Disclaimer: This is an example of a student written assignment.
Click here for sample essays written by our professional writers.

Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of UKEssays.com.

Persistence of Overreaction in the Stock Market

Paper Type: Free Assignment Study Level: University / Undergraduate
Wordcount: 4666 words Published: 21st Oct 2020

Reference this

Abstract

A puzzling feature of stock returns is evidence of returns reversals at horizons of three to five years. One explanation for this evidence comes from behavioural models that assume investors overreact to short runs of recent earnings news. The objective here is to investigate

Introduction

With no doubt the development of one of the most famous theories in financial economics is attributed to Eugene Fama with his Efficient Market Hypothesis (EMH). According to EMH, markets are efficient and rational as securities’ prices fully reflect all available public and private information. Although, there are different forms of efficient markets depending on the relevant information, the main ideas are that the price movement is random, stocks always trade at their fair value and it is impossible to beat the market since any superior returns are arbitraged away immediately. However, it is common for the prices to temporarily deviate from their fundamental values and have delays in incorporating the information.

Get Help With Your Assignment

If you need assistance with writing your assignment, our professional assignment writing service is here to help!

Assignment Writing Service

Such opportunities to earn abnormal returns have challenged the Efficient Market Hypothesis and its weak form in particular. The vast research focused on the equity market identified many patterns in historical stock returns that signal either a market inefficiency or misspecification problems with asset-pricing models. The focus of this paper is on the Overreaction hypothesis proposed by DeBondt and Thaler (1985) that aims to explain market anomalies with the use of behavioural principles. The theory is based on the tendency of investors to “overreact” to news causing stocks’ prices to deviate temporarily from their fundamental values.

Originally tested on the New York Stock Exchange (NYSE), the hypothesis has been found to be robust across time and international markets. However, after the publication of the paper many studies pointed out the weakening of the overreaction in some markets (McLean and Pontiff (2016), Cotter and McGeever (2018)). This in turn raises interesting questions: How strong is the overreaction in the market after the publication of the paper? Does its disappearance reflect the actions of players who learn about the anomaly and trade so that profitable opportunities vanish?

The objective of this study is to investigate the persistence of the overreaction in the market over the time. The analysis is performed using historical return data from NYSE by employing the original methodology developed by DeBondt and Thaler (1985). The weakening of the Overreaction hypothesis is evaluated by comparing the results from two sample periods: the original sample and the period from 1985 to 2017 following the paper’s publication.

Overall, the results indicate a downward trend in the overreaction within NYSE. The paper thus makes an important contribution to the literature on cross-sectional asset pricing and provides evidences on the persistence of the anomaly in the stock exchange over time.

The remaining sections are arranged in the following fashion. Overview of the Overreaction hypothesis and its critique are presented in Section two. The hypothesis is detailed in Section 3. Details of the methodology used and data description are presented in Section 4, and in Section 5, I present and discuss the results of this study. The final section provides the conclusions.

Literature review

Overview of the Overreaction Hypothesis

Experiments conducted by Kahneman and Tversky [1980] demonstrated that man can hardly be viewed as a Bayesian decision-maker. It appears that, contrary to Bayes' Rule, when revising beliefs, subjects systematically overweight singular information about the specific case at hand and underweight distributional or base-rate information. People take, so-to-speak, an "internal" approach to intuitive prediction. They tend to rely on the representativeness, or the degree to which an event is similar in characteristics to its parent population when making judgements. This work has motivated De Bondt and Thaler to investigate whether such deviation from Bayesian optimum when forming and updating beliefs matters at the market level.

De Bondt and Thaler [1985] have first suggested a simple behavioural model, where it is argued that stock prices behave as if the representative agent systematically "overreacts" to new information, particularly as it relates to earnings. Their study has focused on investigating the relative test period performance of 35 stock (50 stock or decile portfolios) of "long-term" winners and losers. They use monthly return data for NYSE stocks between December 1925 and December 1982. Initial performance is measured over formation periods ranging between one and five years. The price run-ups within formation periods of the winners and losers are considered as a proxy for “excessive” market optimism and pessimism.  Consistent with overreaction bias and the profitability of contrarian investment strategies, De Bondt and Thaler reported predictable long-term stock price reversals for prior (extreme) stock market winner and losers. On average during the 36 month test period, an "arbitrage" strategy that buys losers by selling winners short earns an annual return of 24.6%.  The Looser portfolio outperforms the market portfolio by 19.6% over the next 3 to 5 years, while the Winner portfolio underperforms by 5%. 

