Investors’ Reactions to Analysts’ Forecast Revisions and Information Uncertainty

Abstract

This study examines the relationship among analysts’ earnings forecast revisions, information uncertainty, and stock returns and provides new evidence that stock price drift occurs after analysts’ earnings forecast revisions. Using data from the Australian stock market over the period of 1992 to 2009, we find that the stocks with upward earnings revisions experience positive returns, while stocks with downward revisions have negative returns. The effect is more prominent in stocks with high information uncertainty. The results are robust after controlling for market conditions, seasonality, and risks. Our evidence supports the conservatism bias model that investors tend to underweight the public information, such as analysts’ earnings forecast revisions. Importantly, our evidence provides possible explanations about the violation of the efficient market hypothesis. Our results suggest that the conservative bias causes investors not to sufficiently update their beliefs and eventually results in subsequent return continuation as investors’ underreaction to analysts’ earnings forecast revision is stronger with higher information uncertainty.

Keywords

analysts’ forecast revisions investor behavior information uncertainty stock price drift Australian stock market

Introduction

Equity analysts play an important role as information intermediaries (Guan, Lu, & Wong, 2012). Analysts are generally believed to be sophisticated users of financial information who have superior ability in providing high-quality information (Chava, Kumar, & Warga, 2010). A recent study by Jung, Sun, and Yang (2012) suggests that analysts facilitate more effective monitoring of firms’ activities and, thereby, reduce agency costs and increase shareholder value. It has been widely argued that analyst forecasts improve the informational efficiency of the stock markets as they often represent market expectations for a firm’s future stock performance (Beaver, Cornell, Landsman, & Stubben, 2008; Clement, Hales, & Xue, 2011). There is ample evidence that the information provided by analysts through their earnings forecasts, recommendations, and reports is used by market participants and that such output influences stock prices (e.g., Abarbanell & Lehavy, 2003; Asquith, Mikhail, & Au, 2005; Athanasakou, Strong, & Walker, 2009; Beaver et al., 2008; Brav & Lehavy, 2003; Dontoh, Ronen, & Sarath, 2003; Francis & Soffer, 1997; Lys & Sohn, 1990; Stickel, 1995). Market participants use analysts’ forecasts because analysts process and transform the information set in financial statements along with additional information about the industry, firm strategy, and economy into future earnings predictions (Wieland, 2011).

Research on analysts’ forecast revisions has attracted a lot of interest in the literature since the 1970s. One stream of the research shows that the information embedded in analysts’ forecast revisions is helpful in predicting stock returns. A number of reasons and factors have been identified to explain this phenomenon. For example, Chan, Jegadeesh, and Lakonishok (1996) believe that this pattern exists because the market responds gradually to recently released information. Hong and Stein (2003) and Hong, Lim, and Stein (2000) show that price continuations are attributed to the gradual diffusion of firm-specific information. In contrast, Daniel, Hirshleifer, and Subrahmanyam (1998, 2001) attribute this phenomenon to investors’ overconfidence as firms with higher uncertainty are difficult to value; thus, investors tend to be overconfident when there is positive news for a particular stock. Chan et al. (1996); Barberis, Shleifer, and Vishny (1998); and Daniel et al. (1998, 2001) find that the stock price drift followed by firm-specific news announcement as either the market underreaction to recently released information or investors’ overconfidence to specific news.

Prior research also suggests that delayed reaction to information is natural in a fully Bayesian-rational world as long as there is some unsystematic noise in the market framework. Bray (1981) finds that uncertainty about spot market production implies the futures price will only partially reflect the information available to the spot price. Dontoh et al. (2003) further demonstrate post-announcement earnings drifts and find a positive correlation between the unexpected component of current public signals and futures price changes, suggesting that the market is slow to respond to earnings announcements even though traders act rationally and information is processed in a timely and efficient manner. Lewellen and Shanken (2002) document that returns are negatively related to past dividends and prices over the subsequent period. They suggest this phenomenon is inherent in a model with parameter uncertainty because investors’ mistakes eventually reverse as they learn more about the economy. They argue that predictability can take the form of either reversals or continuations or neither, depending on investors’ prior beliefs and the cash-flow process. Therefore, whether there is a positive or negative correlation between an analyst’s forecast revisions and the corresponding in successive stock price changes is ultimately an empirical question.

This article attempts to address this question. We empirically examine how investors’ reactions to analysts’ forecast earnings revisions (ERs) and information uncertainty, using data from the Australian stock market over the period of 1992 to 2009.¹ Our main argument is that if the post-price drift is due to investors’ delay in reactions to analysts’ forecast earnings news, the underperformance should be greater when there is a favorable revision. Moreover, this phenomenon should be stronger when the information about the firm is scarcer. Specifically, the difference between favorable and unfavorable revisions should be more profound among stocks with higher uncertainty.

The analysis of our data consistently shows that Australian stocks with upward-revised forecasts receive higher positive future returns than other similar stocks. The spread in returns between the portfolio of stocks with upward revision and the portfolio of stocks with downward revision converges to zero when the holding period increases. This suggests that the revised information is gradually incorporated into the prices through time. Moreover, the spread is almost dominated by the underperformance of the upward-revision portfolio, and the year-by-year performance provides a robust result for this phenomenon. In addition, our analysis reveals that the spread in return between the upward-revision portfolio and the downward-revision portfolio exhibits a strong January effect; however, the January effect does not drive our results because spreads among other months are also statistically significant. Our results suggest that investor’s conservatism bias delays the incorporation of information about analysts’ forecast revisions into firms’ valuation. We also find that the spreads of analysts’ forecast revision are stronger when the level of information uncertainty is high and vice versa. The spreads of analysts’ forecast revision remain significant even after controlling for market conditions, seasonality, and risks. Our findings are consistent with the explanation of Barberis et al. (1998) that investors do not update their beliefs sufficiently to reflect public information, and this behavior leads to post-forecast revisions price drift. Our results are also in agreement with the findings of Zhang (2006b) who demonstrates that stocks with higher information uncertainty experience higher price drift.

Our article contributes to the literature in several ways. First, our study extends the literature on the relationship among analysts’ earnings forecast revisions, information uncertainty, and stock returns by providing new evidence that stock price drift occurs after the analysts’ earnings forecast revisions. The evidence supports the underreaction hypothesis and is statistically significant for stocks with favorable news. Our finding renders new support to the finding of earlier studies by Aitken, Frino, and Winn (1996) based on data from 1985 to 1992 and Lim and Kong (2004) based on data from 1994 to 1996 that analysts’ forecast revisions in Australia affect stock prices. However, both previous studies consider a relatively short period of time. Our study uses data covering a longer period of time from 1992 to 2009. Second, our evidence provides possible explanations about the violation of the efficient market hypothesis. The traditional economic theory is based on the belief that individuals behave rationally and have perfect knowledge when making their investment decisions; however, some investors may have psychological bias that leads to predictable stock mispricing. Our study proposes a combo index as a joint information proxy to test the impact of uncertainty on stock returns. Our results suggest that the conservative bias causes investors not to sufficiently update their beliefs and eventually results in subsequent return continuation. Moreover, we find investors’ underreaction to analysts’ earnings forecast revisions becomes stronger with higher information uncertainty and the market downturn condition. Our evidence suggests that the persistence of mispricing of stocks depends on the degree of uncertainty to firm-specific information and market conditions.

The rest of the article is organized as follows. The next section briefly reviews relevant literature and develops our hypotheses. “Data, Descriptive Statistics, and Trading Strategies” section describes our sample and trading strategies. “Information Uncertainty and Returns” section investigates the role of information uncertainty and stock returns. “Robustness Check” section provides robustness checks and cross-sectional analysis. The final section concludes the article.

