors are also controlled for a larger number of firm specific factors, next to ROA or industry.

Bhojraj and Lee (2003) do take into account a number of firm characteristics and find country effects to be negligible for G7 countries (including France, Germany, Italy, and the U.K.) in the period 1990 – 2000.

Combining the results of Bhojraj and Lee (2003) with Dittmann and Weiner (2005) and Liu, Nisssim, and Thomas (2002); differences in worldwide multiple valuation can be explained by country information when firm specific factors are not taken into account. However, these effects diminish on a global scale when other firm characteristics are added to the analysis, implying the country effect is to some extend a substitute variable for fundamentals. Results of the country effect as explanatory variable should be interpreted with care though, since some countries heavily depend on a small number of industries. In such a scenario, the found country effect is closely related to the industry effect.

My study contributes to prior research by examining three different types of multiples that are all based on enterprise value, within 15 European countries. Small European countries are also included in the analysis, because the necessity of searching for comparable companies abroad is highest in small countries. The majority of currently available literature searches for country effects on a global scale between large countries, or only searches for explanatory variables within the U.S.. In contrast to current valuation literature, both a univariate and a multivariate analysis are presented. This results in information concerning the country effect both with and without taking into account firm specific factors.

Data and Methodology

The European companies included in the dataset originate from the Amadeus database. This database contains 307,717 firms with their office registered in one of the EU15 countries of 19963. The sample is therefore comparable to the European part of the sample in Dittmann and Weiner (2005). The companies have to be listed on a stock exchange in order to value their shares (-301,350 firms). Since the assets of financial companies are relatively liquid, these firms are more easy to value than non-financial firms (Lie and Lie, 2002). Combining financial and non-financial companies in the dataset could influence the results and therefore all financial firms are omitted from the dataset (-1,657 firms). With all requirements met, the dataset contains 4,710 European firms.

Financial information concerning annual sales, EBITDA, EBIT, and Enterprise Value over the period 1998 – 2008 is provided by Datastream. EBITDA (item WC18198) represents the earnings of a firm before depreciation is taken into account, and the firm’s interest expenses and income taxes are paid. The earnings of a firm before interest expenses and income taxes is called EBIT (item WC18191). Annual sales of a company are defined as net sales (item WC01001): gross sales plus other operating revenue minus discounts, allowances and returns. Finally, Enterprise Value (EV, item WC18100) is defined as a firm’s market capitalization plus preferred stock, minority interest, and total debt minus cash. In both practice and scientific studies, forward looking financial figures are preferred to historical data. Unfortunately, forecasts of growth rates and additional financial information are not available in Datastream, and therefore not used in this paper.

TABLE II

Descriptive statistics multiples

This table shows how each type of multiple is distributed over the period 1998 – 2008. All available multiples are included. EV is defined as market capitalization + preferred stock + minority interest + net debt. EBITDA, EBIT, Sales, and EV are collected on an annual basis.

3 _{Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Portugal, Spain, Sweden, The}

Netherlands, United Kingdom.

EV/EBITDA EV/EBIT EV/Sale s

Me dian 8.1 11.9 1.2

Me an 11.7 17.7 3.9

Harmonic Me an 3.0 3.7 0.3

Number of obse rvations 23,332 21,687 28,081

9

When including all ratios of respectively EV/EBITDA (EBITDA multiple), EV/EBIT (EBIT multiple), and EV/Sales (Sales multiple), I find that all types of multiples in the dataset are not normally distributed given the high Jarque-Bera values in Table II. Therefore I will use median multiples and non-parametric statistical tests in my study. The mean values are structurally higher than medians, implying a number of large positive outliers are included in the set.

The sample contains a total of 33,987 unique firm years. Outliers are dealt with in line with prior research: all negative values on sales, EBITDA, and EBIT are deleted because estimating firm value with a negative accounting statistic results in a negative value. Furthermore, the upper and lower 1% of all observed financial figures is omitted from the analysis in line with prior studies (Bhojraj and Lee, 2002; Dittmann and Weiner, 2005; Henschke and Homburg, 2009).

TABLE III

Country descriptive statistics

This table shows the number of listed companies in the dataset for each country. The median multiples for each country are presented over the period 1998 – 2008. The Kruskal Wallis test is run to determine whether the country classification is relevant in explaining differences between multiples.

