Looking Forward: Enterprise
Multiples and Finding the Optimal
Peer Group Selection Method
Master thesis
Angelina BouwmanGroningen, January 2014
University of Groningen
Uppsala University
Faculty of Economics and Business
Faculty of Social Sciences
MSc International Financial Management
MSc Business and Economics
Student number: 1714643
2
Looking Forward: Enterprise
Multiples and Finding the Optimal
Peer Group Selection Method
Angelina Bouwman
1Abstract
This study focuses on the absolute valuation error of selecting European peer group companies. The selection based on comparable profitability and growth within the same industry (two-digit SIC industry classification) will be compared with the absolute valuation error of a peer group selected solely based on industry classification. Furthermore, the optimal size of the peer group is investigated by comparing the absolute valuation error of peer groups with sizes of 6 and 4 companies. The research sample contains 320 listed companies. The analysis is performed on the Enterprise Value/EBITDA and Enterprise Value/EBIT multiple. The t-test is applied in order to determine significant differences between the absolute valuation error of the peer groups and the absolute valuation error of the industry. The results of these tests are insignificant.
Keywords: corporate valuation, forward-looking multiples, enterprise value JEL Classifications: G19, M41
1
3
Table of Contents
Abstract ... 2 Table of Contents ... 3 1. Introduction ... 4 2. Literature Review ... 72.1 Different multiples for different purposes ... 7
2.2 Relevant value measures ... 8
2.3 Construction of the peer group ... 9
2.4 The optimal size of the peer group ... 11
3. Methodology ... 12
3.1 Sample Data ... 12
3.2 Research design ... 17
3.3 Peer group selection analysis ... 18
4. Results ... 21
5. Conclusion ... 26
6. References ... 29
7. Appendix ... 32
7.1 Key value driver formula ... 32
4
1. Introduction
In the field of corporate valuation there are different methods to assess the value of a company. Among the different methods, a multiple valuation method based on multiples is a commonly used valuation approach (Bhojarj and Lee, 2001). In practice, multiples are widely used in merger and acquisition activities, seasoned equity offerings, initial public offerings, leveraged buy-outs, investment bankers’ fairness opinions and analysts reports (Deangelo, 1990). The method is often used to cross check the results of other valuation methods, for instance to complement the Discounted Cash Flow method (Schreiner and Spremann, 2007).
When applying the multiple valuation method, the first step is the selection of peer group companies. These companies are considered to be comparable to the target firm. According to Koller et al. (2010), peer group companies need to be comparable in terms of profitability, growth and their underlying business model. With these peer group companies, an aggregated multiple is calculated based on the preferred measure of central tendency. This aggregated multiple is used to determine the equity or enterprise value of the target company by multiplying the target company’s value driver by the aggregated peer group multiple (Suozzo et al., 2001).
Besides the optimal type of multiple and the aggregated peer group multiple, another factor affecting the accuracy of the multiple valuation method is the identification of the comparable companies that form the peer group. However, there is only little existing empirical research on this topic (Dittmann and Weiner, 2005). Alford (1992) showed that industry membership or a combination of return on equity (ROE) and total assets (TA) are significant factors for selecting comparable companies in a sample of U.S. companies. Alford (1992) used ROE as measure for profitability and TA as measure for size. In addition, Cheng and McNamara (2000) and Bhojraj and Lee (2002) improved the results of Alford (1992) by using a combination of industry membership with total assets and variables for profitability, growth and risk as identification criteria. However, these results are based on U.S. data only. Most of the empirical studies are exclusively dealing with U.S. datasets (Alford, 1992; Cheng and McNamara 2000; Bhojraj and Lee, 2002 and Liu et al., 2002).
5 Evidence of the existing literature shows, that the use of forward-looking multiples instead of traditional backward-looking multiples increases the accuracy of the value estimation. This result was found by Kim and Ritter, (1999), Lie and Lie, (2002), Liu et al., (2002) and Schreiner and Spremann, (2007). These papers showed that valuation accuracy increased, when later year forward-looking multiples where used instead of earlier year forward-looking multiples. However, these findings are only based on the P/E multiple. Dijkstra (2013) was the first who examined this for the EV/EBITDA and EV/EBIT multiple and concluded that a forward-looking multiple of t+1 leads to the lowest absolute valuation error.
This paper contributes to the existing literature threefold. Firstly, the current literature is mainly U.S. focused. Only three authors (Herrmann and Richter, 2003; Dittmann and Weiner, 2005; Schreiner and Spremann, 2007) investigated a European based sample of companies. Of whom only Schreiner and Spremann (2007) focused on forward-looking multiples. Furthermore, this is the first paper that focuses on EV/EBITDA and EV/EBIT multiples investigating one year ahead forward-looking multiples based on the consensus forecast of professional analysts. Finally, this paper contributes to the literature related to the optimal size of the peer group. This paper compares the inclusion of 4 and 6 peer group companies based on the results of Schreiner (2007) and the research of Bhojraj and Lee (2002).
Considering the previous theory, the following research question is the focus of this paper:
Does the selection of peer group companies within an industry, based on comparable expected profitability and expected growth lead to significantly lower absolute valuation errors for the EV/EBIT(DA) multiples?
In order to select the comparable companies to form the peer group, the methodology of Bhojrai and Lee (2002) will be applied to identify peer group companies based on the variables expected profitability and expected growth, instead of solely based on industry membership. Bhojrai and Lee (2002) came up with a new approach which included a multiple regression model to create a so called ‘warranted’ multiple. The companies with the closest comparable warranted multiple are selected to form the peer group. Bhojrai and Lee (2002) showed in their paper that selecting peer group companies by this method leads to lower absolute valuation errors than selection based on industry classification. It is important to notice that the warranted multiple is only used in order to select the peer group companies.
6 take place. The difference between the actual absolute error and the absolute industry error is investigated. The absolute average industry valuation error will be compared with the average absolute valuation error of a peer group selected based on their warranted multiple. This will be done for both the EV/EBITDA and the EV/EBIT multiple, assessing peer groups of 6 and 4 companies over different years. The lower the absolute valuation error, the more accurate the estimate is. The accuracy of the value estimate of the multiples is tested by assessing the average of the medians of the absolute valuation error (Kim and Ritter, 1999; Lie and Lie, 2002; Schreiner and Spremann, 2007). A lower average median implies an increase in the accuracy of the valuation. The significance will be tested by the t-test as applied by Bhojrai and Lee (2002). The expectation is that the warranted multiple method of Bhojrai and Lee (2002) will lead to significantly lower absolute valuation errors, compared with the absolute industry error, for the investigated European sample.
