1 The Determinants of the Marketability Discount: a Multiple Approach with Evidence from
The American Market By:
Tom Bekkering1 University of Groningen Faculty of Economics and Business
MSc Finance
Supervisor: Dr. ing. N. Brunia Date: January 9, 2020
Abstract
This paper examines the determinants of the marketability discount on transaction multiples, using an improved model of Rodríguez López and Rubio Martín (2019). Several variables have been added to the model. For example, cyclical behaviour is measured by the S&P500 price-to-earnings trading multiple. Based on several regression, I find that the type of buyer and the performance of the target significantly impact the differences between transaction multiples of public targets and private targets. Furthermore, I find a negative marketability discount. This provides evidence for further discussion of the application of the marketability discount in general.
JEL Classification: G34
Keywords: Marketability discount Word count: 10,280
2 1. Introduction
The valuation of companies includes various challenges. According to Koller et al. (2015), the discounted cash flow (DCF) method is the most accurate and flexible method for the valuation of companies. The practitioner evaluates the future cash flows of the company and discounts the cash flows with a certain discount factor towards a present value. However, the outcome could differ substantially among practitioners due to the subjectivity of certain assumptions, estimates and industry figures. Besides the DCF method, Koller et al. (2005) also recommend a multiple approach, which concerns the comparison of the ratio between the enterprise value and a specific item on the financial statements of several similar companies. The multiple approach provides critical insights regarding the valuation of companies.
Public companies are obliged to disclose financial statements. Because of this obligation for public companies, the financial information is widely available to carry out the valuation of a public company based on a multiple approach. This is in contrast with the valuation of a private company based on a multiple approach because it is more difficult to get access to financial statements of private companies. In general, to overcome this issue in valuing private companies, the valuation of a similar public company is used as a proxy and a discount will be applied. This discount is defined as the private company discount or the marketability discount. A lack of marketability occurs when a transaction includes a private target and results in a value adjustment (Paglia and Harjoto, 2010). The marketability discount is considered a downward adjustment and possibly the largest monetary impact on the value of a company (Glazer, 2005).
However, there is a lack of consensus on the application of the marketability discount. In a legal setting in the United States for example, Emory (1995) finds that state courts in multiple states did not agree on applying a marketability discount in dissenters’ rights cases. To determine the fair value of the shares in a private company, four state courts rejected the idea of applying a marketability discount to these shares. On the contrary, seven state courts justify the use of a marketability discount to determine the fair value of private company shares.
3 In this paper, the following research question will be addressed: What determines the marketability discount in the American market? First, to control for several factors that influence a transaction multiple, variables are added to the regression model to isolate the marketability discount. The marketability discount is measured as the remaining difference in transaction multiples between a public target and a private target. Out of three transaction multiples, one transaction multiple is determined as the dependent variable based on the goodness of fit. The interaction between several independent variables and the marketability discount is measured with interaction variables. An interaction variable between an independent variable and the marketability discount will be added when the mean transaction multiple of public targets and private targets significantly differs because of this independent variable. The considered time frame for this paper is January 2005 till January 2019.
The results show a negative marketability discount. This provides evidence for further discussion regarding the application of the marketability discount in general. Furthermore, the type of buyer and the performance of the target have a significant impact on the differences between transaction multiples of private targets compared to transaction multiples of public targets.
The remainder of this paper is organized in the following way: after the introduction, the literature overview and the hypotheses are presented. The third section contains the methodology of this paper. In the fourth section, a description of the population and the sample will be discussed, including summary statistics of the sample. The fifth section contains the results of the model presented in the third section and robustness tests. The last section contains conclusions and the discussion.
2. Literature overview
This section contains the literature considered for this paper. In the first part of this section, three different approaches to determine the marketability discount are discussed. The second part contains the factors that influence the marketability discount. In the last part, the hypotheses are presented.
2.1 The marketability discount
4 for a specific period of time. When the employee does not meet certain goals during this period, the employee must return all or some stocks back to the employer. These studies state that the difference between the stock price on the exchange and the stock price that employees pay for their restricted stocks, is mainly due to the lack of marketability. For example, Abudy and Benninga (2016) research the value of restricted stock grants to non-executive employees. They find that restricted stocks for non-executive employees are issued at an average 30.3% discount. This discount is dependent on firm characteristics like size and performance. Additionally, Bajaj et al. (2002) study the marketability discount for private placements occurring from 1990 until 1995. An average marketability discount of 14.09% is found on these private placements. However, by definition, restricted stocks are restricted and non-marketable for a limited time period. This marketability in the near future does not apply to privately held companies. Therefore, the determined marketability discount based on restricted stock studies is not an appropriate marketability discount to use in valuing privately held companies.
5 Table I
Literature overview
This table contains an overview of the literature regarding the three approaches to determine the marketability discount. The first column specifies the author. The second column presents the considered time frame. In the third column, one of the three approaches is mentioned. The fourth column contains the dependent variable. In the fifth column, the independent variables are presented. The last column summarizes the results regarding the marketability discount.
Author Period Approach Dependent
variable Independent variables Results Abudy and Benninga
2000 - 2009 Restricted stock Marketability discount
Vesting period, volatility, stock return, size and financial crisis
Average discount of 30.3%
Bajaj 1990 - 1995 Restricted stock Difference between market price and private placement price Proceeds of the issue, fraction of shares issued, volatility, Z-score and registered shares Average discount of 14.09%
Emory 1980 - 1997 Pre-IPO Average
6 In recent years, the multiple approach is a broadly researched method to determine the marketability discount. Several studies focus on the marketability discount with an Earnings Before Interest, Taxes, Depreciation and Amortization (EBITDA) transaction multiple, Earnings Before Interest and Taxes (EBIT) transaction multiple or sales transaction multiple. Paglia and Harjoto (2010) study the marketability discount based on a multiple approach with a matched-pair analysis. They focus on the EBITDA and sales transaction multiples of transactions that occurred in United States from 1994 to 2008. Paglia and Harjoto (2010) find an average marketability discount for sales multiples and EBITDA multiples of 75% and 50%, respectively. Moreover, Koeplin and Saphiro (2000) match public transactions and private transactions based on industry, size, country and year, to determine a marketability discount. The considered time frame is between 1984 and 1998. A distinction between transactions occurred in the United States and foreign transactions is made. Koeplin and Saphiro (2000) focus on EBITDA transaction multiples, EBIT transaction multiples, sales transaction multiples and net assets at book value transaction multiples. They find an average discount of 28% on EBITDA multiples in the American market. However, the matched-pair analysis includes limitations. For example, it is probable that after matching the transactions on several factors, there are still substantial differences between the public target and the private target because the companies are not identical. A part of the observed discount might be attributable to these differences instead of a lack of marketability. This could lead to biased coefficients.
