This dissertation has been submitted by a student. This is not an example of the work written by our professional dissertation writers.

Behavior Around Earnings Announcement Events For Emerging Markets


1.1 Background of Research Question

Stock prices show a tendency to behave in a manner not consistent with what current finance theory proposes or expects. This gap may be the result of flawed assumptions presently used as a basis in existing theory. Conventional modern finance theory rests on the assumption of investor rationality. Efficient market hypothesis (EMH), representing an integral part of conventional finance theory, further assumes security prices reflect (to varying degrees) all available information as processed by rational investors. Modern portfolio theory (MPT) suggests [1] , a stock’s price represents the present value of future expected cash flows and therefore, is dependant on investor’s expectations of estimated forecasts of earnings growth rates into the future. Actual short-term stock price behavior indicates anomalies and significant divergences of prices away from fundamental intrinsic values, long-term averages, or expectations as implied by MPT. Behavior finance takes the approach of understanding these price anomalies through a study of cognitive and emotional biases. This thesis investigates the possibility of investor’s making mistakes (through irrational behavior) in their forward expectations of future corporate cash flows, resulting in short-term overreaction to earnings information releases due to influence of representative bias (a cognitive bias).

This proposal attempts to examine the existence, relationship and, impact of overreaction as a determinant of securities price behavior in emerging markets of Far East. The aim of this study is to discover representative bias, a tendency of investors to overweigh most recent information in making future forecasts, as one possible cause leading to overreaction in securities prices. This thesis tests investor responses to corporate earnings announcements, specifically surprises, to determine overreaction behavior and to identify representative bias as the cause of such overreaction. The results may contribute, by offering a missing piece of the puzzle, of understanding stock price behavior (towards the search for a unified theory), into existing research work for behavior finance. In addition, a better understanding of what drives stock prices would be a highly useful forecasting and policy tool for participants concerned with asset pricing.

1.2 Motivation for Research

The field of behavior finance focuses on the question; what drives investor behavior?. It is divided into two main groups. Cognitive and emotional biases, which are further sub-divided into two sub-groups, individual and collective biases. Behavior finance has been seeking to discover the causes of investor irrationality within the investment decision-making framework. Significant empirical and theoretical studies have been conducted, which suggest cognitive and emotional biases affect investor rationality. Indeed, the field of behavior finance directly challenges the conventional finance framework, which uses within its paradigm the assumption that investors are rational decision makers, and securities prices reflect all available information (EMH [4] .

According to Fung (2006), it seems clear that EMH and CAPM [5] (pillars of the current financial theory), despite mounting evidence against their validity, remain widely in use. One reason for this is the fact that these models cannot be empirically falsified due to their dependence on layers of assumptions, which support each other. The other reason is a lack of an alternative asset-pricing model taking cognitive biases into consideration, which does not exist so far. Behavior finance offers just such an alternative, and after observing price anomalies as a trader in the financial markets, I have become interested in pursuing empirical work in this direction to understand and discover a better way to price financial assets. Furthermore, only limited research exists for capital markets covered by this thesis namely Malaysia, Thailand, and Singapore, despite the fact these markets have outperformed western markets recently and continue to offer potential for future growth.

1.3 Statement of Problem

This thesis focuses on the following problem:

“What is the relationship between representative bias and overreaction, as it relates to individual as well as a series of earnings surprise announcements, on investor behavior in the stock markets of Far East? [6] ?

1.4 Research Questions & Objectives

This proposal consists of three components. First, the study aims to discover existence of investor overreaction (derived and tested from stock price behavior) in the stock markets of Malaysia, Thailand, and Singapore based on overreaction hypothesis (ORH) as proposed by Thaler (1985). Thereafter, the second objective of this thesis is to test for overreaction in response to corporate earnings surprises (positive and negative). Third and last objective of the study is to determine if representative bias (cognitive bias) is a source of this investor irrationality, in response to earnings surprises, as demonstrated through price behavior in the respective stock markets.

In other words, to contribute an answer to the key question behavior finance is seeking; what drives investor behavior in the stock markets?, this study tests emerging stock markets of far east for investor overreaction. Subsequently, this thesis focuses on representative bias as one cognitive attribute of investor behavior, which may cause overreaction to occur. Following are some of the research questions this study will attempt to answer.