Assessing the valuation accuracy for the EV/EBITDA and EV/EBIT multiple, will be conducted for a European sample of 320 listed companies. The years 2007-2010 are included with the forward-looking multiples t+1. The sample will be sub divided into industries based on two-digit SIC Codes. The median of these industries will function as a benchmark.
7
2. Literature Review
This section reviews the theory related to the multiple valuation method. The first part provides an outline of the relevant distinctions within the multiple valuation method, which are relevant for this paper. The Second part elaborates on the different value drivers and provides the underlying argumentation for this research. Finally, the theory related to the selection of peer group companies and the optimal size of a peer group will be discussed.
2.1 Different multiples for different purposes
In order to provide a clear view of the multiple valuation method, it is important to make a distinction between valuation methods in general. Firstly, the direct valuation method, this method estimates the company’s value directly out of its expected cash flows, without incorporating the current value of other companies (Bhojraj and Lee, 2001). These models need a net present value calculation of the forecasts of the cash flows of a company. These direct valuation models include the Dividend Discount Method, the Discounted Cash flow Method and the Residual Income Method. The second approach is the indirect valuation method, this approach estimates the company value based on the market value of comparable companies. This valuation approach incorporates market multiples in order to determine the value of the company.
The focus of this paper is on the multiple valuation method. The investigated multiples in this method are summarized ratio’s that provide information about the company’s market value, compared to its peers (Schreiner, 2007). The numerator of a multiple is the enterprise value or the equity value plus debt, minus the cash of a company. In general, practitioners have a preference for enterprise multiples over equity multiples, due to the fact that the cash flow of a company is created by both debt and equity. Enterprise multiples, moreover, estimate the value of a company in total, which results in a more comprehensive value estimation than equity multiples, due to incorporating the debt of the company. Secondly, enterprise multiples are less affected by capital structures, because the unlevered value of a company is included and can therefore directly be used to compare companies (Suozzo et al., 2001).
8 this problem, the sample used in this paper only includes listed companies, therefore this problem will not occur and makes the enterprise multiple appropriate for this research.
The main contribution of this research to the existing literature is the use of forward-looking multiples over trailing multiples. Trailing multiples, also known as backward-looking multiples, use historical company data to calculate the multiple. Forward-looking multiples capture the forecasts of analysts for a number of years into the future. The argumentation behind the use of forward-looking multiples is, firstly, that forward-looking multiples are consistent with the principals of valuation stating that the company’s value equals the present value of future cash flows, not past cash flows. Secondly, the forecasted earnings are a better reflection of the long term cash flows, because they do not incorporate one-time events. Therefore, the forecasted earnings are more normalized compared with historical earnings. Research conducted by Kim and Ritter (1999) and Liu et al. (2002), show that forward-looking multiples increase the predictive accuracy and decrease the variance of multiples within an industry.
Though, there is existing evidence of increased predictive accuracy of the use of forward-looking multiples, however, these findings are mostly based on the Price/Earnings multiple for U.S. based companies. Contrary, this research focuses on the EV/EBITDA and EV/EBIT multiple for European companies. The first research regarding the EV/EBIT and EV/EBITDA multiple and their related absolute valuation errors (Dijkstra, 2013) concluded that the median of the absolute valuation errors decreased when t+1 forecasted data was used, instead of t+2 and t+3. The reason for this can be found in the use of normalized earnings, implicating that less information about one-time costs is known in latter year forecasts, instead of earlier year forecasts.
The previous part of this section investigated the differences in multiple valuation methods and the importance of the use of forward-looking multiples over backward-looking multiples. Based on the results of Dijkstra (2013), forward-looking multiples t+1 are used in this research. In the following part, the relevance of the choice of the EBITDA and EBIT multiple for this research will be discussed.
2.2 Relevant value measures
Basically, corporate valuation is a process of turning information into value (Flostrand, 2006). Therefore it is interesting to determine what drives value.
9 specific target company, results in the value of the target company. The underlying formula (1) for the theory of the multiple valuation method is:
(1)
k = Cost of Capital g = Growth r = Profitability
2.3 Construction of the peer group
In order to apply the multiple valuation approach in practice, a peer group needs to be constructed. Appropriate selection of peer group companies has a significant positive effect on the accuracy of the multiple valuation approach (Schreiner, 2007). According to Arzac (2005) and Koller et al. (2010), in an ideal world, peer group companies have the same cash flows, growth potential and are facing the same risks as the target company. However, in practice these identical companies are often hard to find.
In the study of Alford (1992), peer groups were investigated for P/E multiples based on industry membership, risk (measured by company size) and earnings growth. Alford (1992) concluded that the identification of companies with similar expected fundamental variables can also be done directly. He found that a peer group of companies with similar expectations of risk and growth did not significantly increase the absolute valuation error, compared with a peer group solely based on industry classification. Regarding to industry classification, Alford showed that the valuation accuracy increased when the precision of the used industry definition to identify peer group companies was narrowed from broad one-digit Standard Industrial Classification (SIC) codes to two-digit and three-digit SIC codes. However, the accuracy was not improved when the three-three-digit SIC code was narrowed to a four-digit SIC code.
10 classifications significantly improve the explanation of stock return movements, forecasted and realized growth rates, research and development expenditures and various key financial ratios. The results of the three other classification systems do not differ a lot among each other.
Research of Schreiner (2007) investigated the valuation accuracy of multiples based on industry classification. The results of this study showed that forming a smaller, but more homogeneous peer group by narrowing the industry definition from a one-digit to a two-digit or three-digit industry classification improves the valuation accuracy of multiples.