Prior Literature and the Research Hypothesis

Earlier studies by Griffin (1976); Givoly and Lakonishok (1979, 1980); Elton, Gruber, and Gultekin (1981); Imhoff and Lobo (1984); Stickel (1991); and Lim and Kong (2004), among others, have documented stock price drift following analysts’ forecast revisions. For example, Stickel (1991) finds that stocks with favorable analysts’ forecast revisions consistently outperform stocks with downward revisions. Aitken et al. (1996) show that analysts’ forecast ERs are associated with stock returns in Australia. Furthermore, the attribution of stock price continuation is often categorized into investor and analyst behavioral biases, such as investors underreacting to new information or investors’ being overconfident toward recent positive news for a particular stock (e.g., Abarbanell, 1991; Abarbanell & Bernard, 1992; Ali, Klein, & Rosenfeld, 1992; Kasznik & McNichols, 2002; Lys & Sohn, 1990). Chan et al. (1996) argue that the drift following analysts’ revisions is a part of a general type of return continuation effect whereby the market responds gradually to recently released information, so prices exhibit predictable drift patterns. Barberis et al. (1998) also find that investors tend to underestimate analysts’ forecast revisions, particularly for the favorable revision. They propose investor conservatism bias model and show that price continuation is the result of investors who do not update their beliefs adequately when new information initially emerges in the market. Easterwood and Nutt (1999) discover that analysts overreact to positive information but underreact to negative information.

However, Hirshleifer (2001) argues that uncertainty of stocks and absence of correct feedback mechanism about fundamentals of stocks may cause the psychological biases. Daniel et al. (1998, 2001) show that return predictability should be more pronounced for firms with greater uncertainty because investors tend to be more overconfident when a firm’s business is hard to value. Gu and Xue (2007) argue that analyst’s overreaction to good news is a reasonable response when the uncertainty is high, and this overreaction should not be interpreted as a psychological bias. As the degree of uncertainty varies across stocks, it follows that information should be more quickly embedded in the prices of lower uncertainty stocks than the prices of higher uncertainty stocks. Similarly, errors in stock prices due to investors’ biased expectations on its future earnings stream should be corrected more quickly for stocks with lower uncertainty than stocks with higher uncertainty. For example, Zhang (2006b) shows that stock with greater information uncertainty experiences relatively higher expected returns following upward revisions and relatively lower expected returns following downward revisions. These findings suggest that information slowly flows into prices when the level of uncertainty is higher, thereby supporting the underreaction hypothesis. Other studies (e.g., Abarbanell & Bushee, 1997, 1998) show that analysts’ forecast revisions do not completely impound the information in financial statements and trading strategies based on the information in financial statements but without considering analysts’ forecasts generate abnormal returns.

However, in addition to the behavioral explanation, several theoretical studies suggest that under an assumption of rational expectation, the prior information can still have the ability to predict future price changes, that is, against market efficiency. For example, Bray (1981) shows that the futures price may not be an unbiased estimator of the spot price. He suggests that some traditional benchmarks of market efficiency do not hold under the assumption of the rational expectations. Dontoh et al. (2003) extend Bray’s model to consider public information. They show how a post-announcement drift can arise naturally within a Bayesian framework. Under their model, investors will rationally postpone investing until more information becomes available, and this trading behavior will cause a post-announcement earnings drift. Moreover, Lewellen and Shanken (2002) argue that if investors learn about a structural parameter, it is possible to have negative correlation in price changes even though all investors are Bayesian-rational. They show that price changes reflect investors’ beliefs about future payoffs. If investors’ beliefs are rational, stock prices will reflect all the available information and thus be unpredictable. However, if there is a rational market with Bayesian-rational investors, a biased prior belief will also induce reversal or continuation patterns in the market. As a result, the so-called “anomalies” are actually part of the true data-generating process (Dontoh et al., 2003).

In this study, we investigate how investors react to analysts’ earnings forecast revisions and to what extent information uncertainty contributes to the future stock returns in the Australian stock market. More specifically, we want to examine whether the changes of analysts’ earnings forecast have an influence on stock returns and test whether the level of uncertainty has an impact on investors’ investment decisions. Our main research proposition is that if the post-price drift is due to investors’ delay in reactions to analysts’ forecast earnings news, the underperformance should be greater when there is a favorable revision. We posit that this phenomenon should be stronger when the information about the firm is scarcer, and the difference between favorable and unfavorable revisions should be more profound among stocks with higher uncertainty.

Data, Descriptive Statistics, and Trading Strategies

Variable Definitions

The earnings forecast revision ratio is defined as the average monthly earnings forecast change in expected earnings per share as a percentage of the absolute mean value of the prior consensus forecasts. To measure information uncertainty and its impact on stock returns, we adopt five information uncertainty proxies: (a) firm size, (b) book-to-market (B/M) ratio, (c) analyst coverage, (d) earnings forecast dispersion, and (e) absolute prediction errors of cash flow.

Firm size is used to measure uncertainty because smaller firms tend to have higher uncertainty than larger firms. Firm size (MV) is measured as the market capitalization at the beginning of month t. The B/M ratio is used as a proxy for information uncertainty because growth stocks generally distribute low dividends as earnings are reinvested in the business to support high growth, and value stocks generally pay more dividends. Thus, we can argue that value stocks represent lower uncertainty about future dividends. The B/M ratio is computed by dividing the book value by market capitalization at month t; therefore, B/M ratio is updated each month. Furthermore, analyst coverage (NAF) is considered as a proxy of information uncertainty because the more the analysts analyze a firm, the better the investors understand the particular firm and less uncertainty they face. NAF is defined as the total number of analysts covering the firms for the fiscal period. The dispersion (DISP) in analysts’ forecast earnings is another proxy for information uncertainty. It means that analysts have different views about the firm’s future earnings. If the discrepancy is larger, certainly it could intensify the uncertainty sentiment for investors about the firm’s future earnings power. Dispersion is defined as the ratio of the standard deviation of analysts’ current-fiscal-year annual earnings per share forecasts to the absolute value of the mean forecast, as reported in the Institutional Brokerage Estimate System (I/B/E/S) Summary History file. We also compute absolute prediction errors (ABSE) for firm-level quarterly cash flows and use it as an alternative information uncertainty variable. Following Nam, Brochet, and Ronen (2012), we use a time series model of estimating future cash flows using past cash flows to predict out-of-sample cash flows and determine the prediction error on a firm-by-firm basis. An out-of-sample prediction-based approach was argued to be preferable to one based on fit (Lev, Li, & Sougiannis, 2010). Nam et al. (2012) also observe that the firm-level quarterly cash-flow time series exhibit purely seasonal characteristics. For this reason, we adjust our data to account for these normal seasonal patterns. We then compute ABSE.

We use the reciprocal values of MV, B/M, and NAF to measure information uncertainty, which are represented by INVMV, MB (market-to-book ratio), and INVNAF. We do this for ease of comparison between variables in the analysis. For example, when a firm is smaller or has less analyst coverage than other firms, the uncertainty level for that particular firm will be reflected through a higher INVMV or INVNAF value.

Data and Sample Selection

Our sample data come from two sources. The primary data come from the I/B/E/S International Summary database. Prior studies have argued that I/B/E/S earnings forecasts should result in a more precise proxy for market expectations of earnings (e.g., Claus & Thomas, 2001; Easton, 2004; Frankel & Lee, 1998). Ramnath, Rock, and Shane (2005) compare the accuracy of forecasts of a single forecaster (Value Line) to consensus forecasts (I/B/E/S) and find that I/B/E/S is less biased and more accurate. Beaver et al. (2008) show that the I/B/E/S database has an extended coverage over time, which makes the data more comprehensive ensuring a consistency between the forecast and the realization of earnings, as well as a consistency across analysts in the earnings being forecasted. Our initial sample includes all firms traded on the Australian Securities Exchange (ASX) with at least one I/B/E/S consensus forecast available for the period from July 1992 to June 2009.

In this article, an analyst is defined as an individual or a department at a research organization providing forecast data to I/B/E/S. An analyst also provides guidance to businesses and individuals making investment decisions. They assess the performance of stocks, bonds, and other types of investments. Our data set consists of monthly data of stock prices and returns in Australian dollar, number of shares outstanding, number of analysts covered, 1-year forward earnings forecasts, and standard deviation, mean, and actual reported earnings per share. Individual stocks’ B/M value and other financial data are from the DataStream data set. Each stock in the data set has a market capitalization data available at the beginning of each month with a minimum of 12-month returns history. From the data, we find that analysts tend to cover larger and more actively traded firms, and our sample ends up with 711 Australian stocks for the whole testing period.

