BANKING UNIVERSITY OF HO CHI MINH CITY NGUYEN THIEN NHAN EXAMINING HERDING BEHAVIOR IN VIETNAMESE STOCK MARKET GRADUATION THESIS MAJOR FINANCE – BANKING CODE 52340201 Ho Chi Minh City, April 2022 THE.
THE STATE BANK OF VIETNAM MINISTRY OF EDUCATION AND TRAINING BANKING UNIVERSITY OF HO CHI MINH CITY NGUYEN THIEN NHAN EXAMINING HERDING BEHAVIOR IN VIETNAMESE STOCK MARKET GRADUATION THESIS MAJOR: FINANCE – BANKING CODE: 52340201 Ho Chi Minh City, April 2022 THE STATE BANK OF VIETNAM MINISTRY OF EDUCATION AND TRAINING BANKING UNIVERSITY OF HO CHI MINH CITY NGUYEN THIEN NHAN EXAMINING HERDING BEHAVIOR IN VIETNAMESE STOCK MARKET GRADUATION THESIS MAJOR: FINANCE – BANKING CODE: 52340201 ACADEMIC ADVISOR DUONG THI THUY AN, PHD Ho Chi Minh City, April 2022 DECLARATION I honestly declare that this graduation thesis entitled “Examining herding behavior in Vietnamese stock market” is the final result of my original research work under the guidance of PhD Duong Thi Thuy An, my academic advisor This thesis has never been submitted for a master's degree at any anywhere else before This thesis is the author's own research and the results of the research are trustworthy The thesis does not consist of any previously published content or content made by others except forcitations which are fully cited in the thesis Ho Chi Minh City, April 2022 Nguyen Thien Nhan i ACKNOWLEDGEMENT First of all, I would like to express my special thanks to my academic advisor, PhD Duong Thi Thuy An Thank you for always supporting me whenever I need your help and all the knowledge you have taught me during my university journey Sedconly, I would like to sincerely thank to my parents, who always take care and encourage me during the time doing this thesis I will never forget the love you gave me! Last but not least, I want to say thanks to my friend, Bao Phuong, who has appeared at the hardest time to help me ii LIST OF ABBREVIATIONS Abbreviations Definition CCK Chang, Cheng and Khorana Model CH Christie and Huang’s Model CSAD Cross-sectional absolute deviation EMH Efficient Market Hypothesis GDP Gross Domestic Product HNX Ha Noi Stock Exchange HS Hwang and Salmon Model HSX Ho Chi Minh Stock Exchange NS Nofinger and Sias Model OLS Ordinary Least Squared OTC Over-the-Counter market QREG Quantile Regession TIP Trading Imbalance Picture VND Vietnam Dong iii LIST OF TABLES Name of tables Table 2.1 A summary of relevant empirical research about herding behavior Table 3.1 Summary of data observations used in the study Table 4.1 Summary statistics of cross-sectional absolute deviation and absolute market return Page 13 19 26 Table 4.2 T-test result 28 Table 4.3 Correlation among variables 29 Table 4.4 Regression result of herding behavior in Vietnamese stock market Table 4.5 Regression result of herding in up and down market Table 4.6 Regression result of herding in high and low trading volume 29 30 31 Table 4.7 Table summary of regression results 32 Table 4.8 Summary of regression results 36 iv TABLE OF CONTENTS DECLARATION i ACKNOWLEDGEMENT ii LIST OF ABBREVIATIONS iii LIST OF TABLES iv CHAPTER 1: INTRODUCTION 1.1 Research background 1.2 Research gap 1.3 Research question 1.4 Research objectives 1.5 Research scope and methodology 1.6 Research structure CHAPTER 2: LITERATURE REVIEW 2.1 Theoretical literature review 2.3 Herding in different conditions 10 2.2 Empirical literature review 11 2.3 Hypothesis development 19 CHAPTER 3: RESEARCH METHODOLOGY 21 3.1 Data collection and sample description 21 3.2 Regression model for testing the hypotheses 22 3.2.1 Regression model 22 3.2.2 Regression model for estimation the degree of herd in rising and falling market: 24 v 3.2.3 Regression model for estimation the degree of herd in high and low volume trading market: 24 3.3 Regression methodology 25 3.3.1 Research process when using OLS 25 3.3.2 Quantile regression analysis 25 CHAPTER 4: EMPIRICAL RESULT 28 4.1 Descriptive statistics 28 4.2 Testing for mean different of CSAD 30 4.3 Correlation analysis among variables 30 4.4 Regression result 31 4.4.1 Evidence of herding behavior in Vietnamese stock market 31 4.4.2 Herding in up and down market 32 4.4.3 Herding behavior and trading volume 