Lakonishik et al. (1994) have also observed long-run return reversals when using the firms’ past financial ratios as an indicator of their performance. Each year during 1968 -1990 two portfolios of “value” and “growth” stocks were constructed. The growth stocks were defined as the one traded at low price-to earnings ratio, low price-to-book ratios or high dividend yield. Growth stocks, on the other hand, were traded at high price-to-earnings ratios, high price-to-book ratios or low dividend yields. On average, value stocks outperformed growth stocks by 7-11% per year over 5 year testing period. One of the potential explanations of such pattern is the overreaction of investors to the individual firms’ characteristics such as past earning. Typically, growth stocks have had high earnings growth over past few years.  This results in over extrapolating investors expecting high earnings growth to continue for many years. However, superior earnings growth for growth stocks doesn’t continue for long. As a result investors are negatively surprised by subsequent earnings growth for growth stocks, that results in inferior performance of growth stocks relative to the value stocks.

Another potential explanation as to why investors “overreact” on the stock market is due overconfidence in their abilities to acquire more accurate information about the firm’s value. Investors tend to overestimate the precision of their private info and underestimate the forecast’s errors. Weinstein (1980) suggests that such overconfidence arises due to self-attribution bias, where people tend to attribute the success of their actions due to their own abilities and the failure due to bad luck. As a result, when a positive private signal arrives, investors tend to overweight this information and push stock prices too high compared to its fundamental value. Due to the self-attribution bias, this investor’s overconfidence increases following the arrival of confirming news that pushes prices even higher. The overreaction in prices is eventually corrected in the longer run as investors observe future news and realize their mistakes leading to long-run reversals (Daniel (1998))

Criticism of Overreaction hypothesis: rational explanations

According to DeBondt and Thaler (1985), the results observed were broadly consistent with Overreaction hypothesis and were indicative of an irrational behaviour of players in the stock market. Yet, many researchers disagreed with such conclusion and presented other explanations for the differences in performance of Loser and Winner portfolios. Some of the criticism is presented below.  

One of the assumptions in the original methodology was the identical risk levels between test and formation periods. However, in their papers Chan (1988), Ball and Kothari (1989) have argued that the beta of the Loser portfolio experiences a rapid increase after the period of substantial loss.  On the other hand, the beta for the past extreme winners decreases in the next period. By removing the constant risk level assumption, it was found that the winner-looser strategy yields only small abnormal returns.

Moreover, the Loser portfolio was more likely to contain smaller-sized firms, as they have lost the market value relative to the winners and were more likely to be the bottom performers. Therefore, the size effect or a tendency of firms with relatively smaller market capitalisation to outperform larger firms was another rational proposition to explain abnormal returns of contrarian strategy.  According to Sweeney et al (1996), this effect arises from smaller firms having a bigger ratio of capital budget to market value than larger firms do. As a result, small firms earn greater abnormal returns since their relative excess capital budget returns are greater. Alternatively, it has also been suggested that higher returns come from smaller firms being riskier than the large ones (Jensen 1997). By controlling for the size, effect when applying the original research design Zarowin (1990) found that past losers only outperformed past winners in months of January.

The most discussed feature observed by DT in their results was the extraordinary large positive abnormal returns realised by the Loser portfolio in January. Many researchers pointed out that the January effect may be driving the reversal in stock prices. Additionally to controlling for the size effect Zarowin (1990) modified the beginning of his formation and testing periods to start in July to test whether January returns of the contrarian strategy is due to investors’ overreaction in the initial month or dare due to the January effect. The results suggested that the January returns are mostly driven by investors’ tax-loss selling behaviour than by overreaction.

De Bondt and Thaler addressed the above criticism in their 1987 paper. Firstly, by accounting for varying risk they found that despite the 0.22 difference in CAMP betas between Loser and Winner portfolios it was too small to justify the 9.2% average annual return of the arbitrage portfolio. Further analysis of the firms size within portfolios led them to conclude that firms within the Loser portfolio were not unusually small. As a result, De Bond and Thaler rejected the size effect and the difference in risk explanations for Winner-Loser effect, although no satisfactory explanation was suggested for the January effect.

Why anomalies disappear?

 A central argument in the literature is that the discovery of a profitable anomaly should lead to its gradual disappearance. According to Schwert (2003), one of the primary reasons for the disappearance of anomalies over time is the sample selection bias. However, the long-run price reversals have been previously shown to be robust out of sample (Da Costa (1994), Leung and Li, (1998)).This is strong evidence against data mining and in line with the view of Fama (1991) that economically meaningful anomalies should not disappear in out-of-sample testing.

If the stock return predictability is in fact, an anomaly then the nature of the anomaly, whether it is behavioural or risk-related should directly influence its persistence as suggested by Fama (1998), Cochrane (1999). In line with the market learning theory (Timmermann and Granger, 2004), true behavioural anomalies might appear on the market for a short period of time, but as investors become aware of them and start exploiting them, they should be slowly arbitraged away (Fama, 1998). In contrast, if a stock return reveals an important systematic risk factor priced on the market, it should not disappear after it is documented (Cochrane, 1999).