Table 1 provides the descriptive statistics for the period from July 1992 through June 2009 for all samples. The table reports the number of I/B/E/S firms, their mean and median sizes (in AUD millions), mean and median B/M ratios, and the number of analysts per firm at each coverage percentiles ranging from 10% to 90%. Our sample shows that total numbers of firms covered by analysts increased with time until the subprime mortgage crisis, but the analyst coverage per firm seemed to decrease with time. Generally speaking, the mean size of the sample firms also increased with time, except for the period of the financial tsunami. Mean B/M ratio was relatively stable during this time period, but the financial crisis depressed the market value of individual firms, which caused the mean B/M ratio to increase. The mean B/M ratio was similar as those in the U.S. market, and the number of analyst covered per firm was similar to the U.S. counterpart as well.²

Table 1.

Descriptive Statistics.

						Coverage percentiles
Year	Number of firms	Mean size (AUD millions)	Median size (AUD millions)	Mean B/M	Median B/M	10	20	30	40	50	60	70	80	90
All	273	2,632.90	449.54	0.67	0.52	1	2	3	4	6	7	9	11	13
1992	122	1,447.77	455.27	0.91	0.74	2	3	5	8	10	11	12	13	14
1993	140	1,734.18	570.91	0.76	0.62	2	4	6	9	11	12	13	14	16
1994	185	1,725.82	497.97	0.66	0.61	1	2	4	5	9	11	13	14	16
1995	216	1,415.16	299.86	0.70	0.68	1	2	3	4	6	8	11	13	15
1996	221	1,518.58	348.43	0.62	0.61	1	2	3	5	6	8	10	12	13
1997	227	1,775.26	407.35	0.56	0.52	2	3	4	5	7	8	10	12	13
1998	249	1,910.56	373.51	0.61	0.52	2	4	5	6	8	9	12	14	16
1999	267	2,207.59	403.10	0.60	0.55	2	3	5	6	7	9	11	13	15
2000	327	2,254.62	390.04	0.64	0.49	1	2	3	5	6	7	9	11	14
2001	319	2,099.87	246.05	0.75	0.55	1	2	3	4	6	8	9	11	13
2002	311	2,393.10	368.10	0.69	0.54	1	2	3	4	6	8	9	11	13
2003	257	2,644.57	519.70	0.64	0.57	1	2	3	4	6	7	8	10	11
2004	280	3,014.11	616.21	0.55	0.50	1	2	3	4	5	7	8	9	11
2005	304	3,471.76	665.97	0.50	0.42	1	1	2	3	4	5	6	8	9
2006	344	3,764.67	674.46	0.52	0.41	1	1	2	3	4	5	6	8	10
2007	405	4,353.31	697.75	0.46	0.38	1	1	2	3	4	6	7	10	12
2008	398	3,380.92	406.86	0.74	0.56	1	1	2	3	4	5	7	9	11
2009	350	2,588.81	250.05	1.53	0.95	1	2	2	3	4	6	8	11	13

Note. This table reports descriptive statistics for all listed stocks in Australia between July 1992 and June 2009. For each year, we report the number of I/B/E/S firms, the mean and median sizes of firms (in AUD millions), mean and median B/M ratios, and the number of analysts per firm for each coverage percentiles ranged from 10% to 90%. I/B/E/S = Institutional Brokerage Estimate System; B/M = book-to-market.

Portfolio Strategies

We first examine whether the analysts’ forecast revision is a good predictor for future stock returns. At the beginning of each month t, all stocks are sorted into three different portfolios based on the sign of its analysts’ forecast revisions. Portfolio P1 (P3) contains stocks with negative (positive) revisions and P2 contains stocks without any revision. All portfolios are equally weighted at the formation month and held for 1, 3, 6, 9, 12, 18, 24, and 36 months. The forecast revision is the average revisions of individual analysts who cover the firm in both month t− 1 and t.

Table 2 reports the average monthly portfolio returns for three different forecast revision groups for the period from July 1992 to June 2009. It shows that portfolios without revisions account for 36.3% of the sample followed by 33.7% and 30% for portfolios with downward revisions (P1) and upward revisions (P3), respectively.³ It demonstrates that the analysts’ forecast revisions positively impact stock prices at the portfolios formation period. For example, the stocks with downward revisions (upward revisions) experience significantly negative (positive) returns of −0.87% (2.74%) at formation month t. There is also a monotonic increase in returns with time in negative forecast revision portfolios from the formation period to 6-month holding period, and then the return becomes positive and significant until the 36-month holding period. The positive revision has a significant positive impact on the returns not only for the formation periods but also for the subsequent holding periods for which all of the returns are statistically significant. We also observe a longer underreaction adjustment process for new information. For example, the significant positive returns of portfolios with positive forecast revisions decrease with time until 18 months later. The P2 portfolio has a similar returns pattern as P3. These findings are interesting and support earlier research on the conservatism theory of investors (e.g., Ajayi & Mehdian, 1994). It shows that investors tend to overreact to unfavorable news and underreact to favorable news; therefore, the movements of future stock prices continue to respond to the information.

Table 2.

Forecast Revisions and Return Continuation.

Analysts’ forecast revision	% of sample	Formation period RET_t	Holding periods
			1 month	3 months	6 months	9 months	12 months	18 months	24 months	36 Months
P1 (REV < 0)	33.7%	−0.0087*** (−2.32)	0.0007 (0.19)	0.0032 (1.30)	0.0043** (2.30)	0.0048*** (3.28)	0.0054*** (4.30)	0.0067*** (6.89)	0.0079*** (9.35)	0.0097*** (13.28)
P2 (REV = 0)	36.3%	0.0061* (1.66)	0.0093** (2.04)	0.0097*** (3.56)	0.0102*** (4.78)	0.0092*** (5.37)	0.0091*** (6.22)	0.0107*** (9.10)	0.0120*** (11.79)	0.0153*** (13.30)
P3 (REV > 0)	30.0%	0.0274*** (5.40)	0.0144*** (4.03)	0.0120*** (4.98)	0.0117*** (6.25)	0.0107*** (7.12)	0.0106*** (8.16)	0.0108*** (10.36)	0.0110*** (12.32)	0.0123*** (15.54)
P2 − P1 / P3 − P1		0.41	0.63	0.74	0.81	0.75	0.70	1.00	1.31	2.17
P3 − P1		0.0361*** (5.72)	0.0136*** (2.62)	0.0088*** (2.58)	0.0074*** (2.82)	0.0060*** (2.85)	0.0052*** (2.91)	0.0040*** (2.83)	0.0031** (2.54)	0.0026** (2.39)

Note. This table reports average monthly portfolio returns for the period of July 1992 to June 2009 for all listed stocks in Australia. The forecast revision ratio (ER) is defined as the average monthly earnings forecast change in expected earnings per share as a percentage of the absolute mean value of prior consensus forecasts. Each month, we sort stocks into three categories based on analysts’ forecast revision ratio in month t, which stocks in Portfolio P1 (P3) are those that contain negative (positive) revision, and stocks in P2 contain no revision (zero). The forecast revision is the average of individual revisions by analysts who cover the firm in both months t− 1 and t. The portfolio is equally weighted at the formation month and held for 1, 3, 6, 9, and 12 months. Mean forecast revision ratio, analyst coverage, and firms’ average market capitalization for each formed portfolio are reported in the last column. RET_t is the portfolio return during t period and t statistics are reported in parentheses. ER = earnings revision.

, **, and *** indicate a significant level of 10%, 5%, and 1%, respectively.

We further investigate whether the difference in returns between positive (P3) and negative (P1) revisions is significant in different holding periods. We conduct this analysis by buying the positive-revision portfolio and shorting the negative-revision portfolio (P3 − P1). The results at the bottom of Table 2 illustrate significant positive returns across the formation and different holding periods. We also observe that returns tend to decrease as the holding periods are extended. Furthermore, we investigate the cumulative return patterns for these three ER portfolios. The results in Figure 1 indicate that, except for the downward-revision portfolio (P1), all portfolios could generate positive cumulative returns. It is clear that the zero-cost portfolios (P3 − P1) earn rising cumulative returns in the long run. This evidence implies that analysts’ forecast revisions do provide valuable information to investors and investors react to the information that reflects the future return drift of the portfolio.