The country variable yields the country where the firm has its registered office address. Table III shows that the number of available listed companies in Luxembourg, Denmark, and Ireland is limited. Constructing a peer group for a firm in one of these countries is tough when the search for comparable firms is limited by geographical boundaries. Moreover, the availability of listed firms in general does not cover the topic of comparability when searching for comparable companies. It merely provides us the size of the pool of possible

EV/EBITDA EV/EBIT EV/Sale s

# companies MEDIAN MEDIAN MEDIAN

Austria 120 7.5 11.1 1.0 Be lgium 95 7.1 12.0 1.1 De nmark 27 7.7 13.3 0.9 Finland 400 8.7 12.2 1.2 France 675 7.9 12.0 1.0 Ge rmany 835 7.4 11.6 1.0 Gre e ce 232 9.7 14.6 1.5 Ireland 40 8.2 11.5 1.0 Italy 188 8.1 12.3 1.4 Luxe mbourg 5 6.0 12.7 1.5 Portugal 116 7.6 11.8 1.1 Spain 130 8.9 12.7 1.7 Swe den 141 7.7 12.6 1.1

The Ne the rlands 133 7.5 10.8 0.9

Unite d Kingdom 1,573 8.3 11.3 1.3

Poole d 4,710 8.1 11.9 1.2

Kruskal Wallis Te st statistic 335.8 291.4 587.5

candidates. Highest median EBITDA and EBIT multiples are recorded in Greece, whereas the median Spanish firm has the highest sales multiple. Countries with high earnings multiples such as Greece do not necessarily have high sales multiples as well. The other way around, high sales multiples in a country do not mean that earnings multiples are high too (Luxembourg). The Kruskal Wallis test indicates that country classification is relevant in explaining differences in multiples, but it does not provide information on whether there are significant differences for all countries.

Next I discuss the different firm specific factors that are used in the multivariate analysis to explain differences in multiples along with the country effect. All factors are found to be relevant in explaining differences between multiples. The results of the tests performed to examine the relevance are presented in Appendix I. Table IV provides an overview and definition of the variables that are discussed in the following paragraphs.

Growth is defined as the mean sales growth over the past three years (Koller et al., 2005). Since growth rates fluctuate over time for each firm, growth rates are normalized in the sample to obtain a future estimate. On the one hand, growth differs due to the life cycle stage: new companies in emerging markets grow faster than matured companies do. On the other hand, growth depends on the firm’s current amount of sales: growing at a rate of 30% is more easy with current sales of 50K than with a sales total of 50 million.

TABLE IV

Definition of dependent variables

This table shows all dependent variables used in the multivariate analysis to explain differences in multiples

A firm’s profitability is defined as EBIT divided by sales. Profitability is included in the analysis to proxy ROIC. To avoid the effect of outliers caused by exceptional high or low earnings in a certain year, the estimate for future profitability is defined as the average profitability rate over the past three years.

Size is measured as annual sales, instead of the often used market capitalization or total assets. The risk of the firm is proxied by firm size. From a corporate valuation perspective, firm size should be represented by the firm’s entity value. However, since the entity value is also an input variable in the formula used to calculate the multiples, sales is chosen to proxy firm size. Descriptive statistics concerning the size, growth, and profitability from the firms included in the sample can be found in Table V.

Variable Variable de scription

Growth The average past three year sales growth

Size Size is proxied by the sales of the firm

Profitability The average past three year profitability rate

Industry NACE Rev.2 industry code

Year Year in which the multiple is recorded

11

Growth Size Profitability

Me dian 9.5% € 74.6m 9.7%

Me an 18.2% € 980.2m 19.5%

Harmonic Me an n.a. € 0.8m 7.2%

Numbe r of obse rvations 26,299 34,198 18,067

Variance 13.9% ∞* 62.2% Standard De viation 37.2% € 3,309.9m 78.8% Q1 1.0% € 14.3m 5.9% Q3 23.0% € 412.1m 16.9% Ske wne ss 3.2 6.2 38.1 Kurtosis 14.2 45.3 1,892.0 Jarque-Be ra 180,792 2,768,320 ∞*

Since the research focuses on European companies, the industry classification used is also originated from Europe: NACE4. Industry is defined as the two-digit NACE Rev 2. code of a company, resulting in ten broadly defined industry classes. Alford (1992) suggests industry classification should be based on a code of three digits when you want to fully explain differences in multiples by industry. My objective however, is to control for industry effects in my analysis instead of explaining these differences by the industry variable. Moreover, industry classification on two digits results in 10 industries. Adding an extra digit increases the number of industries to 100. With 4,710 firms in the dataset, the chance that a small number of firms represent a whole industry increases substantially when adding an extra digit.

TABLE V

Growth, size and profitability descriptive statistics

This table shows the descriptive statistics concerning the variables growth, size and profitability over the period 1998 – 2008. Growth is defined as the mean past three year sales growth, size is measured by the firm’s annual sales in millions of euro’s, and profitability is defined as the past three year average profitability rate (EBIT/Sales) of the firm.

* larger than 1 billion

Finally, the year in which the multiple is recorded is also taken into account. Apart from all firm characteristics in explaining differences in multiples, the state of the economy is important as well. Valuations differ over time for all countries, because the drivers of value are to some extend dependent on the economy’s state. To capture time effects that are implicitly included in multiples, I include year dummies in my analysis. Appendix II provides information on how EBITDA, EBIT, and sales multiples evolve over time, and Appendix I shows that the classification of multiples into years is relevant for explaining differences in valuations.

Value =NOPLAT 1 − g_{WACC − g}ROIC ℎ NOPLAT = EBITA1 − T

TABLE VI

Number of observed multiples in each country per year

This table shows the number of observations for each country per year. The number represents the sum of all EBITDA, EBIT and sales multiples in a country per year.

