sign in sign up
<<
Home , Long (finance), Net asset value, Noise trader, Stock, Trader (finance), Yield (finance)

Revista Argentina de Clínica Psicológica 317 2020, Vol. XXIX, N°4, 317-324 DOI: 10.24205/03276716.2020.831

Noise Trading and Abnormal Return in

Chiu-Lan Changa, Ming Fangb*

Abstract This study analyzes the relationship between noise trading and abnormal returns on . This study first validates the positive correlation between noise trading and abnormal returns on assets. In markets with high noise trading, the higher the abnormal returns on its assets, and vice versa. In exploring the relationship between noise trading and , this study uses abnormal return as a new indicator proposed in recent years for analysis. When analyzing the impact of fundamentals on return on assets, the variables of company fundamentals are integrated using clustering analysis and principal component analysis. This suggests that in the run there is a mutual causal relationship between noise trading and abnormal returns on assets. The results of this study show that the deviation from the interaction of noise trades and abnormal returns can be corrected and reach the equilibrium in the long run. Keywords: noise trade, abnormal return, long-term equilibrium relationship, causal relationship.

1. Introduction In analyzing the impact of the company's Noise trading research is an important research fundamentals on the return on assets, the financial in behavioral finance. Scholars have carried out indicators reflecting the fundamentals of the research and discussion on the influence of noise securities and applied with clustering analysis and trading and related problems such as noise trading principal component analysis. measurement indicators. In recent years, with the vigorous development 2. Literature Review of China's , more and more scholars Noise traders are defined as who do not based on China's special policy restrictions on have access to inside information, irrationality, use China's noise trading related research error messages or even information as correct and empirical analysis. information and trade on that basis (De Long et al, 1989, De Long et al, 1990). Therefore, the noise of the This study explores the relationship between capital market also mainly refers to the above- noise trading and excess return on assets. This study mentioned investors as the basic role of transaction not only uses new indicators developed in recent years, but also uses newer research methods. In the decision-making, but not related to the capital market-related products in the capital market specific exploration of the relationship between relevance or incorrect information. noise trading and asset returns, this study uses excess yield, a new indicator proposed in recent In fact, how to measure noise and how to select the measure of noise trading has always been an years, for analysis. important topic in behavioral finance. aProfessor, Department of Finance, Fuzhou University of Some literatures categorize the investors into International Studies and Trade, China *bAssociate Professor, Department of Finance, Fuzhou University of two types which are rational investors and noise International Studies and Trade, China traders. The rational investors have rational *Corresponding Author: Ming Fang expectations about returns and noise traders Email: [email protected] trades according to their sentiment. Both types of investors trade in the market depending on their own beliefs. Therefore, the trading behaviors may be very different. However, most of the investors are