Rodríguez López and Rubio Martín (2019) also study the marketability discount based on a multiple approach but use a regression model to determine the marketability discount. The focus is on EBITDA transaction multiples and sales transaction multiples. They isolate the marketability discount by controlling for several variables that affect a transaction multiple. These control variables are for example measures for industry, size, control of the buyer after the transaction and the type of buyer. The marketability discount is determined as the coefficient of the dummy variable that measures if the target is a privately held company. Rodríguez López and Rubio Martín (2019) base their findings on transactions that occurred in the Spanish market between 2006 and 2017. They find a negative relationship between a private target and an EBITDA and sales transaction multiple. These results are in line with the findings of Paglia and Harjoto (2010) and Koeplin and Saphiro (2000).
2.2 Factors that influence the marketability discount
7 labels do not matter for a multiple approach. Based on empirical research, they find that valuation multiples within industries vary significantly due to different performance among companies. However, using a measure of industry effect for a multiple approach is in line with Koeplin and Saphiro (2000), Paglia and Harjoto (2010) and Rodríguez López and Rubio Martín (2019). Additionally, Paglia and Harjoto (2010) and Rodríguez López and Rubio Martín (2019) also include a measure of performance in their model. Paglia and Harjoto (2010) or Koeplin and Saphiro (2000).
Various studies have examined additional characteristics of a transaction and they concluded that several variables affect the valuation of a company and therefore the magnitude of a transaction multiple. For example, Kooli et al. (2003) investigate the impact of size on the private company discount with a matched-pair analysis, based on transactions that occurred between 1995 and 2002. They match private transaction multiples with a portfolio of public transaction multiples. The composition of the portfolio is based on industry, country and year. They determined the private company discount as the percentual difference between a private company transaction multiple and a portfolio of public transaction multiple. Their findings show that small private companies are transacted at a lower price. Hence, size has a positive impact on the transaction multiple and influences the marketability discount.
Another additional variable that affects the magnitude of a transaction multiple considers the control of the buyer after the transaction. Hanouna et al. (2000) conducted research to value majority control of the buyer. They measure the value of control for both the United States and foreign transactions by calculating the difference between the offer premium for minority transaction and similar majority transactions. In the United States, they find a control premium of approximately 30%. Hence, according to Hanouna et al. (2000), if a buyer obtains a majority control after the transaction, the value of the target increases with 30% on average. Hanouna et al. (2000) also concluded that cross-border transactions exhibit a lower control premium. Hanouna et al. (2000) base their findings on 6,119 domestic transactions and 3,447 foreign transactions that occurred between January 1986 and September 2000. Jarrel and Poulsen (1989) confirm these results. Jarrel and Poulsen (1989) study the returns of shareholders in tender offers. In total, the final sample included 770 tender offers that occurred between 1963 and 1980. These tender offers generated an average premium of approximately 29%.
8 insignificant results regarding the impact of public buyer on the marketability discount and on the transaction multiple. They do find significant results regarding the impact of financial institutions on the marketability discount and on the transaction multiple.
The last variable that affects the magnitude of a transaction multiple is the state of the economy. Rodríguez López and Rubio Martín (2019) assess the impact of cyclical behaviour by adding year dummy variables for each year to their regression. They find a significant negative impact in 2009 and 2011 on EBITDA and sales transaction multiples. Furthermore, they find a significant positive impact in 2013, 2014, 2015, 2016 and 2017 on EBITDA and sales transaction multiples. The methodology is in line with Paglia and Harjoto (2010). Nonetheless, using year dummies to check for cyclical behaviour might be problematic due to the number of observations each year.
To sum up the empirical literature, a multiple approach based on a matched-pair analysis is the most common method to determine the marketability discount. However, this approach could result in an observed discount that is not fully attributable to the lack of marketability. Therefore, a regression model is more convenient to determine the magnitude and the determinants of the marketability discount. Furthermore, while there is consensus on the control variables to isolate the marketability discount, empirical literature is not consistent about the factors that influence the marketability discount.
2.3 Hypotheses
Previous studies find evidence of several factors that influence the marketability discount. However, a recent study on the marketability discount in the American market is absent. The considered time frame of this research is from January 2005 till January 2019. The objective of this paper is to determine the factors that influences the marketability discount on a transaction multiple in the American market. Consequently, this research addresses the following research question:
“What determines the marketability discount in the American market?”
This research question leads to the following hypotheses:
Null hypothesis I: The marketability discount is not related to the size of the target.
9 Alternative hypothesis II: The marketability discount is positively related to the performance of
the target.
Null hypothesis III: The marketability discount is not related to majority control of the buyer. Alternative hypothesis III: The marketability discount is positively related to majority control of
the buyer.
Null hypothesis IV: The marketability discount is not related to a publicly traded buyer.
Alternative hypothesis IV: The marketability discount is positively related to a publicly traded
buyer.
By testing these alternative hypotheses, inference will be drawn regarding the factors that influence the marketability discount. Several robustness tests will check these inferences.
3. Methodology & Data
In this section, the methodology and the data will be discussed. First, the model of Rodríguez López and Rubio Martín (2019) will be discussed. The second part of this section contains the improved model. In the third part, a description of the data will be presented. The last part contains the descriptive statistics.