Figure 1.

Cumulative forecast revision returns (for the period from July 1992 to June 2009).

Several important features after the formation period are summarized as follows. First, all three types of portfolios tend to earn positive returns lasting for 36 months after the formation period. Meanwhile, there exists a monotonically increasing pattern in returns from downward- to upward-revision portfolios. Our result in Table 2 suggests that the upward-revision portfolios seem to continually provide significantly positive post-formation returns; however, the downward revisions do not experience negative post-formation returns. In fact, the downward-revision portfolios provide positive returns after the formation period, and the returns become significant after the 6-month holding period. This result shows, in the short term, analysts’ upward revisions are informative in predicting future stock returns. Our findings are consistent with the findings of Stickel (1991) and Gleason and Lee (2003).

Although all three portfolios experience positive returns after the formation period, the returns on portfolios with upward revisions still consistently outperform portfolios with downward revisions. When we set up a zero-cost trading strategy of long stocks with upward revision (P3) and short stocks with downward revisions (P1), this strategy generates significant positive returns across all different holding periods. This result becomes even more obvious when we consider the cumulative returns. Except for the downward forecast revision portfolio, all portfolios could generate consistent positive cumulative returns during the sample period.

Sample Characteristics of Forecast Revisions by Year

In this section, we examine the year-by-year performance of portfolios based on forecast revisions. The results, as presented in Table 3, indicate that nearly half of downward-revision portfolios generate negative returns and only two of them are not statistically significant. In contrast, the upward-revision portfolios have 14 out of 18 years with positive returns. The outcome in Table 3 shows a consistency in which negative-revision portfolios (P1) always have lower returns than positive-revision portfolios (P3), except for the year 2008. This directly explains the positive returns for the zero-cost portfolios (P3 − P1) in which 11 out of 18 sample periods have significant positive returns. These outcomes show that analysts’ forecast revisions impact investors’ sentiment.

Table 3.

Year-by-Year Performance.

Year	Analysts’ forecast revision			P3 − P1
	P1REV < 0	P2REV = 0	P3REV > 0
1992	0.0045 (0.79)	0.0195*** (2.77)	0.0083 (1.54)	0.0038 (0.48)
1993	0.0268*** (6.11)	0.0393*** (8.69)	0.0411*** (10.74)	0.0144** (2.47)
1994	−0.0210*** (−7.05)	−0.0172*** (−4.69)	−0.0199*** (−7.18)	0.0011 (0.27)
1995	0.0106*** (4.15)	0.0122*** (3.61)	0.0214*** (7.24)	0.0108*** (2.76)
1996	0.0088*** (3.41)	0.0140*** (4.45)	0.0209*** (6.06)	0.0121*** (2.80)
1997	−0.0025 (−0.76)	0.0178*** (5.01)	−0.0017 (−0.46)	0.0007 (0.15)
1998	−0.0064* (−1.83)	0.0119*** (2.91)	0.0149*** (2.91)	0.0213*** (3.54)
1999	−0.0005 (−0.15)	0.0113*** (2.65)	0.0040 (1.37)	0.0045 (1.02)
2000	−0.0131*** (−3.45)	0.0297 (0.94)	0.0130*** (2.87)	0.0261*** (4.43)
2001	0.0010 (0.20)	−0.0086* (−1.78)	0.0355** (2.04)	0.0345** (2.10)
2002	−0.0224*** (−5.97)	−0.0052 (−1.12)	−0.0120*** (−3.74)	0.0104** (2.11)
2003	0.0155*** (4.71)	0.0466** (2.18)	0.0298*** (8.75)	0.0144*** (3.04)
2004	0.0200*** (5.26)	0.0228*** (8.47)	0.0310*** (12.75)	0.0110** (2.56)
2005	0.0069 (1.11)	0.0074*** (2.62)	0.0140*** (4.82)	0.0071 (1.09)
2006	0.0167*** (4.34)	0.0235*** (8.99)	0.0395*** (8.05)	0.0228*** (3.56)
2007	−0.0106*** (−3.07)	−0.0143*** (−2.76)	0.0052 (1.56)	0.0159*** (3.29)
2008	−0.0474*** (−8.81)	−0.0615*** (−11.87)	−0.0648*** (−10.51)	−0.0174** (−2.04)
2009	0.0855*** (10.68)	0.0872*** (7.92)	0.0893*** (7.19)	0.0039 (0.27)

Note. This table reports the year-by-year forecast revision returns between July 1992 and June 2009 for all listed stocks in Australia. Each month, we sort stocks into three categories based on analysts’ forecast revision ratio in month t. The stocks in Portfolio P1 (P3) are those that contain negative (positive) revision, and stocks in P2 contain no revision (zero). The stocks in the portfolio are equally weighted at the formation month and held for 1 month and t statistics are reported in parentheses.

, **, and *** indicate a significant level of 10%, 5%, and 1%, respectively.

Turn-of-the-Year Effect

Seasonality is an important issue in pricing stock returns. Particularly the January effect has been studied extensively in the literature (e.g., Chan, Karceski, & Lakonishok, 1998; Davis, 1994; Jegadeesh & Titman, 2001; Loughran, 1997; Moller & Zilca, 2008; van Dijk, 2011). Most of these studies reveal the evidence that there is a January effect in explaining stock returns. For the same reason, we analyze whether there is a return of the year–month effect in the forecast revision portfolios.

Table 4 shows that almost three fourths of months in a year have negative returns for downward-revision portfolios (P1), in contrast, five sixth of months in a year have positive returns for upward-revision portfolios (P3). Consequently, three fourths of months in a year have significantly positive monthly returns for the zero-cost portfolio (P3 − P1). The shifting pattern in returns from downward-revision portfolios to upward ones is consistent with the pattern presented in Table 3. The evidence aligns with our previous findings that earnings forecast revisions could assist investors in the investment decision in the Australian stocks market.

Table 4.

Seasonality of Forecast Revisions.

Month	Analysts’ forecast revision			P3 − P1
	P1REV < 0	P2REV = 0	P3REV > 0
January	−0.0235*** (−7.36)	−0.0275*** (−7.35)	0.0044 (1.30)	0.0279*** (5.95)
February	0.0335*** (7.92)	0.0307*** (3.69)	0.0241*** (6.89)	−0.0094* (−1.69)
March	0.0140*** (3.73)	0.0273** (2.14)	0.0207*** (6.01)	0.0067 (1.29)
April	−0.0122*** (−2.76)	−0.0098*** (−3.00)	−0.0004 (−0.12)	0.0118*** (2.02)
May	−0.0021 (−0.76)	0.0237 (1.20)	0.0136*** (3.89)	0.0157*** (3.49)
June	0.0177*** (4.93)	0.0244*** (6.80)	0.0300*** (6.13)	0.0123*** (2.07)
July	−0.0085** (−2.48)	−0.0033 (−0.94)	0.0100*** (3.49)	0.0185*** (4.16)
August	−0.0031 (−0.80)	0.0064 (0.87)	0.0090 (1.26)	0.0121 (1.13)
September	−0.0296*** (−6.72)	−0.0112*** (−2.85)	−0.0107*** (−3.30)	0.0188*** (3.44)
October	−0.0127*** (−3.74)	0.0007 (0.19)	0.0077* (1.66)	0.0204*** (3.64)
November	0.0304*** (8.72)	0.0269*** (8.12)	0.0376*** (12.30)	0.0072 (1.47)
December	−0.0126*** (−2.95)	−0.0007 (−0.21)	0.0097* (1.84)	0.0222*** (3.32)
Months other than January	0.0027** (2.29)	0.0101*** (3.77)	0.0132*** (8.44)	0.0105*** (5.46)

Note. This table reports the forecast revision returns partitioned relatively to months to examine seasonality in returns between July 1992 and June 2009 for all listed stocks in Australia. Each month, we sort stocks into three categories based on analysts’ forecast revision ratio in month t, which stocks in Portfolio P1 (P3) are those that contain negative (positive) revision, and stocks in P2 contain no revision (zero). The portfolios are equally weighted at the formation month and held for 1 month and t statistics are reported in parentheses.