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA 318 Chiu-Lan Chang, Ming Fang* avoiding . Previous researches consider the market values. Meanwhile, the return of rational individual investors are noise traders and the investors comes from the losses of irrational institutional investors are rational investors. Because investors. If the irrational traders continue to lose institutional traders generally mine information and , they will gradually exit the market or become are relatively knowledgeable about the information; rational investors. Therefore, under the efficient while individual investors cannot dig deep into market hypothesis, irrational noise traders cannot be effective information due to financial, material, and squeezed out of the trading market for long periods manpower constraints, and even trade based on of time due to the natural selection of the market and gossip, so the researchers believe that noise trading the behavior of rational and informed Generally speaking, they are individual investors. traders. However, practically it is difficult to ensure that all The efficient market hypothesis remains traders are completely rational. Some institutional controversial after the amendment. Opponents investors may be irrational investors sometimes and have questioned the efficient market hypothesis trade with a strong noise (De Long et al, 1988, from previous references to final conclusions. First Hirshleifer and Luo, 2001). of all, noise trading is widely present in the market Some researches use the quantitative methods to and does not correspond to the small number of defined the noise trading. There is no consensus in presences expressed in the hypothesis (Lee et al, academia on how to quantify noise trading. In 1991, Baker and Stein, 2004). Second, investors have general, the metrics are divided into two broad a high probability of expected the same preference categories: cross-disciplinary psychological indicators for choice, which is statistically significantly different and directly related financial transaction indicators. from the objective probability in the expected utility Among them, the psychological substitution index is (Tversky and Kahnemand, 1974). Third, in an mainly confidence index, climate index, etc. environment of information asymmetry, Financial trading indicators, on the other hand, focus sentiment interacts with each other, rather than as on liquidity-related indicators, including the rate of stated in the hypothesis that irrational investors' signature, the growth rate of new account openings, decisions are independent of each other. the turnover rate, etc. First, scholars proposed the Finally, due to the mutual influence of investor use of liquidity, information and trading frequency to sentiment, coupled with the problem of principal distinguish between informed traders, unsuspecting agent and the shortage of arbitrage assets, arbitrage traders and market makers of the EKOP model, and does not have the ability to completely eliminate built a framework for calculating the probability of pricing errors caused by irrational behavior. informed trading (Easley,1996). Dynamic Specifically: First, the issue of proxy, in the fund index is another measurement of noise managers and other found that the capital market and proves that the turnover rate is currently a better asset pricing errors and arbitrage activities, because standard for reflecting noise trading. (Ramiah, 2003). the reverse forces in the market is more powerful, The related research on noise trading focuses on the fund managers manage the of the existence of noise trading and the influence of the fund is likely to have a larger drawdown or even noise trading on the effectiveness of capital markets. reach the warning line situation. Fund managers From the efficient market hypothesis (EMH), the cannot reverse this mispricing, whether based on existence of noise trading noise trading does exist in fund contract rules or under pressure from fund a period of time and later on the market will investors. Second, it is also likely that the average become rational (Fama, 1970). The EMH believes free-trader will be unable to prevent mispricing and that the vast majority of traders in capital markets generate large losses because the amount of assets are rational. These people have all the information cannot compete with the amount of assets of the they have and can make accurate judgments about reverse forces in the market. Based on these two the value and price of assets based on the considerations, most arbitrage traders choose to information. The revised efficient market hypothesis trade in the same way as noise traders in order to acknowledges that there are indeed a small number obtain positive returns. And since the arbitrage will of irrational people in the market who trade in a way operate later than the , the noise trader that causes a deviation in the instinct value of the will not be squeezed out of the market. The other asset. The modified efficient market hypothesis researches suggest that the return of assets to their states that assuming arbitrage is risk-free, rational instinct value is a long process. It is difficult for investors will quickly adjust their portfolios to the fair anyone who continuously lose money can still