Table VI shows that the number of available multiples in a country differs over time. Most multiples are recorded in the years 2006 and 2007. This is due to the number of listed firms in a certain period and to the policy concerning outliers in the study: when a firm has negative sales or earnings, the multiple is omitted from the analysis. This policy decreases the number of available multiples when the number of firms reporting losses increases. Appendix II shows that median EBITDA, EBIT, and sales multiples in the set are lowest in the years 2002 and 2008, which are both known to be years of crisis in the economic world. Appendix III provides insight into the number of observations that are collected for each type of multiple.

What drives multiples

The multiples used in this study are derived from the basic formula of calculating firm value in corporate valuation, also called The Zen of Corporate Finance (Koller et al., 2005). The formula relates a company’s value to the fundamental drivers of economic value: ROIC, growth, and the cost of capital.

(1)

With

• Value Enterprise value

• NOPLAT Net Operating Profits Less Adjusted Taxes • g Growth

• ROIC Return on Invested Capital • WACC Weighted Average Cost of Capital • T Corporate tax rate

13 Valueit EBITAit ! =1 − T 1 − g_{WACC− g}ROIC Valueit EBITDAit ! =_EBITDAEBITAit it∗ 1 − T 1 − g_ROIC WACC− g Valueit EBITit ! =EBITA_EBITit it ∗ 1 − T 1 − g_ROIC WACC− g Valueit Salesit ! =EBITA_Salesit it ∗ 1 − T 1 − g_ROIC WACC − g

The formula in equation 1 assumes ROIC and growth to be constant in the future. Given equation 1, EBITA multiples of firm i in year t are calculated as follows:

(2)

The EBITDA, EBIT and sales multiples used in this study are derived from equation 2 straightforward.

(3)

(4)

(5)

Differences in EBITDA multiples (3) are driven by depreciation charges of firms. ‘Asset-light’ firms have relatively low depreciation charges, and therefore higher EBITDA multiples than ‘asset-heavy’ firms. Countries that rely heavily on manufacturing industries will have lower EBITDA multiples than countries housing more service-based industries.

When EBIT multiples (4) differ between firms, these differences are due to differences in amortization. Amortization will show up in a financial report when a firm acquires another company. The number of mergers and acquisitions differs between countries, resulting in amortization differences between countries. From the 1st of January 2005 on, all European listed firms are obligated to calculate and report amortization following specific rules (IFRS). Nevertheless, the amount that is amortized by firms differs and causes EBIT multiples to differ is well: higher amortization results in higher EBIT multiples.

Finally, differences in sales multiples (5) are driven by the firm’s profitability rate: EBITA/Sales. Higher profitability leads to higher sales multiples. Since some countries house a higher amount of profitable firms than others, sales multiples can differ systematically among countries.

Methodology

I want to find out whether the country where a firm is headquartered influences the valuation of that firm. To do so, I will perform a univariate and a multivariate analysis over the period 1998 – 2008 in 15 European countries.

(6)

section I already found the country classification to be relevant. The Mann-Whitney test provides information on which countries actually differ from others in terms of valuation.

The multivariate analysis searches for valuation differences between European countries, when firm fundamentals (growth, size, and profitability), industry membership, and the year in which the multiple is recorded are taken into account. Since both sales growth and profitability are measured over the past three years with a research period of 1998 – 2008, sales and EBIT information is collected for the period 1995 – 1997 as well. To combine both time-series and cross-sectional information contained in the sample, a panel estimation technique is used with the multiple itself being the dependent variable. This method ensures that cross-sectional effects in the sample are captured while taking into account the effect of time differences. Another major advantage of using a regression-based model, is that it allows explanatory variables to simultaneously affect the dependent variable.

A panel regression can be run by allowing for either fixed or random effects. Both types of effects add a constant to the regression analysis for each firm. On the one hand, fixed effects add an individual constant to each firm’s regression formula. Random effects on the other hand, add a random constant that is uncorrelated with the estimation error term to each regression formula. I use random effects in my panel estimation because a number of explanatory variables in the regression is time invariant: a firm’s country and industry do not change over time.

The multiple of firm i in year t is estimated as follows:

M'(= C'(+ β+∗ Growth'(+ β1∗ Size'(+ β3∗ Profitability'(+ β7∗ Industry'+

β:∗ Year + β<∗ Country'+ ε'(

With

• Mit Multiple of the firm i in year t

• Growthit Average sales growth of the firm in the years t-1, t-2, and t-3 • Sizeit Firm size in year t

• Profitabilityit Average sales growth of the firm in the years t-1, t-2, and t-3 • Industryi Firm industry classification

• Year The year in which the multiple is recorded • Countryi The firm’s country

Equation 6 is used to estimate EBITDA, EBIT, and sales multiples. Definitions of the dependent variables in the regression analysis are shown in Table VII.

15

the year 1998 from the regression, so that all possible differences from time dummies are relative to the first year in the dataset. Furthermore, the ‘Arts, Entertainment, and Recreation’ industry is omitted because the number of firms from this industry (132) is the lowest in the dataset.

