Effect of Overconfidence that Happens Within the Market

Odean (1998) theoretically investigated the effects of overconfidence in different market structures which differ in information distribution and in price determination. He presented overconfidence effects on different market measures such as trading volume, market depth, volatility, expected utility and market efficiency taking overconfidence of different traders: price takers in the market where information is broadly disseminated, strategic-trading insiders in markets with concentrated information and market-makers. Odean assumed that investors were rational in all respects except valuing the information. They overestimate the precision of their information signals. The most robust result presented is that trading volume increases when price-takers, insiders or market-makers are overconfident, however market depth also increases but expected utility of overconfident traders decrease. He also presented that overconfidence of price takers worsens the price quality but overconfidence in insiders improves it. Overconfident traders increase volatility though overconfident market makers may decrease it. Overconfident price takers reduce market efficiency because they are overconfident about their ability to interpret publicly disclosed information while overconfident insiders temporarily increase the market efficiency. In the presence of many overconfident traders market tends to under-react to the information of rational traders. Markets also under-react to highly relevant information and overreact to salient but less relevant information.

Daniel, Hirshleifer and Subrahmanyam (1998) developed theoretical models based on investor overconfidence and self attribution biased. They made distinction between public and private information and defined investor overconfidence as overestimation of the precision of private information signals, but not of information signals that are publicly received. In their model, investors are quasi-rational in a way that they are Bayesian optimizers except for over valuing valid private information and biased updating of precision. Overconfidence makes investors to overestimate precision of their own valuation abilities and as a result they overestimate precision of private information signals. The theory predicts stock market under and over reaction on the basis of investor overconfidence and changes in confidence resulting from biased self-attribution of investment outcomes. Primary theme of the study was that stock prices overreact to private information signals and under react to public information signals. Unlike Odean (1998b) who presented trading volume implications of investor overconfidence, Daniel et. al. focus was on return implications of overconfidence. They showed that under certain circumstances security returns are positively auto-correlated in the short run (momentum) but negatively auto-correlated in the long run and overconfidence cause excess volatility.

Odean (1998b) examined the disposition effect, the propensity of investors to sell winning investments too early and hold losing investments too long, by analyzing trading records of 10,000 randomly selected accounts from January 1987 to December 1993 at a nationwide discount brokerage house. The data file included 162,948 records of all trades made in 10,000 accounts. Accounts that were closed during the window period were not replaced which made data subject to survivorship bias to some extent in favor of more successful investors. For each day Odean constructed a portfolio of securities for which the purchase date and prices were known and compared the selling price for each stock sold to its average purchase price to determine whether that stock is sold for a gain or a loss. Securities not sold on a particular day were considered to be as unrealized (paper) gain or loss. He constructed two proportions, proportion of gain realized "PGR" (realized gain divided by realized gain plus paper gain) and proportion of loss realized "PLR" (realized loss divided by realized loss plus paper loss). He developed two hypothesis: 1. PGR is greater than PLR (for the entire year), 2. (PLR - PGR) in December is greater than (PLR - PGR) in January-November. The second hypothesis was for analyzing tax motivated selling in December. He also analyzed PLR and PGR by partitioning the data in time periods and in frequent traders and infrequent traders. He showed the results by comparing proportions using t-test and concluded that investors exhibited a strong preference for selling winners and holding losers throughout the year except December when tax-motivated selling was prominent. Further, desire to rebalance or to avoid the higher trading costs of low priced stocks did not seem to motivate investor behavior.