3.1. Rodríguez López and Rubio Martín’s (2019) model
The model of Rodríguez López and Rubio Martín (2019) contains several control factors to isolate the marketability discount. The measures for size, year, industry and country are in line with Koeplin and Saphiro (2000) and Kooli et al. (2003). Other measures are not based on previous literature. Rodríguez López and Rubio Martín (2019) isolate the marketability discount with the following equation:
𝐸𝑉 _𝑅𝑎𝑡𝑖𝑜 = 𝛼0𝑖+ 𝛼1𝑖𝑌𝑒𝑎𝑟𝑖+ 𝛼2𝑖𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑖 + 𝛼3𝑖𝑆𝑖𝑧𝑒𝑖 + 𝛼4𝑖𝑃𝑟𝑜𝑓𝑖𝑡𝑖+ 𝛼5𝑖𝐺𝑅𝑖 +
𝛼6𝑖𝐿𝑒𝑣𝑒𝑟𝑖+ 𝛼7𝑖𝑆𝐺𝐸𝑖+ 𝛼8𝑖𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑖 + 𝛼9𝑖𝐵𝑢𝑦𝑒𝑟𝑖+ 𝛼10𝑖𝑀𝐷𝑖 + 𝜀𝑖 (1), where:
𝐸𝑉 _𝑅𝑎𝑡𝑖𝑜 = sales or EBITDA transaction multiple, 𝑌𝑒𝑎𝑟𝑖 = the year dummy variables,
10 𝑆𝑖𝑧𝑒𝑖 = the book-value of the assets of the target. A categorical dummy variable with four levels,
𝑃𝑟𝑜𝑓𝑖𝑡𝑖 = marketability discount. A dummy variable: 1 if the target is private, 0 otherwise, 𝐺𝑅𝑖 = EBITDA growth momentum of the target. A dummy variable: 1 if last year’s EBITDA growth is greater than the average EBITDA growth of previous two years, 0 otherwise, 𝐿𝑒𝑣𝑒𝑟𝑖 = the leverage ratio of the target,
𝑆𝐺𝐸𝑖 = standard deviation of the target’s EBITDA, calculated over the last three years, 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑖 = minority control. A dummy variable: 1 if minority control, 0 otherwise,
𝐵𝑢𝑦𝑒𝑟𝑖 = type of buyer dummy variables for private buyer, public buyer and financial institution,
𝑀𝐷𝑖 = The ratio between EBITDA and sales of the target.
To explain the factors that influence the marketability discount, Rodríguez López and Rubio Martín’s (2019) incorporated interaction variables between the marketability discount and several variables. These variables are size, EBITDA growth momentum, leverage, standard deviation of the target’s EBITDA, EBITDA margin, minority control and type of buyer. However, a reasoning for choosing these interaction variables is absent.
3.2. Improved model
11 The transaction multiple with the lowest mean squared error is the dependent variable for this paper. The transaction multiple is determined as follows:
𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑛𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑒 = VT
𝐼𝑡𝑒𝑚 𝑓𝑖𝑛𝑎𝑛𝑐𝑖𝑎𝑙 𝑠𝑡𝑎𝑡𝑒𝑚𝑒𝑛𝑡𝑠 (2), where VT is the total value of the target and the denominator is sales, EBITDA or EBIT.
The explanatory variables of Rodríguez López and Rubio Martín (2019) also allow for improvement. For example, incorporating year dummies to capture potential cyclical behaviour might be problematic due to the number of observations per year. Rodríguez López and Rubio Martín (2019) do not provide an overview of the number of observations per year. However, their model includes 824 observations in total and the considered period is 11 years. This results in an average number of observations per year of approximately 75. The interpretation of the results of these year dummies based on an average of 75 observations is not reliable. To capture potential cyclical behaviour, it is more appropriate to use the S&P500 price-to-earnings trading multiple. The S&P500 price-to-earnings trading multiple is the stock performance related to earnings of the 500 American companies with the highest market capitalization. It is a proxy for the state of the economy.
Another potential improvement of the model is a different measure for size. While using sales as a measure for size is in line with Paglia and Harjoto (2010) and Koeplin and Saphiro (2000), the coefficients are insignificant. Furthermore, multicollinearity is a potential issue if the dependent variable is a sales transaction multiple. I include the deal value in billions as a measure of size.
Rodríguez López and Rubio Martín (2019) uses two performance variables in the model. These performance variables are EBITDA growth momentum and EBITDA margin. Because the EBITDA growth momentum variable is based on three years, the interpretation of the results regarding this variable is not reliable. Therefore, to account for performance in the model, I only use a variable that measures the EBITDA margin. This is also in line with Paglia and Harjoto (2010). The EBITDA margin is measured as follows:
𝐸𝑀 = EBITDA
12 Risk variables are also added to the model of Rodríguez López and Rubio Martín (2019). These risk variables are the standard deviation of the target’s EBITDA and the leverage ratio of the target. The use of these variables is not in line with Paglia and Harjoto (2010) or Koeplin and Saphiro (2000). While the coefficients are significant for both risk variables, the economic interpretation of the standard deviation of the target’s EBITDA is questionable because it is calculated over the last three years. Therefore, I exclude risk variables from the model.
The inclusion of other variables is in line with Paglia and Harjoto (2010) and Koeplin and Saphiro (2000) or other literature. The inclusion of industry dummies in the regression is in line with Rodríguez López and Rubio Martín (2019). Additionally, Koeplin and Saphiro (2000) and Paglia and Harjoto (2010) use industry for the matched pair analysis. The industries of this paper are based on the major industries categorized by Zephyr. To ensure the interpretability of the results, a dummy variable was created for 17 out of the 18 industries. The interpretation of the industry dummy variables is as follows:
the industry dummy = 1 if target operates in that specific industry, 0 otherwise.
Adding a control variable to the model is in line with Rodríguez López and Rubio Martín (2019). Furthermore, the presence of a control premium in transactions is concluded by Hanouna et al. (2000) and Jarrel and Poulsen (1989). The interpretation of majority control variable (MC) is as follows:
MC = 1 if the final stake of the buyer >0.5, 0 otherwise.
The inclusion of a variable that captures the type of buyer is also in line with Rodríguez López and Rubio Martín (2019) and Paglia and Harjoto (2010). While Paglia and Harjoto (2010) consider public and private buyer, Rodríguez López and Rubio Martín (2019) add financial institutions to the type of buyer. Due to the absence of this data, I stick to the distinction of public and private buyers. The interpretation of the buyer variable (PB) is as follows:
PB = 1 if the buyer is public, 0 otherwise.
13 is in line with Rodríguez López and Rubio Martín (2019). The interpretation of this dummy variable (PT) is as follows:
PT = 1 if the target is a private company, 0 otherwise.