33 4.5 Quantile regression result 34 CHAPTER 5: CONCLUSION AND IMPLICATIONS 39 5.1 Conclusion 40 5.2 Implication for herding in Vietnamese stock market 41 5.3 Limitations and further research direction 42 REFERENCES 43 APPENDICES 48 Descriptive Statistic 48 Appendix 1: Correlation between variables 51 Appendix 2: Regression results to test for the presence of herding behavior 51 vi Appendix 3: Regression results to test for the level of herding behavior in up and down markets 51 Appendix 4: Regression results to test for the level of herding behavior in high and low market trading volume 52 Appendix 5: Quantile regression analysis to test for the presence of herding behavior 53 Appendix 5: T-test 64 vii CHAPTER 1: INTRODUCTION This chapter depicts the introduction of the research It comprises the research background, research question, research objective, research gap, research methodology, scope, and research structure 1.1 Research background The Vietnamese stock market has experienced 21 years from its foundation in 1998, including Ho Chi Minh Stock Exchange (HSX) and Ha Noi Stock Exchange With only two listed companies at the early stage in 2000, the Vietnamese stock market has undergone many ups and downs with lots of memorable milestones On September 30, 2021, the Vietnamese stock market had 2,133 listed stocks The market value has reached over 8.3 million billion VND, equivalent to 133.83% of GDP (according to the State Securities Commission of Vietnam) In the recent two years, this emerging market has entered a new period of development with many impressive mileposts The value of the VN-Index has increased significantly, especially since March 2020; the Vietnamese stock market has broken many records in terms of liquidity as well as the number of new accounts At the end of 2021, there were 52 listed companies having capitalization reaching more than billion USD, VN-Index increased from about 1120 points to 1498.28 points for gaining 378 points in a year, and the average daily trading volume was more than 26.560 trillion Dong in 2021 The first outbreak of the Covid-19 pandemic occurring in March 2020 has affected the market index adversely VN-Index dropped from 1000 to 650 in March 2020, equivalent to losing 35% points Afterwards, the market has recovered from the bottom and experienced am impressive growth until 2021 at 1498.28 points During the period from March 2020 to December 2021, investors also experienced many strong downward, unpredictable trading sessions and were difficult to explain Then, at each milestone where the VN-Index was about to break the new record in value, the market experienced strong declines from 100 points to 200 points So, is the market really efficient and does any mispricing or bubbles exists in our equity market? After a Jarque-Bera 859.3029 403.0920 Probability 0.000000 0.000000 Sum 13.36934 0.925751 Sum Sq Dev 0.008311 0.067102 Observations 736 736 Appendix 1: Correlation between variables CSAD ABS_RET SQUARED_RET CSAD 0.729655 0.600722 ABS_RET 0.729655 0.896884 SQUARED_RET 0.600722 0.896884 Appendix 2: Regression results to test for the presence of herding behavior Dependent Variable: CSAD Method: Least Squares Date: 03/15/22 Time: 09:50 Sample: 1/05/2016 12/31/2021 Included observations: 1500 Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.444145 -3.083936 0.015869 0.017899 0.441805 0.000111 24.81411 -6.980308 143.3190 0.0000 0.0000 0.0000 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.547137 0.546532 0.002501 0.009364 6859.708 904.3180 0.000000 Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.018856 0.003714 -9.142278 -9.131651 -9.138319 1.012028 Appendix 3: Regression results to test for the level of herding behavior in up and down markets Herding in down market Dependent Variable: CSDA Method: Least Squares Date: 03/15/22 Time: 09:02 Sample: 634 51 Included observations: 634 Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.485782 -4.305972 0.015693 0.026663 0.588412 0.000172 18.21951 -7.317960 91.24843 0.0000 0.0000 0.0000 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.577465 0.576126 0.002665 0.004482 2859.921 431.1836 0.000000 Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.018951 0.004094 -9.012368 -8.991302 -9.004188 1.143549 Herding in up market Dependent Variable: CSDA Method: Least Squares Date: 03/15/22 Time: 09:12 Sample: 866 Included observations: 866 Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.324723 1.695330 0.016293 0.027450 0.882650 0.000149 11.82975 1.920728 109.5927 0.0000 0.0551 0.0000 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.535010 0.533932 0.002328 0.004677 4023.028 496.4763 0.000000 Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.018787 0.003410 -9.284130 -9.267627 -9.277814 1.090113 Appendix 4: Regression results to test for the level of herding behavior in high and low market trading volume Herding in high market trading volume Dependent Variable: CSDA Method: Least Squares Date: 03/15/22 Time: 09:55 Sample: 764 Included observations: 764 Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.460290 -3.943196 0.016268 0.025247 0.563029 0.000169 18.23118 -7.003541 96.17257 0.0000 0.0000 0.0000 R-squared Adjusted R-squared S.E of regression Sum squared resid 0.524737 0.523488 0.002700 0.005549 Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion 52 0.019522 0.003912 -8.987055 -8.968841 Log likelihood F-statistic Prob(F-statistic) 3436.055 420.1092 0.000000 Hannan-Quinn criter Durbin-Watson stat -8.980043 0.995528 Herding in low market trading volume Dependent Variable: CSDA Method: Least Squares Date: 03/15/22 Time: 10:03 Sample: 736 Included observations: 736 Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.313768 2.225888 0.015840 0.027986 0.900380 0.000142 11.21143 2.472167 111.7273 0.0000 0.0137 0.0000 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.596776 0.595676 0.002138 0.003351 3481.947 542.4245 0.000000 Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 0.018165 0.003363 -9.453660 -9.434905 -9.446427 1.045186 Appendix 5: Quantile regression analysis to test for the presence of herding behavior Overall market Dependent Variable: CSAD Method: Quantile Regression (tau = 0.1) Date: 03/23/22 Time: 08:34 Sample: 1/05/2016 12/31/2021 Included observations: 1500 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.030225 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.451417 -4.120082 0.013294 0.021723 0.410245 0.000138 20.78054 -10.04297 96.15590 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression 0.271995 0.271022 0.003656 Mean dependent var S.D dependent var Objective 53 0.018856 0.003714 0.511217 Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.015023 0.009078 0.000000 Restr objective Quasi-LR statistic 0.702216 467.5358 Dependent Variable: CSAD Method: Quantile Regression (tau = 0.25) Date: 03/23/22 Time: 08:34 Sample: 1/05/2016 12/31/2021 Included observations: 1500 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.058782 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.429165 -3.076586 0.014379 0.018756 0.353722 0.000118 22.88101 -8.697758 121.6514 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.307208 0.306283 0.002974 Mean dependent var S.D dependent var Objective 0.018856 0.003714 0.977201 0.016299 0.005567 0.000000 Restr objective Quasi-LR statistic 1.410527 830.2212 Dependent Variable: CSAD Method: Quantile Regression (Median) Date: 03/23/22 Time: 08:35 Sample: 1/05/2016 12/31/2021 Included observations: 1500 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.084873 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.406023 -1.464275 0.015572 0.023817 0.776504 0.000123 17.04758 -1.885728 126.3713 0.0000 0.0595 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var 0.337882 0.336997 0.002547 Mean dependent var S.D dependent var Objective 0.018856 0.003714 1.334180 0.018149 Restr objective 2.015019 54 Sparsity Prob(Quasi-LR stat) 0.004753 0.000000 Quasi-LR statistic 1145.900 Dependent Variable: CSAD Method: Quantile Regression (tau = 0.7) Date: 03/23/22 Time: 08:35 Sample: 1/05/2016 12/31/2021 Included observations: 1500 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.066914 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.362456 0.267154 0.016888 0.041072 1.635353 0.000177 8.824894 0.163361 95.28526 0.0000 0.8703 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.353737 0.352873 0.002696 Mean dependent var S.D dependent var Objective 0.018856 0.003714 1.284947 0.019801 0.006705 0.000000 Restr objective Quasi-LR statistic 1.988272 998.9978 Dependent Variable: CSAD Method: Quantile Regression (tau = 0.9) Date: 03/23/22 Time: 08:35 Sample: 1/05/2016 12/31/2021 Included observations: 1500 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.030225 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.307113 1.615778 0.019386 0.046271 1.261506 0.000257 6.637263 1.280832 75.41331 0.0000 0.2005 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.367137 0.366292 0.004023 Mean dependent var S.D dependent var Objective 0.018856 0.003714 0.787316 0.023721 0.019182 0.000000 Restr objective Quasi-LR statistic 1.244055 529.1344 Up market 55 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.1) Date: 03/15/22 Time: 10:14 Sample: 866 Included observations: 866 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.036299 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.395403 -2.291212 0.013618 0.041502 1.312363 0.000200 9.527328 -1.745867 68.21895 0.0000 0.0812 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.259964 0.258249 0.003486 Mean dependent var S.D dependent var Objective 0.018787 0.003410 0.283191 0.015266 0.008790 0.000000 Restr objective Quasi-LR statistic 0.382671 251.5040 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.25) Date: 03/15/22 Time: 10:15 Sample: 866 Included observations: 866 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.070594 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.330012 0.718471 0.014909 0.097076 6.039640 0.000253 3.399539 0.118959 58.98085 0.0007 0.9053 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.294490 0.292855 0.002744 Mean dependent var S.D dependent var Objective 0.018787 0.003410 0.543246 0.016397 0.005211 0.000000 Restr objective Quasi-LR statistic 0.770005 464.1710 Dependent Variable: CSDA Method: Quantile Regression (Median) Date: 03/15/22 Time: 10:15 Sample: 866 Included observations: 866 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.10193 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic 56 Prob ABS_RET SQUARED_RET C Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.354210 2.273662 0.015710 0.025786 0.908643 0.000146 13.73650 2.502261 107.4654 0.0000 0.0125 0.0000 0.320475 0.318901 0.002370 Mean dependent var S.D dependent var Objective 0.018787 0.003410 0.737721 0.018249 0.004724 0.000000 Restr objective Quasi-LR statistic 1.085643 589.2399 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.7) Date: 03/15/22 Time: 10:15 Sample: 866 Included observations: 866 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.080361 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.303523 3.804329 0.016992 0.046339 1.885417 0.000209 6.550097 2.017766 81.34516 0.0000 0.0439 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.330368 0.328816 0.002467 Mean dependent var S.D dependent var Objective 0.018787 0.003410 0.699695 0.019853 0.005926 0.000000 Restr objective Quasi-LR statistic 1.044895 554.7466 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.9) Date: 03/15/22 Time: 10:16 Sample: 866 Included observations: 866 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.036299 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.168488 7.385814 0.019838 0.072317 2.225549 0.000423 2.329844 3.318648 46.88239 0.0200 0.0009 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.335627 0.334087 0.003813 Mean dependent var S.D dependent var Objective 0.018787 0.003410 0.421986 0.023126 0.019423 0.000000 Restr objective Quasi-LR statistic 0.635165 243.9063 57 Down market Dependent Variable: CSDA Method: Quantile Regression (tau = 0.1) Date: 03/15/22 Time: 10:19 Sample: 634 Included observations: 634 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.040275 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.495166 -4.749798 0.012962 0.032237 0.569301 0.000228 15.36007 -8.343206 56.73258 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.285830 0.283567 0.003819 Mean dependent var S.D dependent var Objective 0.018951 0.004094 0.225357 0.014777 0.009489 0.000000 Restr objective Quasi-LR