Relative Valuation

Page 106 of 162

Page 107 of 162 share price in each scenario is fair. The MMC case is trading almost at the harmonic mean. A harmonic mean of 6.8x results in an implied share price of DKK 320, approximately 2% less than the MMC case. The maximum implied value is DKK 589 and the minimum is DKK 215. These are not values that are seen as realistic for Ørsted in the future.

However, when comparing to peers, their current share price as a percentage of its 52-week high should be taken into account (Rosenbaum & Pearl, 2009). This is a widely used metric that provides perspective on valuation and gauges the current market sentiment and outlook for both the individual company and its broader sector (Ibid.). If a given company’s percentage is significantly out of line with that of its peers, it is generally an indicator of company-specific as opposed to sector-specific issues (Ibid.). That is the case for Centrica with a 64% of its 52-week high. If Centrica is removed, the harmonic mean is 7x, resulting in a DKK 333 share price, only 1% more than the MMC case. In general, the 52-week measure is high for all the peers with Ørsted, Fortum, EDPR, and EDF as best performers. The harmonic mean of these three peers is 7.27x, resulting in a share price of DKK 347.

Figure 70 – Ørsted’s implied share price

Source: Authors’ own creation

In conclusion, when comparing to peers, all DCFs are fair with MMC as the closest to the harmonic mean. But according to Koller et al. (2010), this analysis is flawed:

“The most common flaw in relative valuation is to compare a particular company’s multiple with an average multiple regardless of differences in financials. To choose a peer group only use companies with similar growth and ROIC characteristics.” (Koller et al., p. 316).

411

236 215

318 281

589

359 420

229 446

320 355 329

392

Iber.

E.ON

EDPR Centr. EDF ENEL Fortum Engie RWE SSE Har

Mean

DCF DCF

MC

RDCF

Ø 350 Ørsted Implied Share price

Page 108 of 162 Therefore, to challenge the validity of the relative

valuation based on Ørsted’s closest peers, machine learning is used. K-means clustering is one of the data mining techniques used to obtain groups of objects that have common characteristics in large enough data (Raschka & Mirjalil, 2017). Thus, the goal is to group companies similar to Ørsted into clusters based on their

financial ratios. Bloomberg classifies Ørsted as a member of Europe’s top power generators, which consist of 38 companies. From these 38 companies, the ratios WACC, ROIC/WACC, EV/Sales, EV/EBITDA, EV/Invested capital, Sales growth, EBITDA margin, and net debt/EBITDA are calculated. The ratios are then used as variables to define clusters. To define the numbers of clusters, the so-called elbow method is used (Raschka & Mirjalil, 2017). The goal is to define clusters such that the total within-cluster sum of square (WSS) is minimised (Ibid.). The elbow method looks at the total WSS as a function of the number of clusters.

WSS should be minimised to the point where adding another cluster does not improve the total WSS (Ibid.).

At six clusters, adding one more cluster does not improve the WSS substantially.

Now, K-means can predict the closest cluster for each company based on all the financial ratios. Figure 72, shows the six clusters in 3D plots based on some of the financial ratios. The algorithm sorts Ørsted into a cluster of EDPR, Uniper SE, ACEA SpA, Iren SpA (see the red circles). Only EDPR is a part of the original peer group and mainly operating within wind. The other companies in the cluster are multi-utility businesses, similar to Ørsted’s other two divisions. This could indicate that the market is valuing Ørsted as multi-utility company. Therefore, a sum-of-the-parts analysis could be a more appropriate valuation method for Ørsted.

However, it has been clearly stated in the introduction why the valuation is only based on the wind power division.

Figure 71 – Elbow curve

Source: Authors’ own creation with use of Python

Page 109 of 162 Figure 72 – K-means clustering of Ørsted and its peers

Source: Authors’ own creation with use of Python

The previously defined closest peers to Ørsted are mostly in the same cluster with high EV/EBITDA and net debt/EBITDA. Hence, they are not truly comparable to Ørsted from a financial perspective. Therefore, the EV/EBITDA comparison to EDPR is weighted higher, which is in line with the competitive analysis, highlighting EDPR as the closest peer business-wise. The harmonic mean of Ørsted’s cluster EV/EBITDA is 8.3x, totalling a share price of DKK 401 for Ørsted.

If Ørsted was to trade at EDPR’s EV/EBITDA of 8.5x, the share price would be DKK 411. A significant upside from the MMC and base case. However, it is worth paying attention to the fact that the RDCF is trading almost in line with EDPR, implying that the market is valuing Ørsted and EDPR equivalent. The correlation matrix in the analysis of peers also highlighted their close relationship. The analysis of EDPR highlighted that in EDP owning 77.5% of EDPR, EDP tried in the summer of 2017 to buy the remaining 22.5% stake in EDPR with the objective of gaining larger exposure to renewables energies (EDP, 2017; RS, 2017). The transaction EV/EBITDA multiple was at 8.7x EBITDA, which the remaining shareholders in EDPR declined (RS, 2017).

