Biases and Error Measures: How to Compare Valuation Methods

Abstract

We investigate biases of valuation methods and document that these depend largely on the choice of error measure (percentage vs. logarithmic errors) used to compare valuation procedures. We analyze four multiple valuation methods (averaging with the arithmetic mean, harmonic mean, median, and the geometric mean) and three present value approaches (dividend discount model, discounted cash flow model, residual income model). Percentage errors generate a positive bias for most multiples, and they imply that setting company values equal to their book values dominates many established valuation methods. Logarithmic errors imply that the median and the geometric mean are unbiased while the arithmetic mean is biased upward as much as the harmonic mean is biased downward. The dividend discount model dominates the discounted cash flow model only for percentage errors, while the opposite is true for logarithmic errors. The residual income model is optimal for both error measures.