This article challenges conventional methods used in financial valuation across transactional and litigation domains. We show that conventional valuation methods allow for considerable discretion, making it possible for each side’s experts to submit dramatically varying valuations simply by choosing among facially reasonable values of parameters that must be selected to carry out conventional valuations. We use large-scale empirical simulations powered by real-world data to demonstrate the scope of such discretion. We next consider several alternatives based on data-driven machine learning approaches, and show that they offer both approximately unbiased estimates of valuation and substantially reduced variability in valuation results. Consequently, they reduce the scope of expert or party discretion in valuation. We end by applying this approach to a well-known Delaware valuation dispute.
This paper thus both diagnoses and offers a remedy for the discretion and variability in valuation disputes. Although we focus our principal analysis on the comparable companies approach, many of the insights we develop here lend themselves to other valuation methodologies (such as comparable transactions and discounted cash flow analysis). If adopted, our methods would lead to better performing and more empirically grounded outcomes in legal disputes involving valuation, thus enhancing the fairness and efficiency of the judicial processes in valuation litigation.