Most experienced investors feel that during the past 25 years there have been more “torpedo” stocks—high-expectation stocks that have reported disappointing earnings—than there have been low-expectation stocks that have provided pleasant earnings surprises. Here, the tabulations over 25 years show that, on average, 5.0 percent of the companies fell into the extremely disappointing earnings cell in the lower left-hand corner of the grid. Conversely, only 2.1 percent of the companies fell into the extremely pleasant surprise cell in the upper right-hand corner. Have analysts been able to forecast next year’s changes in earnings per share accurately? One way to answer this question is to use contingency analysis to compare the actual frequencies with the frequencies that we would expect if there were no relationship between forecasted and actual percent earnings changes. The other is to use correlation and regression analysis to compare the forecasted and actual changes.
Yes, there has been a statistically significant relationship between beginning-of-year forecasted and year-later actual earnings changes. Statistically speaking, actually analysts can, indeed, correctly forecast next year’s earnings changes. To determine whether analysts can, on average, correctly forecast earnings changes 12 months into the future, you can compare the average 12-month-ahead earnings estimates in the Institutional Brokers Estimate Service (I/B/E/S) database with the actual percent earnings changes that occurred 12 months later. Even though far from the 20 percent, for example, of the forecasts that would cluster along the diagonal with perfect forecasts, this tendency for more than 4 percent of the forecasted-actual classifications to cluster along the diagonal is indicative that analysts have some degree of ability to forecast year-ahead earnings changes accurately.