Unfortunately, most findings in compiling peer groups have been derived from U.S. data only, except for the results of Herrmann and Richter (2003) and Dittmann and Weiner (2005). Dittmann and Weiner (2005) investigated which peer group identification method led to the most accurate valuation for a European sample investigating the EV/EBIT multiple. They found that peer group companies should be identified from the same country as the target company for the countries U.S., the U.K., Denmark, and Greece. For the remaining European countries, peer group companies needed to be identified from the EU15 or from the OECD. Furthermore, they showed that for all countries (including the U.S.) the valuation is the most accurate when companies are identified on the basis of comparable return on assets (ROA). The selection based on ROA (measure for profitability) outperformed the identification based on SIC code industry membership classification.
11 This paper selects peer group companies based on a SIC industry classification of two-digits, in line with the in insight of Koller et al. (2010) that peer group companies need to have the same underlying business model, which can be found within the same sub industry. Selection within industry classification is in line with Cheng and McNamara (2000) and Dittmann and Weiner (2005).
2.4 The optimal size of the peer group
In 2008, a study of Cooper and Cordeiro introduced the topic of choosing the optimal size of a peer group. They investigated how the performance of the multiple valuation method varies as the number of peer companies in the peer group increases. This research was motivated by the contrast between the approach followed by practitioners and the approach followed by academic researchers. Practitioners typically use a small number of closely related comparable companies, while academics often use the entire industry according to Cooper and Cordeiro (2008). They examined a peer group identification rule with growth rates, which led to the conclusion that using ten comparable companies is, on average, as accurate as using the entire industry, in terms of absolute valuation error. However, a peer group of five comparable companies led to a slightly better performance, due to adding more companies to the peer group, on average, means adding more noise. Schreiner (2007) argues that a peer group with four to eight comparable companies is the optimal size. This conclusion is derived from several interviews with academics and practitioners. Furthermore, Henschke and Homburg (2009) concluded that a peer group of four to eight companies led to the most accurate valuation results in terms of absolute valuation error. Bhojraj and Lee (2002) included the 6 closest peers based on their warranted multiples.
12 profitability and growth. The 6 and 4 closest companies are selected to form the peer group. The decision for these numbers is based on respectively Schreiner (2007) and Bhojraj and Lee (2002). In line with Bhojraj and Lee (2002) the absolute valuation error of the peer groups will be examined by using the median. The investigating of the absolute valuation errors will be conducted for both the EV/EBITDA and the EV/EBIT multiples, because practitioners use these multiples the most Schreiner (2007).
3. Methodology
3.1 Sample Data
In order to test the hypothesis, this section describes the methodology of this study. The data investigated in this paper is retrieved from InFinancials, a database containing company data and stock screening information for companies active in private banking, asset management, private equity and corporate finance such as KPMG, HSBC and others. This database is provided by Holland Corporate Finance, Amsterdam. The data retrieved out of InFinancials contains the Enterprise Value/Sales (EV/SALES), Enterprise Value/EBITDA (EV/EBITDA) and Enterprise Value/EBIT (EV/EBIT) multiples, as well as the country in which the company is listed, and the SIC industry classification code. The sample contains the companies of stock indexes in Europe2. Only listed companies can be included in the sample, because the market capitalization is needed in the analyses. The financial companies (SIC code 6000-6999) are excluded due to differences in valuing these companies, related to the liquidity of their assets (Lie and Lie, 2002). The sample contains all companies with the required valuation information available for the years 2007 until 2010. This research includes forward-looking multiples. The forward-forward-looking multiples include a forecast for three years (t+1, t+2, t+3). The forward-looking multiples incorporate the available consensus (average) forecasts of several analysts. Each forecast is at least built up out of more than 30 analyst forecasts, according to the information available in InFinancials. The enterprise value is the end-year enterprise value as on 31th of December of each investigated year.
In this study, the enterprise value, sales, EBITDA and EBIT need to have a positive value, otherwise this would result in negative valuation multiples (Liu et al., 2002; Schreiner and Spremann, 2007). As a consequence, the average multiple will be biased upwards. In order to solve this problem, the average multiple is adjusted downwards by excluding multiples larger than 30, this restriction is also
2
13 applied by Kohlleppel (2011), and a comparable approach is applied by Dittman and Weiner (2005) for the ROA. They removed the highest observations with the argumentation that practitioners would have removed them otherwise by hand, because they are unrealistically high. When the multiples above 30 would not be excluded, a problem would occur in the next step of the methodology where the regression coefficients are the basis of the selection of the peer group companies. The above mentioned restrictions led to a decrease from 603 companies to 320.
In this research the information of three-year forward-looking multiples is investigated. For each year (2007-2010) a consensus forecast is available for the upcoming three years. Kim and Ritter (1999), Lie and Lie (2002), Liu, Nissim and Thomas (2002, 2007), and Schreiner (2007), show the superiority of forward-looking multiples over trailing multiples in terms of valuation accuracy. The study of Dijkstra (2012) shows that earlier-year forecasts instead of later-year forecasts for enterprise multiples results in a more accurate valuation. He finds the most accurate valuation results for t+1. Therefore, the analyses in this study are performed with one-year forward-looking multiples.
Liu et al. (2002) also compared peer groups with three estimators of central tendency: the simple mean, harmonic mean and the median. Their findings are consistent with the findings of Baker and Ruback (1999), and Beatty, Riffe and Thompson (1999), since they concluded that the valuation accuracy of the simple mean and median is lower compared to the harmonic mean. However, Alford (1992), Bhojraj and Lee (2002) use the median in their research. Despite of the advantage of the harmonic mean, this paper uses the median as well, in order to stay consistent with Bhojraj and Lee (2002), because they included the median in combination with the warranted multiple approach, which is also applied in this paper.
14
Table 1. Descriptive Statistics
Table 2. Median multiples per industry
Table 2 below shows the median multiples per industry for the period 2007-2010. The Kruskal-Wallis test is significant for each multiple at a significance level of 10% indicating that the multiples differ significantly per year and per industry.
