Analysts' use of earnings forecasts in predicting stock returns: Forecast horizon effects

Article history: Received January 25, 2015 Received in revised format 28 March 2015 Accepted 2 May 2015 Available online May 3 2015 This paper presents an empirical investigation to study the effects of quality of earnings forecasts in predicting stock returns on 121 selected firms traded on Tehran Stock Exchange over the period 2009-2013. The study investigates the effects of three year means of earnings forecast accuracy on investors’ investment decisions in terms of volume and time horizon, i.e. short term and long term investment sentiment. Using some regression analysis, the study has determined a positive and meaningful relationship between the quality of earnings forecast and investment time horizon. However, the study did not find any meaningful relationship between earnings forecast accuracy and volume of investment. Growing Science Ltd. All rights reserved. 5 © 201


Introduction
One of the primary methods for making investment decisions is to have a good forecast on firms' performance in terms of future earnings (Stickney et al., 2009).A good earnings forecast helps investors make better investment decisions and build a portfolio with better performance (De Livera et al., 2011).It also helps reduce stock price fluctuation and investors do not have to worry about the day to day stock market events (Kasznik, 1996).A good earnings forecast may also be considered as good signal for initial public offering investment decisions (Firth, 1997).There are many studies in learning more about the effects of earnings quality on investors' decisions.Coën et al. (2009) analyzed earnings forecasting errors made by financial analysts for 18 developed countries from 1990 to 2006.They applied the Heston-Rouwenhorst method to unravel country-, industry-, and firm-specific impacts as a source of variation in financial analysts' earnings forecast errors.They estimated each effect using a dummy variable regression, and then decomposed the variance of forecast errors into various effects.They explained that the differences between countries, industrial sectors, and analyst-following could provide a weak explanation for differences in forecast errors.Country effects nevertheless largely dominated industry and analyst-following effects.On the contrary, the type of earnings and variations in earnings played essential role in the forecast accuracy of financial analysts.Siougle (2007) studied the credibility of earnings forecasts disclosed by firms listed on the Athens Stock Exchange.The primary concern of the study was on earnings forecasts disclosed in the prospectus of some listed companies, which were issuing equity through a rights issue.They reported that the earnings forecasts were overestimated and indicated low accuracy, a finding implying that firms were optimistic when they developed their forecasts.The level of forecast errors could be described by valuation uncertainty, which was measured by the standard deviation of stock returns prior to the equity offering, and the market-to-book ratio.Goodman et al. (2013) studied whether the quality of managers' externally reported earnings forecasts could be implemented to infer the quality of their corporate investment decisions.Depending on the intuition that managers may draw on similar skills when building external earnings forecasts and internal payoff forecasts for their investment decisions, they forecasted that managers with higher quality external earnings forecasts could make better investment decisions.They reported that forecasting quality was positively related to the quality of both acquisition and capital expenditure decisions.Payne (2008) investigated audit firm specialization in settings where managers had incentives to modify earnings to reach analysts' earnings forecasts.Their results indicated that audit firms that had a large market share of clients within a specific industry, and audit companies that received a significant portion of their firm revenues from a particular industry, were related to audited financial statement earnings that increased absolute levels of analysts' forecast error and were less likely to just meet or beat analysts' forecasts.Rees and Sivaramakrishnan (2007) investigated the effect of meeting or beating revenue forecasts on the relationship between quarterly returns and earnings forecast errors.Fang (2009) investigated the role of management forecast precision in predicting management forecast error.

