[Paper Review] A Bayesian Framework for Combining Valuation Estimates

Kenton K. Yee|ArXiv.org|Jul 24, 2007

Forecasting Techniques and Applications35 references3 citations

TL;DR

This paper proposes a Bayesian framework that combines multiple equity valuation estimates—such as discounted cash flow, comparable companies, and market price—into a single, more accurate point estimate. By treating each estimate as a probability distribution and updating beliefs using Bayes' theorem, the method provides a statistically principled way to average conflicting valuations, significantly improving estimation accuracy over individual methods.

ABSTRACT

Obtaining more accurate equity value estimates is the starting point for stock selection, value-based indexing in a noisy market, and beating benchmark indices through tactical style rotation. Unfortunately, discounted cash flow, method of comparables, and fundamental analysis typically yield discrepant valuation estimates. Moreover, the valuation estimates typically disagree with market price. Can one form a superior valuation estimate by averaging over the individual estimates, including market price? This article suggests a Bayesian framework for combining two or more estimates into a superior valuation estimate. The framework justifies the common practice of averaging over several estimates to arrive at a final point estimate.

Motivation & Objective

• To address the persistent problem of discrepant valuation estimates from different methods such as DCF, comparables, and fundamental analysis.
• To develop a statistically rigorous method for combining multiple valuation estimates into a superior, consolidated point estimate.
• To justify and formalize the common financial practice of averaging multiple valuations using Bayesian probability theory.
• To improve stock selection, value-based indexing, and tactical asset allocation by reducing estimation error.

Proposed method

• The framework models each valuation estimate as a random variable with a known or assumed probability distribution.
• It applies Bayes' theorem to update prior beliefs about intrinsic value using evidence from multiple independent valuation methods.
• The posterior distribution of intrinsic value is derived as a weighted average of the individual estimates, with weights proportional to the inverse of their estimated variances.
• The method incorporates market price as one of the input estimates, treating it as a noisy signal of true value.
• It assumes conditional independence of the estimates given the true value, enabling tractable analytical solutions.
• The final point estimate is the mean of the posterior distribution, representing the optimal combination under squared error loss.

Experimental results

Research questions

• RQ1 Can a statistically coherent method be developed to combine multiple, conflicting equity valuation estimates into a single superior estimate?
• RQ2 How should the relative reliability of different valuation methods be quantified and integrated in a unified framework?
• RQ3 To what extent does combining estimates using Bayesian updating reduce estimation error compared to using individual methods?
• RQ4 Does including market price as an input estimate improve the accuracy of the final valuation?
• RQ5 Is the common practice of averaging multiple valuations empirically and theoretically justified?

Key findings

• The Bayesian framework provides a mathematically sound justification for averaging multiple valuation estimates, resolving the inconsistency of ad hoc averaging.
• The optimal combined estimate is a weighted average of the inputs, with weights inversely proportional to their estimated variances, minimizing mean squared error.
• Incorporating market price as one of the estimates improves the final valuation by leveraging market efficiency while correcting for its noise.
• The method reduces estimation error more effectively than relying on any single valuation technique alone.
• The framework is generalizable and can be applied to any set of independent, noisy valuation signals, including fundamental, relative, and market-based methods.