Research Questions:

This study attempts to answer the following research questions:

Research Question 1: Does investor overreaction exist in emerging stock markets?

Research Question 2: Do investors overreact to positive earnings surprises?

Research Question 3: Do investors overreact to negative earnings surprises?

Research Question 4: Is representative bias present during investor overreaction to earnings surprises?

Research Question 5: Does representative bias cause investor overreaction when earnings surprises are positive?

Research Question 6: Does representative bias cause investor overreaction when earnings surprises are negative?

Research Question 7: What is the relationship between representative bias and overreaction as it relates to a series of earnings surprise announcements, both positive and negative?

1.5 Research Significance

The motivation behind this research is to contribute a better understanding of determinants of stock pricing in the context of investor decision making. The results of this study will be useful in furthering current empirical research on cognitive biases affecting stock price behavior, specifically investor overreaction and representative heuristic, (as it relates to earnings surprises), as well as provide useful understanding for investment decision making, forecasting, and policy making for financial market participants in the asset management field. In addition, contribution towards a piece of the stock-pricing puzzle, as well as further research questions may also be discovered.


2.1 Cognitive Biases & Overreaction Hypothesis (ORH)

Behavior finance seeks to understand effects of psychology on financial behavior. However, a unified model, which can replace the conventional mainstream models such as the efficient market hypothesis and CAPM (capital asset pricing model), has yet to be discovered, Fung (2006). Keynes (1973, original publication 1936) wrote in his General Theory:

“ Day-to-day fluctuations in the profits of existing investments... tend to have an altogether excessive, and even absurd, influence on the market? (1973, pp. 153-154).

This became the starting point for a study on overreaction by De Bondt and Thaler (Fung 2006, p.29). Keynes remark infact implied the possibility of systematic mispricing of securities by investors. Indeed, this argument led to directly challenging the efficient market hypothesis, a mainstream concept used for financial asset pricing invented by Fama (1970). Fama created an empirically testable model to price securities based on this concept called the “fair game? which implied the notion that markets cannot have expected returns in excess of equilibrium expected returns (ibid., p.385). This particular notion is questioned in De Bondt and Thaler’s, 1985 paper, proposing instead the stock market overreaction hypothesis (ORH). They suggest, stock prices fluctuate from their intrinsic values (PV of expected future cash flows of the firm) due to optimism and pessimism prevailing amongst investors. In addition, De Bondt and Thaler (1985) suggested two other hypotheses:

1. “Extreme movements in stock prices will be followed by subsequent price movements in the opposite direction?

2. “The more extreme the initial price movement, the greater will be the subsequent adjustment. Both hypotheses imply a violation of weak form market efficiency? (1985 p.795).

In other words, their hypothesis suggested the existence of the possibility of earning excess returns above equilibrium returns. This may be accomplished by investing in stocks, which have performed poorly relative to the average (a new contrarian strategy).

In addition to casting a doubt on the EMH, De Bondt and Thaler’s paper (1985) also brought into question the validity of the CAPM (Fung, 2006). According to CAPM, an assets return is a function of the asset risk premium and the risk free rate. In other words, higher the systematic risk (Beta) of an asset, higher the assets return (represented by a linear relationship between risk and return). However, the findings of De Bondt and Thaler discovered low Beta (low systematic risk) portfolios (L) generating higher returns and high beta portfolios (W) producing low returns. This result contradicts the basic risk-return relationship as proposed by CAPM. Fama and French in their 1992 paper were the first to confirm this CAPM contradiction through empirical study based on discovering the (positive) relationship between size and beta. They suggested CAPM did not fully measure and adjust for the higher risk of smaller(size) firms.

In summary, empirical research has discovered that consistent anomalies exist between risk-return relationship as proposed by the widely used EMH and CAPM (the traditional financial theory paradigm). However, no comprehensive theory exists to explain why some stocks do better than others. Lakonishok et al. (1994) confirm in their paper that value stocks outperform glamour stocks. Thereby they suggest cognitive bias as a possible explanation for this anomaly.