EV(07)/EBIT (08) EV(08)/EBIT (09) EV(09)/EBIT (10) EV(10)/EBIT (11) EV(07)/EBIT DA(08) EV(08)/EBIT DA(09) EV(09)/EBIT DA(10) EV(10)/EBIT DA(11) Mean 12.34 9.23 10.14 10.75 8.69 5.96 6.99 7.60 Median 11.28 8.62 9.75 10.41 7.86 5.56 6.83 7.21 Maximum 29.88 29.39 26.31 27.56 24.15 18.29 16.02 23.43 Minimum 2.65 0.46 1.81 1.85 2.05 0.25 0.97 2.04 Std. Dev. 5.02 4.60 3.88 4.23 3.41 2.63 2.47 2.94 Skewness 1.08 1.37 0.98 0.88 1.13 1.09 0.61 1.28 Kurtosis 4.23 6.13 4.57 4.47 4.55 5.95 3.72 6.27 Jarque-Bera 82.57 230.27 83.93 69.87 100.37 180.16 26.90 230.07 Probability 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Observations 320 320 320 320 320 320 320 320 S IC
code Industry # Obs
EV(07)/EBIT (08) EV(08)/EBIT (09) EV(09)/EBIT (10) EV(10)/EBIT (11) EV(07)/EBIT DA(08) EV(08)/EBIT DA(09) EV(09)/EBIT DA(10) EV(10)/EBIT DA(11) 1311 Crude petroleum and natural gas 6 11.63 8.77 8.81 8.23 7.97 5.76 5.42 6.43
1500
Building construction general
contractors and operative builders 6 11.08 8.15 8.26 10.00 7.91 5.13 7.48 7.57
1600 Heavy construction 12 11.28 8.57 10.05 13.63 7.00 5.43 6.49 8.06
2000 Food and beverages 17 13.12 9.90 9.29 11.64 8.49 6.62 7.48 9.09
2800 Pharmaceuticals and chemicals 40 11.39 9.40 10.60 10.02 8.91 6.61 6.69 7.33
2900 Petroleum refining 6 8.75 7.68 9.99 10.06 6.16 4.52 5.08 5.68
3241 Cement and hydraulic 8 8.70 8.31 12.03 13.19 6.00 5.18 8.02 7.59
3500 Machinery and equipment 25 10.30 7.20 7.70 9.91 7.83 5.21 6.52 7.93
3600
Electronic and other electrical equipment and components, except
computer equipment 44 12.07 8.62 9.52 11.07 7.68 5.85 7.59 7.93
3700
Motor vehicle and aircraft parts and
Accessories 8 10.22 7.14 9.06 7.08 6.28 3.98 5.57 5.29
4700 Transportation services 6 13.65 10.69 13.80 12.51 10.64 7.04 8.06 8.13
4800
Telephone communications and
television Broadcasting Stations 27 11.91 8.84 10.23 11.26 7.17 5.44 6.08 6.26
4900 Electric and gas services 24 13.97 10.80 9.68 11.84 9.17 7.23 7.53 7.50
5000 Wholesale trade-durable goods 11 17.19 10.56 9.78 10.09 10.29 6.16 5.47 7.03
5411 Grocery Stores 6 11.99 9.62 9.88 10.59 8.47 5.11 6.33 6.55
7300 Business services 55 9.41 6.32 11.95 8.22 7.72 5.11 6.53 6.55
8700 Engineering services 18 9.12 5.45 8.32 7.30 7.17 3.78 5.73 6.77
Test statistic 38.20 38.83 37.30 60.77 23.61 37.94 24.92 26.72
15 Country # Obs EV(07)/EBIT (08) EV(08)/EBIT (09) EV(09)/EBIT (10) EV(10)/EBIT (11) EV(07)/EBIT DA(08) EV(08)/EBIT DA(09) EV(09)/EBIT DA(10) EV(10)/EBIT DA(11) Austria 6 9.62 13.08 11.26 9.12 7.76 7.08 6.64 8.34 Belgium 16 10.75 8.14 10.40 9.91 6.56 5.20 7.71 6.31 Denmark 6 17.25 9.76 12.18 13.25 12.19 6.93 9.72 10.71 Finland 8 13.07 9.79 11.50 13.94 9.55 6.51 8.91 9.78 France 83 10.03 8.20 9.29 9.30 6.95 5.55 6.17 6.30 Germany 46 12.68 8.78 9.95 11.26 8.34 5.61 7.26 8.26 Greece 7 13.80 8.54 13.09 10.43 9.50 5.10 6.69 6.51 Hungary 3 13.71 6.01 10.05 10.05 7.92 3.62 6.31 5.80 Italy 37 11.14 8.84 9.83 10.48 6.97 5.88 6.84 7.03 Netherlands 17 9.62 7.23 11.24 9.32 7.10 5.05 8.15 8.07 Norway 8 12.51 5.75 8.71 9.33 9.83 3.51 6.05 6.46 Poland 4 14.22 12.91 11.68 11.54 7.85 6.17 6.65 6.81 Portugal 9 13.59 11.50 13.77 15.51 8.71 6.38 8.05 7.50 Spain 15 11.91 10.76 10.35 11.95 9.17 7.94 8.28 7.51 Sweden 10 15.67 9.05 8.76 11.95 10.02 6.07 6.83 7.94 Switzerland 21 11.03 9.72 7.31 10.57 8.48 6.83 5.82 8.29 United Kingdom 21 13.02 6.41 7.96 8.56 9.98 5.12 6.47 6.93 Test statistic 31.92 31.61 50.42 44.19 32.47 25.88 31.39 32.61 Kruskal-Wallis p-value 0.0102 0.0112 0.0000 0.0002 0.0087 0.0557 0.012 0.0083
Table 3. Median of country multiples
Table 3 shows the median of the multiples for each year and for each country. The county of the company is based on the registered head-office of the company.
The smallest EV/EBIT multiple can be found in the United Kingdom, the average of four years is 8.99 and the highest average can be found in Portugal (13.59). The smallest EV/EBITDA multiple can be found in Hungary (average of 5.99) and Denmark has on average the highest multiple with a value of 9.89. The Kruskal-Wallis p-values at the bottom of the table are significant at 10% for all countries, implying the relevance of classification in countries in explaining the value of the multiples.
16
Table 5. EBITDA Country dummy
Table 5 shows the coefficients and the probabilities of the country dummy variables in the regression analysis. The bold values are significant at a 5% level. ‘Dum’ refers to dummy variable.
Table 6. EBIT Country dummy
Table 6 shows the coefficients and the probabilities of the country dummy variables in the regression analysis. The bold values are significant at a 5% level. ‘Dum’ refers to dummy variable.
Variable Coefficient Prob. Coefficient Prob. Coefficient Prob. Coefficient Prob.