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA

319 Chiu-Lan Chang, Ming Fang* restrain the anxiety of the state of mind and point collection is the direction vector sequence, maintain the original rational judgment. Therefore, where the vector is the direction of the line that best rational investors cannot completely eliminate fits the data and mates with the first vector. The core noise. In addition, Chinese and foreign scholars have of PCA is to find out a number of "common tested the effectiveness of different markets, information" hidden behind the variables among the indicating that noise traders' behavior is not many variables. In practical application, in order to effectively controlled, such investors are likely to understand the characteristics of data in all aspects, dominate the entire market, at which point it is the value of multiple variables will be collected from rational traders who are squeezed out of the market. multiple angles. More variables do reflect the characteristics of data more completely, but it also 3. Methodology creates some difficulties for data analysis such as 3.1 Cluster Analysis degrees of freedom reduced and a high degree of Cluster analysis is a technique for grouping collinearity. Therefore, PCA can not only grasp the similar observations into clusters based on important information of the original data, but also observations of multiple variables for each individual reduce the number of variables and simplify the (Ketchen, and Shook, 1996). Clustering is actual application. conceptually similar to differential analysis. The 3.3 Panel Regression group membership of the observation sample is The parameters of a linear model with a fixed known in advance in the latter, while any single effect can be estimated with the fewest observation of the former is not known. Cluster squares in the ordinary. In panel data analysis, linear analysis divides the subjects into different categories fixed and random effect models undoubtedly according to certain criteria or scales. In general, provide most of the ideas. Panel datasets are clustering is used to classify data into categories. becoming increasingly popular because of the Major clustering approaches include hierarchical widespread use of computers, making it easy to clustering algorithms and partitioning clustering organize and generate such data. Panel data is algorithms. Hierarchical clustering algorithms are observational multi-dimensional data that is in two methods which are agglomerative measured repeatedly over time. Considering the method and divisive method. The clustering dataset of this study are cross-sectional, therefore, approaches are adapted differently due to the data it would be more appropriate to use the panel selected. Currently, the hierarchical agglomerative regression to analyze the relationship between noise clustering approach is broadly adopted in academic trading and abnormal returns . researches. However, no matter which clustering 3.4 Causality analysis method is used, the core principle is to analyze the Granger first analyzed the causal relationship similarity between the investigated objects and then between variables from the perspective of classify them. Therefore, the key to clustering is how probability theory and the statistical nature of the to measure the similarity between two objects. In data. When the two variables have a lead-lag general, the distance function is used to d(x,y) relationship, it can be statistically observed that the represent the distance between x and y. Supposed relationship is one-way causality relationship or there are two points x = (x1, x2, x3, … , xk) and y = mutual causality relationship. If the performance of (y1, y2, y3, … , yk). The block distance between x and the previous issue y of the series x helps explain the 푘 y can expressed as 푑(푥, 푦) = ∑푖=1|푦푖 − 푥푖|. Then in changes in the current period of y the sequence, the Granger cause for the series y. If the above help ( ) √∑푘 ( )2 a general form 푑 푥, 푦 = 푖=1 푦푖 − 푥푖 . interpretation behavior is primarily one series to Minkowski distance can be expressed as 푑(푥, 푦) = another sequence, the two sequences are one-way 1 휆 relationship. There is a two-way relationship 푘 휆 (√∑푖=1(푦푖 − 푥푖) ) with λ>0. between the two sequences if both parties have been acting in the past to help explain each other's 3.2 Principal Component Analysis current behavior Primary Component Analysis (PCA) is a technique that reduces the size of such datasets, 4. Empirical Analysis improves interpretability, and minimizes information 4.1 Data and Model Setting loss. It does this by creating new uncorrelated The listed in SSE50 index are selected. The variables to continuously maximize variance. The samples are from January 2012 to December 2018. principal components of the actual p-space mid-

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA 320 Chiu-Lan Chang, Ming Fang*

This study selects the liquidity-related indicators of performance and the industry performance. Ci is the stock turnover rate to describe the noise trading. The characteristics of individual company and APi is the stock prices reflect four perspectives of information analysts’ prediction of the individual company. The including the market, industry, characteristics of abnormal returns should consider these four factors individual company and the investor sentiment. The can affect the expected price of individual company. abnormal return can be views as the reflection of Hang Seng SOE Index is used as the benchmark investor sentiment. According to the CAPM model, of the market performance which reflects the there is a certain relationship between market yield fundamental information of the Chinese market. and individual stock yield. Combined with the Shen Wanhongyuan Security Industry Index of each methodology of the APT model, it can be concluded industry are used as the benchmark of industry that market information and industry information performance. The cluster analysis performs cluster will affect the price of individual stocks at the same validation for the above classification. The clustering time. Considering that macroeconomic information tree is shown in Fig. 1. The relevant indicators under will also affect the development of an industry in -paying ability, profitability, growth ability and many cases, that is to say, macroeconomic operational capability are clustered into groups, and information will enlarge or narrow the impact of the results of clustering analysis are consistent with industry information on the company's price, expectations. then there should be some interaction between The relevant indicators under debt-paying market factors and industry factors. ability, profitability, growth ability and operational 퐴푅푖 = 푟푖 − 퐹(푅푚,푖, 푅퐼,푖,, 퐶푖, 퐴푃푖) (1) capability are analyzed separately, and after these where ri is the return of individual company i, four major categorized factors are obtained, the ARi refers to the abnormal returns of individual principal component analysis are used to capture company i, 푅푚,푖 and 푅퐼,푖 are the market the characteristics of the corporate fundamental information.