statistic 0.315551 211.2220 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.25) Date: 03/15/22 Time: 10:19 Sample: 634 Included observations: 634 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.078327 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.468352 -3.792680 0.014105 0.027789 0.545878 0.000170 16.85367 -6.947852 82.95441 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.329795 0.327671 0.003137 Mean dependent var S.D dependent var Objective 0.018951 0.004094 0.428682 0.016197 0.005515 0.000000 Restr objective Quasi-LR statistic 0.639629 408.0225 Dependent Variable: CSDA Method: Quantile Regression (Median) Date: 03/15/22 Time: 10:20 Sample: 634 Included observations: 634 Huber Sandwich Standard Errors & Covariance 58 Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.11309 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.415513 -1.983320 0.015511 0.040252 1.210869 0.000208 10.32277 -1.637932 74.60671 0.0000 0.1019 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.366049 0.364040 0.002736 Mean dependent var S.D dependent var Objective 0.018951 0.004094 0.588943 0.018102 0.004943 0.000000 Restr objective Quasi-LR statistic 0.929004 550.4140 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.7) Date: 03/15/22 Time: 10:20 Sample: 634 Included observations: 634 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.089163 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.363477 -0.982953 0.016910 0.037489 1.037673 0.000229 9.695688 -0.947267 73.87487 0.0000 0.3439 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.391500 0.389572 0.002840 Mean dependent var S.D dependent var Objective 0.018951 0.004094 0.573703 0.019709 0.006860 0.000000 Restr objective Quasi-LR statistic 0.942815 512.4370 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.9) Date: 03/15/22 Time: 10:20 Sample: 634 Included observations: 634 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.040275 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.429210 -2.010341 0.019087 0.138298 4.086313 0.000479 3.103518 -0.491969 39.83746 0.0020 0.6229 0.0000 Pseudo R-squared 0.396203 Mean dependent var 59 0.018951 Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.394289 0.004289 S.D dependent var Objective 0.004094 0.360379 0.024849 0.022940 0.000000 Restr objective Quasi-LR statistic 0.596854 229.0761 High market trading volume Dependent Variable: CSDA Method: Quantile Regression (tau = 0.1) Date: 03/15/22 Time: 10:21 Sample: 764 Included observations: 764 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.037848 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.457442 -4.312134 -4.312134 0.029566 0.536644 0.000202 15.47172 -8.035366 67.67307 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.266318 0.264390 0.003834 Mean dependent var S.D dependent var Objective 0.019522 0.003912 0.276887 0.015356 0.009369 0.000000 Restr objective Quasi-LR statistic 0.377394 238.3999 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.25) Date: 03/15/22 Time: 10:22 Sample: 764 Included observations: 764 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.073606 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.456358 0.456358 0.014639 0.025987 0.545351 0.000155 17.56084 -6.888240 94.37251 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.322725 0.320945 0.003158 Mean dependent var S.D dependent var Objective 0.019522 0.003912 0.523274 0.016722 0.005359 0.000000 Restr objective Quasi-LR statistic 0.772618 496.3075 60 Dependent Variable: CSDA Method: Quantile Regression (Median) Date: 03/15/22 Time: 10:22 Sample: 764 Included observations: 764 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.10628 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.447868 -2.882867 0.015718 0.054150 1.950160 0.000241 8.270857 -1.478272 65.09124 0.0000 0.1397 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.338649 0.336911 0.002765 Mean dependent var S.D dependent var Objective 0.019522 0.003912 0.727972 0.018792 0.005012 0.000000 Restr objective Quasi-LR statistic 1.100733 594.9410 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.7) Date: 03/15/22 Time: 10:22 Sample: 764 Included observations: 764 