EV07/EBITDA08 EV08/EBITDA09 EV09/EBITDA10 EV10/EBITDA11
DUM _AUSTRIA -0.7478 0.6609 -0.6404 0.6753 0.3360 0.7977 -0.3633 0.7928
DUM _BELGIUM -1.6195 0.2113 -0.1731 0.8806 0.6374 0.5196 -1.1074 0.2903
DUM _CZECH_REPUBLIC -4.9316 0.1860 -1.4335 0.6681 -0.5681 0.8431 -2.2536 0.4574
DUM _DENM ARK 2.4530 0.1739 1.3137 0.3895 3.4549 0.1186 1.8845 0.1736
DUM _FINLAND 1.6283 0.2773 1.0896 0.4177 2.2898 0.0477 1.6209 0.1854
DUM _FRANCE -1.5612 0.1538 -0.5772 0.5577 0.4976 0.5554 -1.2667 0.1556
DUM _GERM ANY -1.3075 0.2557 -0.8698 0.4001 0.3797 0.6680 -0.2386 0.7980
DUM _GREECE 1.6269 0.6625 5.3089 0.1239 13.6348 0.0000 1.8249 0.5468 DUM _HUNGARY -1.5966 0.4895 -2.1753 0.2938 0.5687 0.7488 -1.4886 0.4279 DUM _IRELAND -1.3119 0.3709 -1.7496 0.1837 0.7859 0.4845 -2.0819 0.0803 DUM _ITALY -2.3686 0.0442 -1.1225 0.2882 0.6349 0.4827 -1.6519 0.0854 DUM _NETHERLANDS -2.1848 0.0951 -1.1760 0.3175 1.8410 0.0695 -0.1789 0.8667 DUM _NORWAY -0.0590 0.9689 -1.3809 0.3083 0.1516 0.8971 -0.0465 0.9698 DUM _POLAND -1.2390 0.5501 -1.6636 0.3747 0.6444 0.6877 -0.5072 0.7650 DUM _PORTUGAL -0.1256 0.9302 0.0733 0.9547 3.2303 0.0036 1.5849 0.1748 DUM _SPAIN -0.6912 0.6091 2.1603 0.0751 2.5935 0.1280 0.4074 0.7102 DUM _SWITZERLAND 0.9038 0.4647 0.5761 0.6046 0.4499 0.6375 1.5823 0.1165 DUM _UNITED_KINGDOM 0.0088 0.9944 -1.3998 0.2147 0.2539 0.7932 -1.0702 0.2964
Variable Coefficient Prob. Coefficient Prob. Coefficient Prob. Coefficient Prob. EV07/EBIT08 EV08/EBIT09 EV09/EBIT10 EV10/EBIT11
DUM _AUSTRIA -2.5720 0.2930 3.9183 0.0893 3.6317 0.0606 0.4626 0.8233
DUM _BELGIUM -3.6454 0.3860 1.6063 0.3684 1.8520 0.2123 -0.1541 0.9224
DUM _CZECH_REPUBLIC -6.3197 0.2143 1.6054 0.7464 1.5889 0.7071 1.4979 0.7415
DUM _DENM ARK 0.5206 0.8320 1.0309 0.6690 5.3517 0.1558 2.5032 0.2275
DUM _FINLAND -0.5345 0.7986 4.5095 0.2750 3.7451 0.0281 2.8566 0.1194
DUM _FRANCE -4.1491 0.1570 -0.0537 0.9718 1.0245 0.4104 -1.2805 0.3376
DUM _GERM ANY -3.4863 0.2700 -0.0818 0.9591 1.1784 0.3665 0.2525 0.8565
17 Table 5 and table 6 show between 0 and 4 significant country dummies per year out of the 17 countries in the sample. Dittmann and Weiner (2005) found significant differences between the USA, the UK, Denmark and Greece for the EV/EBIT multiple. This study observes a country-effect for Finland, Greece, Italy, and Portugal for the EBITDA multiple for only single years. For the EBIT multiple the significant countries are: Finland, Ireland, and Spain for only one year (except for Ireland, 2 years) Due to the significant values of specific countries in only single years and the relative low number of companies out of these countries that are present in the sample, the results are not further included in the methodology.
3.2 Research design
The methodology of Bhojraj and Lee (2002) is applied in this paper. Firstly, annual regressions are estimated for both the EBITDA and EBIT multiple on the variables expected profitability and expected growth.
The applied regression formula is:
Multiple it = c β0 + β1 * growth it + β2 * profitability it + εit (2)
In which c β0 = a constant, β1 * growth it =expected growth β2 * profitability it =expected profitability
and εit = the error term.
The estimated coefficients of the regressions with the appropriate sign are multiplied with each firm’s expected profitability and expected growth variables and are combined in order to create a warranted multiple for each company. After that, the closest comparable companies are selected based on their warranted multiple.
18
3.3 Peer group selection analysis
Step 1. Identification of comparable companies
The more similar the comparable companies in the peer group are related to the target company, the more information they provide in order to accurately valuate the target company (Eberhart, 2004). The first condition for the comparable peer companies is industry membership. By nature, companies with a finer industry classification are more similar in their current operating characteristics. The sample is divided based on industry classification by using a two-digit SIC Code. This provides a broader range of companies than the three-digit SIC code and therefore leaves more available companies to select based on their the closest warranted multiple. This is in line with the approach of Bhojraj and Lee (2002). After that, an aggregated industry multiple is calculated for the multiples in the same industry, this is the median of the industry multiple. Industries with less than four companies are excluded. The aggregated industry multiple will function as a bench mark.
Step 2. Forming the peer group
The selection variables after industry classification are: Expected growth (sales growth)
Expected profitability (EBIT(DA) / Sales)
19 The influencing sign of growth is expected to be positive in general. Though, the values of profitability and the cost of capital of a company will have important consequences for the effect on the multiple when growth changes. For instance, if growth increases due to a decrease in profitability, this could result in a negative multiple. Conversely, prior studies concluded that growth has a positive sign for every multiple (Bhojraj and Lee, 2002; Hermann and Richter 2003). For profitability prior studies find positive and negative signs for different multiples in different years. In this paper the expected sign of profitability is positive.