, **, and *** indicate a significant level of 10%, 5%, and 1%, respectively.

Further analysis provides the evidence that there is a significant January effect for the zero-cost portfolio that yields a significant return of 2.79% with t value of 5.95. However, for Portfolios P1 and P2, there are significant losses in terms of returns in January. If excluding the month of January, numbers at the bottom of Table 4 show that the forecast revision effect is not driven by the January effect.

Information Uncertainty and Returns

In this section, we investigate whether information uncertainty and analysts’ forecast revision could predict future stock returns and how these two factors influence one another. First, all stocks are sorted into three groups based on the earnings forecast revision ratios at the beginning of each month in the formation period. Then, within each forecast revision group, stocks are sorted into additional three groups based on the information uncertainty variables, which U1 contains the lower 30% uncertainty stocks and U3 contains higher 30% uncertainty stocks. All portfolios are equally weighted at a formation month and held for 1 month. Table 5 reports the average monthly portfolio returns sorted by analysts’ forecast revisions and information uncertainty variables for the period from July 1992 through June 2009.

Table 5.

Portfolio Returns by Information Uncertainty and Analysts’ Forecast Revision.

Panel A: INVMV
Analysts’ forecast revision	U1	U2	U3	U3 − U1
P1 (REV < 0)	0.0038 (1.22)	0.0004 (0.11)	−0.0020 (−0.39)	−0.0058 (−0.97)
P2 (REV = 0)	0.0111** (2.25)	0.0053 (1.24)	0.0127* (1.64)	0.0016 (0.18)
P3 (REV > 0)	0.0084*** (2.81)	0.0137*** (3.73)	0.0211*** (4.07)	0.0127** (2.12)
P3 − P1	0.0046 (1.07)	0.0133** (2.43)	0.0231*** (3.19)	0.0185*** (3.61)
Panel B: MB
Analysts’ forecast revision	U1	U2	U3	U3 − U1
P1 (REV < 0)	0.0088* (1.88)	0.0022 (0.58)	−0.0042 (−1.01)	−0.0130** (−2.08)
P2 (REV = 0)	0.0333** (2.04)	0.0094 (1.49)	0.0070 (1.46)	−0.0263 (−1.55)
P3 (REV > 0)	0.0182*** (4.00)	0.0131*** (3.75)	0.0123*** (2.89)	−0.0060 (−0.95)
P3 − P1	0.0094 (1.45)	0.0109** (2.12)	0.0165*** (2.77)	0.0070* (1.68)
Panel C: INVNAF
Analysts’ forecast revision	U1	U2	U3	U3 − U1
P1 (REV < 0)	0.0041 (1.19)	−0.0001 (−0.02)	−0.0014 (−0.32)	−0.0055 (−0.97)
P2 (REV = 0)	0.0076** (2.09)	0.0048 (1.05)	0.0158** (2.03)	0.0082 (0.95)
P3 (REV > 0)	0.0092*** (2.84)	0.0115*** (2.94)	0.0230*** (4.76)	0.0139*** (2.39)
P3 − P1	0.0051 (1.08)	0.0116** (2.03)	0.0245*** (3.69)	0.0194*** (4.13)
Panel D: DISP
Analysts’ forecast revision	U1	U2	U3	U3 − U1
P1 (REV < 0)	0.0014 (0.41)	0.0012 (0.32)	0.0004 (0.07)	−0.0010 (−0.17)
P2 (REV = 0)	0.0047 (1.46)	0.0087** (2.12)	0.0051 (0.91)	0.0003 (0.05)
P3 (REV > 0)	0.0108*** (3.49)	0.0128*** (3.63)	0.0166*** (3.32)	0.0058 (0.98)
P3 − P1	0.0094** (2.10)	0.0116** (2.26)	0.0162** (2.30)	0.0068 (1.63)
Panel E: ABSE
Analysts’ forecast revision	U1	U2	U3	U3 − U1
P1 (REV < 0)	0.0017 (0.42)	0.0029 (0.76)	−0.0023 (−0.51)	−0.0040 (−0.66)
P2 (REV = 0)	0.0146* (1.92)	0.0097** (2.29)	0.0183 (1.30)	0.0037 (0.59)
P3 (REV > 0)	0.0117*** (2.86)	0.0139*** (3.48)	0.0184*** (4.43)	0.0067 (1.14)
P3 − P1	0.0099 (1.71)	0.0110** (1.97)	0.0207*** (3.38)	0.0107*** (2.61)
Panel F: CI
Analysts’ forecast revision	U1	U2	U3	U3 − U1
P1 (REV < 0)	0.0043 (1.27)	0.0040 (1.01)	−0.0059 (−1.24)	−0.0103* (−1.74)
P2 (REV = 0)	0.0092** (2.50)	0.0091** (1.96)	0.0095* (1.66)	0.0003 (0.04)
P3 (REV > 0)	0.0095*** (2.89)	0.0121*** (3.21)	0.0205*** (3.98)	0.0110* (1.78)
P3 − P1	0.0052 (1.09)	0.0081 (1.47)	0.0264*** (3.75)	0.0213*** (4.90)

Note. This table reports average monthly portfolio returns for the period from July 1992 to June 2009 for all listed stocks in Australia. Each month, we sort stocks into three categories based on whether the forecast revision ratio is negative, zero, or positive. The forecast revision ratio is defined as the average monthly earnings forecast change in expected earnings per share as a percentage of the absolute mean value of prior consensus forecasts. For each category, we further sort stocks into three groups based on information uncertainty proxy: U1 contains lower 30% uncertainty stocks and U3 contains higher 30% uncertainty stocks. Firm size (MV) is the market capitalization (AUD in millions) at the end of month t. The B/M ratio is computed by matching the yearly BE value figure for all fiscal years ending in calendar year t to returns starting in July of year t. This figure is then divided by market capitalization at month t to form the B/M ratio, so that the B/M ratio is updated each month. Analyst coverage (NAF) is the total number of estimates covering the firms for the fiscal period. DISP is defined as the ratio of the standard deviation of analysts’ current-fiscal-year annual earnings per share forecasts to the absolute value of the mean forecast, as reported in the I/B/E/S Summary History file. ABSE are absolute prediction errors of the predict out-of-sample cash flows and the actual value. INVMV, MB, and INVNAF are the reciprocals of MV, B/M, and analyst coverage. The portfolios are equally weighted at formation and held for 1 month. Mean size is in AUD million and t statistics are reported in parentheses. B/M = book-to-market; DISP = Dispersion; BE = book equity; I/B/E/S = Institutional Brokerage Estimate System; MB = market-to-book ratio; CI = Combo Index.

, **, and *** indicate a significant level of 10%, 5%, and 1%, respectively.

Panel A of Table 5 indicates that within the positive forecast revision portfolios (P3), high-uncertainty small firms (U3) earn significantly higher returns than low-uncertainty large firms (U1), and the difference in returns between the two is significantly positive at 1.27% per month (t value = 2.12). On the contrary, neither high-uncertainty small firms nor low-uncertainty large firms produce significant results in the downward-revision group (P1); in the meantime, the variation in returns between U3 and U1 is insignificantly negative. Within each uncertainty group, we find upward forecast revision portfolios produce significant positive returns and the return spread between the upward and downward forecast revision portfolios is significantly positive as well, with the exception of the low-uncertainty group. The picture becomes clearer when returns from the zero-cost portfolios increase as the stocks uncertainty level increases. The portfolio with high information uncertainty (small stocks) can produce a monthly average return of 2.31% (t value = 3.19), while the low-uncertainty portfolio provides insignificant results.

We also analyze the impact on returns from other information uncertainty variables such as MB, INVNAF, DISP, and ABSE. The results are reported in Panels B, C, D, and E of Table 5. Those results are quite similar to the results from Panel A. Under each uncertainty group, the upward-revision portfolios produce higher significant returns and the differences in returns between up- and down-revision portfolios (P3 − P1) are all significantly positive. The returns within each uncertainty group also increase with the forecast revision ratios. For example, for the high-uncertainty group U3 in Panels B, C, D, and E, the returns in upward-revision portfolios are higher than portfolios with no revisions, and portfolios with downward revisions do generate lower returns than portfolios that had no revisions.

To examine the aggregate effect of information uncertainty on stock returns, we then constructed a joint information uncertainty proxy using the equally weighted average ranking of the five information uncertainty proxies as tested in Panels A, B, C, D, and E. Specifically, the joint measure for stock j in month t is expressed as

{C O M B O}_{t}^{j} = \frac{\sum_{i = 1}^{5} {R a n k}_{i, t}^{j}}{5},

where ${Rank}_{i, t}^{j}$ is the cross-sectional rank of information uncertainty proxy i for stock j. Stocks are ranked into three groups in each month based on ${C O M B O}_{t}^{j}$ . Firms with the lowest and highest ${C O M B O}_{t}^{j}$ measures are denoted as U1 and U3 stocks, respectively. Results in Panel E are consistent with results reported in other Panels in Table 5, indicating the level of information uncertainty can affect investors’ investment decisions. In summary, evidence from the previous section shows that positive forecast revision portfolios produce higher returns than negative ones. After combining with all uncertainty factors, the difference in the returns pattern becomes even clearer. Hence, we can argue that analysts’ forecast revisions do provide important information to investors and stock uncertainty levels do have influence on investors’ investment decisions.

Robustness Check

Characteristics of Portfolios Sorted by Forecast Revisions and Information Uncertainty

In this section, a robustness check was conducted for information uncertainty properties after controlling risk factors. Based on various forecast revisions and uncertainty levels, we use Carhart’s (1997) four-factor model to accomplish the robustness check. Table 6 reports the intercepts of the four-factor regression model for monthly excess returns of the four information uncertainty portfolios for three analyst forecast revision portfolios. The return in excess of the risk-free rate in each portfolio is regressed against four factors. We investigate performance persistence using Carhart’s four-factor model as follows:

r_{pt} = α_{p} + b_{p} R M_{t} + s_{p} S M B_{t} + h_{p} H M L_{t} + u_{p} U M D_{t} + ε_{pt},

where r_pt is the excess return in portfolio p in month t. RM_t is the excess return on the market, SMB_t is the return on the mimicking size portfolio, HML_t is the return on the mimicking B/M portfolio, and these portfolios are constructed in the same way as in Fama and French (1996). UMD_t is the return on the mimicking momentum factor and constructed as Carhart’s (1997) model. A significantly positive abnormal return (α) indicates performance persistence and vice versa for a significantly negative abnormal return.⁴

Table 6.

Risk-Adjusted Monthly Returns on Portfolios Sorted by Analysts’ Forecast Revisions and Information Uncertainty.

Analysts’ forecast revision	All		Information uncertainty
			INVMV	MB	INVNAF	DISP	ABSE	CI
P1 (REV < 0)	−0.0032 (−0.78)	U1 (low)	−0.0007 (−0.21)	0.0043 (0.83)	−0.0002 (−0.04)	−0.0035 (−0.97)	−0.0026 (−0.56)	0.0002 (0.06)
		U2	−0.0037 (−0.84)	−0.0012 (−0.29)	−0.0047 (−1.04)	−0.0025 (−0.61)	−0.0009 (−0.20)	−0.0001 (−0.04)
		U3 (high)	−0.0051 (−0.92)	−0.0106** (−2.34)	−0.0041 (−0.81)	−0.0033 (−0.61)	−0.0051 (−1.06)	−0.0097 (−1.84)
P2 (REV = 0)	0.0081 (1.61)	U1 (low)	0.0050 (0.92)	0.0596*** (3.57)	0.0022 (0.54)	−0.0010 (−0.27)	0.0074 (0.89)	0.0034 (0.85)
		U2	0.0007 (0.16)	0.0038 (0.54)	−0.0001 (−0.01)	0.0043 (0.95)	0.0060 (1.27)	0.0043 (0.85)
		U3 (high)	0.0207** (2.53)	0.0031 (0.57)	0.0217*** (2.59)	0.0004 (0.07)	0.0062*** (2.99)	0.0066 (1.05)
P3 (REV > 0)	0.0112*** (2.87)	U1 (low)	0.0047 (1.42)	0.0146*** (2.92)	0.0063* (1.78)	0.0059* (1.72)	0.0087* (1.94)	0.0051 (1.43)
		U2	0.0103** (2.53)	0.0098** (2.54)	0.0071* (1.64)	0.0091** (2.33)	0.0114*** (2.59)	0.0091** (2.19)
		U3 (high)	0.0189*** (3.34)	0.0118** (2.53)	0.0213*** (4.02)	0.0148*** (2.69)	0.0168*** (3.69)	0.0205*** (3.64)
P3 − P1	0.0145*** (2.78)	U1 (low)	0.0054* (1.76)	0.0103** (2.24)	0.0065* (1.95)	0.0094*** (2.93)	0.0113*** (3.54)	0.0050* (1.90)
		U2	0.0141*** (3.63)	0.0110*** (3.01)	0.0118*** (2.93)	0.0116*** (3.18)	0.0123*** (4.01)	0.0093*** (3.07)
		U3 (high)	0.0241*** (4.72)	0.0223*** (5.35)	0.0254*** (5.42)	0.0181*** (3.64)	0.0220*** (6.63)	0.0303*** (7.85)

Note. This table reports the intercepts of the regression model for monthly excess returns of the six information uncertainty portfolios for three analyst forecast revision portfolios. The return on each portfolio is then taken in excess of the risk-free rate and regressed against a number of factors. We investigate persistence using Carhart’s (1997) four-factor model as follows:

r_{pt} = α_{p} + b_{p} R M_{t} + s_{p} S M B_{t} + h_{p} H M L_{t} + u_{p} U M D_{t} + ε_{pt},

where r_pt is the excess return on portfolio p in month t. RM_t is the excess return on the market, SMB_t is the return on the mimicking size portfolio, HML_t is the return on the mimicking book-to-market portfolio and constructed in the same way as in Fama and French (1996). UMD_t is the return on the mimicking momentum factor and constructed as Carhart’s (1997) model. A significantly positive α indicates performance persistence and vice versa for a significantly negative α; t statistics are reported in parentheses. MB = market-to-book ratio; DISP = Dispersion; CI = Combo Index.

, **, and *** indicate a significant level of 10%, 5%, and 1%, respectively.

Table 6 provides support for our previous findings. First, we find that the abnormal returns are significantly positive (1.12% per month with t value = 2.87) in the upward-revision portfolios. Second, the upward-revision portfolios bring up a very interesting observation that the abnormal returns increase with uncertainty levels and 16 out of 18 are significant. On the contrary, for the downward-revision portfolios, the alphas decrease with the uncertainty level, and most alphas are not significant. This implies that information uncertainty does affect a portfolio performance persistency, and higher information uncertainty often worsens the performance of downward-revision portfolio in terms of alpha. Finally, similar to the return patterns in upward-revision portfolios, all of the zero-cost portfolios (P3 − P1) generate significantly positive abnormal returns, and there is an obvious pattern that alphas increase as the information uncertainty level increases. Our robustness results confirm our earlier findings that forecast revisions and uncertainty level do contain important information in determining investors’ investment decisions. An investor may underreact information when there is high degree of uncertainty in the market; therefore, this may lead to returns continuation in the next periods.