“Putting excessive weight on recent past history, as opposed to a rational prior, is a common judgment error in psychological experiments and not just in the stock market? (ibid., p.1575).

It seems clear that EMH and CAPM (pillars of the current financial theory), despite mounting evidence against their validity, remain widely in use. Behavior finance seems to offer an alternative to the current financial theory, and therefore it is imperative to pursue empirical work in this direction in order to understand and discover a better way to price financial assets.

2.2 Investor Overreaction & Earnings Surprises

Research in Behavior Finance has taken different approaches to discovering cognitive biases and their impact on asset pricing. Odean (1998) and Daniel, Hirshleifer and Subrahmanyam (1998) focused on overconfidence, while Hong and Stein (1999) investigated mispricing by positive feedback trading.

De Bondt and Thaler (1985) have empirically investigated overreaction, and concluded investors overreact to information. Their research focused on NYSE stocks and cumulative returns for three years (event window). They concluded specifically that stocks with previous abnormally low returns performed better than those with previous abnormally high returns. This return reversal indicated investor reaction to be over-weighted in response to information, which was later corrected in prices over the longer term.

The overreaction anomaly has been empirically established in finance through multiple studies. Kaestner (2006) points out that Poteshman (2001) tests overreaction in the options market; Cutler, Poterba, and Summers (1991) for gold market; Chui, Titman and Wei, (2000) and Bhojraj and Swaminathan (2001) for international stock markets. In addition Chopra, Lakonishok, and Riter (1992) as well as De Bondt and Thaler (1987) have confirmed overreaction in stock markets.

The main gap, which remains in empirical research, is establishing the determinant or driver of overreaction (Kaestner 2006, pp.3). This particular cognitive bias has recently been tested as a driver of overreaction in a few markets only.

Kaestner (2006) further points out that Poteshman (2001) has researched for representative bias and overreaction in the options markets through investor responses to changes in variance of the underlying asset. De Bondt and Thaler, (1985 and 1987), Chopra, Lakonishok, and Ritter, (1992) look at representative bias as the potential driver of overreaction basing their studies on current earnings, forecasted changes in earnings and past performance without directly testing for it. Kaestner (2006) directly tests for this link between representative bias and overreaction for NYSE stocks related to earnings surprises.

2.3 The Representative Heuristic

Representative heuristic belongs to the family of cognitive biases within the field of behavior finance. Tversky and Kahneman (1974) suggest this bias may affect an investor’s decision-making framework by causing the investor to over weight most recent fundamental information regarding a stock while making estimations of future earnings forecasts. Because of representative bias, investors may over/under estimate a stock’s intrinsic value [7] resulting in subsequent anomalous security price behavior. According to Tversky and Kahneman (1974), representative bias involves estimating:

“... the probability of an uncertain event, or a sample, by the degree to which it is similar in its essential properties to the parents’ population...? (ibid.).

In other words, investors place too much weight on recent small sample datasets (law of small numbers) when projecting future cash flows for a stock. Kaestner (2006) suggests that when a series of such recent earnings data (in the same direction) is presented to the investor, this series is interpreted as a pattern and thereby becomes the basis for future projections of a stock’s performance. Such projection thus leads to over/under estimation of future projected probability distributions of expected price performance. Kaestner (2006) argues that if representative bias affects investors then evidence of two related phenomenon would be present.

“First, statistical results would indicate a market’s overreaction to some disclosed information. Second, the overreaction will be increasing in the extent to which the series of similar information is long? (ibid., pp.10).

Although De Bondt and Thaler (1987 and 1990) only tested for overreaction in their paper, they do mention the representative bias as a possible reason. Poteshman (2001) tests for representative bias and relates it to overreaction in the options markets. However, Kaestner (2006) for the first time directly tests the representative heuristic and links it to overreaction in the U.S. stock markets.


3.1 Methodology Overview

The literature has identified a variety of research methodologies, used to test for cognitive biases. This proposal extends the methodology of testing for overreaction used by De Bondt and Thaler (1985) for U.S markets, to Asian markets. Subsequently, the main focus of this study, “representative bias? as a driver of investor overreaction in response to earnings surprises is tested based on Kaestner’s paper (2006) which also only explored the US stock markets.