All firm characteristics and time effects are added to the country variable to find whether country information matters when more information about the firm is available. I expect growth to positively influence all types of multiples, because growth is a major value driver in corporate valuation (Koller et al., 2005). The more growth, the higher firm value will be. In contrast, I expect size to negatively influence multiples based on findings of Lie and Lie (2002) and the fact that larger companies have lower growth prospects than smaller firms. Profitability proxies the firm’s ability to generate returns from revenue, and therefore I expect profitability to positively influence firm valuation.

TABLE VII

Variables and descriptions used in the multivariate analysis

This table provides information on the variables included in the multivariate analysis. The hypothesized signs represent the expected relationship of the variable with the multiple, derived from prior research.

To actually test the country effect in the multivariate analysis, all firm specific factors and year dummies are included in the equation together with one country dummy. For each country, a separate regression analysis is run to assess the relevance of the country variable. To control for possible heteroskedasticity in the dataset, I use White’s consistent standard errors.

Variable Hypothe size d Sign Variable description

Mit N/A Enterprise valuation multiples:

EV/EBITDA, EV/EBIT and EV/Sales

Growthit + Sales growth in percentages

Sizeit - Size in billions of euro's

Profitabilityit + Profitability in percentages

Industryi +/- Series of industry dummies

Year +/- Series of year dummies

Results

To determine whether the country effect is important in explaining differences between European firm valuations, both a univariate and a multivariate analysis are run. The multivariate analysis controls for growth, size, profitability, industry classification, and the year wherein the multiple is recorded. The results of the univariate analysis indicate that country differences do matter when valuing a firm from 10 of the 15 countries. However, the number of countries that differ significantly from each other decreases to approximately 5 of the 15 countries in the multivariate analysis. Nevertheless, the presence of country effects in the multivariate analysis implies that not all differences between European multiples can be explained by firm fundamentals, industry classification and the year in which the multiple is recorded.

TABLE VIII

Results of univariate and multivariate analysis

This table shows the results on the country variable for both the univariate and the multivariate analysis in 1998 – 2008. The left sign represents the comparison of median country multiples to all other multiples in the set by using the Mann-Whitney test. The right sign shows the relationship between country and multiple when firm fundamentals (growth, size, profitability), industry classification, and the year in which the multiple is recorded are also taken into account by estimating a panel regression. - means multiples in the country are lower than in other countries at a p-value of 0.05, whereas 0 means that no differences with multiples from other countries are found. Finally, + says that the country multiples are higher than multiples in other countries.

The results and implications of the univariate analysis are discussed first. After that, the results from the multivariate analysis are dealt with. Both discussions refer to the results presented in Table VIII.

EV/EBITDA EV/EBIT EV/Sales

17

Univariate analysis

In the univariate analysis, the country where the firm is headquartered is the only variable that is used to differentiate between a long list of multiples. Whether the median multiples from a country are above, below or indifferent from the European median is presented in Table VIII on the left of the dash. Because the effect of 15 countries is tested for three multiple types, the results for 45 country dummies is presented. 17 of the 45 country effects are found to be positive, implying firms in 17 countries are valued significantly above European firms from other countries at a p-value of 0.05. There are no differences in firm valuation for 14 countries, and for another 14 countries the valuation multiples turn out to be below the EU15 median.

Using EBITDA multiples results in the highest number of significant country dummies (11). Both EBIT and sales multiples indicate that country differences are present for 10 of the 15 countries. In Austria and The Netherlands, all three types of multiples are below the EU15 median. In contrast, firms from Spain, Italy, Greece, and Finland are above the EU15 median for each type of multiple.

EBITDA multiples are driven by depreciation charges. Differences in EBIT multiples are caused by amortization differences, and sales multiples are driven by profitability. Consider Ireland and Sweden. Both countries have indifferent EBITDA multiples, but EBIT multiples are significantly higher in Ireland and significantly lower in Sweden. This result indicates that differences in firm valuation between Sweden and Ireland are due to differences in amortization. Since higher amortization results in higher EBIT multiples, I indicate mergers and acquisitions to be more common in Ireland than they are in Sweden. Now consider Belgium and Portugal. Sales and EBIT multiples in both countries are indifferent from a EU15 perspective, but Belgian EBITDA multiples are lower and Portuguese EBITDA multiples are indifferent. Given that higher depreciation results in lower EBITDA multiples, my findings suggest that Belgian firms have higher depreciation charges than Portuguese companies have.

When searching for comparable firms within the geographical boundaries of the target firm, the most problems concerning availability occur when the number of listed companies within a country is limited. In my sample, this problem will be most likely for target firms headquartered in either Luxembourg, Ireland or Denmark. The univariate analysis indicates that Luxembourgian EBITDA multiples are not different from the European median whereas EBIT and sales multiples from Luxembourg are valued above the European median firm. Apparently, the value driver of EBITDA multiples, depreciation, is not significantly different in Luxembourg compared to other European countries. The higher EBIT multiples indicate that amortization of Luxembourgian firms is above the European average, given the positive relationship between amortization and EBIT multiples. Finally, the sales multiple indicates that profitability rates are high in Luxembourg from a European perspective, since higher profitability leads to higher sales multiples. In Ireland, firm valuations only differ negatively from the set when using EBIT based multiples. The EBITDA multiple in contrast, is the only type of multiple that differs significantly from other European firms in Denmark.