퐶푃 = 0.98 × (0.88퐶푅 + 0.91푄푅 + 0.69퐷퐴) + 0.86 × (0.90퐺푃푀 + 0.45푅푂퐸) + 0.54 × (0.64훥푂퐼 + 0.58훥푇퐴) − 0.62 × (0.31푅푇 + 0.52퐼푇) (2)

This study will first collect all the analysis reports This study uses the panel regression to (including Internet articles) for the last 3 months of understand the relationship between noise trading all the months of each individual , and and abnormal returns . give standardized weights on the numbers of reports in the range from 0 to 1. ퟑ ퟑퟎ ퟖퟑ

푨푹풊,풕 = 휶ퟎ,풊 + 휷ퟏ,풊풕푻푹풊,풕 + ∑ 푾풌ퟏ풕푫풌ퟏ풕 + ∑ 푼풌ퟐ풕푭풌ퟐ풕 + ∑ 푽풌ퟑ풕푻풌ퟑ풕 + 풖풊.풕 풌ퟏ=ퟏ 풌ퟐ=ퟏ 풌ퟑ=ퟏ (3) where TR represents the turnover rate of the the panel differences and the dummy variable T individual company, the dummy variable F captures captures the time differences.

∑3 푟푖푡 = 휶ퟎ,풊 + 훽1,푖푡푇푅푖푡 + 훽2,푖푡푅푚,푡 + 훽3,푖푡푅퐼,푖푡 + 훽4,푖푡(퐶푃 × 퐴푆)푖푡 + 푘1=1 푊푘1푡퐷푘1푡 + ∑30 ∑83 푘2=1 푈푘2푡퐹푘2푡 + 푘3=1 푉푘3푡푇푘3푡 + 푢푖푡 (4) during September 2014 to June 2015. Therefore, this Fig.2 shows the turnover rate of each individual study divided the whole sample period into three company. The Chinese is in a bull time periods to distinguish whether if there is any market due to lots of companies listed in ChiNext difference between the bull market and the bear and lots of hot money invested in the stock market market.

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA

321 Chiu-Lan Chang, Ming Fang*

Fig.1 Cluster Analysis of the characteristics of companies

Fig.2 Turnover rate of individual companies

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA 322 Chiu-Lan Chang, Ming Fang*

4.2 Empirical Results (CP×AS) all significantly positively affect the stock Using the Hausman test, panel data exhibits a return. The results indicate that the fundamental fixed effect on time and cross-section in the first performance and market analyst sentiment all period (January 2012 to August 2014) and the third positively related with the stock return. The period (July 2015 to December 2018), and the data turnover rate (TR) all significantly positively affect of the second period (September 2014 to June 2015) the stock return which indicate that the noise shows the random effect. Table 2 shows the results trading affects the stock return. of panel regression. The stock returns significantly To understand the relationship between the positively affected by the market performance (Rm). abnormal return and the noise trading, the Granger Compared with the coefficient of Rm, in bull market causality test is applied. The correlation coefficient (the second period) the marginal effect of the between the abnormal return and the turnover rate market performance on stock returns is 0.782 which is 0.8254 significantly. The Granger causality test is is five times larger than the marginal effect in bear reported in Table 3. With the Granger causality test, market. The stock returns also significantly the mutual causal interactions with noise trading positively affected by the industry performance (RI). and abnormal returns are detected. The results Compared with the coefficient of RI, the marginal indicate that the higher the noise trading, the higher effects are very close with the values of 0.826, 0.840 the abnormal return. The noise trading positively and 0.817. The cross product of characteristics of affects the abnormal return of the company. individual company and the analyst’s reports

Table 1 Financial Performance Indicators Financial Performance Variable Current ratio(CR) debt-paying ability Quick Ratio (QR) Debt Asset ratio (DA) Gross Profit (GPM) Profitability Return on (ROE) Operating income growth rate (ΔOI) Growth ability Total Assets Growth Rate (ΔTA) Accounts receivable turnover rate (RT) Operational capability Turnover Rate (IT)

