New Insights into Equity Valuation Using Multiples PhD Thesis submitted to the Faculty of Economics and Business Information Management Institute University of Neuchâtel For the degree of PhD in Statistics by Jian KANG Accepted by the dissertation committee: Prof Catalin Starica, University of Neuchatel, thesis director Prof Kilian Stoffel, University of Neuchatel Prof Alain Schatt, University of Lausanne Prof Lucie Courteau, University of Bozen-Bolzano, Italy Prof Jacob Thomas, Yale University, USA Defended on 28th, Jan, 2016 Faculté des Sciences Economiques Avenue du 1er-Mars 26 CH-2000 Neuchâtel www.unine.ch/seco New insights into equity valuation using multiples JIAN KANG Abstract The thesis focuses on the equity valuation using multiples. Based on the notion of stochastic dominance, a relative valuation framework for comparing accuracy is devel- oped and employed in the thesis. In the first paper, we investigate the feasibility of a relative valuation framework for tracking developments of value relevance of earnings and book values across time. The second paper focuses on the performance of EV/EBITDA, a multiple that gained in popularity with practitioners during the last decade, relative to that of the traditional multiple P/E. The last paper explores the possibility of applying non-linear methods to improve the precision of multiple valuation. Keywords: Equity valuation, multiples, accuracy, stochastic dominance, non-linear, P/E, P/B, EV/EBITDA Acknowledgment The experience of my PhD study at the University of Neuchatel is one of the largest assets in my life. First and foremost, I am grateful to my supervisor Catalin Starica for his guidance, supervision and support for last 66 months. His enlightening comments and advises guided me to discover and solve interesting questions, not only in academic, but also in life. It is my great honor and fortune to have met Catalin. The thesis can never be completed without his help and contributions, and I can never stand where I am today without his lead. I would like to thank my co-author Alain Schatt, who inspired me how to compose pieces of evidence from a perspective of accounting. The paper we worked on together started my journey in the field of accounting. Thanks to Kilian Stoffel for having accepted to be the president of my dissertation committee, Lucie Courteau and Jacob Thomas for have accepted to review and comment this thesis. I also thank the University of Neuchatel for providing me a terrific research environ- ment. I would like to express my gratitude to my colleagues in the university. Finally, I want to thank my wife, Li Gao, and my parents for their continuous support and selfless dedication. Contents 1 Summary of papers1 1.1 Paper I.....................................1 1.2 Paper II....................................2 1.3 Paper III....................................2 2 Included Papers5 2.1 Paper I: Have earnings and book values lost relevance? A relative valua- tion framework................................6 2.2 Paper II: Is P/E always more accurate than EV/EBITDA?........ 54 2.3 Paper III: Price-to-earnings valuation - imperfect heuristic? Maybe, but then a precise one!.............................. 99 vi CONTENTS 1 Summary of papers 1.1 Paper I The aim of this paper is to investigate the feasibility of a relative valuation framework for tracking developments of value relevance of earnings and book values across time. In such a framework, the measure of relevance of earnings or book values is the precision of the equity valuation using multiple P/E, respectively P/B. A loss of relevance of the ac- counting variable would be documented by a decrease through time of the corresponding multiple valuation precision. To render such an approach operational, we optimize the implementation of the val- uation by multiples with respect to precision of valuation (characteristics and number of peers, aggregation of peer multiples). Extending the existing literature that compares accuracy using particular statistics of the distribution of valuation errors, we develop a framework for comparing accuracy based on the notion of stochastic dominance. The relative valuation framework offers an alternative to the linear regression approach im- paired by a number of econometric weaknesses when employed in the frame of value relevance studies. Using this framework, we document a reduction of valuation accuracy of both P/E and P/B multiples that affect mostly the large firms after 1990. In the relative valuation framework for value relevance, we interpret this finding as evidence for a significant reduction of the value relevance of earnings and book values for large firms in the last 1 2 1. SUMMARY OF PAPERS twenty five years. 