Choice of comparable firms for multiple valuation. A paper by Jens Overgaard Knudsen, Simon Vesterby Kold and Thomas Plenborg

Size: px

Start display at page:

Download "Choice of comparable firms for multiple valuation. A paper by Jens Overgaard Knudsen, Simon Vesterby Kold and Thomas Plenborg"

Victor Wilkerson
5 years ago
Views:

1 Choice of comparable firms for multiple valuation A paper by Jens Overgaard Knudsen, Simon Vesterby Kold and Thomas Plenborg 1

2 Agenda 1 Comparable firm selection for multiple valuation 2 Our idea 3 How did we test it and how does it work? 4 Questions 2

3 Derivation of multiples from the present value relation For simplicity, consider a firm which has a constant growth rate and discount factor forever. EV t MVE r WACC FCFF g D r e g D FCFF P/E P/B EV/EBIT EV/EBITDA g r g r ROIC g r e ROIC g WACC WACC D g r g e D g g D 1 D FCFF FCFF FCFF FCFF 1 ROIC 1 ROIC (1 TR) (1 TR) (1 R% Ideal peer-group companies for multiple valuation purposes should be truly identical on the basis of these properties - But how can we identify such firms? DEPR ) EV/SALES ROIC g r WACC g FCFF FCFF 1 ROIC (1 TR) (1 R% DEPR ) EBITDAmargin 3

4 Drawbacks Rationale for the approach Two obvious approaches for comparable firm selection: Industry affiliation and fundamentals Industry Affiliation Fundamentals Companies within the same industry are subject to similar market characteristics Companies within the same industry tend to produce similar products, leading to similarities of their supply and demand curves Easy to use and widely applied in practice Current fundamentals can be used to approximate future profitability, growth and risk Independent of subjectively defined industry classes Can incorporate, in principle, an infinite amount of peers Used for comparable firm selection in a range of empirical studies, including combinations with industry Theoretical difficulties in defining an industry Industry classes are subjectively defined and subjectively chosen Companies within the same industry class can be very different from each other Which fundamentals are best at projecting future levels of profitability, growth and risk? Gathering fundamental data for a lot of companies can be a cumbersome process 4

5 Agenda 1 Comparable firm selection for multiple valuation 2 Our idea 3 How did we test it and how does it work? 4 Questions 5

The Idea: Sum of Rank Differences (SARD) SARD proposition Our method selects comparable firms based on the least sum of absolute rank differences across a range of variables which are expected to

6 The Idea: Sum of Rank Differences (SARD) SARD proposition Our method selects comparable firms based on the least sum of absolute rank differences across a range of variables which are expected to affect multiples SARD i,j = r X,i r X,j + r Y,i r Y,j + + r Z,i r Z,j where SARD is the sum of absolute rank difference between firm i and firm j, r x,i is the rank of firm i in terms of variable x, r x,j is the rank of firm j in terms of variable x and so forth. 6

7 A 5-step implementation guide to the SARD approach 1 Select base sample 2 Estimate relevant value drivers 3 Convert into ranks 4 Calculate rank difference ( Penalty points ) 5 Find the firms with the lowest SARD 7

8 A 5-step implementation guide to the SARD approach 1 Select base sample Construct a sample from which to select peergroup from. Could be: - A large sample (like our empirical study where we use S&P 1500) - A qualitative selected sample (e.g. subjective perception of industry affiliation) 2 Estimate relevant value drivers Select and calculate relevant value drivers. Our tests include: - Return on equity (profitability) - Net debt/ebit (risk) - Size (risk and different levels of multiples) - Forecasted EPS growth t+2 (growth) - EBIT-margin (relevant for EV/sales multiples) 8

A 5-step implementation guide to the SARD approach 3 Convert into ranks Transform all value drivers into ranks Enables you to combine value drivers regardless of scale Possibility of assigning