Multivariate analysis

presented in Table VIII on the right of the dash. The actual coefficients of the country dummies can be found in Appendix IV. Similar to the univariate analysis, the effect of 15 countries is tested for three multiple types, resulting in 45 panel estimations. Of these 45 results, 5 regressions contain positively different country dummies. 11 country dummies are significantly lower, and 29 country effects are found to be insignificant.

The results indicate that when using EBITDA multiples, the country effect is significant for multiples from 8 of the 15 countries. Using EBIT or sales multiples results in respectively 3 and 5 countries that differ significantly from other countries. From the 15 European nations, 4 have the same sign on EBITDA, EBIT, and sales multiples. This sign represents that differences are insignificant for each of these 4 countries, meaning no country is valued above or below other European firms on each of the three types of multiples. 5 out of 15 countries have both negative country dummies on EBITDA and insignificant EBIT dummies. Consider France and Germany. Although both countries do not differ from other European firms on EBIT multiples, their firms do differ when looking at sales multiples. Given that French firms are valued lower than other EU15 companies on sales multiples, and German firms are indifferent given its sales multiples, the results indicate that German firms are more profitable than French firms. Dittmann and Weiner (2005) used the EBIT multiple to examine country effects within the EU15, and found English, Danish, and Greek multiples to differ significantly from others. My results are in line with Dittmann and Weiner for the U.K. and Greece, but I do not find Danish multiples to differ significantly from other European firm valuations. Instead, Swedish EBIT multiples are significantly different in my study whereas this country effect is not found by Dittmann and Weiner. The difference can be explained by the fact that Dittmann and Weiner search for country effects in the period 1993 – 2002, while my sample covers the years 1998 – 2008. Furthermore, the U.S. and Japanese firms in their dataset are not included in my research.

Both Liu, Nissim, and Thomas (2002) and Dittmann and Weiner (2005) first control multiples for either industry classification and ROA, and then search for country effects. Both studies find country effects to be present for certain countries in the EU15. Bhojraj and Lee (2003) control multiples for 9 firm specific factors before searching for the explanatory power of country effects, and find that country differences are negligible for G7 countries after controlling for firm specific factors. My study controls multiples for 5 factors and is therefore positioned in between the studies of Dittmann and Weiner (2005), Liu, Nissim, and Thomas (2002) on the one hand, and the study of Bhojraj and Lee (2003) on the other. I find country effects to be significant for a number of countries, just like Liu, Nissim, and Thomas (2002) and Dittmann and Weiner (2005), even when the multiples are controlled for more than one explanatory variable. However, I do not use as many explanatory variables as Bhojraj and Lee (2003), which could explain why I find significant country effects in my study in the first place.

19

Next I determine how fundamentals are affecting the different multiples in the multivariate analysis. The results of the found signs on fundamentals in the panel regression are presented in Table IX. I find sales growth to positively affect EBITDA, EBIT, and sales multiples. This finding is in line with both Alford (1992) and Koller et al. (2005), who both argue sales growth to be a major value driver in multiple valuation. Firm size is negatively related to all three types of multiples in the study, and therefore in line with findings of Lie and Lie (2002). The size of a firm is represented by its annual sales, implying that higher sales results in lower growth forecasts and therefore lower multiples. The effect of profitability on the multiple however, differs for each type of multiple. The positive effect of profitability on the sales multiple can be explained by the fact that firm sales influences both profitability (EBIT/Sales) and sales multiples (EV/Sales) directly. Relatively high sales will negatively affect both the multiple and profitability ratio, resulting in a positive relationship. Profitability is directly linked to the EBIT multiple as well, but in the opposite direction compared to sales: all other numbers being equal, higher EBIT results in a lower EBIT multiple and higher profitability. EBITDA multiples are not significantly influenced by profitability rates.

TABLE IX

Results of fundamentals in multivariate analysis

This table provides the signs of the fundamentals growth, size and profitability that are found in the multivariate analysis for each multiple. - = the relationship between the fundamental and the multiple is negative at a p-value of 0.05, 0 = no differences between multiples with different fundamental values + = positive significant relationship between the fundamental and the multiple.

Apart from firm fundamentals, the multivariate analysis also accounts for time and industry effects. Although a number of industry and year dummies are significant, the results are relative to the omitted dummies of the ‘Arts, Entertainment, and Recreation’ industry and the year 1998. The signs of both industry and year dummies is presented in Appendix V.

The chosen firm specific factors in the multivariate analysis might not explain differences in each multiple equally well. Industry classification, size, growth, year, and profitability could be suitable for explaining differences in EBITDA multiples and suboptimal when explaining differences in sales multiples. To proxy whether the chosen factors explain differences between EBITDA, EBIT, and sales multiples, the average R-squared values of the 15 panel analyses for each type of multiple are presented in Table X.