1.2 Paper II This paper focuses on the equity valuation using multiples P/E and EV/EBITDA. We compare the accuracy of firm valuation based on EV/EBITDA, a multiple gained in popularity with practitioners during the last decade, with that of the traditional multiple P/E. The paper is motivated by the evolution in analysts' preferences. Recent surveys denote that the preferences of choosing multiples of practitioners are changed in the United States (Block, 2010) and in Europe (Bancel & Mittoo, 2014), i.e. more analysts tend to use the multiple EV/EBITDA in equity valuation. Our analysis shows that valuations using P/E is more accurate than EV/EBITDA when aggregating over all observations. The result is robust to the implementation of the valuation method (characteristics and number of peers, aggregation of peer multiples), the type of multiple drivers (historical numbers or forecasts), the period (recent years or not), and the industry type (capital intensive industry or not). When different firm characteristics are taken into account, the relation becomes more intricate. For compa- nies with low debt, EV/EBITDA is at least as precise as P/E. Moreover, EV/EBITDA is significantly more accurate than P/E when valuing firms that report largely negative special items and/or non-operating items. This finding is in line with the idea that not all items convey information to the investors. Overall, we conclude that EV/EBITDA leads to more accurate valuation in some specific cases. 1.3 Paper III This paper revisits the price-to-earnings multiple valuation approach. Interpreting the median industry earning ratio as a quantity close to the expected return on equity and heeding the prescriptions of valuation models commonly used in the accounting literature, we motivate the need for a non-linear prediction of the multiple. We propose the use of multiple industry median ratios as predicting variables. 1.3. PAPER III 3 We show that non-linearly forecasting the multiple using three industry median mul- tiple constructed according to different depth definition of industry (one, two or three SIC digits) sharply improves the accuracy of the prediction of the multiple. Since the relative pricing error for multiples that are ratios of prices to value driver is equal to the relative prediction error for the multiple, the non-linear prediction of the multiple greatly enhances pricing precision. We report a reduction of 47% of the median absolute pricing error, of 45% for the 75%-ile, and of 40% for the 90%-ile of the absolute error with respect to plain-vanilla multiple pricing. Our findings reposition the multiple pricing with respect to the more comprehensive valuation approaches. Although not evaluated on the same sample, the performance of the non-linearly enhanced price-to-earnings multiple pricing compares favorable with that of more comprehensive valuation methods as reported in the literature. We argue that, while price-to-earnings valuation might be an 'imperfect heuristic', when strengthened by the understanding of accounting valuation models, it becomes a precise pricing method. 4 1. SUMMARY OF PAPERS 2 Included Papers 5 2.1 Paper I: Have earnings and book values lost rel- evance? A relative valuation framework 2.1. PAPER I 7 Have earnings and book values lost relevance? A relative valuation framework Jian Kang and Catalin Starica School of Economics and Business, University of Neuchˆatel Abstract We inquire into the suitability of a relative valuation framework for track- ing developments of value relevance of earnings and book values across time. In such a framework the measure of relevance of earnings or book values is the precision of P/E, respectively P/B, valuation. A loss of relevance of the accounting variable would be documented by a decrease through time of the corresponding multiple valuation precision. To render such an approach operational we optimize the implementation of the valuation by multiples with respect to precision of valuation. We argue that the concept of stochastic dominance gives the right statistical frame for measuring and comparing the accuracy of equity valuation. We document a reduction of valuation accuracy of both P/E and P/B multiples that affect mostly the large firms after 1990. In the relative valua- tion framework for value relevance, we interpret this finding as evidence for a significant reduction of the value relevance of earnings and book values for large firms in the last twenty five years. Keywords: Valuation, multiples, P/E, P/B, accuracy, stochastic domi- nance, value relevance. 8 2. INCLUDED PAPERS 1 Introduction Temporal changes have been a recurrent issue in the extensive literature on value relevance of accounting information. In a thorough discussion on interpreting value relevance in academic research, Francis and Schipper(1999) identifies four main alternatives. In the frame of studying developments in value relevance, the most common among the four is that of association between financial information (i.e. accounting variables) and share prices or returns (Interpretation 4). This specific interpretation of value relevance allows for different operationalisations based on two main choices. First, the researcher needs to chose a way to formalize the relationship between value and financial information.