De Bondt and Thaler (1985) tested the overreaction hypothesis by constructing winner(W) and loser (L) portfolios of NYSE stocks. These were selected based on past three year’s residual stock performance, which was, defined as monthly returns minus monthly market returns. The hypothesis test was run for the period January 1926 through December 1982. Positive results constituted the (W) and negative results the (L) portfolios. The two residual return results of prior and post three-year event periods (total of six years) were tested using the cumulative residual returns (CRS), a sum of the post formation monthly residual returns over three years. CRS was calculated for all sets of portfolios from 1930-1982 and each portfolio’s component residual returns were averaged to determine the cumulative average residual returns (CARS). Average of the CARS was calculated as the next step for all (L) and (W) sets of portfolios. CAPM was used in determining the market portfolio return. Overreaction hypothesis suggested that (L) portfolios would perform better than market and vice versa. Results supported the hypothesis. This finding directly challenged the EMH implicit assumption that arbitrage gaps are filled rapidly by rational investors, since markets find an equilibrium price where excess returns are not possible and occur only as a result of luck. In essence, De Bondt and Thaler (1985) discovered the possibility of earning better than market returns.

3.2 Sample Data & Characteristics

Data for stocks will be acquired for Malaysia, Thailand, and Singapore stock exchanges directly, and/or from other information providers for the period 1990-2008. A total of eighteen years will cover three event windows and two economic cycles. This would also include the 1990 U.S. recession, 1997 Asian crises, and current 2008 global market fallout data. Daily stock prices in the form of open, high, low, and close would be used to determine daily returns and same data set would be required for the market index to get a proxy for market returns. In addition, quarterly EPS (earnings per share), earnings announcements by the company, and earnings consensus estimates of analysts would be gathered. For time line analysis, EPS estimates and actual EPS values will need to be obtained for preceding 4 quarters. To adjust for size, company capitalization data (shares x stock price) will be recorded at the beginning of each year in the sample set through the company’s financial statements. In addition, Beta estimates for all companies would also be required for the sample period. Most of this data is available in digital form through research sources including online internet subscriptions and the stock exchange itself. Stock selection criteria are detailed in the following sections.

3.3 Dependant and Independent Variables

Dependant variable is the excess return derived as the difference between stock raw daily return and market daily return as an indicator of investor reaction. Independent variable is the data of actual earnings surprises in relation to analyst’s estimates.

3.4 Overreaction Test

Overreaction will be tested using sort-ranking procedure, as used by De Bondt and Thaler (1985) and repeated by Kaestner (2006) based on past stock return performance. Stocks will be ranked according to their past performance over three year event windows as a proxy for prior information and will be included into portfolios based on this performance. Post-formation performance will be assessed to test the ORH.

3.5 Portfolio Construction

Portfolios of best performing (W) and worst performing (L) stocks will be constructed using past performance as a criteria. Each stock’s historical monthly closing prices will be used to determine past performance over three-year event windows. The monthly return (Rist ) for stock i, based on its monthly closing stock price minus the monthly market return (Rmt ) derived from the closing market index price (for the same period) will generate the excess return (ARiet ) for stock i.

ARiet = Rist - Rmt

This excess return (ARiet) will be defined as a performance measure for stock i, at time t, where the “e? represents excess and “s? represents stock specific return. Positive excess return (+ARiet) stocks would indicate best performers (W) while negative excess return (-ARiet) stocks would become part of the worst performers (L). The next step will calculate the cumulative excess (abnormal return denoted by AR) returns (CARi(p,q)) by summing the monthly returns for the 36 month formation event window for all stocks.


CARi(p,q) = ∑ ARiet


p = time at beginning of event window (t=0)

q = time at end of event window (t=3 yrs)

Thereafter, stocks will be ranked from high to low, based on their cumulative excess returns of previous three years (formation period), and subsequently be added to either the best performers (W) portfolio or the worst performers (L) portfolio. To provide for progressive multiple tests, this portfolio formation process will be repeated for all non-overlapping event windows covering the entire test period with new portfolios formed every 3 years to be tested against their performance in the post-formation 3-year periods. Overreaction hypothesis suggests that (L) stocks should perform better than the market in the subsequent three-year post-formation event window as compared with the previous three years, whereas the reverse should be true for (W) stocks. This implies (L) stocks should have positive excess returns (+ARiet) while (W) stocks should generate negative excess returns (-ARiet) in the post event three-year period. A total of six years data will be required for the stocks under consideration. For the market return (Rmt), an equally weighted monthly arithmetic average of stock returns in the sample will be used as proxy for market return (De Bondt and Thaler 1985).

