Comparison of selection methods using different peer pools

Page 68 of 177

within industries, in general yields the greatest accuracy across all INDSARD- and SARD combinations, however, the results are not significant against SARD4 and SARD5 for neither EV/EBITDA nor EV/EBIT indicating an ambiguous pattern of whether SARD within or across industries is most optimal.

Page 69 of 177

There are however modifications to the EU peer pool’s superiority, as SARD1 and SARD2 using an EU dataset do not yield greater accuracy than any SARD-combinations using a pure Danish peer pool. SARD1 and SARD2 using an EU peer pool lead to higher prediction errors when compared directly to SARD1 and SARD2 using a Danish peer pool. Nonetheless, those results are not consistently significant across all multiples. However, comparing SARD1 and SARD2 on the EU peer pool to broader variations of SARD using a Danish peer pool, the lower prediction accuracy with EU is statistically significant. This means, that if ROE (SARD1) or ROE and ND/EBIT (SARD2) are applied as the only selection variables when using a cross-country peer pool, the accuracy of multiple predictions is significantly lower compared to simply using peers within Denmark’s borders and instead applying more selection variables including Size (SARD3), Growth (SARD4) and EBIT-margin (SARD5). This counts for all three multiples applied.

Page 70 of 177

A further modification to the EU peer pool’s superiority is seen for EV/Sales as the first four combinations of SARD (SARD1-4) using an extended peer pool in EU leads to lower prediction accuracy compared to SARD5 solely using a Danish peer pool. The results for EV/Sales are significant at a 1%-level for both statistical tests, indicating that applying the EBIT margin as a selection variable in the full combination (SARD5) is of greater importance for the prediction accuracy compared to extending the peer pool for Denmark to EU in any combination up to and including SARD4. Ultimately, SARD5 using an EU dataset is, however, more accurate in predicting multiples for Danish targets than SARD5 using a pure Danish dataset. Hence, once the EBIT margin is applied in SARD5 using a larger, cross-border peer pool in EU there is sufficient statistical evidence to conclude that it is better than using the smaller peer pool focused on the home-country solely.

To summarize, when comparing different peer pools under the SARD approach, the overall picture indicates that finding peers for Danish target firms is more accurate when using a larger peer pool consisting of cross-country firms than using a smaller, home-country peer pool.

However, some modifications to the superiority of an EU peer pool appear, as when only applying a limited number of selection variables (SARD1-2), predictions are in fact more accurate using a Danish peer pool with more selection variables (SARD3-5). Furthermore, a modification specifically related to EV/Sales occurs as SARD5 using a Danish peer pool yields greater accuracy than SARD1-4 with an EU peer pool, however, SARD5 using an EU peer pool ultimately outperforms all combinations.

5.4.2 SARD

within

Ultimately, the accuracy of the two peer pools is examined by comparing the findings from Section 5.2.2 and 5.3.2 using INDSARD. Table 5.9 shows the Wilcoxon signed-rank test for the median of pairwise differences between absolute percentage errors reached by a peer pool of solely Danish firms relative to a peer pool of all EU firms. The paired t-tests can be found in Appendix 9, Table 9.2, and yield similar results with the exception of some significance levels as well as minor differences for EV/EBITDA using INDSARD1 and INDSARD2. However, the Wilcoxon examining the median is preferred to be reviewed due to the large differentiation in dataset sizes between the Danish and EU peer pools, as observed in the descriptive statistics in Section 5.1.

Page 71 of 177

For SARD used within industries, applying the fundamentals ROE and ND/EBIT as the only selection variables (INDSARD1 and INDSARD2), the prediction accuracy is lower when a larger EU peer pool is used rather than a smaller, home-country peer pool. Hence, when the number of proxies is too few in INDSARD, a smaller peer pool without cross-country differences is preferred to a larger peer pool looking across borders. These results are significant for EV/Sales, while it only counts for INDSARD1 for EV/EBIT. For EV/EBITDA, these findings are not verified statistically.

Overall, the empirical results demonstrate that the greatest accuracy is achieved when using INDSARD3 with a large, cross-border peer pool consisting of all firms in EU for both EV/EBIT and EV/EBITDA as seen in the relative ranking of all combinations in Appendix 9, Table 9.1. There is statistical evidence of such results on a 1%-level against all combinations of INDSARD using a pure Danish peer pool as seen in Table 5.9. For EV/Sales, however, INDSARD5 applied on a home-country peer pool is preferred when predicting multiple valuations relative to all other combinations of INDSARD using an EU peer pool. This implies

Page 72 of 177

that the underlying value drivers of EV/Sales multiples are captured better using a pure Danish peer pool for Danish targets rather than a larger EU peer pool.

Overall, the INDSARD selection method which applies SARD within a 2-digit industry level shows rather ambiguous results in regard to the optimal peer pool used for predicting multiples for Danish target firms. In general, whether each combination of INDSARD should be applied using a cross-border or a home-country peer pool variable, e.g. INDSARD1 (DK) vs INDSARD1 (EU), varies for each of the five INDSARD combinations and across the three multiples. These ambiguous results could presumably be linked to a trade-off between applying a larger peer pool and cross-border differences, which will be discussed further in Section 6.2.3.

However, it can ultimately be concluded that INDSARD3 containing ROE, ND/EBIT, and Size, on an EU peer pool is the most accurate compared to all other combinations on either Denmark or EU for EV/EBIT and EV/EBITDA, while for EV/Sales INDSARD5 using a Danish peer pool brings the most accurate predictions.