Essays of Three Fundamental Biases in Behavioral Finance from the Capital Market of Bangladesh

Abstract

The thesis empirically investigates three essays on core behavioral biases such as the disposition effect, overconfidence bias, and herding behavior within the framework of frontier equity markets like Bangladesh across distinct market states: bearish, bullish, crisis, extended crisis, and COVID-19. The developed and emerging equity markets extensively reported the evidence of biases in investor decision-making, but behavioral finance research in frontier markets remain underdeveloped and underexplored. By addressing this gap, the thesis systematically examined these three biases, applying econometric and advanced machine learning tools. The first essay of the thesis empirically investigates whether investors have the inclination to realize the stock gains promptly rather than holding on to the stock that has declined in value under different market conditions in the Bangladesh equity market. The analysis estimates the disposition coefficients with the evaluation of the responsiveness of the current market trading volume to lagged market index prices. The findings indicate strong and ubiquitous effects of the disposition effect in the overall market and in all other markets conditions except for bearish periods. It is quite remarkable that the strength of the behavior increases in times of crisis and prolonged periods of crisis, thereby indicating that the loss aversion behavior of investors continues to increase under greater uncertainty. The results are an extension of the knowledge on behavioral trading patterns in frontier markets as they show that the disposition effect is time-varying and state-dependent. The study, therefore, adds new information on the influence of emotional and cognitive bias in the responses to the trading volume in various market setups. The second essay discusses the nature, severity, and dynamics of herding behavior in a Bangladesh stock market in both different market states and situations. Based on the Cross-Sectional Standard Deviation (CSSD) model and Cross-Sectional Absolute Deviation (CSAD) model developed by Christie and Huang (1995) and Chang et al. (2000), respectively, in addition to a quantile regression framework, the study offers a complete picture of the investor herding concept with different market regimes. The findings demonstrate that the herding behavior in Bangladesh is widespread but also state-specific where the bearish and the protracted crisis periods depict the most prominent state-dependent herding behavior, especially in the extreme down markets. The fact that asymmetric herding happens in the high-volatility and high-trading-volume conditions supports the sensitivity of the CSSD model to explain the extreme collective movements as compared to the CSAD approach. The analysis at the industry level also shows that the herding behavior is not evenly distributed but rather concentrated in certain sectors during the periods of market stress. In addition, macroeconomic variables and monetary policy instruments have been found to have a considerable effect on herding, particularly in bearish environment, and in crisis environment. In general, the results contribute to the behavioral finance literature since they reveal the contingency of state-based, asymmetric, and policy-dependent herding behavior in frontier market setting. The third essay explores the expression and behavior of investor overconfidence in Bangladesh stock market with a special focus on how it has been changing under different market states. The vi problem of overconfidence investors overestimating their ability to predict and trading too much due to the existing returns is investigated using a combined framework comprised of Vector Autoregressive (VAR), Transfer Entropy (TE), and Long Short-term Memory (LSTM) models. The analysis of the total and unexpected elements of market returns during bullish, bearish, crisis, extended crisis, and COVID-19 periods unveils that the behavioral phenomenon of overconfidence is persistent and clearly state-dependent. One of the main discoveries is a new type of behavior, which is called defensive overconfidence: when the market is under intense stress, especially in equity market downturn, investors still show overconfident behavior, yet their behavior corresponds to defensive responses to uncertainty, instead of optimism of making future gains. The analysis at the industry level also proves the prevalence of this bias in all major and dominant industries. Overall, the findings are expected to make contributions to the behavioral finance literature by adding a new conceptual dimension to the notion of overconfidence, as well as evidencing the applicability of hybrid econometric-machine learning paradigms in order to capture the non-linear and adaptive nature of investor psychology in frontier markets.