Predictive Accuracy of Relative Valuation Models

According to many researchers, relative valuation models are becoming increasingly widespread in practice every year, even if their predictive power is not fully studied. However, current results are highly contradictory regarding the effectiveness of each multiple and the accuracy of the comparative approach in identifying undervalued and overvalued firms. The purpose of this work is to test whether relative stock valuation models have predictive accuracy in identifying undervalued and overvalued stocks in the Russian and Chinese stock markets, verify these models on historical data, and explore the possibility of using these models for effective investment strategies. The most commonly used multiples in practice are analyzed - P/E, P/B, EV/S and EV/EBITDA. The data was collected from the Digital Financial Analyst application database (www.difan.xyz) for the period from 2002 to 2021 in one-month increments. The methodological tools are based on regression analysis of estimated coefficients, empirical testing of the CAPM model, construction of portfolios of various types with overvalued and undervalued stocks based on sorting by multiples. Based on the results of the study, it was revealed that valuation models based on multiples have predictive power and help to find undervalued companies on the stock market. Using relative valuation models, a market player will be able to receive abnormal returns over time. A close connection between the multiple and its fundamental variables was revealed, which confirms the necessity and effectiveness of using a fundamentally adjusted indicator. For the Russian market, it is recommended to use fundamentally adjusted P/E, P/B and EV/EBITDA multiples and a raw EV/S ratio. For Hang Seng Index companies - use of adjusted P/E, EV/EBITDA, EV/S and a simple P/B indicator.