Filtering down the comparable companies by their absolute closeness to the target company is based on their warranted multiple. This method has proven to be the most optimal peer group company identification method for a dataset of European companies in the study of Dittmann and Weiner (2005). The peer group size that is investigated is built up out of 6 and 4 companies, in line with Schreiner (2007) and Bhojraj and Lee (2002). The analysis will be conducted for both the EV/EBITDA and EV/EBIT multiples.
Step 3. Comparing the different peer group selection methods
In this step the peer group selection method is compared with the industry multiple. This is investigated by comparing the actual multiple of a company with the median of the industry (target company multiple – median of industry multiple, scaled by the target company multiple). This is the industry absolute valuation error. Secondly, the actual multiples are compared with the average of the medians of the peer group multiple of peer groups of 6 companies and peer groups of 4 companies (target company multiple – peer group multiple, scaled by the target company multiple). The absolute valuation error is used in the analysis (Kim and Ritter, 1999; Lie and Lie, 2002 and Schreinder and Spremann, 2007). This is done for all companies for each of the four years. In line with Bhojraj and Lee (2002) the absolute valuation error of the peer groups will be examined using the average median. The investigating the absolute valuation errors will be conducted for both the EV/EBITDA and the EV/EBIT multiples.
20 Step 4. Determining the optimal peer group size
21
4. Results
This paper investigates the absolute valuation error of selecting European peer group companies based on comparable expected profitability and expected growth within the same two-digit SIC industry classification. In order to assess this objective, first a regression analysis is performed. This regression is conducted for both the EV/EBITDA and EV/EBIT multiple with the independent variables expected profitability and expected growth.
Table 7 shows the estimated signs and coefficients of the explanatory variables profitability and growth and the corresponding adjusted R². The regression is performed for all years separately. Only the significant (bold) coefficients are used to calculate the warranted multiple. For a year with insignificant coefficients, the average ratio of profitability or growth is used of the significant coefficients of the other years of the specific variable to calculate an appropriate value to replace the insignificant value.
Table 7. Estimated coefficients of expected profitability and expected growth
This table shows the coefficients of the explanatory variables of the Enterprise Value/EBITDA and Enterprise Value/EBIT for the years 2007-2010 with the EBITDA and EBIT values of t+1. The bold marked values are significant at a 5% level. The corresponding adjusted R² can be found in the last row.
The adjusted R² values in table 7 are between 0.03 and 0.15 for the EV/EBITDA multiple and for the EV/EBIT multiple between 0.04 and 0.24. The average of the EV/EBITDA adjusted R² is 0.08 and the average of the EV/EBIT multiple is 0.16. Due to the small number of explanatory variables the R² and the adjusted R² are almost the same for each year. The average R² of Kohlleppel (2011) is 11.6% for the EV/EBIT multiple. However, Bhojraj and Lee (2002) presented an average R² of respectively 72% for their EV/SALES multiple. The R² of the EV/EBITDA in this sample is relative low, however, the explanatory variables in the regression make sense from an economic point of view. These factors are based on the formula of enterprise value presented by Koller et al (2010) (see appendix). This formula combines the fundamental drivers of value: profitability, growth and cost of capital. Furthermore, prior literature also incorporates these variables (e.g. Bhojraj and Lee (2002).
22 The second step in the selection of comparable companies is to calculate a ‘warranted’ multiple in line with the approach of Bhojraj and Lee (2002). Bhojraj and Lee (2002) used the regression coefficients of the explanatory variables of the entire sample to calculate the warranted multiple. The calculated warranted multiples are only used to select the closest peer group companies. The selection is made for the EV/EBITDA and EV/EBIT multiple for peer groups of 6 and 4 companies. The absolute valuation errors are used in order to determine the valuation accuracy (Lie and Lie, 2002). This absolute valuation error is compared with the absolute valuation error of the industry median. This compares a valuation of the target company with a peer group selected based on comparable expected profitability and expected growth, with a valuation of the target company based on the industry median. Table 8 and 9 show the outcomes of this analysis.
Table 8. Average Absolute valuation error EV/EBITDA t+1
This table shows the average absolute valuation error of the selection method of 6 and 4 peer group companies based on their warranted multiple. The t-test is applied to compare the average absolute valuation error of the peer group median, with the average absolute valuation error of the industry median. This analysis is performed on the EV/EBITDA multiple, for peer groups consisting of 6 and 4 peers. The bold values are significant at a 5% level. The t-test results are insignificant. The last row shows the average standard deviation of the absolute valuation error.
The t-test is performed in order to assess whether there are significant differences between the average absolute valuation errors in case of selecting peer group companies based on comparable expected profitability and expected growth and average absolute valuation errors based on the industry median. The t-test is performed on the EV/EBITDA multiple for the years 2007-2010 with the EBITDA values of t+1. Comparing the industry average absolute valuation errors with the peer group errors of 6 peers shows for each year higher absolute valuation errors than the average absolute valuation error of the industry medians. Investigating the valuation errors for peer groups of 4 companies shows a similar result in table 8. The differences between the industry absolute valuation error for peer groups of 6 companies and peer groups of 4 companies are due to small industries that did not include enough companies to make peer groups of 6 companies. These insufficient industries are therefore excluded
23 in the analysis of the peer groups of 6 companies. In order to determine the quality of the absolute valuation errors, the standard deviation is included in the last row of the table 8. Firstly, comparing the standard deviation of the industry absolute valuation error with the standard deviation of the absolute valuation error of a peer group of 6 companies shows a decrease of 0.7657 to 0.7251 for the peer groups of 6 companies. The standard deviation of the industry absolute valuation error of 4 companies and the absolute valuation error of peer groups of 4 companies decreases from 0.7466 to 0.7261 This result implies that the dispersion decreases by downsizing the industry to 6 or 4 companies, however the standard deviations of peer groups of 6 and 4 companies are almost similar. A low standard deviation indicates that the data points tend to be very close to the mean.
Table 9. Average Absolute valuation error EV/EBIT t+1
This table shows the average absolute valuation error of the selection method of 6 and 4 peer group companies based on their warranted multiple. The t-test is applied to compare the average absolute valuation error of the peer group median with the average absolute valuation error of the industry median. This analysis is performed on the EV/EBIT multiple, for peer groups consisting of 6 and 4 peers. The bold values are significant at a 5% level. The t-test results are insignificant. The last row shows the average standard deviation of the absolute valuation error.
