What Matters from the Past: Market Value, Book Value, or Earnings? Earnings Valuation and Sufficient Statistics for Prior Information

Abstract

A number of recent analytical and empirical papers seek to identify the variables that best explain stock prices. We derive the relation between prices and earnings for three one-parameter “excess earnings” evolution processes that describe three different ways in which current period shocks to earnings persist in future. In each case, we show that price is a weighted average of capitalized current earnings and a sufficient statistic for all past information. The sufficient statistics are three amounts from last year: book value, market value, and capitalized earnings. Using time-series data over the 1969-1988 period for a sample of 511 firms, we estimate firm-specific excess earnings regressions and price regressions for the three cases. Although the book value model provides the best fit for a majority of firms for the excess earnings regressions, the market value model is far superior for the price regressions. We argue that this latter result is due to prior period price (included in the market value model alone), reflecting other information that is not explicitly modeled here. Despite this bias in favor of the market value model for the price regressions, we find a positive cross-sectional association between the relative explanatory power of the three models in the excess earnings regressions and the corresponding relative explanatory power in the price regressions. That is, if a firm's excess earnings series is best described by a particular model, its price series is also likely to be best described by the valuation relation derived from the same model.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 41 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.