The above-mentioned variables combined lead to the following equation:
𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑒 = 𝛼0𝑖+ 𝛼1𝑖𝑆𝑖𝑧𝑒𝑖 + 𝛼2𝑖𝐼𝑛𝑑1𝑖 + 𝛼3𝑖𝐼𝑛𝑑2𝑖 + 𝛼4𝑖𝐼𝑛𝑑3𝑖 + 𝛼5𝑖𝐼𝑛𝑑4𝑖 + 𝛼6𝑖𝐼𝑛𝑑5𝑖 + 𝛼7𝑖𝐼𝑛𝑑6𝑖+ 𝛼8𝑖𝐼𝑛𝑑7𝑖 + 𝛼9𝑖𝐼𝑛𝑑8𝑖 + 𝛼10𝑖𝐼𝑛𝑑9𝑖+ 𝛼11𝑖𝐼𝑛𝑑10𝑖 + 𝛼12𝑖𝐼𝑛𝑑11𝑖 + 𝛼13𝑖𝐼𝑛𝑑12𝑖 + 𝛼14𝑖𝐼𝑛𝑑13𝑖+ 𝛼15𝑖𝐼𝑛𝑑14𝑖 + 𝛼16𝑖𝐼𝑛𝑑15𝑖 + 𝛼17𝑖𝐼𝑛𝑑16𝑖+ 𝛼18𝑖𝐼𝑛𝑑17𝑖 + 𝛼19𝑖𝐼𝑛𝑑18𝑖 + 𝛼20𝑖𝑀𝐶𝑖 + 𝛼21𝑖𝑃𝐵𝑖+ 𝛼22𝑖𝑃𝑇𝑖 + 𝛼23𝑖𝑇𝑀𝑡+ 𝛼24𝑖𝐸𝑀𝑖 + 𝜀𝑖 (4), where 𝛼0𝑖 is a transaction multiple with the following characteristics:
- The target of the transaction operates in the Chemicals, rubber, plastics, non-metallic products industry.
- The buyer does not have a majority control. - The buyer is a publicly traded company. - The target is a publicly traded company.
The estimation of this regression model is done using ordinary least square (OLS) estimators. To review the presence of heteroskedasticity, a White’s test was performed. The outcome of the White’s test was very significant with a p-value of 0.0001. Hence, the regression model suffers from heteroskedasticity. Consequently, robust standard errors were used to overcome the heteroskedasticity issue. Moreover, a Spearman test is conducted to check for multicollinearity. The Spearman correlation test is considered because it is appropriate for categorical variables. In Appendix A, the Spearman correlation matrix is displayed. The correlation coefficient of MC and
Size, PT and Size and PT and MC show signs of multicollinearity. This multicollinearity potentially
affects the results by less precise coefficients. Therefore, a reduced regression model (Model II) takes this multicollinearity into account by removing variable Size.
14 Additionally, a two-samples t-test will be conducted for several independent variables to determine the relevant interaction variables for the regression analysis. This two-samples t-test also assumes equal variances. Hence, the Levene’s test is conducted to check for equal variances. The independent variables considered for an interaction variable with the marketability discount are in line with Rodríguez López and Rubio Martín (2019). These variables are Size, MC, PB and EM. If one subgroup of a variable shows significant differences between mean EBITDA transaction multiples with public targets and mean EBITDA transaction multiples with private targets, the variable is determined as relevant input for an interaction variable. If all variables show significant results regarding the two-samples t-test, the result will be the following equation:
𝑇𝑟𝑎𝑛𝑠𝑎𝑐𝑡𝑖𝑜𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑒 = 𝛼0𝑖+ 𝛼1𝑖𝑆𝑖𝑧𝑒𝑖 + 𝛼2𝑖𝐼𝑛𝑑1𝑖 + 𝛼3𝑖𝐼𝑛𝑑2𝑖 + 𝛼4𝑖𝐼𝑛𝑑3𝑖 + 𝛼5𝑖𝐼𝑛𝑑4𝑖 + 𝛼6𝑖𝐼𝑛𝑑5𝑖 + 𝛼7𝑖𝐼𝑛𝑑6𝑖+ 𝛼8𝑖𝐼𝑛𝑑7𝑖 + 𝛼9𝑖𝐼𝑛𝑑8𝑖 + 𝛼10𝑖𝐼𝑛𝑑9𝑖+ 𝛼11𝑖𝐼𝑛𝑑10𝑖 + 𝛼12𝑖𝐼𝑛𝑑11𝑖 + 𝛼13𝑖𝐼𝑛𝑑12𝑖 + 𝛼14𝑖𝐼𝑛𝑑13𝑖+ 𝛼15𝑖𝐼𝑛𝑑14𝑖 + 𝛼16𝑖𝐼𝑛𝑑15𝑖 + 𝛼17𝑖𝐼𝑛𝑑16𝑖+ 𝛼18𝑖𝐼𝑛𝑑17𝑖 + 𝛼19𝑖𝐼𝑛𝑑18𝑖 + 𝛼20𝑖𝑀𝐶𝑖 + 𝛼21𝑖𝑃𝐵𝑖+ 𝛼22𝑖𝑃𝑇𝑖 + 𝛼23𝑖𝑇𝑀𝑡+ 𝛼24𝑖𝐸𝑀𝑖 + 𝛼25𝑖𝑆𝑖𝑧𝑒𝑖 ∗ 𝑃𝑇𝑖+
𝛼26𝑖𝑀𝐶𝑖 ∗ 𝑃𝑇𝑖 + 𝛼27𝑖𝑃𝐵𝑖∗ 𝑃𝑇𝑖+ + 𝛼28𝑖𝐸𝑀𝑖∗ 𝑃𝑇𝑖 + 𝜀𝑖 (5). Lastly, a regression analysis will be performed based on equation (4) to determine the magnitude
of the marketability discount. A regression analysis based on equation (5) will be conducted to assess the determinants of the marketability discount and to test the hypotheses.
3.3. Data description
The data sample used for the regressions is obtained from mergers and acquisitions database Zephyr, owned by Bureau van Dijk. The observations contain transactions that occurred between January 2005 till January 2019. The monthly data of the S&P500 price-to-earnings trading multiple is retrieved from Macrotrends2. Macrotrends is a company that provides a research platform for long term investors. To overcome potential issues with country effects, the buyer and target are located in the United States of America. To ensure the interpretability of the results, a substantial amount of observations is removed due to the absence of values from one or more variables. This results in an initial sample that contains of 5,546 observations, without checking for missing values of the transaction multiples.
15 Table II
Total number of observations per multiple
This table contains an overview of the considered multiples with a distinction between private and public targets. The first column specifies the item on the financial statement considered for the multiple. The second column presents the number of observations per multiple. In the third column, the number of observations including a private target are presented. The last column presents the number of observations including a public target.
Multiple N Public target Private target
Sales 5,160 1,702 3,458
EBITDA 3,611 1,389 2,222
EBIT 3,452 1,281 2,171
Table II shows the number of observations per multiple within the initial sample. The distinction between public and private targets shows that the sample contains approximately 60% of transactions with a private target and 40% of transactions with a public target. Furthermore, the sample size is considered large enough for each multiple considered to assume a normal distribution based on the central limit theorem.