For the EV/EBIT t+1 multiple, the t-test is performed in order to assess whether there are significant differences between the average absolute valuation errors comparing a selection of peer group companies based on comparable profitability and growth, and the average absolute valuation error based on the industry median. For all years the t-test shows insignificant results, implying that a selection of peer group companies based on comparable profitability and growth are not significantly performing better than the valuation of a target company based on the industry median, in terms of average absolute valuation errors. The t-test is insignificant for both peer groups of 6 and 4 companies. In examining the peer group of 6 and 4 companies, the industry absolute valuation error and the peer group absolute valuation error are comparable, except for EV2008 with peer group error of 0.3224 and
24 an industry absolute valuation error of 0.2333. Although the results are not statistically significant, in terms of economical significance, a practitioner would prefer using the industry median instead of the median of a peer group, because both yield almost the same results and the industry median is easier to determine than to compile a peer group with a regression methodology.
Investigating the standard deviation of the absolute valuation errors for the EV/EBIT t+1 multiple, for a peer group of 6 companies, shows an increase from 0.80 to 0.81. For a peer group of 4 companies it increased from 0.7185 to 0.7765. This implies that the precision of the measurement of the absolute valuation error decreased.
The results regarding to the standard deviation of table 8 and 9 are in line with the results of Schreiner and Spremann (2007) and Cooper and Cordeiro (2008), who concluded that the dispersion of the absolute valuation errors increased by adding more companies to the peer group.
Related to the optimal size of the peer group, the peer group of 6 companies’ average absolute valuation error is compared with the average absolute valuation error of a peer group of 4 peers. This leads to an average absolute valuation error of 0.26 and 0.25 for the average absolute error of EV/EBITDA t+1 for respectively 6 and 4 peers, as can be seen in table 10. Investigating the EV/EBIT t+1, the absolute valuation error is 0.27 for 6 peers, compared with the absolute valuation error of a peer group of 4 companies, which has a value of 0.28. For both multiples, the values are almost the same.
Table 10. Average absolute valuation error for peer groups with 6 and 4 companies
This table shows the average absolute valuation error for peer groups consisting of 6 and 4 companies, selected based on their warranted multiple. The average absolute valuation error is presented for the EV/EBITDA t+1 and the EV/EBIT t+1 multiple. The t-test is insignificant.
The values in the table 4 are the average absolute valuation errors of all years for the peer groups built up out of 6 and 4 companies and for both multiples. The t-test in insignificant, which implies that there are no significant differences between the peer groups of 6 companies and the peer groups of 4
EV/ EBITDA t+1
6 peers Absolute valuation error
4 peers
Absolute valuation error p-value t-test
0.26 0.25 0.47
EV/EBIT t+1
6 peers Absolute valuation error
4 peers
Absolute valuation error p-value t-test
25 companies and therefore a peer group of 6 companies performs statistically not better than a peer group of 4 companies and vice versa.
26
5. Conclusion
According to Koller et al. (2010), peer group companies need to be comparable in terms of profitability, growth and underlying business model. This paper examines whether peer group selection based on comparable expected profitability and expected growth, leads to lower absolute valuation errors, compared to average absolute valuation errors based on the median of the industry. The analysis is performed for the Enterprise Value/EBITDA and the Enterprise Value/EBIT multiple, because these two multiples are commonly used by practitioners. Previously, Alford (1992) showed that selection of peer group companies based on industry membership or a combination of return on equity’s (ROE) as a measure for profitably, and total assets (TA) as a measure for size, turned out to be significant factors for the selection of peer group companies. Cheng and McNamara (2000) and Bhojraj and Lee (2002) improved the results of Alford (1992) by using a combination of industry membership, total assets and profitability, growth and other risk characteristics (e.g. leverage). These variables are based on the value drivers, as presented in formula (1), which is driven by growth, profitability and the cost of capital. Based on this insight, the absolute valuation error of peer groups selected on comparable profitability and growth of peer groups consisting of 6 and 4 companies, are compared with the absolute valuation error of the industry.
The research method used in this study is based on the methodology of Bhojraj and Lee (2002). In order to select peer group companies, a so-called ‘warranted multiple’ is calculated for each company. This paper investigates the accuracy of the multiple valuation method by examining the absolute valuation error of a peer group multiple in comparison with the absolute industry median error. In this method, a comparison is made for valuing a target company based on the median of the industry, and a valuation of the target company based on the median of the peer group. The selection of these peer companies is firstly based on industry membership, using the two-digit SIC Classification and secondly, on comparable expected profitability and expected growth, which is included in the warranted multiple.
This study contributes to the existing literature by investigating forward-looking multiples based on the consensus forecast of analysts. Furthermore, this research uses a European based sample, and contributes with respect to the optimal size of the peer group in practicing the multiple valuation method.
27 forward-looking multiples include a forecast for three years (t+1, t+2, t+3). Dijkstra (2013) showed that earlier-year forecasts instead of later-year forecasts for enterprise multiples resulted in a more accurate valuation, he found the most accurate valuation results for t+1. Therefore, the analyses in this study are performed with one-year forward-looking multiples.
In order to investigate the valuation accuracy the t-test is applied. This test compares the average median of the peer group absolute valuation errors with the average absolute valuation error of the median of the industry. These tests are performed on the enterprise multiples EBITDA t+1 and EBIT t+1. The t-test investigates whether there are significant differences between the absolute valuation errors over the years 2008–2011.
The results of the t-test are insignificant, which implies that there is no statistical evidence to conclude that compiling a peer group based on a warranted multiple with forward-looking data leads to lower absolute valuation errors than valuing the same company with the median multiple of the industry. This result is found for EV/EBITDA t+1 and EV/EBIT t+1 multiples. Despite the statistical insignificance, the results can provide some economical insights. Being a practitioner and compiling a peer group is not that attractive when the target company can also be valued by the industry median solely, resulting in a similar result. The results of the t-test related to the last part of the research question, the optimal size of the peer group, are also insignificant, which implies that there are no significant differences in absolute valuation errors between the peer groups of 6 companies and the peer groups of 4 companies. This is contrary to the results of Schreiner and Spremann (2007) and Cooper and Cordeiro (2008), who showed that including more companies in a peer group leads to significantly lower absolute valuation errors.