Cross-Section of Stock Returns

We further use univariate regressions and multivariate regressions to examine the cross-section of stock returns in the Australian market. The dependent variable is the individual stock return and the independent variables are those information uncertainty proxies and ERs ratio. Models 1 to 6 are used to test individual information proxy on the stock returns. The equations are as follows:

Model 1 : R_{i, t} = α_{1} + b_{1} I N V M V_{i, t - 1} + ε_{i, t},

Model 2 : R_{i, t} = α_{1} + b_{2} M B_{i, t - 1} + ε_{i, t},

Model 3 : R_{i, t} = α_{1} + b_{3} I N V N A F_{i, t - 1} + ε_{i, t},

Model 4 : R_{i, t} = α_{1} + b_{4} D I S P_{i, t - 1} + ε_{i, t},

Model 5 : R_{i, t} = α_{1} + b_{4} A B S E_{i, t - 1} + ε_{i, t},

Model 6 : R_{i, t} = α_{1} + b_{5} E R_{i, t - 1} + ε_{i, t},

\begin{matrix} Model 7 : R_{i, t} = α_{1} + b_{1} I N V M V_{i, t - 1} + b_{2} M B_{i, t - 1} + b_{3} I N V N A F_{i, t - 1} \\ + b_{4} D I S P_{i, t - 1} + b_{5} E R_{i, t - 1} + ε_{i, t}, \end{matrix}

\begin{matrix} Model 8 : R_{i, t} = α_{1} + b_{1} I N V M V_{i, t - 1} + b_{2} M B_{i, t - 1} + b_{3} INVNA F_{i, t - 1} \\ + b_{4} D I S P_{i, t - 1} + b_{5} A B S E_{i, t - 1} + b_{6} E R_{i, t - 1} + ε_{i, t} . \end{matrix}

The results of univariate regressions in Table 7 show that MB, INVNAF, and ER have significant impacts on the stock’s returns and their signs represent different directional influence toward the stock’s returns. For example, the significant positive sign of analysts’ coverage INVNAF, 0.0148 with t value 1.90, denotes that the less the stock coverage by analysts, the greater the stock’s return. This result is consistent with the findings of Hong et al. (2000) and suggests that uncertainty factors have significant effects on stock returns. The result of a negative relation between returns and MB indicates value firms outperform growth firms, which are consistent with the findings of Fama and French (1992). Table 7 also shows that Australian stock returns are positively related to the ERs (with a t value of 2.91). It implies that when one analyst upgrades a firm that firm expects to have a 122-basis point increase in its return over the next month.

Table 7.

Fama–MacBeth Regression Analysis.

	Univariate regressions						Multivariate regressions
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
INVMV	0.0429 (1.07)						−0.0035 (−0.10)	−0.0266 (−0.65)
MB		−0.0079** (−2.29)					−0.0034** (−1.99)	−0.0032 (−1.58)
INVNAF			0.0148* (1.90)				0.0133 (1.39)	0.0189* (1.70)
DISP				−0.0008 (−0.29)			0.0074 (1.11)	0.0133 (1.43)
ABSE					0.0002 (0.03)			0.0091** (2.10)
ER						0.0122*** (2.91)	0.0357** (2.29)	0.0445** (2.49)
R ²	.0219	.0268	.0152	.0176	.0206	.0129	.1449	.1648

Note. This table reports Fama–MacBeth regression coefficients (Fama & MacBeth, 1973) for all Australian stocks over the period from July 1992 to June 2009. We regress monthly stock returns over the period (t − 1, t) on the following set of explanatory variables: INVMV, MB, INVNAF, DISP, ABSE, ER ratio: $R_{i, t} = α_{1} + b_{1} I N V M V_{i, t - 1} + b_{2} M B_{i, t - 1} + b_{3} I N V N A F_{i, t - 1} + b_{4} D I S P_{i, t - 1} + b_{5} A B S E_{i, t - 1} + b_{6} E R_{i, t - 1} + ε_{i, t}$ . Newey–West adjusted t statistics are reported in parentheses. MB = market-to-book ratio; DISP = Dispersion; ER = earnings revision.

, **, and *** indicate a significant level of 10%, 5%, and 1%, respectively.

Models 7 and 8 in Table 7 present the Fama–MacBeth regression results using all stocks for the whole testing period (Fama & MacBeth, 1973). The multivariate results are similar to those of the univariate regressions, indicating that Australian stock returns are positive related to INVNAF. The multivariate results also show that absolute prediction errors of cash flows are statistically significant in predicting Australian stock returns. The result is consistent with the argument that information is integrated faster into prices for firms with less prediction errors. Moreover, the results from Model 8 in Table 7 also show that the average forecast revision ratio has a positive significant influence on stock returns, indicating that the information contains in earnings forecast revisions are important in determining cross-section returns of Australian stocks. This result conforms to our earlier findings and shows that investors’ underreaction to the information can produce the observed returns pattern.

Conclusion

This article provides empirical evidence that stock price drift emerges after the analysts’ earnings forecast revisions. Our results consistently show that stocks with favorable ERs substantially experience positive stock returns; however, stocks with downward forecast revisions do not experience negative returns. We also find that when the level of information uncertainty increases, the positive returns for upward-revision stocks increase. Moreover, the spread between upward- and downward-revision portfolios is positively significant for higher information uncertainty stocks. Our results are robust after controlling for market condition, seasonality, and risks. We argue that analysts’ forecast revisions do contain important information that can help investors make investment decisions.

Our results are consistent with the existing literature concerning both rational and behavioral explanations for the empirically observed correlation pattern in successive stock price changes. Lewellen and Shanken (2002) reveal that predictability is parameter uncertainty and argue that investors are Bayesian-rational but have imperfect information about the parameters of the distribution of fundamental value. It provides a possible explanation for our findings as investors learn about structure parameters, and it is possible to have a negative correlation in price changes even they are Bayesian-rational with updates of analysts’ earnings forecast revisions. Dontoh et al. (2003) provide an alternative explanation, documenting that rational investors have no parameter uncertainty but with experience of noise trading. Investors try to stay away from this risk (price fluctuations) by not fully acting all available information. Moreover, Barberis et al. (1998) present that stock price drift is the result of investors who update their belief in the right direction but by too little on its weight, relative to a rational Bayesian. Consistent with the investor conservatism bias model proposed by Barberis et al., we show that investors tend to underweight public information, such as analysts’ earnings forecast revisions.

We extend the literature on the relationship among analysts’ earnings forecast revisions, information uncertainty, and stock returns by providing evidence that stock price drift occurs after analysts’ earnings forecast revisions in the Australian stock market. Importantly, our evidence provides possible explanations about the violation of the efficient market hypothesis. Our results suggest that the conservative bias causes investors not to sufficiently update their beliefs and eventually results in subsequent return continuation as investors’ underreaction to analysts’ earnings forecast revision is stronger with higher information uncertainty.

Footnotes

Acknowledgements

The authors would like to thank Dr. Bharat Sarath, Editor in Chief, Journal of Accounting, Auditing and Finance, and an anonymous reviewer for their constructive comments and suggestions that have helped to improve the quality of the article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abarbanell

(1991). Do analysts’ earnings forecasts incorporate information in prior stock price changes? Journal of Accounting and Economics, 14, 147-165.

Abarbanell

Bernard

(1992). Tests of analysts’ overreaction/underreaction to earnings information as an explanation for anomalous stock price behavior. Journal of Finance, 47, 1181-1207.

Abarbanell

Bushee

(1997). Fundamental analysis, future earnings, and stock prices. Journal of Accounting Research, 35, 1-24.

Abarbanell

Bushee

(1998). Abnormal returns to a fundamental analysis strategy. The Accounting Review, 73, 19-45.

Abarbanell

Lehavy

(2003). Biased forecasts or biased earnings? The role of reported earnings in explaining apparent bias and over/underreaction in analysts’ earnings forecasts. Journal of Accounting and Economics, 36, 105-146.

Aitken

Frino

Winn

(1996). Consensus analysts’ earnings forecasts and security returns. Asia Pacific Journal of Management, 13, 101-110.

Ajayi

R. A.

Mehdian

(1994). Rational investors’ reaction to uncertainty: Evidence from the world’s major markets. Journal of Business Finance & Accounting, 21, 533-545.

Ali

Klein

Rosenfeld

(1992). Analysts’ use of information about permanent and transitory earnings components in forecasting annual EPS. The Accounting Review, 67, 183-198.

Asquith

Mikhail

(2005). Information content of equity analyst reports. Journal of Financial Economics, 75, 245-282.

10.