In table III, all variables that are relevant for the regression are presented. The sales, EBITDA and EBIT variables are used to calculate the transaction multiple.
16 Table III
Variables
This table contains an overview of the variables used in this paper. The first column specifies name of the variable. The second column presents the symbol of the variable. In the third column, the exact name used in Zephyr for the variable or the formula is presented. The last column contains remarks if needed.
Variable name Symbol Zephyr name/
formula
Remarks
Sales Rev Target operating
revenue/turnover millions USD last avail. yr
In millions
EBITDA EBITDA Target EBITDA
USD last avail. yr
EBIT EBIT Target EBIT USD
last avail. yr Total value of the target VT Deal total target
value USD Sales transaction multiple SalesM VT / Rev EBITDA transaction multiple EBITDAM VT / EBITDA
EBIT transaction multiple EBITM VT / EBIT
Deal size Size Deal value USD In billions
Banking industry Ind1 Target major sector
Construction industry Ind2 Target major sector Education, Health industry Ind3 Target major sector Food, Beverages Tobacco industry Ind4 Target major sector Gas, Water, Electricity industry Ind5 Target major sector Hotels & Restaurants industry Ind6 Target major sector
Insurance industry Ind7 Target major sector
Machinery, Equipment, Furniture, Recycling industry
Ind8 Target major sector
Metals & Metal products industry Ind9 Target major sector Other Services industry Ind10 Target major sector Post and Telecommunications industry Ind11 Target major sector Primary Sector (Agriculture, Mining,
etc.)
Ind12 Target major sector
Public Administration and Defense industry
Ind13 Target major sector
Publishing, Printing industry Ind14 Target major sector Textiles Wearing Apparel, Leather
industry
Ind15 Target major sector
Transport industry Ind16 Target major sector
Wholesale & Retail Trade industry Ind17 Target major sector Wood, Cork, Paper industry Ind18 Target major sector
EBITDA margin EM EBITDA / Rev EM * 100
Majority control MC Final stake (%)
Public buyer PB Acquiror listed
Private target PT Target listed
17
4.2. Descriptive statistics
Table IV displays the descriptive statistics of the regression variables. The skewness and kurtosis values of the dummy variables are not presented because these values are non-informative. The non-binary variables are positively skewed. Skewness indicates that outliers are still present. Skewness of the sample leads to substantial differences between the mean and the median. Therefore, both the mean and median will be presented in the descriptive statistics. Skewness is also an indicator of a non-normal distribution and normality is an issue for small samples. However, the sample used in this paper is considered large enough to overcome normality issues based on the central limit theorem. Notable is substantial difference between the minimum value and maximum value of the variable TM. The highest values are observed at the end of 2008 and 2009. This time period is also considered the start of the financial crisis.
Table IV Descriptive statistics
Table IV contains the descriptive statistics of the variables. The first column specifies the name of the variable. In the second column, the number of observations is presented. The third column contains the mean. In the fourth column, the median of the variable is presented. The fifth column contains the standard deviation. The sixth column presents the minimum value of the variable. The seventh column consists of the maximum value. The eighth column presents of the skewness of the variable. The last column contains the kurtosis of the variable.
N Mean Median Std.dev Min Max Skewness Kurtosis
EBITDAM 2,887 11.50 9.81 6.63 2.97 33.16 1.09* 3.64* Size 2,887 1.66 0.12 5.61 0 108.7 8.85* 111.62* Ind2 2,887 0.01 0 0.09 0 1 Ind3 2,887 0.03 0 0.16 0 1 Ind4 2,887 0.03 0 0.17 0 1 Ind5 2,887 0.03 0 0.16 0 1 Ind6 2,887 0.04 0 0.21 0 1 Ind7 2,887 0.01 0 0.08 0 1 Ind8 2,887 0.19 0 0.39 0 1 Ind9 2,887 0.02 0 0.14 0 1 Ind10 2,887 0.29 0 0.46 0 1 Ind11 2,887 0.03 0 0.17 0 1 Ind12 2,887 0.05 0 0.22 0 1 Ind13 2,887 0.00 0 0.06 0 1 Ind14 2,887 0.04 0 0.19 0 1 Ind15 2,887 0.01 0 0.11 0 1 Ind16 2,887 0.03 0 0.17 0 1 Ind17 2,887 0.1 0 0.31 0 1 Ind18 2,887 0.01 0 0.11 0 1 MC 2,887 0.51 1 0.5 0 1 PB 2,887 0.78 1 0.41 0 1 PT 2,887 0.62 1 0.49 0 1 TM 2,887 21.74 21.18 9.74 13.01 122.41 7.03* 61.4* EM 2,887 21.33 17 16.44 0.32 105.72 1.65* 6.16*
18 Table V presents the distribution of the transactions with a distinction between public and private targets per year. Remarkable is the high number of transactions with a public target in 2016. Approximately 24% of the transaction with a public target occurred in 2016. Furthermore, the difference in mean EBITDA transaction multiples between public and private target is significant at the 5% level in 9 out of the 14 years. The difference in mean EBITDA transaction multiples between public and private target regarding the entire sample is also significant at the 5% level. This table shows that the mean EBITDA transaction multiple of private targets is significantly larger than the mean EBITDA transaction multiple of public targets, all other factors held constant. Therefore, the presence of a negative marketability discount or a marketability premium is found.
Table V
Sample distribution per year
This table presents data regarding the EBITDA transaction multiple per year. In the first column, the year of the transaction is presented. The second column presents the number of transactions including a public target. The third column consists of the number of transactions including a private target. The fourth column contains the mean EBITDA transaction multiple of public targets. The median EBITDA transaction multiple of public targets is presented in the fifth column. In the sixth column, the mean EBITDA transaction multiple of public targets is presented. The seventh column contains the median EBITDA transaction multiple of public targets. The eighth column consists of the difference between the mean EBITDA transaction multiple of public targets and the mean EBITDA transaction multiple of private targets. The last column contains the difference between the median EBITDA transaction multiple of public targets and the median EBITDA transaction multiple of private targets.
N Public targets Private targets Difference
Year Public target Private target Mean Median Mean Median Mean Median
2005 3 135 6.26 4.79 13.65 11.86 -7.39* -7.07 2006 9 148 10.73 10.98 13.4 12.67 -2.67 -1.69 2007 18 201 10.11 6.18 14.41 13.63 -4.3** -7.45 2008 10 92 6.78 5.23 13.51 11.36 -6.73*** -6.13 2009 9 52 7.2 5.92 9.77 8.78 -2.57 -2.86 2010 7 103 5.87 5.95 11.4 10.05 -5.54*** -4.1 2011 8 121 8.02 7.54 12.36 10.96 -4.35* -3.42 2012 45 129 8.46 6.7 11.55 9.92 -3.09*** -3.22 2013 29 107 8.94 7.53 12.65 10.94 -3.7*** -3.41 2014 155 123 9.67 7.38 11.76 10.78 -2.09*** -3.4 2015 221 160 9.52 8.71 13.81 12.36 -4.29*** -3.65 2016 484 195 8.87 7.35 11.48 9.99 -2.61*** -2.64 2017 100 125 10.8 8.26 13.6 12.51 -2.8*** -4.25 2018 8 90 13.06 10.87 15.29 14.34 -2.23 -3.47 Total 1,106 1,781 9.27 7.59 12.89 11.29 -3.62*** -3.7 *** p<0.01, ** p<0.05, * p<0.1
19 Noteworthy is the absence of observations of Ind1, which is the banking industry. This means that the banking industry will not be considered as industry dummy for the regression analysis. Most transactions occurred in Ind10. This is the ‘other services’ industry. This industry contains for example, gaming companies, entertainment companies and real estate companies. Additionally, the difference between mean EBITDA transaction multiples with public targets and mean EBITDA transaction multiples with private targets is significant at the 5% level for 13 out of the 18 industries considered for regression analysis. Table VI shows that the mean EBITDA transaction multiple of private targets is significantly larger than the mean EBITDA transaction multiple of public targets, all other factors held constant. This is in line with the results of Table V.
Table VI
Sample distribution per industry
This table presents data regarding the EBITDA transaction multiple per industry. In the first column, the industry is specified. The second column presents the number of transactions including a public target. The third column consists of the number of transactions including a private target. The fourth column contains the mean EBITDA transaction multiple of public targets. The median EBITDA transaction multiple of public targets is presented in the fifth column. In the sixth column, the mean EBITDA transaction multiple of public targets is presented. The seventh column contains the median EBITDA transaction multiple of public targets. The eighth column consists of the difference between the mean EBITDA transaction multiple of public targets and the mean EBITDA transaction multiple of private targets. The last column contains the difference between the median EBITDA transaction multiple of public targets and the median EBITDA transaction multiple of private targets.
N Public targets Private targets Difference Industry Public target Private target Mean Median Mean Median Mean Median
20 In Table VII, the distribution of the transactions per subgroup of Size, MC, PB and EM is presented. Additionally, the results of the two-samples t-test regarding these subgroups are displayed. In panel A, the variable Size is presented in three categories. The category small considers all transactions with a deal value below 100 million euros. medium contains all transactions with a deal value between 100 million euros and 1,000 million euros. The category large consists of all transaction with a deal value higher than 1,000 million euros. The variable EM is also presented in three categories. The category small concerns an EBITDA margin of 10% or lower. The medium category considers EBITDA margins between 10% and 20%. In the high category, EBITDA margins of 20% or higher are presented.
21 Table VII
Sample distribution per interaction variable input
This table presents data regarding the EBITDA transaction multiple per interaction variable input. In the first column, the name of the variable is presented. The second column contains the number of
transactions including a public target. The third column consists of the number of transactions including a private target. The fourth column contains the mean EBITDA transaction multiple of public targets. The median EBITDA transaction multiple of public targets is presented in the fifth column. In the sixth column, the mean EBITDA transaction multiple of public targets is presented. The seventh column contains the median EBITDA transaction multiple of public targets. The eighth column consists of the difference between the mean EBITDA transaction multiple of public targets and the mean EBITDA transaction multiple of private targets. The last column contains the difference between the median EBITDA transaction multiple of public targets and the median EBITDA transaction multiple of private targets.
N Public targets Private targets Difference Variable Public
targets
Private targets
Mean Median Mean Median Mean Median
Panel A: Size
Small 910 467 9.17 7.54 10.8 9.37 -1.63*** -1.83
Medium 156 642 9.59 7.52 13.1 11.31 -3.51*** -3.79
Large 40 672 10.36 10.19 14.15 13.21 -3.79*** -2.85
Panel B: Control of the buyer
Minority control 1,079 323 9.25 7.57 10.41 8.96 -1.16*** -1.39 Majority control 27 1,458 9.91 9.74 13.44 12.07 -3.53*** -2.33
Panel C: Type of buyer
Public buyer 67 571 8.94 7.89 13.94 12.78 -5.01*** -4.89 Private buyer 1,039 1,210 9.29 7.55 12.39 10.84 -3.1*** -1.55
Panel D: EBITDA margin
Low 219 468 10.22 7.62 13.08 11.16 -2.86*** -3.54
Medium 372 611 9.16 7.53 13.08 11.49 -3.92*** -3.96
High 515 702 8.95 7.58 12.6 11.2 -3.65*** -3.62
*** p<0.01, ** p<0.05, * p<0.1
4. Results
This section contains the results of the regression analysis. In the first section, the empirical results will be discussed. The last section contains the robustness tests.
4.1. Empirical Results
22 because the inclusion of insignificant parameters possibly affects the significance of other parameters. Based on a 5% significance level, variables Size, Ind2, Ind7, Ind8, Ind10, Ind13, Ind15,
Ind16 and TM are excluded from Model II. In Model III, the interaction variables discussed in
section three are added to Model II. First, the results of Model II will be discussed. This model presents the magnitude of the marketability discount and the control variables to isolate the marketability discount.
4.1.1 Results of model II
The intercept is an EBITDA transaction multiple with a target that operates in the chemicals, rubber, plastics, non-metallic products industry, a buyer that does not have a majority control, a buyer and a target that are publicly traded companies. Therefore, the interpretation of the coefficient of the industry dummy variables is similar. All these coefficients result in a decrease or increase in the EBITDA transaction multiple compared with an EBITDA transaction multiple with a target that operates in the chemicals, rubber, plastics and non-metallic products industry.
The coefficients of the industry dummy variables present significant and insignificant coefficients. 11 out of the 18 industry dummy variables have statistically significant coefficients and are presented in model II. The coefficient of Ind3 is statistically significant at the 1% level. Therefore, this coefficient indicates that when the target operates in the education or health industry, the EBITDA transaction multiple decreases with 2.36. Furthermore, the coefficient of
Ind4 is statistically significant at the 5% level. The coefficient of Ind4 therefore indicates that when
23 of Ind14 therefore indicates that when the target operates in the publishing or printing industry, the EBITDA transaction multiple increases with 1.96. The coefficient of Ind16 is statistically significant at the 1% level. Therefore, this coefficient indicates that when the target operates in the transport industry, the EBITDA transaction multiple decreases with 2.12. The coefficient of Ind17 is statistically significant at the 1% level. Hence, this coefficient indicates that when the target operates in the post or telecommunications industry, the EBITDA transaction multiple decreases with 2.14. The last significant coefficient regarding the industry dummy variables concerns the wood industry. It is significant at the 1% level. When the target operates in the wood industry, the EBITDA transaction multiple decreases with 3.64.
Table VIII
Results regression model I, II and III
Table VIII contains regression model I, II and III. The explanatory variables in the first column are discussed in section 2. The second column contains the output of model I. In the third column, the results of reduced model II are presented. The fourth column consists of the results of interaction model III. The heteroskedasticity issue is overcome by using robust standard errors. 𝑆𝑖𝑧𝑒 measures the size of the target, the industry dummy variable measures the industry effect, 𝑀𝐶 measures the effect of majority control, PB is a measure of the effect of a private buyer, 𝑃𝑇 is a measure of the impact of a private target.
Model I Model II Model III VARIABLES Full model Reduced model Interaction model
24 Ind14 2.508*** 1.966*** 1.963*** (0.8) (0.682) (0.678) Ind15 -1.426 (1.048) Ind16 -1.61** -2.116*** -2.182*** (0.851) (0.741) (0.743) Ind17 -1.688*** -2.14*** -2.2*** (0.548) (0.354) (0.355) Ind18 -3.077*** -3.645*** -3.631*** (0.833) (0.724) (0.72) MC 2.514*** 2.64*** 0.817 (0.375) (0.37) (1.268) PB -0.915*** -0.919*** 0.226 (0.321) (0.316) (0.617) PT 1.147*** 1.162*** 1.485* (0.352) (0.35) (0.827) TM -0.017 (0.011) EM -0.029*** -0.023*** -0.042*** (0.009) (0.008) (0.01) SizePT 0.006 (0.017) MCPT 1.977 (1.324) PBPT -1.262* (0.717) EMPT 0.029** (0.014) Constant 11.401*** 11.373*** 10.788*** (0.668) (0.435) (0.646) Observations 2,887 2,887 2,887 R-squared 0.14 0.136 0.139 F-statistic 22.93 33.32 29.35 Prob > F 0.000 0.000 0.000
Root mean squared error 6.173 6.181 6.175
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
The coefficient of the variable MC, which determines the effect of majority control of the buyer, is statistically significant at the 1% significance level. This indicates that majority control of the buyer results in a 2.64 increase in the EBITDA transaction multiple compared to minority control of the buyer. This is in line with the findings of Hanouna et al. (2000) Jarrel and Poulsen (1989). They also find evidence of a control premium.
25 is in contrast with Paglia and Harjoto (2010). They find a positive relation with public buyers and transaction multiples.
The coefficient of the variable PT, which determines the effect of a private target and the marketability discount, is statistically significant at the 1% level. This statistical significance indicates that a private target in a transaction results in an increase of the EBITDA transaction multiple of 1.16. Additionally, this indicates the presence of a negative marketability discount. With regards to the mean EBITDA transaction multiple of 11.50, the marketability premium is approximately 10%. This in contrast with Paglia and Harjoto (2010), Rodríguez López and Rubio Martín (2019) and Koeplin and Saphiro (2000). They find evidence of a marketability discount of approximately 30%.
The last significant coefficient is the parameter of the variable EM, which measures the effect of the EBITDA margin of the target. The coefficient is significant at the 1% level. Therefore, a 1% increase in EBITDA margin results in a 0.02 decrease in EBITDA transaction multiple. This is in contrast with Rodríguez López and Rubio Martín (2019). They find a positive relation between EBITDA margin and transaction multiples.
4.1.2 Results of model III
The results of model III are presented in Table VIII, fourth column. By adding the interaction variables to the model, it is possible to test the hypotheses discussed in section two and to answer the research question: “What determines the marketability discount in the American
Market”. The first alternative hypothesis is formulated as follows:
Alternative hypothesis I: The marketability discount is positively related to the size of the target. This hypothesis is tested by the interaction variable of Size and PT. The fourth column in Table VIII shows a coefficient of this interaction variable, which is insignificant at the 10% level. Hence, we cannot reject null hypothesis I and it is not possible to draw inference. This is in contrast with Rodríguez López and Rubio Martín (2019). They find a significant positive interaction variable, which means that an investor is willing to pay more for size of private targets compared to size of public targets.
26 Alternative hypothesis II: The marketability discount is positively related to the performance of
the target.
This hypothesis is tested by the interaction variable of EM and PT. The positive coefficient of this interaction variable is significant at the 5% level. The EBITDA transaction multiple increases with approximately 0.03. Hence, an investor is willing to pay more for performance of private targets compared to performance of public targets. Because this shows a positive relation between the performance of the target and the difference in EBITDA transaction multiples between public and private targets, we can reject null hypothesis II and accept the alternative hypothesis II. This is in contrast with Rodríguez López and Rubio Martín (2019). They find a significant negative interaction variable, which means that a buyer pays less for performance of private targets compared to public targets.
The third alternative hypothesis considers the potential influence of majority control of buyer on the marketability discount. This alternative hypothesis is formulated as follows:
Alternative hypothesis III: The marketability discount is positively related to majority control of
the buyer.
This hypothesis is tested via the interaction variable of MC and PT. At a 10% significance level, the coefficient of this interaction variable is insignificant. Hence, we cannot reject null hypothesis III and any inference regarding this interaction variable is not possible. This is in line with Rodríguez López and Rubio Martín (2019). While they focussed on minority control instead of majority control, the results are also insignificant regarding this interaction variable.
The last alternative hypothesis concerns the potential influence of a publicly traded buyer on the marketability discount. This alternative hypothesis is formulated as follows:
Alternative hypothesis IV: The marketability discount is positively related to a publicly traded
buyer.
27 relation between the type of buyer and the difference in EBITDA transaction multiples between public and private targets, we cannot reject null hypothesis IV.
4.2 Robustness
To review the results of the abovementioned regression model, two robustness tests have been performed. The first test of robustness is a regression model with smaller outliers of the dependent variable. A robustness test with smaller outliers is chosen because the determination of the 10% outliers is not in line with Rodríguez López and Rubio Martín (2019). While trimming the transaction multiple is in line with Rodríguez López and Rubio Martín (2019), They determine the outliers at 5% at each tail. Therefore, the smaller outliers for this robustness test are determined at 5% at each tail. Because of the change in outliers, the number of observations increase to 3,249. The model includes robust standard errors to overcome potential heteroskedasticity issues. The results of this robustness test are presented in Appendix D.
