Thirty years ago, predicting consensus beats and misses was a profitable endeavour. Today, the payoff is much reduced.Getty Images/iStockphoto
If you read many research reports from sell-side analysts, you will have noticed that, no matter how long the report, the valuation section tends to be quite rudimentary. After an in-depth analysis of the company, its product line and market opportunity, the analyst typically forecasts earnings per share for the next fiscal year. An arbitrary multiple is applied to this earnings forecast to derive a target price, which in turn leads to a buy recommendation 90 per cent of the time.
Once this basic research report has been published, subsequent quarterly earnings reports from the company are parsed in great detail. Deviations from the consensus earnings forecast are thought to be highly significant and sometimes result in a dramatic stock price response.
All of this research directed toward earnings forecasts and the financial-media coverage suggests that the effort must add value to the stock selection process. But is this really the case? If a reliable source offered to tell you the next quarter's earnings per share (EPS) two months before the end of the quarter, how much should you pay for this advance information?
An article in the latest issue of the Financial Analysts Journal by professors Feng Gu and Baruch Lev tackles this question with some surprising results.
Using a 30-year database covering 1986 to 2015, the authors created a portfolio of companies that met or beat analysts' consensus earnings estimates and shorted those that missed the consensus estimate. With the benefit of perfect foresight, these investments were made 60 days before the end of the quarter in question and held for 30 days after, which approximates the date of the earnings release.
The return in excess of the market for these 90 day portfolios was smoothed and charted over the entire time frame. As the authors point out: "The most striking feature is the sharply declining curve." The average excess return from investing in companies that met or beat the consensus estimate fell from 6 per cent in 1989-91 to 2 per cent in 2013-15. Thirty years ago, predicting consensus beats and misses was a profitable endeavour. Today, the payoff is much reduced.
To be fair, the 2-per-cent excess return is from a 90 day, or three-month, portfolio, so the annualized reward is not trivial. But, the strategy is based on 100-per-cent predictive accuracy (the researchers knew beforehand which stocks would meet or exceed their target earnings) and does not include transaction costs, so the real-world payoff would be much reduced.
What is the cause of this downward trend in value-added from traditional valuation models based on earnings forecasts? The authors' hypothesis sounds familiar: It is the same explanation for the underperformance of value managers over the past few years. When companies manufactured and distributed physical products, current earnings and tangible assets were a reasonable proxy for corporate value creation. In a world where software and service companies dominate the market, current earnings may be depressed because of high R&D spending, new subscriber acquisition costs or a host of similar expenditures. Investors in knowledge-based companies presumably are willing to look through the short-term negative impacts on reported earnings in the belief that future value is being created by these expenses. As a result, quarterly earnings play a lesser role in establishing stock price targets.
This leaves investors with the problem of determining whether a company is creating and sustaining a competitive niche or simply fighting a defensive action. The authors point to Dell Inc. as an example where a competitive advantage – build to order – was gradually eroded by others in the industry. Unfortunately, the authors' alternative analytical framework involves more due diligence than any individual investor would undertake. It requires a deep understanding of market share trends, customer acquisition costs, customer churn ratios and many other metrics in order to derive "total customer lifetime value." Maybe they hope that readers will be prompted to invest $60 to buy their recent book, The End of Accounting and the Path Forward for Investors and Managers, to explore this in more detail, but I resisted that temptation.
They do suggest, however, a simple ratio that can provide useful insight for an investor comparing companies within a single industry: Just add back R&D spending to a company's cash flow before calculating an adjusted price-to-cash-flow ratio. This will normalize the ratio across the industry and highlight stocks that otherwise appear to be cheap on a simple price-to-cash-flow basis because they are underinvesting in R&D at the expense of future profitability compared with those that appear to be expensive because they are spending heavily on the future. In this situation, you might prefer the company with the higher ratio.