Survey of Recent Research on Accounting Anomalies

What is the state, from an investor’s perspective, of research on the power of accounting and fundamentals to predict stock returns? In their September 2010 paper entitled “Accounting Anomalies and Fundamental Analysis: A Review of Recent Research Advances”, Scott Richardson, Irem Tuna and Peter Wysocki present an overview of post-2000 research on accounting anomalies and fundamental analysis geared toward forecasting future earnings and stock returns. They include results from matched 2009 surveys of 201 investment practitioners and 63 accounting academics on relevant beliefs about this research. They also present a new analysis of how expected risk and expected transaction costs affect the accrual and post-earnings announcement drift (PEAD) anomalies. Using for this new analysis accounting data (lagged three months) and stock returns for 1,000 relatively liquid U.S. stocks over the period 1979 through 2008, they find that:

• Citation analysis of post-2000 research on accounting anomalies and fundamental analysis suggests four main streams:
  1. Fundamental Analysis
  2. Accrual Anomaly
  3. Underreaction to Accounting Information
  4. Pricing Multiples and the Value Anomaly.
• Academic research generally falls short of practitioner requirements by failing to: (1) estimate risk prospectively rather than calculate it retrospectively; (2) apply industry affiliation as a risk factor; and, (3) include realistic prospective trading frictions to forecast net outcomes.
• There are large differences between practitioners and academics on the success of various strategies over the past decade, perhaps due to disparities in approaches to measuring net returns and risks:
  • 61% (22%) of practitioners (academics) believe that earnings/cash flow momentum worked.
  • 57% (22%) of practitioners (academics) believe that growth worked.
  • 56% (52%) of practitioners (academics) believe that value worked.
  • 41% (74%) of practitioners (academics) believe that accounting quality worked.
• There are large differences between practitioners and academics on valuation methods, perhaps due to the difficulty of assembling quality accounting data:
  • 16% (71%) of practitioners (academics) use residual (discounted) income.
  • 74% (54%) of practitioners (academics) use earnings multiples.
  • 52% (38%) of practitioners (academics) use book value multiples.
• Complex backtests that incorporate rolling forecasts of risk (via mean-variance analysis) and forecasts of trading frictions into the process of monthly long-short portfolio formation from a universe of the 1,000 U.S. stocks with the largest market capitalizations over the period 1979 through 2008 indicate that:
  • The negative relationship between accruals (measured as lagged change in net operating assets) and future stock returns is robust, but has diminished greatly in recent years.
  • The relationship between PEAD (measured based on standardized unexpected earnings, calculated as changes in seasonally adjusted earnings as a percentage of average assets divided by the rolling 12-quarter past standard deviation of these changes) and future stock returns is only marginally significant, and is greatly diminished in recent years.
  • Results are consistent with an adaptively efficient market.

In summary, evidence from recent research on accounting anomalies and fundamental analysis suggests that markets adapt over time and what was once mispriced becomes correctly priced, reflective of an “inefficiency bias” in older data. But while forecasting returns is increasingly competitive, the rewards are potentially still substantial.

Note that the complex backtest methodology in the new analysis of accruals and PEAD is likely out of reach for all but the most sophisticated institutional investors.

The paper includes considerable discussion of including and estimating trading frictions to prove exploitability of anomalies. Some amplifications are:

• In long-term backtesting, it is very difficult to model the large empirical variations in broker fees and bid-ask spreads of the U.S. stock market (see “Trading Frictions Over the Long Run”). Determination of exploitability of past trades should apply then-prevailing trading frictions, not current trading frictions.
• Relatively illiquid stocks may drive the gross profitability of anomalies, but these stocks tend to have comparatively high trading frictions. Modeling cross-sections of trading frictions is complex.
• Trading frictions (including the cost of shorting) may spike during brief intervals of high interest to trading strategies (for example, around the time of earnings release or other firm news events) due to temporary stock supply/demand mismatches.
• Trading frictions vary by type of investor according to, for example, position sizes.