This has added a premium to EDPR’s EV/EBITDA, implying that it could be artificially high compared to its underlying operations.

8.4.1. Comparable Transaction Analysis (CTA)

Transaction multiples reflect actual payments for real-life deals, rather than traded multiples that are subject to supply and demand pressure. They provide guidance to assess what a buyer may be willing to pay for Ørsted.

Page 110 of 162 When acquiring, companies usually do it to gain full control over the company. More often than not, the buying party are willing to pay a larger amount for a controlling stake as it puts them in the “front-seat” of decision making. The extra price paid for the company is called the control premium (Rosenbaum & Pearl, 2009). This makes the multiple higher than the trading multiple. Table 11 shows the comparable transactions to Ørsted.

Table 11 – CTAs

Source: Authors’ own creation from (EDP, 2017; Bloomberg, 2018; RWE, 2018)

The interesting thing about the transactions is that 80% of them are cases where the holding company decide to buy back the renewable subsidiary. This reflects that they want to make the renewable business a part of their core business, perhaps to develop an integrated business model like Ørsted. For example, Iberdrola bought back its renewable subsidiary at a profit in 2011, and now almost exclusively invests in renewables (Iberdrola, 2017). If Ørsted’s Wind Power division was a subsidiary, a potential acquisition would be based on the multiples in this table, which EDP did when trying to acquire EDPR (EDP, 2017). This results in a share price of DKK 472 for Ørsted. As described above, the deal did not go through, as EDPR wanted a higher price for the shares (RS, 2017). However, EDP already owned the controlling stake of 77,5% and consequently the premium for the deal was 9,70%. As the table above is dominated by core businesses acquiring or trying to acquire its renewable businesses, it potentially skews the premiums to a low, consequently resulting in artificially low EV/EBITDA multiples. Even though the multiples collected are higher than those of the peers trading in the market, it could be assumed that if an acquirer targeted Ørsted, the premium would be larger than the above, resulting in an even higher EV/EBITDA multiple.

The harmonic mean of all CTAs is 11.11x, totalling a share price of DKK 548 for Ørsted. In summary, the transaction multiples yield a much higher share price range than seen from the intrinsic valuation and market-based multiples.

8.4.2. Market Regression

Finally, as a last resort, a multiple regression is done across the 38 European top power generators. The advantage of including this valuation approach is that the estimates become more precise as the number of firms increases, thus decreasing the impact of accounting differences (Damodaran, 2012). EV/EBITDA is again chosen as the dependent variable and proxies for profitability, while risk and growth are used as independent variables. After running regressions across different variables, testing their t-statistics and

p-Date Target company Acquirer Type of offer Premium EV/EBITDA

03-08-2011 Iberdrola Renovables Iberdrola Shares 11,80% 12,58

04-08-2011 EDF Energies Nouvelles EDF Shares or Cash 9,20% 13,13

18/11/2015 Enel Green Power Enel Shares 1,90% 10,47

27/03/2017 EDP Renováveis EDP Cash 9,70% 9,65

03-12-2018 Innogy SE E.ON Asset Swap 28,00% 10,5

Page 111 of 162 values, the following two independent variables are used: 1) a function of CAPEX, Depreciation and EBITDA, which is proxy for the reinvestment rate in the business and 2) the one-year sales growth. The variables are tested for multicollinearity, which would distort the statistics. The two inputs’ correlation is -4.2%. The resulting R-squared is 27% (Appendix 19). According to Damodaran (2012) the R-squared in a relative valuation will almost never be higher than 70%, most common are levels of 30-35%. Instead, it is important to use variables that are true drivers of the multiple and accept the wider range of possible forecasts (Ibid.).

EV/EBITDA = 5.8461 + 5.4554 * (CAPEX - Depreciation)/EBITDA * 1Y Revenue growth

The EV/EBITDA for Ørsted, after plugging in the variables, is 7.15x. This is in line with the earlier computed multiples from the peers. From the equation, it can be discussed whether 1Y revenue growth is too volatile a measure. Table 12 shows the equation’s sensitivity to changes in the 1Y revenue growth. The multiple is stable across the different input variables, meaning the reinvestment rate is a driver of the multiple. Therefore, 7.15x is a reliable measure for Ørsted’s multiple. The multiple equals a share price of DKK 341, close to both the DCF base case and MMC case, giving more confidence in the sell recommendation and the DCF.

Table 12 – Market regression sensitivity analysis

Source: Authors’ own creation

-5% -4% -3% -2% -1% 0% 1% 2% 3% 4% 5%

7.15x 7.15x 7.15x 7.15x 7.15x 7.15x 7.16x 7.16x 7.16x 7.16x 7.16x

Page 112 of 162