9 A 5-step implementation guide to the SARD approach 3 Convert into ranks Transform all value drivers into ranks Enables you to combine value drivers regardless of scale Possibility of assigning weights to value drivers Size (bn. $) r Abbott Labs 10% Coca-Cola 22% Estee Lauder 34% r size General Mills 24% Johnson & Johnson 23% Kellogg Co 20% Mondelez 7% Pepsico 31% Procter & Gamble 17% Reynolds America 30% Implementation tip 9

differences in the table is based on Abbott Labs Coca-Cola Estee Lauder Abbott Labs n/a 9-6 + 6-3 =6 9-1 + 6-9 =11 Coca-Cola

10 A 5-step implementation guide to the SARD approach 4 Calculate rank difference ( Penalty points ) Calculate the difference in ranks between each firm in the base sample and the firm subject to the valuation Repeat for each value driver The rank differences in the table is based on Abbott Labs Coca-Cola Estee Lauder Abbott Labs n/a = =11 Coca-Cola n/a =11 Estee Lauder General Mills Johnson & Johnson Kellogg Co Mondelez Pepsico Procter & Gamble Reynolds America n/a General Mills

11 A 5-step implementation guide to the SARD approach 5 Find the firms with the lowest SARD Calculate the sum of absolute rank differences across the value drivers Identify the firms with the least sum of absolute rank differences Decide on a cut-off (i.e. number of firms to include as peers). In the example below we use six as cut-off (i.e. select six peers) Pe er 1 Abbott Labs Mondelez (2) Coca-Cola Procter & Gamble (3) Estee Lauder General Mills (4) General Mills Reynolds America (2) Johnson & Johnson Coca-Cola (3) Kellogg Co General Mills (5) Mondelez Abbott Labs (2) Pepsico Reynolds America (4) Procter & Gamble Coca- Cola (3) Reynolds America General Mills (2) Procter & Gamble (5) Kellogg Co (6) Coca-Cola (6) General Mills (7) Reynolds America (7) Johnson & Johnson (3) Pepsico (5) Abbott Labs (6) Mondelez (6) Reynolds America (7) Reynolds America (4) Pepsico (6) Kellogg Co (7) Abbott Labs (11) Coca- Cola (11) Estee Lauder (4) Kellogg Co (5) Pepsico (6) Abbott Labs (7) Coca- Cola (7) Procter & Gamble (4) Pepsico (6) General Mills (8) Reynolds America (8) Abbott Labs (9) Abbott Labs (6) Reynolds America (7) Estee Lauder (7) Mondelez (8) Coca- Cola (8) Procter & Gamble (5) Coca-Cola (6) Kellogg Co (8) Reynolds America (9) Pepsico (9) Coca-Cola (5) General Mills (6) Estee Lauder (6) Johnson & Johnson (6) Procter & Gamble (8) Johnson & Johnson (4) Mondelez (5) Abbott Labs (5) Pepsico (8) Kellogg Co (9) Pepsico (4) Estee Lauder (4) Coca-Cola (7) Abbott Labs (7) Kellogg Co (7)

12 Agenda 1 Comparable firm selection for multiple valuation 2 Our idea 3 How did we test it and how does it work? 4 Questions 12

13 Random example from S&P 1500: Six peers selected for IFF on the basis of SARD Firm being valued Industry EV/EBIT Size International Flavors & Fragrances (IFF) Materials % 2.7 3,356 Peers Industry EV/EBIT Size KB Home Consumer Durables & Apparel % 2.3 3,175 Beckman Coulter Inc Health Care Equip. & Services % 2.0 3,354 Pioneer Natural Resources Co Energy % 3.6 3,878 McCormick & Co Inc Food, Beverage & Tobacco % 2.0 4,593 TCF Financial Corp Banks % 4.5 3,601 Knight-Ridder Inc. Media % 2.6 5,777 Actual EV/EBIT of IFF Prediction of the EV/EBIT multiple (harmonic mean of peers) Valuation error ( )/ % 13