The results of this robustness test indicate similarities and differences compared to the results of model III. The negative coefficient of the interaction variable of PB and PT is significant at the 10% level. The EBITDA transaction multiple decreases because of this interaction variable. This is in line with the results of model III, that a public buyer is willing to pay less for a private target compared to a public target. Regarding the interaction variable of EM and PT, the results of the robustness test differ. While the interaction variable is significant at the 5% level in model III, the interaction variable is not even significant at the 10% level. This is in contrast with the results of model III. Hence, the claim that a buyer pays less for performance of private targets compared to performance of public targets is not confirmed by this robustness test.
28 5. Conclusion
This paper aims to examine the determinants of the marketability discount of transaction multiples. The results of model II show the presence of a negative marketability discount or marketability premium. This finding contradicts the findings of Koeplin and Saphiro (2000), Paglia and Harjoto (2010) and Rodríguez López and Rubio Martín (2019). They find a marketability discount. The findings of this paper contribute to the discussion if a discount on private companies should be applied.
The results of model III show a significant relation between the performance of the target and the differences in EBITDA transaction multiples between public and private targets. Regarding the performance of the target, an investor is willing to pay more for performance of private targets compared to public targets. This is in contrast with the findings of Rodríguez López and Rubio Martín (2019). They conclude that investors are willing to pay less for performance of private targets compared to public targets. This result shows proof that the performance of the target is a determinant of the differences in EBITDA transaction multiples between public and private targets.
Moreover, the results of model III show a significant relation between the type of buyer and the differences in EBITDA transaction multiples between public and private targets. A public buyer is willing to pay less for a private target compared to a public target. This relation has been confirmed by the first robustness test. However, it contradicts the findings of Rodríguez López and Rubio Martín (2019). They find that a publicly traded buyer is willing to pay more for a private target compared to a public target. The proof of this relation contributes to the literature regarding the determinants of the differences in EBITDA transaction multiples between public and private targets.
29 References
Abudy, M., Benninga, S., 2016. Valuing restricted stock grants to non-executive employees. Journal of Economics and Business 86, 33-51.
Bajaj, M., Denis, D. J., Ferris, S. P., Sarin, A., 2001. Firm value and marketability discounts. Journal of Corporation Law 27, 89-115.
Bartlett, M. S., 1937. Properties of sufficiency and statistical tests. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences 160, 268-282.
Bothra, A., Williams, Z., 2019. Multiples analysis: industry labels don’t matter, performance does. McKinsey & Company.
Emory, J. D., 1995. The role of discounts in determining “fair value” under Wisconsin’s dissenters’ rights statutes: the case for discounts. Wisconsin Law Review 5, 1155-1176. Emory, J. D., 1997. The value of marketability as illustrated in initial public offerings of common stock (eighth in a series). Business Valuation Review 16, 123-131.
Glazer, R. T., 2005. Understanding the valuation discount for lack of marketability. The CPA Journal 75, 60.
Hanouna, P., Sarin, A., Shapiro, A. C., 2000. Value of corporate control: some international evidence. Unpublished working paper. Purdue University, West Lafayette.
Jaffe, J. F., Pedersen, D. J., Voetmann, T., 2019. Do unlisted targets sell at discounts? Journal of Financial and Quantitative Analysis 54, 1371-1401.
Jarrel, G. A., Poulsen, A. B., 1989. The returns of acquiring firms in tender offers: evidence from three decades. Journal of Financial Management 18, 12-19.
30 Kooli, M., Kortas, M., L’Her, J., 2003. A new examination of the private company discount: the acquisition approach. Journal of Private Equity 6, 48-55.
Koller, T, Goedhart, M, Wessels, D., 2015. Valuation: Measuring and Managing the Value of Companies. New Jersey: Wiley.
Levene, H., 1960. Robust tests for equality of variances. Stanford University Press. Stanford. Paglia, J., Harjoto, M. A., 2010. The discount for lack of marketability in privately owned companies: a multiples approach. Journal of Business Valuation and Economic Loss Analysis 5, 1-26.
31 Internet references
Mactotrends. (2019, December 10). Retrieved from
32 Appendix A
Table A.1: Spearman correlation matrix
This table below contains the Spearman correlation table of the dependent and independent variables.
34 Appendix B
Table B.1: Regression output sales transaction multiple
This table presents the regression output of the sample that includes a sales transaction multiple as dependent variable. Only significant parameters and interaction variables are presented.
VARIABLES Full model
35 Ind16 -1.61** (0.851) Ind17 -1.688*** (0.548) Ind18 -3.077*** (0.833) MC 2.514*** (0.375) PB -0.915*** (0.321) PT 1.147*** (0.352) TM -0.017 (0.011) EM -0.029*** (0.009) SizePT MCPT PBPT EMPT Constant 11.401*** (0.668) Observations 2,887 R-squared 0.14 F-statistic 22.93 Prob > F 0.000
Root mean squared error 6.173 Robust standard errors in parentheses
36 Appendix C
Table C.1: Regression output EBIT multiple
This table presents the regression output of the sample that includes an EBIT transaction multiple as dependent variable. Only significant parameters and interaction variables are presented.
37 Observations
R-squared 0.12
F-statistic 24.14
Prob > F 0.000
Root mean squared error 11.61 Robust standard errors in parentheses
38 Appendix D
Table D.1: Regression output EBITDA transaction multiple 5% outliers
This table presents the regression output of the sample that includes an EBITDA transaction multiple as dependent variable. The sample is trimmed at 5% at each tail. Only significant parameters and interaction variables are presented.
VARIABLES Full model
39 PT 3.04** (1.394) TM -0.036** (0.0149) EM -0.091*** (0.0136) SizePT -0.001 (0.03) MCPT 2.544 (1.768) PBPT -2.151* (1.214) EMPT 0.011 (0.021) Constant 13.72*** (1.192) Observations 3,249 R-squared 0.15 F-statistic 32.52 Prob > F 0.000
Root mean squared error 9.861 Robust standard errors in parentheses
40 Appendix E
Table E.1: Median regression output EBITDA transaction multiple
This table presents the median regression output of the sample that includes an EBITDA transaction multiple as dependent variable. Only significant parameters and interaction variables are presented. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
VARIABLES Full model