For each (W) and (L) portfolio, cumulative excess return (CARi(p,q)) for three-year post-formation event window will be calculated and repeated for all sets of portfolios. Finally, a mean of all member stock CARi(p,q)’s in each portfolio will be computed and referred to as MCARi(p,t). Thus, two MCARi(p,t)’s will be obtained for each of these portfolios during the formation period of 36 months, and this will be repeated for all progressive portfolios over different event windows of the entire test period.

In summary, CARi(p,q) and MCARi(p,t) will be calculated for the post-formation 3-year event window to test with formation event portfolios. If overreaction exists, then it is expected worst performing portfolios (L) should generate positive excess returns and vice versa for the best performing portfolios (W), i.e. negative excess returns. This return reversal would imply initial prices had overshot rational values (investor overreaction) as a short-term reaction and therefore adjusted back to rational values subsequently.


W-MCARi(b,t) = best performer portfolios

L-MCARi(w,t) = worst performer portfolios

The overreaction hypothesis (De Bondt and Thaler, 1985) expects post-formation results as follows:

W-MCARi(b,t) < 0

L-MCARi(w,t) > 0

Such that,

L-MCARi(w,t) – W-MCARi(b,t) > 0

3.6 Overreaction Due to Earnings Surprise Events Test

Overreaction in response to earnings surprises will be tested based on sort-rank for single event and selection-ranking event method, for series analysis, as used by Kaestner (2006). The study will use portfolio-study approach as proposed by Ball and Brown (1968). Portfolios will be constructed based on stock earnings announcement events, ranked on highest to lowest surprises, for stocks in the sample. Highest earnings positive surprise stocks will be added to portfolio (+Ĥp ) and highest negative surprise stocks will constitute portfolio (-Ĥn ).

Standardized Earnings Surprise criteria

Quarterly earnings surprises will be computed based on standardized surprise earnings (ŞSĖq) represented by the difference between the actual released earnings (AĖq) and the one month prior to announcement consensus estimate for earnings (EĖq), scaled by the standard deviation of the individual estimates for each stock (σest-q). Therefore, quarterly standardized surprise would be as follows:

Standardized Surprise measurement:

ŞSĖq = (AĖq - EĖq) / σest-q

Excess Returns Measurement

Excess returns (ARiet) for each stock will be computed on daily basis using closing stock prices as described in section 3.5. These returns will use size-adjusted approach [8] (Kaestner 2006) and will be calculated as the difference between stock daily return (Rist ) and the equally weighted daily return of the stock’s own portfolio (Rmt ). Thus excess return will be, as explained earlier in section 3.5:

ARiet = Rist - Rmt

Thereafter, post-formation cumulative excess returns (CARi(p,q) ) will be derived based on event windows of 0 (0 represents the announcement date) to 1, 3, 10, 30 and 60 days for each portfolio Ĥp and Ĥn . The cumulative excess return for an event window will be:


CARi(p,q) = ∑ ARiet


where; q= 1,3,10,30 or 60 day post-formation event window

Overreaction to earning surprises expects stock prices to overshoot in the direction of the surprise, represented as follows:

Positive cumulative excess returns for positive surprise portfolios and vice versa.

CARi(p,q) > 0 for +Ĥp

CARi(p,q) < 0 for -Ĥn

Such that,

Positive surprises (+ŞSĖq >0): CARi(p,q) > 0

Negative surprises(-ŞSĖq <0): CARi(p,q) < 0

Null Surprises ( ŞSĖq = 0): CARi(p,q) = 0

3.7 Representative Bias a Driver of Overreaction to Earnings Surprises

According to Kaestner (2006), two expected phenomenon should be present, if representative bias plays a role in investors’ overreaction to earnings surprises. First, a confirmation of overreaction to released information events, as tested earlier in this thesis, must exist. Second, this overreaction phenomenon should be increasing in relation to a similar series of earnings surprises over consecutive event windows [10] of future expected cash flows). This tendency to misjudge future prospects of a company causes investors to overreact in response to earnings surprises (both positive and negative), and would be reflected through securities prices. As a company’s actual earnings information is released, investors would have to readjust their expectations (if they had overreacted) of cash flows based on the new real (actual) information and therefore the stock price must correct itself over time, if overreaction had occurred in the first instance.

Representative bias causing overreaction due to earnings surprises will be tested by measuring investor reaction to a series of same sign surprises (Kaestner 2006). A non-parametric significance test will be used as proposed by Foster, Olsen, and Shevlin (1984) and later reviewed by Lyon, Barber, and Tsai (1999) [11] . This test relaxes the assumption of normality, constant variance of security returns over time, and cross-sectional independence in residuals (Kaestner 2006). The test focuses on establishing a companion empirical sample distribution, and then comparing its cumulative excess return (CARi(p,q)) with the observed CARi(p,q) to assess for statistical significance. Empirical distribution is generated by randomly selecting one event in the parent population for each event, computing equally weighted CARi(p,q)’s for the companion sample, and ranking the companion sample CARi(p,q)’s from highest to lowest based on a repetition of the first two steps 2,500 times in order to obtain the empirical distribution.

If representative bias exists and investors rely heavily on most recent earnings surprise information, then in the event of positive surprises, they will project overly optimistic future surprises and overreact immediately after the event on the positive side, subsequently reversing the price direction upon next earnings release (should the actual surprise not meet their expected higher surprise, (ŞSĖq =0).

To test a series of events of earnings surprises, the sequential sort-ranking procedure will be used (Kaestner 2006). The initial portfolio is constructed for all individual null-surprise event stocks (ŞSĖq,t=0) and then ranked based on their most recent surprise events (ŞSĖq,t-1) relative to this null-surprise, as compared with the stock’s last surprise (t-1). In fact all subsequent sorting will be in relation to a starting most recent null-surprise (ŞSĖq =0), so that post null-surprise period can be studied. These rankings are used to construct three equal sized portfolios labeled as positive, neutral, and negative (ŞSĖq,t-1) surprise portfolios. These three portfolios are then each sub-divided into three more portfolios based on rankings of stock surprises at (ŞSĖq,t-2) which means 2 quarters behind in a series. Similarly, these are sub-divided into three more portfolios at t-3, and t-4, giving rise to a total of 5 quarters backwards surprise series-related portfolios. In this way, series of similar earnings surprises in sequence are separated for testing while keeping the current surprise at zero (null). For each family of series of portfolios (including one null surprise) which gives rise to different number of stocks in the sample, CARi(p,q)’s are computed for different event windows (i.e 0;1, 0;3, 0;30 and 0;60). This will allow for a study of the impact of similar preceding surprises on overreaction to the most recent earnings surprise. Portfolios will be tested for investor reactions for four event windows. Representativeness hypothesis (Kaestner 2006), suggests investors will overreact initially to a surprise due to overweighting of this information in future projections. And in the case of a series of such surprises, the overweighing of future expectation will be stronger, giving rise to greater overreaction. However, when the actual earnings number results in no surprise, investors readjust their view in the opposite direction, causing a correction to the initial overreaction. This hypothesis suggests then, that a correction will ensue post the current null-surprise event causing CARi(p,q)’s to be of the opposite sign of the surprise and also, this reversal should be stronger for longer series of such similar surprises as opposed to shorter series, implying that investors overreaction is affected by the representative bias. Thus, for positive surprises, CARi(p,q)’s should be negative in relation to the empirical distribution generated by the null-surprise sample portfolio and positive for the negative surprise portfolios. For the empirical distribution sample and null-past surprise series, the CARi(p,q) should be = 0.