Gervais and Odean (2001) developed a multi-period equilibrium model that describes the process by which traders learn about their abilities and how a bias in this learning makes them overconfident. The model argued that traders initially do not know their ability but they infer it from their success and failure. Traders are biased in assessing their ability when they take too much credit for their success. This leads them to become overconfident. This biasness is more prevalent in early stages of trades' careers. With more experience, traders better assess their abilities. Further, overconfident traders are wealthy as a result of success but wealth is not a function of overconfidence. However, process of becoming wealthy can make traders overconfident. Due to wealth of overconfident traders, they are not in immediate danger of being driven out of the marketplace. Overconfidence of a particular trader would not grow indefinitely; it would decrease gradually with time and experience. But, in a market in where new traders are born every minute, overconfidence will flourish. The model showed that higher returns from general market increase make investors overconfident due to their biased attribution of returns to their abilities and therefore they trade more actively. The primary implications of model were: periods of market increase would tend to be followed by periods of increased aggregate trading. Aggregate trading would likely to rise late in a bull market and to fall late in a bear market. Volatility would also increase with the degree of traders' learning bias. Due to suboptimal behavior overconfident would lower their profit.

Odean (1999) tested the proposition that overconfident investor would trade too much in the market. The data included ten thousand customer accounts randomly selected from accounts that were active in 1987. It was provided by a large discount brokerage house and had 162,948 records of all trades made in sample accounts from 1987 to 1993. Odean (1999) examined that whether the trading profits of discount brokerage customers were enough to cover their trading costs. Return horizons were taken as four months (84 trading days), one year (252 trading days), and two years (504 trading days). The first null hypothesis was established as, the difference in returns (average returns to securities bought minus the average returns to security sold) was greater than or equal to the average total cost of a round-trip trade (i.e about 5.9 percent) over all the return horizons. The second null hypothesis was that average returns to securities bought were greater than or equal to those sold over the same horizons. Due to lack of independence between overlapping periods statistical test that require independence were not employed. Instead, bootstrapping an empirical distribution under the assumption that returns were drawn from the same distribution was used to test the significance of differences in returns. The results showed that investors trade too much in that their returns were reduced through trading even after controlling for trades motivated by tax-loss selling, liquidity demands, portfolio rebalancing and a move towards lower-risk securities. Overconfident investors actually lowered their returns through trading even when trading costs are ignored.

De Bondt and Thaler (1995) reviewed the literature on behavioral finance. They argued that "the key behavioral factor needed to understand the trading puzzle is overconfidence". They further pointed out that overconfidence can explain why portfolio managers trade too much, why active equity managers are hired by pension funds, and why financial economists often hold actively managed portfolios. Moreover, high trading volume and the pursuit of active investment strategies look to be inconsistent with common knowledge of rationality.

Benos (1998) developed a theoretical model of overconfidence. He examined the overconfidence where some risk neutral investors overestimate the precision of their private information. He analyzed a strategic model of trading in a call auction market with many informed traders. Investors overestimate their precision in a way that when they receive imperfect information on asset characteristics, some treat it cautiously, realizing that it may contain irrelevant noise, while, others think signals are perfect. In his model some informed investors were overconfident about their estimates of unknown variables or about their valuation abilities. And also all market participants knew about the beliefs of all their opponents and about the overconfidence of traders and reacted accordingly. And overconfident investors compete in market orders with informed traders who have rational expectations. The model concluded that presence of overconfident traders in the market results in higher trading volume, more depth, higher volatility and increased market efficiency. Striking result presented was that overconfident traders may make higher expected profits. This is due to first-mover in competition setting.

Gallant, Rossi and Tauchen (1992) investigated price and trading volume co-movement. Their empirical study was mainly a data-based effort. It was not organized around the specification and testing of a particular model or class of models. To analyze the relationship between contemporaneous trading volume and volatility and the relationships among prices, volatility, and volume over time were the two main objectives among four. A bi-variate time series of 16,127 daily observations on the S&P composite index and total NYSE trading volume from 1928 to 1987 was used. They employed seminonparametric estimation approach, which was a nonparametric estimation strategy, to analyze the data. The nonparametric choice was adopted to avoid risk of specification errors with parametric techniques. Initially data was adjusted for both long term trends and known calendar effects using different dummy variables. Data was further partitioned into three sub-periods to analyze the stability of the findings. The results explained that there is positive and nonlinear relationship between daily price change and daily trading volume. Conditional volatility and trading volume are also positively related. Positive relationship also revealed between risk and return after conditioning on lagged trading volume.