14 A distinct pattern: Combining fundamentals seems to work Absolute percentage errors and ranks (brackets) of valuations based on each selection method EV/EBIT Industry Size Size Growth Size Growth EBITmargin Median (4) (6) (5) (3) (2) (1) Mean (3) (6) (5) (4) (2) (1) IQ (3) (6) (5) (4) (2) (1) EV/Sales Median (2) (6) (5) (4) (3) (1) Mean (2) (6) (5) (4) (3) (1) IQ (2) (6) (5) (3) (4) (1) P/B Median (6) (5) (4) (3) (2) (1) Mean (6) (5) (4) (3) (2) (1) IQ (6) (5) (4) (3) (1) (2) P/E Median (5) (6) (4) (3) (2) (1) Mean (6) (5) (4) (3) (2) (1) IQ (6) (5) (4) (3) (1) (2) 14

15 The pattern remains intra-industry Absolute percentage errors and ranks (brackets) of valuations based on each selection method Industry (same industry) (s.industry) Size (s.industry) Size Growth (s.industry) Size Growth EBIT-% (s.industry) Median (6) (5) (4) (3) (2) (1) EV/EBIT Mean (6) (5) (4) (3) (2) (1) IQ (6) (5) (4) (3) (2) (1) Median (6) (5) (4) (3) (2) (1) EV/Sales Mean (6) (5) (4) (3) (2) (1) IQ (6) (5) (4) (3) (2) (1) P/B Median (6) (4) (5) (3) (1) (2) Mean (6) (4) (5) (3) (2) (1) IQ (6) (4) (5) (3) (2) (1) P/E Median (6) (5) (4) (3) (2) (1) Mean (6) (5) (4) (3) (2) (1) IQ (6) (3) (4) (5) (2) (1) 15

16 Robustness checks Robustness checks performed in this study Number of firms in peer group Across time Across size 16

Conclusions Current environment Most analysts and investors tend to use industry classification as proxy for perfect substitutes However, firms within the same industry do not necessarily have the

Our solution The SARD offers a promising alternative in that: The SARD approach is significantly more accurate than the industry approach The SARD approach is able to cater for, in principle, an

17 Conclusions Current environment Most analysts and investors tend to use industry classification as proxy for perfect substitutes However, firms within the same industry do not necessarily have the same profitability, risk or growth characteristics and they should therefore not be traded at the same multiple. Our solution The SARD offers a promising alternative in that: The SARD approach is significantly more accurate than the industry approach The SARD approach is able to cater for, in principle, an infinite number of proxies for profitability, growth and risk The SARD approach is less sensitive to sample size than the industry approach The SARD approach is able to tailor the selection variables to fit the need of any desired multiple Less accurate valuation estimates More accurate valuation estimates> 17

18 Agenda 1 Comparable firm selection for multiple valuation 2 Our idea 3 How did we test it and how does it work? 4 Questions 18

19 Number of firms in peer group The ranking of selection methods remain stable across various numbers of peers Incrementally increasing accuracy when peers are added Similar results for all evaluated multiples EV/Sales 19

20 Across time The ranking of selection methods remain stable over the sample time period Increase in valuation errors around the dot-com-bubble and the financial crisis Similar results for all evaluated multiples P/B 20

21 Across size The ranking of selection methods remain stable over the sample time period Seems to be less estimation error in S&P 500 compared to the other two indices Similar results for all evaluated multiples P/B: Industry Size Size Growth Size Growth EBIT-margin S&P (6) (4) (5) (3) (2) (1) S&P (6) (5) (4) (2) (3) (1) S&P (6) (5) (4) (3) (2) (1) 21

Prior research has shown that the most widely used technique

Prior research has shown that the most widely used technique Research Stick to the Fundamentals and Discover Your Peers Jens Overgaard Knudsen, Simon Kold, and Thomas Plenborg Jens Overgaard Knudsen is an associate at Carnegie Investment Bank, Copenhagen, Denmark.