The most recent studies on long-run reversals have shown that despite the anomaly being discovered more than 30 years ago it is still present on many international markets including emerging and developed markets. Evidence of overreaction have been found in Chinese stock market (He and Tan, 2009), Greek stock market (Antoniou et al (2005)), US stock market (Hribar, McInnis, 2012) etc. However, the finding of mean reversion is not uniform across studies.

Research question and hypotheses

Although much research have been conducted using NYSE return data after the publication of the DeBondt and Thaler (1985) paper, the majority of it employed modified methodology. As a result, there are mixed outcomes with regard to persistence of the profitability of the contrarian strategy. Therefore, the following hypotheses were developed to compare the long-term price reversals in the original sample and samples after publications, ceteris paribus.

Hypothesis 1: Following the paper’s publication and further research the profitability of the contrarian strategy had decreased but remains statistically significant.

Hypothesis 2: The rate of change of contrarian strategy’s profitability was greater during the 1985 -1999 (first 15 years after the paper publication) than during 2000-2018.

Data and Methodology

Empirical test procedure description:

To analyse the persistence of Overreaction hypothesis the original methodology developed by DeBondt and Thaler was applied with relation to the New York Stock Exchange common stocks. Monthly data on securities’ returns was obtained from the Center for Research in Security Prices (CRSP).  For the purpose of this study, the analysis is run from the January 1926 until December 1982, which is the original data sample used by DeBondt and Thaler, and then extended to include the period from the January 1983 up until December 2016.

The model is based on examining the monthly market-adjusted returns of each stock, where the market index is calculated as an equally weighted arithmetic average rate of return of all stocks on the record. The whole time period in the dataset is split into the 27 distinct formation periods each equalling to 36 month and starting from January 1930 up to December 1978 and from January 1982 to December 2017, as well as 27 testing periods of equal size that run from January 1932 to December 2017. In order to be qualified for the analysis each security has to  have at least 84 months of continuous return data prior to the end of each formation period and an additional one month of the subsequent testing period qualify for the analysis. Starting from January 1930 the abnormal return of each individual security is then obtained up to December 1932 as:                               

 

Where:

 - Abnormal return for stock i at month t

        - Return of a stock at month t  ­

         -Market portfolio return at month t

This process is repeated every 36 months from January 1930 and up to January 2016(except for 1981).

  1. After calculating the abnormal return for each month, I then obtained the cumulative excess return over each of the formation period:

Where: 

         - Cumulative excess return for each stock

  t=0 - Formation month (December 1932,…, December 2014)

Within each of the formation periods, the stocks are ranked according to their CUi from the lowest to the highest. The 35 worst performing stocks and 35 best performing stocks in each period form the equally weighted looser and winner portfolios respectively. In this way, the securities’ abnormal returns prior to formation month determines portfolios’ composition.

  1. Cumulative average residual returns(CARpnt) are calculated within each of the portfolio and period for every month within the testing periods:

                                                            

       Where:

           - Average cumulative residual return for test period n at month t for portfolio p

          p - Winner (W) and Loser (L) portfolios

          K - Number of securities in the portfolio

Whenever any stock within the portfolio has a missing value for the abnormal return for any month t, it is dropped from the portfolio.

According to Overreaction hypothesis, for any month throughout the testing period the difference between the average cumulative residual returns (ACARpt) across all test periods for Loser portfolio and Winner portfolio should be greater than 0. Particularly, ACARLt in each month is greater than zero and ACARWt is less than zero. On the other hand, if markets are efficient then such difference should equal zero.

The ACAR for each portfolio is obtained in the following way:

                        

       Where:

          ACARpt - Average cumulative residual return at month t for portfolio p

The statistical significance of the difference between portfolios’ performance is evaluated with the t-test, as follows:

To calculate the estimate for the population variance I used the formula below.

              

  1. Lastly, I test whether at any month t CARpnt is significantly different from zero in order to assess its impact on ACARpt.  The following formulas for the sample standard deviation and the t-statistics are used. The process is conducted for both Loser and Winner portfolios.

 

 

For the experiment described in this section, between 281 and 1956 NYSE stocks are used in the various replications.