The proposed study
This paper presents an empirical investigation to study the effects of quality of earnings forecasts in predicting stock returns for 121 selected firms traded on Tehran Stock Exchange (TSE) over the period 2009-2013.The study investigates the effects of three year mean of precision of earnings forecast on investors' investment decisions in terms of volume and horizon.The study determines short term and long term investment through the following equations Short term investment Short term investment= Short term investment + Long term investment , (1) Long term investment Long term investment= Short term investment + Long term investment . (2) If the ratio of Eq. ( 1) is greater than 0.5, the investment policy is considered as short term and long term, otherwise.The change in volume of investment (∆ Volume) is determined as follows, where vt and vt-1 are trading volumes of two consecutive periods of t and t-1, respectively.In order to measure forecasting accuracy, the study measures the error between the earnings forecast from the beginning of the period to the end of period as follows, Actual earning earnings forecast of the beginning of the year Forecasting earnings accuracy error= earnings forecast of the beginning of the year and Forecasting Accuracy = 1 -Forecasting earnings accuracy error.In this study, earnings accuracy is calculated by taking an average from three years of earnings.
In addition, the study uses four control variables including firm Size, Leverage, Growth and Tobin's Q.The first control variable, Size, is calculated by taking natural logarithm of total assets.The second control variable, Growth, is measured as follows, Actual sales Sales at the beginning of the year Growth = Sales at the beginning of the year  . (5) Leverage is also calculated using debt ratio, which is calculated as follows, Total liabilities Leverage = Total assets . (6) Finally, Tobin's Q is calculated as follows, Market value Book value of liabilities Tobin's Q = Book value of total assets There are two hypotheses associated with the proposed study of this paper as follows, 1.There is a relationship between the short/long term investment (Maintenance) and earnings accuracy.
2. There is a relationship between the volume of investment (∆ Volume) and earnings accuracy.
The proposed study of this paper uses the following two regression analysis.

Table 1
The summary of some basic statistics According to the results of Table 1, we may have some idea about the normality of the data.However, we use Kolmogorov-Smirnov test to make sure the data were normality distributed and Table 2 shows the results of our survey.The results of Table 2 indicates that the dependent variable ∆ Volume is normally distributed.Table 3 demonstrates the results of Pearson correlation ratios among some of the variables.According to the results of Table 3, Investment horizon has positive correlation with Accuracy, Size and negatively associated with Tobin's Q.However, ∆ Volume has no correlation with any other control or independent variables.The implementation of Hausman and Limer tests have indicated that the study has to use random and fixed effects for models one and two, respectively.

Results
In this section, we present the results of the implementations of regression technique on Eq. ( 8) and Eq. ( 9), respectively.

The first hypothesis: The relationship between time horizon and earnings accuracy
The first hypothesis of the survey is associated with relationship of time horizon and earnings accuracy.Table 4 demonstrates the results of our survey.According to the results of Table 4, R-Square is equal to 0.15, which means the regression technique can approximately describe 15% of the changes between independent and dependent variables.Moreover, Wald value for the variable Accuracy is positive and significant.Durbin-Watson is within desirable value, which indicates no sign of auto-correlation.There is a positive relationship between earnings accuracy and investment discipline in terms of long or short term decisions.Therefore, the first hypothesis is confirmed.

The second hypothesis: The relationship between volume of investment and earnings accuracy
The second hypothesis of the survey is associated with relationship of volume of trading and earnings accuracy.Table 4 presents the results of our survey.According to the results of Table 5, R-Square is equal to 0.26, which means the regression technique can approximately describe 26% of the changes between independent and dependent variables.Durbin-Watson is within desirable value, which indicates no sign of auto-correlation.However, t-student value for independent variable is not statistically significant, which means there is no meaningful relationship between independent and dependent variable and the second hypothesis of the survey was not confirmed.

Conclusion
In this paper, we have presented an empirical investigation to study the effects of earnings accuracy on investment decisions for selected firms traded on Tehran Stock Exchange (TSE).Using some regression techniques, the study has determined that while there was a positive and meaningful relationship between earnings accuracy and investment discipline in terms of short/long term investment sentiment, there was no meaningful relationship between earnings accuracy and volume of trades.The results of our survey are consistent with findings earlier reported in the literature (Coën et al., 2009;Fang, 2009;Payne, 2008;Rees & Sivaramakrishnan, 2007).The results of this survey suggest TSE officials to build a better rules and regulations, which could help many investors.Unlike US stock market, many retail investors on this exchange are not familiar with details of earnings forecast and firms with good quality earnings forecast may keep novice investors from possible frauds or misleading earnings forecast made by small firms whose shares were publically traded on this exchange.

Table 2
The summary of Kolmogorov-Smirnov test

Table 4
The summary of regression analysis for the first hypothesis

Table 5
The summary of regression analysis for the second hypothesis