Statman, Thorley and Vorkink (2006) empirically tested the theories of overconfidence developed by Odean (1998) and Gervais and Odean (2001). The focus of the study was on trading volume implications of overconfidence. The study incorporated the overconfidence hypothesis as a separate theory of trading activity related to investors' beliefs about trading in general, rather than an attitude about individual stocks they currently hold (disposition effect). The primary hypothesis was that past market returns can explain the current trading volume. In other words, there is long-lag relationship between market returns and trading volume. Database consists of monthly observations on all NYSE/AMEX common stocks, excluding closed-end funds, REITs, and ADRs, from August 1962 to December 2002. Vector autoregressive (VAR) model at market level and individual security level, was used to empirically test the implications of overconfidence. They constructed a market portfolio consisted of all the securities and calculated the variables. Monthly market return and monthly trading volume were endogenous, and monthly volatility based on daily data and value weighted cross sectional mean average deviation from market returns were the control variables for market-wide VAR. Individual security level VAR had only one control variable i.e volatility but had three endogenous variables: security return, market returns and security turnover. Turnover variable was constructed as total shares traded divided by outstanding shares at the beginning of a month. The key findings of study were: there was a significant positive relationship between market-wide turnover and past market returns after controlling for contemporaneous and lagged volatility associations, consistent with the prediction of the overconfidence hypothesis. Individual security turnover was also positively related to both lagged security returns and lagged market returns. The positive security turnover response to own lagged return was interpreted as disposition effect, while positive turnover response to lagged market returns was interpreted as change in investor overconfidence. The phenomena of overconfidence and disposition effect trading were both more prominent in small-cap stocks.

Zaiane and Abaoub (2009) examined empirically the theory of overconfidence in a small Tunisian market using Vector autoregressive (VAR) model and related impulse response functions. Database consisted of monthly observations of common stocks from January 2000 to December 2006. Raw shares-traded was treated as proxy for trading volume. Market returns and trading volume were endogenous variables in VAR. Whereas, volatility and value weighted cross sectional dispersion were treated as exogenous or control variables. Variables used were the same as used by Statman et.al. (2006). Hypothesis of the study was that there is positive significant relationship between past returns and current trading volume. Little evidence of relationship between trading volume and lagged returns was found in Tunisian market which was interpreted as absence of overconfidence. Though, past returns affect trading activity over some months but mostly, relationship is insignificant. However, significant positive relation was found between volume and volatility.

Pisedtasalasai and Gunasekarage (2007) examined the causal and dynamic relationships among stock returns, trading volume and return volatility in five emerging markets of South-East Asia: Malaysia, Indonesia, Philippines, Thailand and Singapore using daily data. Specifically, they investigated contemporaneous relationship between return and volume, effect of trading volume on return volatility and causal relationship between return and trading volume. They used simple regression equation for contemporaneous relationship, EGARCH model for trading volume and return volatility relationship and Vector autoregressive model for Causality. The dataset was collected from Datastream. The selection criteria for equity indices were based on representation of majority of securities and availability of corresponding trading volume. Jakarta Stock Exchange Composite Index (JSECI) for Indonesia, the Manila Stock Exchange Composite Index (MSECI) for the Philippines, theKuala Lumpur Stock Exchange Composite Index (KLSECI) for Malaysia, the Bangkok SET Index (BSETI) for Thailand and the Datastream Market Index (DMI) for Singapore were analyzed. Data series started from January 1990 for MSECI, DMI, and BSETI, from January 1991 for JSECI and from March 1996 for KLSECI. All series ended at December 2004. To check robustness of results dummy variables for Monday effect and for Asian Financial Crisis 1997 were also incorporated. The results showed the evidence of a statistically significant causal relationship from stock returns to trading volume for Indonesia, Malaysia, Singapore and Thailand. Such evidence was not found for Philippines market. Causality from trading volume to stock returns was detected only for Singapore. Evidence of trading volume being useful in predicting return volatility was found only for Singapore and Philippines markets. Striking behavior of Philippines market was suspected as due to lower capitalization of the market.

Cite this page