EV/EBITDA Regression

Based on these thoughts, we will be estimating the related exit EV/EBITDA multiple with the help of a regression model, containing the 1-year forward EV/EBITDA multiple, as the dependent variable, explained by the 3-year forwardly estimated sales growth (CAGR). A forward-looking variable has been used instead

100 of a historical one, to reflect the fact that a company’s value is the sum of its future cash flows (Berk and DeMarzo, 2014). Based on the model results, we expect to substantiate a discussion towards which exit multiple will be applicable to the exit of A&F. In this case, the chosen time and result will be based on the expected growth rates after a holding period of 6 years, as this was the middle of the estimated holding period. Our sample consists of 32 publicly listed companies of similar characteristics that have been provided with a 1-year forward EBITDA estimate, a current Enterprise Value, as well as a 3-year sales CAGR estimate, collected from Bloomberg. The 3-year time horizon for the CAGR has been selected based on the available data, since sales estimates were not available beyond 2019 a cut had to be made. Moreover, any estimates beyond this period are becoming more variable and thus it is assumed that the three-year period selected is the most relevant period for the 1 year forward EV/EBITDA multiple. We are aware that the sample size used in the regression is of small size, but the limited availability of information has set a direct limitation to the regression’s scope in consideration to the value it creates for our LBO model.

Although this limitation exists, we have still made sure that our sample size exceeds 30, enabling us to assume an approximated normal distribution. The sample size, values and other descriptive statistics used in the linear one-factor regression analysis is in appendix 8.

Moreover, we have been conducting multiple regressions where we included multiple variables such as binary variables controlling for the size of the company and if the company during the last 6 months has experienced a share price drop of more than 50%. Additionally, we have also included a variable controlling for if there are any difference between companies for which the share price has increased compared to its share price one year ago. When controlling for these different characteristics, we found that none of them were significant at a 10% level, and as a result we have chosen not to include these factors and only use the 3 year forward CAGR as the independent variable for our regression.

Using an OLS regression method with the data in appendix 8, we have achieved the linear function, as shown below and in figure 23.

𝐸𝑉/𝐸𝐵𝐼𝑇𝐷𝐴 = 4,8342 + 45,525 ∗ 𝐶𝐴𝐺𝑅3𝑦𝑒𝑎𝑟𝑓𝑜𝑟𝑤𝑎𝑟𝑑

Table 20 - Compiled by the authors with regression results

As can be seen in figure 23, our regression shows a positive relationship between the EV/EBITDA and the 3-year forward growth rate, indicating that an expected higher sales growth leads to a higher valuation multiple. With an exit after 6 years of holding period, in 2023, the expected sales growth for the following

101 3-year period can be calculated to 3,5% as derived in the income statement in table 10. By applying the expected sales growth of 3,5% into the regression, a valuation multiple of 6,43 is obtained for A&F and should according to our regression model be applicable for a potential exit in 2023, as seen in table 20. In order to determine if this value is suitable to use as an estimated level for our exit multiple, a t-test has to be made to evaluate if the p-value of each variable is significant at a 95% significance level or more.

Figure 23 - Compiled by the authors

As can be seen in appendix 8, both the intercept as well as the CAGR variable are both significant at a 95%

and 99% level, with their respective p-values both being 0,0000. Moreover, an evaluation of the adjusted 𝑅², which measures how much of our dependent variable that is explained by our independent variable, has to be made to. As can be seen in appendix 8, the adjusted 𝑅² is suggesting that 55,74% of the variation in the 1 year forward EV/EBITDA multiple is explained by variations in the 3 year forward looking CAGR, which indicates that there are more variables that can have an impact on the EV/EBITDA multiple. Such variables could be the forecasted margin levels of the different companies, which should have an impact on the multiple, but have not been included in the model due to lack of data.

Additionally, we have constructed a 95% confidence interval to see what possible values that our estimated exit multiple could lie within. As can be seen from table 20, our exit multiple should according to our regression model, with a 95% confidence, lie between 5,12 and 7,74. Thus our entry multiple of 5,52x is contained within our confidence interval. We have therefore chosen to assume the same exit multiple as used at entry to ensure that a conservative multiple is applied when determining our estimated

y = 45,5252x + 4,8342 Adj. R² = 0,5574

0 2 4 6 8 10 12 14 16

-5% 0% 5% 10% 15% 20%

EV/EBITDA

CAGR

Exit EV/EBTIDA Multiple

102 EV and return at exit. It could however still be argued that the exit-multiple could be increased, based on the significance of the model’s explanation. However, the effects of a different exit multiples will however be discussed in section 15.1 including our sensitivity analysis. In conjunction to this, our chosen exit multiple will be 5,52x, resulting in an Enterprise Value of $2564,98 million.