3.8 Empirical Results & Statistical Significance

Representativeness hypothesis expects:

Non-surprise portfolios:

Where : (ŞSĖq = 0)

CARi(p,q) = 0

(no significant market reaction)

For positive surprise portfolio:

(ŞSĖq > 0)

CARi(p,q) < 0

(negative subsequent post event correction beyond null-surprise)

For negative surprise portfolio:

(ŞSĖq < 0)

CARi(p,q) > 0

(positive subsequent post event correction beyond null-surprise)

Such that:

For positive surprise portfolio:

(ŞSĖq > 0)

CARi(p,q-3) < CARi(p,q-2) < CARi(p,q-1) < CARi(p,q)

For negative surprise portfolio:

(ŞSĖq < 0)

CARi(p,q-3) > CARi(p,q-2) > CARi(p,q-1) > CARi(p,q)

3.9 Summary of Research Objectives

Note: details of methodology appear in relevant sections above. This section indicates only a summary of already discussed objectives for ease of reading.

Objective 1: Does investor overreaction exist in emerging stock markets?

W-MCARi(b,t) < 0

L-MCARi(w,t) > 0

Such that,

L-MCARi(w,t) – W-MCARi(b,t) > 0

Objective 2: Do investors overreact to earnings surprises?

Null Surprises:

(ŞSĖq =0) q

CARi(p,q) = ∑ ARiet = 0


Objective 3: Do investors overreact to positive/negative earnings surprises?

Positive surprises:

(+ŞSĖq > 0) q

CARi(p,q) = ∑ ARiet > 0


Negative surprises:

(-ŞSĖq < 0) q

CARi(p,q) = ∑ ARiet < 0


Objective 4: What is the relationship between representative bias and over- reaction as it relates to earnings surprises? Is representative bias present during overreaction?

Non-surprise portfolios:

(ŞSĖq = 0)


CARi(p,q) = ∑ ARiet = 0


(No significant market reaction)

Objective 5: Does representative bias cause overreaction when earnings surprises are positive?

For positive surprise portfolio:

(ŞSĖq > 0)


CARi(p,q) = ∑ ARiet < 0


(Negative subsequent post event correction beyond null-surprise)

Objective 6: Does representative bias cause overreaction when earnings surprises are negative?

For negative surprise portfolio:

(ŞSĖq < 0)


CARi(p,q) = ∑ ARiet > 0


(Positive subsequent post event correction beyond null-surprise)

Objective 7: What is the relationship between representative bias and overreaction as it relates to a series of earnings surprise announcements, both positive and negative?

For positive surprise portfolio:

(ŞSĖq > 0)

CARi(p,q-3) < CARi(p,q-2) < CARi(p,q-1) < CARi(p,q)

For negative surprise portfolio:

(ŞSĖq < 0)

CARi(p,q-3) > CARi(p,q-2) > CARi(p,q-1) > CARi(p,q)


The contribution this research would make is extending current empirical research on cognitive biases affecting stock price behavior, and testing for investor irrationality, specifically investor overreaction and representative heuristic, as it relates to earnings surprises. This study may also provide further understanding of the drivers of stock prices for investment decision-making, forecasting, and policy making relevant to financial market participants. In addition, contribution towards a piece of the stock pricing puzzle, as well as further research questions may be discovered. It seems clear that EMH and CAPM (pillars of the current financial theory), despite mounting evidence against their validity, remain widely in use. Behavior finance seems to offer an alternative to the current financial theory, and therefore latest empirical research in asset pricing, is towards this direction in order to understand and discover a better way to price financial assets.

This thesis expects, based on Thaler’s (1985) ORH, investor overreaction in the emerging markets (Malaysia, Thailand, and Singapore) as demonstrated through securities prices to be confirmed through empirical testing. In addition, this study expects to identify representative bias as a source of such overreaction.

This study offers to contribute empirical research towards the field of behavior finance in understanding better the notion of investor irrationality and its consequent impact on asset pricing.





Brief History of Research Question

Theoretical Framework

Research Objectives

Statement of Problem

Research Significance

Literature Review

1) Prior Research and Gaps

2) Investor Overreaction

3) Representative Heuristic

4) Research Questions


1) Methodology Overview

2) Research Hypothesis

3) Model Configuration

4) Dependent and Independent Variables

5) Sample Data and Characteristics

6) Construction of Test Portfolios

7) Measuring Abnormal Returns and Earnings

8) Empirical Results and Statistical Significance

9) Does Representative Bias drive Overreaction?

10) Limitations of Study