EV/EBITDA EV/EBIT EV/Sales

Growth + + +

Size - -

TABLE X

R-squared in multivariate analyses

This table provides the average R-squared values from the 45 panel regression analyses that are run. For each type of multiple, the regression is run 15 times with a different country dummy in each formula. The R-squared displayed is the average R-squared from the 15 regression for each type of multiple.

The average R-squared values are based on the 15 country regressions that are run to determine country effects, and indicate that the chosen explanatory variables suit the sales multiple best. Almost 19% of all deviations in sales multiples are explained, whereas this is only 6% for EBIT multiples. Damodaran (Website) also searches for factors with explanatory power in European valuation multiples. Damodaran reports higher R-squared values for the EBITDA (0.245) and sales multiples (0.420) in January 2010. Differences in the found R-squared values can be explained by sample and methodology differences. The estimation formula used by Damodaran is different for each type of multiple. Furthermore, Damodaran adjusts the explanatory variables in the estimation formula for each geographical region (Japan, U.S., Europe) for the same multiple type. This indicates he searches for the highest R-squared by being selective in choosing explanatory variables. Furthermore, the growth rates in Damodaran’s estimation model are forecasts of growth rates. Although the R-squared values in my study are relatively low, the explanatory variables in the panel regression make sense from an economical perspective.

EV/EBITDA EV/EBIT EV/Sales

21

Conclusion

For valuing a firm, a practitioner can use a multiple analysis to proxy firm value based on comparable companies. When the number of comparable companies within the geographical boundaries of the target firm is insufficient, the search for firms has to be expanded abroad. This study focuses on the country effect within EU15 multiple valuation based on the enterprise value of firms.

In the univariate analysis I find country effects to be significant for 11 of the 15 countries when using EBITDA multiples, and 10 of the 15 countries when using sales and EBIT multiples. The number of countries for which the valuation is significantly different from other countries in the set decreases when the firm specific factors growth, size and profitability together with industry classification and the year wherein the multiple is recorded are taken into account. In the multivariate analysis on the EU15 countries in the dataset I find country effects to be significant for 8 countries when using EBITDA multiples, 3 using EBIT multiples, and 5 when the sales multiples are used. The fact that the number of country effects decreases says that for some countries, the difference in the univariate analysis is explained by fundamentals. Nevertheless, since country effects are still present in the multivariate analysis, the selected fundamentals cannot explain differences in firm valuation for all countries.

When valuing a firm located in a country with a limited number of listed firms available, the search for comparable companies abroad is most likely. I find that European EBIT and sales multiples are not significantly different from domestic multiples in the three smallest countries in the sample: Luxembourg, Denmark, and Ireland. EBITDA multiples do differ between European countries and multiples from Luxembourg and Denmark. This means that when using EBIT or sales multiples from EU15 firms to value a Danish firm, geographical boundaries are not systematically influencing the multiples.

Unfortunately, this study comes with a number of limitations. Due to restrictions in gathering data via Datastream, the multiples in the study are based on historical financials. Prior research finds accuracy in estimation errors to increase significantly when using forecasted financial data and firm values. Ideally, financial data from annual reports is corrected for any non-operating items included in the results. Because of the time consuming aspect of doing so for 4,710 companies, I did not correct the numbers for non-operating items. The found R-squared values indicate that the dependent variables used in the multivariate analysis do not explain differences in each type of multiple equally well. Although the fit is best for sales multiples, the R-squared is still relatively low (0.189). When the firm specific variables are chosen such that differences in multiples are explained better, the country effect could lose its explanatory power.

References

Alford, A.W., 1992, “The Effect of the Set of Comparable Firms on The Accuracy of the Price-Earnings Valuation Method”, Journal of Accounting Research, 30 (1), 94-108

Baker, M., Ruback, R.S., 1999, “Estimating Industry Multiples”, working paper, Harvard University, Cambridge Bhojraj, S., Lee, C.M.C., 2002, “Who is my peer? A valuation-based approach to the selection of comparable

firms”, Journal of Accounting Research, 40, 407-444

Bhojraj, S., Lee, C.M.C., 2003, “International Valuation Using Smart Multiples”, working paper, Cornell University, Ithaca

Brooks, C., 2002, Introductory Econometrics for Finance, 1st edition, Cambridge University Press, Cambridge Damodaran, A., Website, visited at: http://pages.stern.nyu.edu/~adamodar/

Dittmann, I., Weiner, C., 2005, “Selecting Comparables for the Valuation of European Firms”, working paper, Humboldt University, Berlin

Grinblatt, M., Titman, S., 2002, Financial Markets and Corporate Stategy, 2nd edition, McGraw Hill, New York Henschke, S., Homburg, C., 2009, “Equity Valuation using Multiples: Controlling for Differences amongst

Peers”, working paper, University of Cologne, Cologne

Herrmann, V., Richter, F., 2003, “Pricing with Performance-Controlled Multiples”, Schmalenbach Business