The difficulty regarding the literature of the multiple valuation method is the diversity in the investigated multiples and methodologies. The uniqueness of this paper in applying warranted multiples is at the same time a weakness, due to the difficulties in the comparability of the results. This methodology, adapted from Bhojraj and Lee (2002) is only used in the papers of Bhojraj and Lee (2002), Bhojraj et al. (2003), Hermann and Richter (2003) and Kohlleppel (2010). However, Bhojraj and Lee (2002), Bhojraj et al. (2003) and Hermann and Richter (2003) did not investigate their results for the EV/EBITDA and EV/EBIT multiple, and never examined a European sample or forward-looking multiples.
28 allowed for unprofitable companies. It is unclear how this influences the performed regressions and R² values.
29
6. References
Alford, A. A., 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.
Arzac, E. R., 2005. Valuation for Mergers, Buyouts, and Restructuring. Wiley, Hoboken, NJ.
Benninga, S., Sarig, O., H., 1997. Corporate finance: a valuation approach. New York: McGraw-Hill.
Bhojraj, S., Charles M. C., L., 2002. Who Is My Peer? A Valuation-Based Approach to the Selection of Comparable Firms. Journal of Accounting Research 40(2), 407-439.
Bhojraj, S., Charles M. C., L., Oler, D. K., 2003. What’s My Line? A Comparison of Industry Classification Schemes for Capital Market Research. Journal of Accounting Research 41, 745-775.
Cheng, C. S., Agnes McNamara, R., 2000. The valuation accuracy of the price earnings and price-book benchmark valuation methods. Review of Quantitative Finance and Accounting 15, 349-370.
Cooper, I., Cordeiro, L., 2008. Optimal Equity Valuation Using Multiple: The Number of Comparable Firms. Working paper, London Business School.
Damodaran, A., 2002. Investment Valuation. Second edition, John Wiley & Sons.
Deangelo, L., 1990. Equity Valuation and Corporate Control. Accounting Review 65, 93-112.
Dijkstra, K.., 2013. Forward-looking Enterprise Value Multiples: An Empirical Research. Master Thesis University of Groningen
Dittmann, I., Ernst M., Kemper J., 2004. How Fundamental are Fundamental Values? Valuation Methods and their Impact on the Performance of German Venture Capitalists. European Financial Management 10 (4), 609-638.
30 Eberhard, A. C., 2004. Comparable Firms and the Precision of Equity Valuations. Journal of Investing 13, 48-54.
Feltham, G,. A., Ohlson, J., A., 1995. Valuation and clean surplus accounting for operating and financial activities. Contemporary accounting research 11(2), 689-731.
Flöstrand, P., 2006. The Sellside: Observations on Intellectual capital Indicators. Journal of Intellectual Capital (6), 385-396.
Henschke, S., Homburg, C., 2009. Equity Valuation Using Multiples: Controlling for Differences Between Firms. Working paper, University of Cologne.
Herrmann, V., Richter, F., 2003. Pricing with Performance-Controlled Multiples. Schmalenbach Business Review 55 (6), 194-219.
Kaplan, S., N., Ruback, R., S., 1995. The valuation of cash flow forecasts: An empirical analysis. The Journal of Finance 50(4) 1059-1093.
Kim, M., Ritter, J.R., 2003. Valuing IPOs. Journal of Financial Economics 53 (3), 409-437.
Kohlleppel, B., A., 2011. Country Effects on European Enterprise Valuation Multiples. Master Thesis University of Groningen
Koller, T., Goedhart, M., Wessels, D., 2010. Valuation, Measuring and Managing the Value of Companies. Fifth edition, University edition. John Wiley & Sons.
Lie, E., Lie, H.J., 2002. Multiples Used to Estimate Corporate Value. Financial Analysts Journal 58 (2), 44-54.
Liu, L., Nissim, D., Thomas, J., 2002. Equity Valuation Using Multiples. Journal of Accounting Research 40(1), 135-172.
Lohnert, P. G., U. J. Bockmann, 2005. Multiplikatorverfahren in der Unternehmensbewertung. Praxishandbuch der Unternehmungsbewertung 3rd edition, 403-428.
31 Penman, S. H., 2004. Financial Statement Analysis and Security Valuation, 2nd edition, McGraw- Hill, NY.
Schreiner, A., Spremann, K., 2007. Multiples and Their Valuation Accuracy in European Equity Markets. Working paper, University of St. Gallen.
Suozzo, P., Cooper, S., Sutherland, G., Deng, Z., 2001. Valuation Multiples: A Primer. UBS Research Report, UBS Investment Bank Department.
Volker, H., Richter, F., 2003. Pricing with performance-controlled Multiples, Schmalenbach Business Review 55,194-219.
Witter, D., 1999. How do we value stocks? Morgan Stanley Research report
32 Sales Growth CAGR EV 07 Sales Growth CAGR EV 08 Sales Growth CAGR EV 09 Sales Growth CAGR EV 10 profit CAGR EV 07 profit CAGR EV 08 profit CAGR EV 09 profit CAGR EV 10 Mean 0.02 0.05 0.05 0.04 0.00 0.06 0.05 0.03 Median 0.02 0.04 0.06 0.04 0.00 0.04 0.05 0.03 Maximum 0.06 0.12 0.08 0.07 0.06 0.15 0.08 0.07 Minimum -0.04 0.01 0.02 0.01 -0.08 0.02 0.01 0.01 Std. Dev. 0.03 0.03 0.02 0.02 0.03 0.04 0.02 0.02 Skewness -0.55 1.45 -0.50 -0.17 -0.54 1.13 -0.11 0.91 Kurtosis 3.17 4.79 3.03 2.91 4.05 3.55 1.99 2.51 Jarque-Bera 0.86 8.27 0.70 0.09 1.61 3.83 0.76 2.51 Probability 0.65 0.02 0.71 0.96 0.45 0.15 0.68 0.29 # Industries 17 17 17 17 17 17 17 17
7. Appendix
7.1 Key value driver formula*
*As presented in Koller et al. (2010) p. 41
7.2 Descriptive statistics expected profitability and growth
Table 11. Descriptive statistics expected profitability and growth