Table 2 Panel Regression Results January 2012 to September 2014 to July 2015 to August 2014 June 2015 December 2018 (Bear Market) (Bull Market) (Bear Market) Intercept -0.010 *** 0.005 -0.005 *** *** *** Rm 0.071 0.782 0.141 *** *** *** RI 0.826 0.840 0.817 CP×AS 0.008 *** 0.008 *** 0.008*** TR 0.092*** 0.097 *** 0.066 *** *** ** *** D1 -0.005 -0.031 0.017 *** * D2 0.002 0.029 0.001 * *** D3 0.009 -0.051 0.012 Notes: ***, **, * represent the significance level at 1%, 5% and 10%, respectively.

Table 3 Granger Causality Test Null Hypothesis: Obs F-Statistic p-value

EXCESS does not Granger Cause TR 2602 53.2952 2 × 10−23 TR does not Granger Cause EXCESS 12.4703 4 × 10−6

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA

323 Chiu-Lan Chang, Ming Fang*

5. Conclusions financial markets. Journal of political Economy, This study applies the cluster analysis to 98(4), 703-738. categorize the financial condition of the company. Fama, E. F. (1970). Efficient capital markets: A After using the cluster analysis, this study uses the review of theory and empirical work. The principal component analysis to compose the factor journal of Finance, 25(2), 383-417. which reflects the characteristics of each individual Easley, D., Kiefer, N. M., O'hara, M., & Paperman, J. company. B. (1996). Liquidity, information, and Then, this study applies the panel data regression infrequently traded stocks. The Journal of to analyze relationship of noise trading and abnormal Finance, 51(4), 1405-1436. returns . This study collects the analysts’ reports to Hirshleifer, D., & Luo, G. Y. (2001). On the survival of represent the institutional investors’ sentiment and overconfident traders in a competitive uses the turnover rate to capture the noise trading of securities market. Journal of Financial Markets, the stock market. To investigate if there is any 4(1), 73-84. difference between the bear market and the bull Ketchen, D. J., & Shook, C. L. (1996). The application market, the samples are divided into three periods. of cluster analysis in strategic The results show that the market performance research: an analysis and critique. Strategic affects the stock returns significantly positively, management journal, 17(6), 441-458. especially in the bull market, the market Lee, C. M., Shleifer, A., & Thaler, R. H. (1991). performance affects the stock returns larger than in Investor sentiment and the closed‐end fund the bear market. The fundamental performance and puzzle. The journal of finance, 46(1), 75-109. the institutional investors’ sentiments positively Lemmon, M., & Portniaguina, E. (2006). Consumer affect the stock returns. The noise trading also affects confidence and asset prices: Some empirical the stock returns. evidence. The Review of Financial Studies, To understand the relationship between the 19(4), 1499-1529. abnormal returns and the noise trading, the Granger Ramiah, V., & Davidson, S. (2003). Noise trader and causality test is applied. The results show that the a behavioural asset pricing model: australian abnormal returns and the noise trading are mutual evidence. Working Papers of RMIT University, causal interacted with each other. The results of this 574. study can be beneficial for the investors to make Tversky, A., & Kahneman, D. (1974). Judgment . under uncertainty: Heuristics and . Science, 185(4157), 1124-1131. Acknowledgment This research was supported by Social Science Fund of Fujian Province, China under Grant number FJ2018B075 and Fuzhou University of International Studies and Trade under Grant number FWKQJ201904 and Grant number FWKQJ202003.

References Baker, M., & Stein, J. C. (2004). as a sentiment indicator. Journal of Financial Markets, 7(3), 271-299. De Long, J. B., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1988). The survival of noise traders in financial markets (No. w2715). National Bureau of Economic Research. De Long, J. B., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1989). The size and incidence of the losses from noise trading. The Journal of Finance, 44(3), 681-696. De Long, J. B., Shleifer, A., Summers, L. H., & Waldmann, R. J. (1990). Noise trader risk in

REVISTA ARGENTINA 2020, Vol. XXIX, N°4, 317-324 DE CLÍNICA PSICOLÓGICA