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.083789 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.371228 -1.059390 0.017254 0.042932 1.405633 0.000247 8.646854 -0.753675 69.77764 0.0000 0.4513 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.345851 0.344132 0.002842 Mean dependent var S.D dependent var Objective 0.019522 0.003912 0.712332 0.020628 0.006868 0.000000 Restr objective Quasi-LR statistic 1.088945 522.2809 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.9) Date: 03/15/22 Time: 10:22 Sample: 764 Included observations: 764 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.037848 Estimation successfully identifies unique optimal solution 61 Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.377263 -1.303789 0.019840 0.097029 2.933464 0.000462 3.888147 -0.444454 42.94591 0.0001 0.6568 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.332637 0.330883 0.004286 Mean dependent var S.D dependent var Objective 0.019522 0.003912 0.439668 0.024986 0.020278 0.000000 Restr objective Quasi-LR statistic 0.658813 240.1587 Low market trading volume Dependent Variable: CSDA Method: Quantile Regression (tau = 0.1) Date: 03/15/22 Time: 10:23 Sample: 736 Included observations: 736 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.038322 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.365173 -0.351309 0.013423 0.032155 1.013369 0.000162 11.35672 -0.346675 82.74089 0.0000 0.7289 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.281655 0.279695 0.003162 Mean dependent var S.D dependent var Objective 0.018165 0.003363 0.226382 0.014635 0.007933 0.000000 Restr objective Quasi-LR statistic 0.315144 248.6376 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.25) Date: 03/15/22 Time: 10:24 Sample: 736 Included observations: 736 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.074527 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.374940 -0.632164 0.014295 0.026869 0.672191 0.000166 13.95440 -0.940454 86.36724 0.0000 0.3473 0.0000 Pseudo R-squared 0.302542 Mean dependent var 62 0.018165 Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.300639 0.002572 S.D dependent var Objective 0.003363 0.434679 0.015964 0.005433 0.000000 Restr objective Quasi-LR statistic 0.623233 370.1903 Dependent Variable: CSDA Method: Quantile Regression (Median) Date: 03/15/22 Time: 10:24 Sample: 736 Included observations: 736 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.10761 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.309754 2.556054 0.015610 0.044623 2.302406 0.000180 6.941520 1.110167 86.49416 0.0000 0.2673 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.333451 0.331632 0.002151 Mean dependent var S.D dependent var Objective 0.018165 0.003363 0.581679 0.017607 0.004568 0.000000 Restr objective Quasi-LR statistic 0.872673 509.6611 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.7) Date: 03/15/22 Time: 10:24 Sample: 736 Included observations: 736 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.084838 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.242645 5.915658 0.016760 0.046392 1.918078 0.000197 5.230284 3.084159 85.15350 0.0000 0.0021 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.370639 0.368922 0.002312 Mean dependent var S.D dependent var Objective 0.018165 0.003363 0.540838 0.019006 0.005533 0.000000 Restr objective Quasi-LR statistic 0.859344 548.2104 Dependent Variable: CSDA Method: Quantile Regression (tau = 0.9) 63 Date: 03/15/22 Time: 10:24 Sample: 736 Included observations: 736 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.038322 Estimation successfully identifies unique optimal solution Variable Coefficient Std Error t-Statistic Prob ABS_RET SQUARED_RET C 0.162896 8.251101 0.018933 0.067244 2.014454 0.000367 2.422459 4.095950 51.54321 0.0157 0.0000 0.0000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.417150 0.415560 0.003440 Mean dependent var S.D dependent var Objective 0.018165 0.003363 0.320796 0.022212 0.018260 0.000000 Restr objective Quasi-LR statistic 0.550393 279.4156 Appendix 5: T-test Between Overall and Up Market t-Test: Two-Sample Assuming Equal Variances Mean Variance Observations Pooled Variance Hypothesized Mean Difference df t Stat P(T