Integrity Researchwatch has moved to
http://www.integrity-research.com/cms/researchwatch
Please update your bookmarks.

New York – In a working paper from Penn State entitled “The Accuracy of Analysts’ Long-Term Earnings per Share Growth Rate Forecasts” (1) the interesting aspect of the over-optimism of equity analysts is studied. The study asserts that there are three potential sources of bias in the EPS estimates:  1) Analysts bias, created by the economic benefits to the analysts and the firms they work for; 2) Selection bias, which means the analysts cover stocks they expect to perform well and drop coverage of stocks they don’t and; 3) behavioral bias, which implies that analysts become attached to the companies they cover.

While analyst over-optimism has been studied before, the studies were developed using short-term quarter-ahead earnings estimates and actuals. This study follows 1-year and long-term EPS forecasts and actuals. Long-term is defined as estimates which are 3 to 5 years ahead.

Of course, is it not sensible for EPS growth forecasts to be consistently above the growth in nominal GDP. This relates to the use of valuation tools, such as dividend discount models that use infinite time series to discount a growing stream of dividends. If EPS growth is in excess of GDP growth for the infinite time series, eventually the stock will be larger than the economy.  

The study includes an assessment of the Global Analyst Settlement’s impact on earnings growth over-optimism. The assumption is that the effect of Analyst biased should be washed out after the settlement.  According to the working paper, the average EPS of the companies covered was 8.25% in the period from 1988 to 2002. Over the same period, the average forecasted EPS was 14.40%, giving a percentage forecast error of 141.65%. For the period after the settlement, the average EPS was 12.33%. Over that time frame, the average analyst forecasted EPS was 16.77% for a forecast error of 66.94%.

The working paper places the reduction of bias is placed at the feet of the Settlement, and indicates that this demonstrates that selection bias and behavioral bias are strong forces in the decomposition of analyst over-optimism.     

      Our take  

     The paper helps to identify the various components of overly optimistic EPS forecasts by analysts. This is an excellent result and seems to suggest that the Settlement may have worked to reduce overall analyst over-optimism.

      Here are some suggested extensions to the analysis, which could potentially be done with the same data set. First, is there a systematic difference between EPS forecast error, depending on whether the research is conducted by sell-side analysts, buy-side analysts or independent shops? This could be done for periods before and after the Settlement.  Second, is there any differential in biased related to the market cap of the stocks covered?  Finally, is there any systematic differential between the bias as it relates to growth vs. value stocks under analyst coverage?

                 (1)The study is authored by Cusatis and Woolridge.