Review, 55, 194-219

Kaplan, S.N., Ruback, R.S., 1995, “The Valuation of Cash Flow Forecasts: An Empirical Analysis”, The Journal

of Finance, L (4), 1059-1080

Kim, M., Ritter, J.R., 2003, “Valuing IPOs”, Journal of Financial Economics, 53 (3), 409-437

Koller, T., Goedhart, M., Wessels, D., 2005, Valuation: Measuring and Managing the Value of Companies, 4th edition, John Wiley, New Jersey

La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998, “Law and Finance”, Journal of Political

Economy, 106 (6), 1113-1155

Lie, E., Lie, H.J., 2002, “Multiples Used to Estimate Corporate Value”, Financial Analysts Journal, 58 (2), 44-54

Liu, J., Nissim, D., Thomas, J., 2002, “International equity valuation using multiples”, Journal of Accounting

Research, 40, 135-172

Liu, J., Nissim, D., Thomas, J., 2007, “Is Cash Flow King in Valuations?”, Financial Analysts Journal, 63 (2), 1-13

Richter, F., 2005, “Using Value Drivers to Identify Peer Group Multiples”, working paper, University Ulm, Ulm Schreiner, A., Spremann, K., 2007, “Multiples and Their Valuation Accuracy in European Equity Markets”,

23

Appendix I: Kruskal Wallis tests on firm specific factors

All multiples are classified into either one of the 10 deciles of Growth, Size, and Profitability. Furthermore, each multiple is assigned to one of the 10 industries and one of the 11 sample years. Multiples are assigned to the deciles on the basis of the firm its characteristics. When the Kruskal Wallis test shows that multiples among classes differ significantly from each other, the classification is considered relevant.

Growth

This table shows how multiples differ among 10 classes of growth. Growth is defined as the average past three year sales growth of the company. Class 1 represents the 10% multiples from the slowest growing firms, whereas class 10 includes the 10% multiples of the fastest growing firms.

EV/EBITDA EV/EBIT EV/Sale s

MEDIAN MEDIAN MEDIAN

Appendix I: Kruskal Wallis tests on firm specific factors (Continued)

Size

This table shows how multiples differ among 10 classes of size. Size is defined as the sales of the company. Class 1 represents the 10% multiples from the smallest firms, whereas class 10 includes the 10% multiples of the largest firms.

Profitability

This table shows how multiples differ among 10 classes of profitability. Profitability is defined as the average past three year EBIT/Sales ratio of the company. Class 1 represents the 10% multiples from the least profitable firms, whereas class 10 includes the 10% multiples of the most profitable firms.

EV/EBITDA EV/EBIT EV/Sale s

MEDIAN MEDIAN MEDIAN

1 12.5 16.6 5.4 2 9.3 13.2 1.7 3 8.5 12.3 1.3 4 8.0 11.7 1.1 5 7.7 11.3 1.1 6 7.6 11.4 0.9 7 7.4 11.5 0.9 8 7.6 11.4 1.0 9 7.5 11.2 0.9 10 7.7 11.6 1.0 Poole d 8.1 11.9 1.2 Kruskal Wallis Te st statistic 816.0 432.0 2,975.0 p-value 0.000 0.000 0.000

EV/EBITDA EV/EBIT EV/Sales

MEDIAN MEDIAN MEDIAN

25

Appendix I: Kruskal Wallis tests on firm specific factors (Continued)

Industry

This table shows how multiples differ among 10 industries. Multiples are classified into one of the 10 industries on the basis of NACE Rev 2. Industry classification.

Year

This table shows how multiples differ among the 11 years included in the dataset.

EV/EBITDA EV/EBIT EV/Sale s

MEDIAN MEDIAN MEDIAN

Administration, he alth care, and e ducation 11.6 15.3 2.7

Agriculture , mining, and oil 8.0 12.2 3.4

Arts, e ntertainme nt, and re cre ation 9.0 14.2 1.8

Construction and whole sale trade 7.7 10.6 0.8

Ele ctricity and wate r supply 7.8 11.6 1.3

ICT, financial institutions, and re al e state 10.7 14.1 2.2

Manufacturing of chemicals and machine ry 7.3 11.1 1.0

Manufacturing of food, clothing, and pape r 7.3 11.8 0.8

Profe ssional and administrative activitie s 7.7 11.7 1.1

Transportation and storage 9.0 13.8 1.7

Poole d 8.1 11.9 1.2

Kruskal Wallis Te st statistic 787.0 556.8 2,322.1

p-value 0.000 0.000 0.000

EV/EBITDA EV/EBIT EV/Sales

MEDIAN MEDIAN MEDIAN

8 9 10 11 12 13 14 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

EV/EBITDA

EV/EBIT

EV/Sales

Appendix II: Multiples per year

27

Appendix III: Number of observations for each multiple per year

This appendix shows how many multiples of each type are collected from each country in the dataset, by year.

