Industry's earnings forecasts and market efficiency.

摘要

The behavior of dividends yields and payout ratios over the last decades shifted the focus of stock price's evaluation towards earnings. The first goal of my dissertation is to forecast this variable at the industry's portfolio level. This level of analysis is important as industry level variables are a frequent benchmark against which individual firms can be compared. The second goal of this thesis is to analyze whether market prices seem to include all relevant information regarding earnings' forecasts.; In the first chapter, I use historical price, accounting and macroeconomic data to construct alternative forecasts. Using the random walk model as the benchmark, I construct alternative forecasts that significantly increase forecast accuracy in a simulated out-of-sample setting. The most successful alternative combines the forecasts for all industries taking into consideration the “economic” distance between industries.; In the second chapter, I expand the information set used to construct the forecasts in the first chapter. In particular, I add to this set the (forward looking) financial analysts' forecasts. The forecasts constructed using this additional information proved to be significantly more accurate than those obtained in the first chapter. This confirms the idea that financial analysts' release relevant information. However, analysts' forecasts of earnings may be improved upon using information from other industries.; Finally, in the third chapter, I use the difference in forecast's accuracy to test for market efficiency. First, I anticipate forecast revisions using the more accurate forecasts developed before. Then, I devise a buy-and-hold strategy that tries to explore the price impact of such revisions. Although I am able to predict strong changes in earnings, I cannot explore their price impact. This result seems to suggest that markets use efficiently the information related to future earnings.