An investigation of asymmetric earnings forecasts of Japanese financial analysts

Multinational Business Review, Spring 2003 by Mande, Vivek, Wohar, Mark E, Ortman, Richard F

ABSTRACT: A number of U.S. studies have documented an optimistic bias in analysts' forecasts of earnings. This study investigates whether the optimistic bias and asymmetric behavior of forecast errors found in most U.S. studies exists in Japan. We find that for firms reporting profits, Japanese analysts' forecasts have much greater accuracy and exhibit a small pessimistic bias in comparison to firms reporting losses, where analysts' forecasts exhibit extremely poor accuracy and an extremely significant optimistic bias. The lack of ability to forecast losses is due to their transitory nature and not due to earnings management. Forecast accuracy and bias are not related to firm size, but are related to the magnitude of reported losses and profits.

INTRODUCTION

Forecasts of corporate profits influence the price of corporate stocks. When a firm announces that earnings will not be as large as the forecasted value, the firm's stock price immediately falls. In a similar way, when a firm earns higher profits than that forecasted, the company's stock price increases. Numerous studies in accounting and finance have found results suggesting that earnings forecasts made by U.S. analysts display an optimistic bias (i.e., forecasted earnings exceed actual earnings).1 A number of explanations have been offered regarding this phenomenon. Some studies focus on the behavior of analysts making the forecasts while others focus on the behavior of the managers of firms whose earnings are being forecasted.

With respect to the analysts, one argument is that those making forecasts may have incentives to overestimate earnings (Philbrick and Ricks, 1991; Dugar and Nathan, 1995), while another explanation is simply that the analysts are not able to make rational forecasts of earnings. With respect to the latter, Dowen (1996) and Hwang et al. (1996) find that analysts have difficulty predicting losses and large profits because of the highly transitory nature of these occurrences.2

A second line of reasoning focuses on the behavior of managers and explores whether the optimistic bias is the result of certain kinds of earnings management or, possibly, the result of firms having the ability to manage analysts' forecasts. With regard to earnings management, Brown (1998) argues that an optimistic bias (forecast > actual earnings) often exists when managers expect to report losses, because they may take efforts to exacerbate the loss (leading to a large optimistic forecast error) hoping to have a more prosperous following year.3 But when managers expect profits, a slight pessimistic bias (forecast

"This is the pattern earnings management creates: companies try to meet or beat Wall Street earnings projections in order to grow market capitalization and increase the value of stock options. Their ability to do so depends on achieving the earnings expectations of analysts. And analysts seek constant guidance from companies to frame those expectations." (Levitt, 1998).

A major objective of this study is to investigate whether the optimistic bias and asymmetric behavior of forecast errors found in most U.S. studies exists in Japan. There are a number of reasons that we focus on Japanese firms. First, while Japan has the second largest stock market in the world with over 343 trillion yen in market value of stocks on the Tokyo Stock Exchange-First Section as of February 2001, there has been little investigation of biases in analysts' forecasts of Japanese firms. Second, large amounts of U.S. and international funds have flowed into the Japanese stock market and, as such, international investor interest in Japanese stocks is high. For example, as of December 31, 1999, the value of U.S. mutual funds invested solely in Japanese stocks was $8.1 billion. This amount does not include balances of funds that only invest in Pacific Rim, global and/or international stocks. The balance of U.S. mutual funds that specialize in Pacific Rim stocks as of December 31, 1999, was $12.4 billion. Third, if the forecast bias exists in Japan, similar to that found in the U.S. studies, we attempt to determine whether the cause is due to analysts' irrationality or to either earnings management or forecast management. If earnings or forecast management is the reason, knowing the extent of this practice in Japan should be of concern to U.S. and international investors and the Japanese Ministry of Finance (MOF) which monitors financial and accounting practices of Japanese firms. The U.S. Securities and Exchange Commission (SEC) has recently begun considering whether to allow foreign firms to list on U.S. exchanges using International Accounting Standards (IAS). The SEC is also interested in knowing the extent of earnings management by foreign firms. Indeed, the Chief Accountant of the SEC has asked academic researchers to provide more evidence on earnings management in other countries using IAS (Turner and Godwin, 1999). While our discussion focuses on Japanese accounting standards rather than IAS, our evidence is suggestive of the strictness with which accounting standards are applied and interpreted in Japan and provides evidence useful to the SEC on this issue.4

Our empirical investigation focuses on the 225 Japanese firms that make up the Nikkei composite stock index. Our results suggest that during periods in which firms experience losses, forecasts tend to be overly optimistic. During periods where firms experience profits, forecasts appear to be reasonably accurate, but exhibit a small pessimistic bias. Our results suggest that analysts who forecast earnings in Japan have a difficult time predicting losses and large profits due to their transitory nature. We also show that firm size is not directly related to forecast bias, but, rather, that forecast bias is related to the magnitude of profit or loss a firm reports. With the exception of a few cases, there is generally a significant improvement in forecast accuracy of annual earnings from the first to the fourth quarter. However, we also find that the extent of analysts' forecast bias for firms that report losses decreases only slightly from their first to their fourth quarter forecasts of annual earnings. Finally, we investigate whether the large negative forecast errors for firms experiencing losses and small positive errors for firms reporting profits result from discretionary accruals being used by managers.