Country type 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 Austria EBITDA 67 81 88 85 85 88 96 98 102 101 94 EBIT 65 79 85 79 71 77 88 90 92 94 83 Sales 68 85 93 99 98 101 104 107 113 114 103 Be lgium EBITDA 40 41 45 46 45 53 52 55 59 65 27 EBIT 35 36 43 37 40 48 49 53 58 65 26 Sales 41 45 45 54 55 57 59 63 65 69 32 De nmark EBITDA 14 16 17 14 19 17 16 15 15 14 7 EBIT 14 15 15 11 16 13 15 15 13 12 6 Sales 14 15 16 14 17 18 17 16 16 15 7 Finland EBITDA 92 103 111 111 126 121 151 158 187 203 137 EBIT 87 93 105 94 109 104 137 149 178 195 135 Sales 103 124 151 172 183 192 193 206 248 276 232 France EBITDA 225 253 293 303 301 295 317 343 393 375 86 EBIT 211 241 277 279 265 255 282 314 378 353 86 Sales 254 295 338 367 373 372 374 393 447 432 109 Ge rmany EBITDA 260 307 364 348 351 370 409 439 484 515 103 EBIT 250 286 337 309 295 313 369 405 459 483 97 Sales 284 355 431 476 476 496 496 523 558 616 120 Greece EBITDA 76 95 145 167 159 163 173 180 177 186 22 EBIT 76 101 144 157 151 156 167 169 164 172 21 Sales 76 80 118 170 182 185 191 201 202 208 25 Ireland EBITDA 15 20 20 15 18 20 20 20 21 23 11 EBIT 15 18 19 15 18 18 19 19 20 21 10 Sales 17 20 23 18 22 22 22 21 22 25 17 Italy EBITDA 55 60 88 94 100 101 119 125 138 156 47 EBIT 55 57 81 88 88 88 104 111 124 145 46 Sales 57 62 88 111 117 122 130 142 159 174 60 Luxembourg EBITDA 4 5 5 4 5 4 4 4 3 2 0 EBIT 5 5 5 3 4 4 3 3 3 2 0 Sales 5 5 5 5 5 4 4 4 3 2 0 Portugal EBITDA 55 54 55 55 60 66 65 65 67 72 46 EBIT 53 52 52 48 57 63 59 59 63 63 41 Sales 60 60 64 71 72 73 74 75 74 89 58 Spain EBITDA 71 76 82 82 84 84 92 89 93 101 81 EBIT 70 75 82 79 82 79 90 88 92 100 83 Sales 72 75 80 86 89 93 96 94 95 106 103 Swede n EBITDA 54 63 66 68 62 68 64 73 84 93 24 EBIT 50 59 60 61 56 63 63 69 83 93 30 Sales 51 63 68 74 73 76 74 82 95 107 36

The Nethe rlands EBITDA 82 85 83 81 82 79 87 92 103 104 46

EBIT 79 82 79 73 71 72 85 86 96 94 46 Sales 88 94 96 97 94 101 103 106 114 115 61

Unite d Kingdom EBITDA 476 500 511 535 542 578 669 742 818 877 666

Appendix IV: Coefficients of country dummies in the multivariate analysis

This appendix presents the actual coefficients of the country dummies in the multivariate analysis, in addition to the table in the results section in the study.

Coefficients of country dummies This table shows the coefficients of the country dummies in the multivariate analysis in the period 1998 – 2008. The multivariate analysis assesses the relationship between country and multiple when firm fundamentals (growth, size, profitability), industry classification, and the year in which the multiple is recorded are also taken into account by estimating a panel regression.

* significant at a p-value of 0.05

EV/EBITDA EV/EBIT EV/Sales

29

Appendix V: Results of industry and year effect

This appendix shows the signs on industry and year dummies from the 45 regression analyses. Since the significance of the dummies did not change when the country dummy changed, a + means that the year or industry is positively influencing the multiple in each country at a p-value of 0.05.

Results on industry dummies

This table shows the relationship between the type of multiple and the industry. All effects are relative to the ‘Arts, Entertainment, and Recreation’ industry. - means multiples in an industry are lower than in the Arts, Entertainment, and Recreation industry at a p-value of 0.05, whereas 0 means that no differences are found. Finally, + says that multiples from the industry are higher than multiples in the Arts, Entertainment, and Recreation industry.

Results on year dummies

This table shows the relationship between the type of multiple and the year. All effects are relative to the year 1998. - means multiples in the year are lower than in 1998 at a p-value of 0.05, whereas 0 means that no differences are found. Finally, + says that multiples from the year are higher than multiples in 1998.

EV/EBITDA EV/EBIT EV/Sales

Administration, he alth care, and e ducation + 0 +

Agriculture , mining, and oil 0 0 0

Construction and whole sale trade 0 - 0

Ele ctricity and wate r supply 0 - 0

ICT, financial institutions, and re al e state + 0 0

Manufacturing of chemicals and machine ry 0 - 0

Manufacturing of food, clothing, and pape r 0 0 0

Profe ssional and administrative activities 0 0 0

Transportation and storage 0 0 0

EV/EBITDA EV/EBIT EV/Sales