Athanasakou

Strong

N. C.

Walker

(2009). Earnings management or forecast guidance to meet analyst expectations? Accounting and Business Research, 39, 3-35.

11.

Barberis

Shleifer

Vishny

(1998). A model of investor sentiment. Journal of Financial Economics, 49, 307-343.

12.

Beaver

Cornell

Landsman

W. R.

Stubben

S. R.

(2008). The impact of analysts’ forecast errors and forecast revisions on stock prices. Journal of Business Finance & Accounting, 35, 709-740.

13.

Brav

Lehavy

(2003). An empirical analysis of analysts’ target prices: Short-term informativeness and long-term dynamics. Journal of Finance, 58, 1933-1967.

14.

Bray

(1981). Futures trading, rational expectations, and the efficient markets hypothesis. Econometrica, 49, 575-596.

15.

Carhart

(1997). On persistence in mutual fund performance. Journal of Finance, 52, 57-82.

16.

Chan

L. K. C.

Jegadeesh

Lakonishok

(1996). Momentum strategies. Journal of Finance, 51, 1681-1713.

17.

Chan

L. K. C.

Karceski

Lakonishok

(1998). The risk and return from factors. Journal of Financial and Quantitative Analysis, 33, 159-188.

18.

Chan

L. K. C.

Narasimhan

Lakonishok

(1995). Evaluating the performance of value versus glamour stocks: The impact of selection bias. Journal of Financial Economics, 38, 269-296.

19.

Chava

Kumar

Warga

(2010). Managerial agency and bond covenants. Review of Financial Studies, 23, 1120-1148.

20.

Claus

Thomas

(2001). Equity premia as low as three percent? Evidence from analysts’ earnings forecasts for domestic and international stock markets. Journal of Finance, 56, 1629-1666.

21.

Clement

M. B.

Hales

Xue

(2011). Understanding analysts’ use of stock returns and other analysts’ revisions when forecasting earnings. Journal of Accounting and Economics, 51, 279-299.

22.

Daniel

Hirshleifer

Subrahmanyam

(1998). Investor psychology and security market over- and under-reactions. Journal of Finance, 53, 1839-1886.

23.

Daniel

Hirshleifer

Subrahmanyam

(2001). Overconfidence, arbitrage, and equilibrium asset pricing. Journal of Finance, 56, 921-965.

24.

Davis

J. L.

(1994). The cross-section of realized stock returns: The pre-COMPUSTAT evidence. Journal of Finance, 49, 1579-1593.

25.

Dontoh

Ronen

Sarath

(2003). On the rationality of post-earnings announcement drift. Review of Accounting Studies, 8, 69-104.

26.

Easterwood

Nutt

(1999). Inefficiency in analysts’ earnings forecasts: Systematic misreaction or systematic optimism? Journal of Finance, 54, 1777-1797.

27.

Easton

(2004). PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital. The Accounting Review, 79, 73-95.

28.

Elton

E. J.

Gruber

M. J.

Gultekin

(1981). Expectations and share prices. Management Science, 27, 975-987.

29.

Fama

E. F.

French

K. R.

(1992). The cross-section of expected stock returns. Journal of Finance, 47, 427-465.

30.

Fama

E. F.

French

K. R.

(1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33, 3-56.

31.

Fama

E. F.

French

K. R.

(1996). Multifactor explanations of asset pricing anomalies. Journal of Finance, 51, 55-84.

32.

Fama

E. F.

MacBeth

(1973). Risk, return and equilibrium: Empirical tests. Journal of Political Economy, 81, 607-636.

33.

Francis

Soffer

(1997). The relative informativeness of analysts’ stock recommendations and earnings forecast revisions. Journal of Accounting Research, 35, 193-211.

34.

Frankel

Lee

C. M. C.

(1998). Accounting valuation, market expectation, and cross-sectional stock returns. Journal of Accounting and Economics, 25, 283-319.

35.

Givoly

Lakonishok

(1979). The information content of financial analysts’ forecasts of earnings. Journal of Accounting and Economics, 2, 165-185.

36.

Givoly

Lakonishok

(1980). Financial analysts’ forecast of earnings: The value to investors. Journal of Banking and Finance, 4, 221-233.

37.

Gleason

Lee

(2003). Analyst forecast revisions and market price discovery. The Accounting Review, 78, 193-225.

38.

Griffin

(1976). Competitive information in the stock market: An empirical study of earnings, dividends and analysts’ forecasts. Journal of Finance, 31, 631-650.

39.

Xue

(2007). Do analysts overreact to extreme good news in earnings? Review of Quantitative Finance and Accounting, 29, 415-431.

40.

Guan

Wong

M. H. F.

(2012). Conflict-of-interest reforms and investment bank analysts’ research biases. Journal of Accounting, Auditing & Finance, 27, 443-470.

41.

Hirshleifer

(2001). Investor psychology and asset pricing. Journal of Finance, 56, 1533-1596.

42.

Hong

Lim

Stein

(2000). Bad news travels slowly: Size, analyst coverage, and the profitability of momentum strategies. Journal of Finance, 55, 265-295.

43.

Hong

Stein

(2003). Differences of opinion, rational arbitrage and market crashes. The Review of Financial Studies, 16, 487-525.

44.

Imhoff

Lobo

(1984). Information content of analysts’ composite forecast revisions. Journal of Accounting Research, 22, 541-554.

45.

Jegadeesh

Titman

(2001). Profitability of momentum strategies: An evaluation of alternative explanations. Journal of Finance, 56, 699-720.

46.

Jung

Sun

K. J.

Yang

Y. S.

(2012). Do financial analysts add value by facilitating more effective monitoring of firms’ activities? Journal of Accounting, Auditing & Finance, 27, 61-99.

47.

Kasznik

McNichols

M. F.

(2002). Does meeting earnings expectations matter? Evidence from analyst forecast revisions and share prices. Journal of Accounting Research, 40, 727-759.

48.

Lev

Sougiannis

(2010). The usefulness of accounting estimates for predicting cash flows and earnings. Review of Accounting Studies, 15, 779-807.

49.

Lewellen

Shanken

(2002). Learning, asset pricing, and market efficiency. Journal of Finance, 57, 1113-1145.

50.

Lim

T. K.

Kong

H. C.

(2004). New evidence on price impact of analyst forecast revisions. International Review of Financial Analysis, 13, 161-190.

51.

Loughran

(1997). Book-to-market across firm size, exchange, and seasonality: Is there an effect? Journal of Financial and Quantitative Analysis, 32, 249-268.

52.

Lys

Sohn

(1990). The association between revisions of financial analysts’ earnings forecast and security-price changes. Journal of Accounting and Economics, 13, 341-363.

53.

Moller

Zilca

(2008). The evolution of the January effect. Journal of Banking and Finance, 32, 447-457.

54.

Nam

Brochet

Ronen

(2012). The predictive value of accruals and consequences for market anomalies. Journal of Accounting, Auditing & Finance, 27, 151-176.

55.

Pettengill

G. N.

Sundaram

Mathur

(1995). The conditional relation between beta and return. Journal of Financial and Quantitative Analysis, 30, 101-116.

56.

Ramnath

Rock

Shane

(2005). Value line and I/B/E/S earnings forecasts. International Journal of Forecasting, 21, 185-198.

57.

Stickel

(1991). Common stock return surrounding earnings forecast revisions: More puzzling evidence. The Accounting Review, 66, 402-416.

58.

Stickel

(1995). The anatomy of the performance of buy and sell recommendations. Financial Analysts Journal, 51, 25-39.

59.

van Dijk

M. A.

(2011). Is size dead? A review of the size effect in equity returns. Journal of Banking and Finance, 35, 3263-3274.

60.

Wieland

M. M.

(2011). Identifying consensus analysts’ earnings forecasts that correctly and incorrectly predict an earnings increase. Journal of Business Finance & Accounting, 38, 574-600.

61.

Zhang

X. F.

(2006a). Information uncertainty and analyst forecast behavior. Contemporary Accounting Research, 23, 565-590.

62.

Zhang

X. F.

(2006b). Information uncertainty and stock returns. Journal of Finance, 61, 105-137.