Rpt-Rft=αp+βpRMt-Rft+δpSMLt+ηpHLNt +εpt

Main Findings

Table 1 presents the results of the tests developed in a previous section. They are consistent with the hypothesis that in the years following the publication of the original paper, the profitability of the contrarian strategy had decreased but remained statistically significant particularly for 1 and 36 months after portfolio formation. Over the three decade after the discovery of the price reversal anomaly still persists in the market and on average outperforms the market by 14.7%. It represents a statistically significant decline in overall strategy’s profitability (t-statistics: 2.93). Figure 1 shows the movement of ACAR for both Winner and Loser portfolios within two sample periods: the original sample period and the period after paper’s publication up to 2017.

 Table 1:

Overall, in the Figure 1 a graphical representation of findings reveal asymmetric levels of reversal within the two portfolios. The reversal of a Loser portfolio is quite prominent throughout the whole period after the portfolio formation, whilst the Winner portfolio tends to experience it at 1 month after the formation and within 20 to 36 months after the formation. This can suggest that the overreaction effect is more acute within the underperforming stocks. Similar reversal discrepancies have also been observed in other studies. (e.g., Bremer & Hiraki, 1999). However, whilst the Loser portfolio from the original sample tends to experience the overreaction mostly during the second and third year after the formation the Loser portfolio after the publication remains relatively stable in this period. For example, the returns from the contrarian strategy in both samples remained statistically significant 36 months after the portfolios’ formations but the magnitude of the return fell considerably from 24% to 14.7%. It is important to note that the January effect still can be observed on the NYSE. The performance of the past Winner portfolio has also relatively flattened out. It underperforms the market by 2.1% 36 months after the formation.

Figure 1:

The results show strong evidences of declining strength for the overreaction in the stock market. However, the persistence of the anomaly in the market raises more questions about its nature. As mentioned before, anomalies that are behavioural in their nature, tend to disappear relatively quickly. Investors are prone to correct their biases after the discovery of market predictability and exploit the arbitrage opportunities. Although in the periods of high sentiment even anomalies that arise because of “irrational” behaviour of the investors may persist if it is not countered by a sufficiently large amount of arbitrage capital. On the other hand, such persistence of the anomaly can represent a systematic risk factor instead of a true behavioural anomaly, and the model does not acknowledge that the true idiosyncratic risk of the portfolio should be higher. One of the main limitations in the DT’s methodology is the use of the market-adjusted returns instead of the risk-adjusted ones. In order to assess whether returns reversals can be explained through the basic models of risk and return, CAPM and the three-factor model of Fama and French (1996) were applied to the Loser and Winner portfolios in both samples.

An alternative explanation for return predictability is that it results from some kind of risk premia—risky assets predictably return more than less-risky ones. This explanation then raises the question of the actual source of risk and whether plausible levels of risk aversion are high enough to explain the size of the predictability, a question we address below

It follows, that if the CAPM is successful in explaining the return in terms of  systematic risk, the alpha will be zero. That is, the excess return of the momentum portfolio is entirely due to a high beta -value. On the other hand, if alpha is positively significant different from zero, the

portfolio has provided a better return than what is expected given the portfolio’s beta value. Likewise, if the observed returns are due to higher risk-bearing, it is expected that the zero-cost portfolio loads heavily on the two additional proxies of risk; SMB and HML in the Fama and

Conclusion

In this paper we examined the winner loser effect in the Tunisian stock market over the period 1974e2013. We ranked stocks based on their cumulative abnormal returns for a formation period of 12e60 months and assigned them to three portfolios. We then tracked the performance of the extreme tercile portfolios during a holding period of 1e60 months. Our findings indicated that the stock market exhibits a significant winner-loser effect supported by a performance reversal beyond 12 months whatever is the period during which the stock past performance is measured. This phenomenon persists for the sub-period 1991e2013. In order to check whether the contrarian profits are due to the change in risk from the formation period to the holding period, we introduced a dummy variable in the three-factor model of Fama and French where returns are adjusted to the size and Book to Market related risks, in addition to the market risk.

Find Out How UKEssays.com Can Help You!

Our academic experts are ready and waiting to assist with any writing project you may have. From simple essay plans, through to full dissertations, you can guarantee we have a service perfectly matched to your needs.

View our services

The returns of the winner, loser and winner minus loser portfolios are regressed on these factors for each 6-year sub-period and the regression results are aggregated for the whole sample period. Three main conclusions can be deduced from our empirical analysis. First, the Tunisian investors can on average realize contrarian profits by exploiting the contrarian strategy. Second, the contrarian profits not only represent a compensation for additional two-dimensional risk related to the firm size and the book to market, but they are also due to the risk change in these two factors from the formation to the holding period. Third, since the excess returns represent a normal compensation for risk, the overreaction hypothesis as an explanation for the winner loser effect does not seem to hold in the Tunisian stock market.

 

Cite This Work

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.

Related Services

View all

DMCA / Removal Request

If you are the original writer of this assignment and no longer wish to have your work published on UKEssays.com then please: