(Luận văn) herding behavior in vietnamese stock market, empirical evidence from quantile regression analysis

UNIVERSITY OF ECONOMICS HO CHI MINH CITY t to International School of Business ng hi ep w n lo Phan Dang Bao Anh ad ju y th yi pl al n ua HERDING BEHAVIOR IN VIETNAMESE STOCK n va ll fu MARKET: EMPIRICAL EVIDENCE FROM oi m at nh QUANTILE REGRESSION ANALYSIS z z ht vb k jm om l.c gm MASTER OF BUSINESS (Honours) n a Lu n va y te re Ho Chi Minh City – Year 2015 UNIVERSITY OF ECONOMICS HO CHI MINH CITY t to International School of Business ng hi ep w Phan Dang Bao Anh n lo ad y th ju HERDING BEHAVIOR IN VIETNAMESE STOCK yi pl al n ua MARKET: EMPIRICAL EVIDENCE FROM n va ll fu QUANTILE REGRESSION ANALYSIS oi m at nh ID: 22130006 z z ht vb k jm MASTER OF BUSINESS (Honours) om l.c gm SUPERVISOR: A.Pro.Dr VO XUAN VINH n a Lu n va y te re Ho Chi Minh City – Year 2015 ACKNOWLEDGEMENT t to Firstly, I would like to express my gratefulness to my supervisor A.Prof Dr.Vo ng Xuan Vinh for his professional guidance, intensive support, valuable suggestions, hi ep instructions and continuous encouragement during the time of research and writing this thesis w n lo I would like to express my deepest appreciation to ISB Research Committee for ad their valuable time as their insightful comments and meaningful suggestions were y th contributed significantly for my completion of this research ju yi My sincere thanks also go to all of all of my lecturers at International Business pl n my Master course ua al School- University of Economics Ho Chi City for their teaching and guidance during va n Last but not least, I would like to thanks my family, whom were always fu ll supporting me and encouraging me with their best wishes oi m at nh z z ht vb k jm om l.c gm n a Lu n va y te re t to TABLE OF CONTENT ng hi ep CHAPTER 1: INTRODUCTION 1.1 Research background w 1.2 Research gap n lo 1.3.Research objectives ad 1.4 Research methodology and scope y th ju 1.5 Research structure yi CHAPTER 2: LITERATURE REVIEW pl al 2.1 Theoretical literature review n ua 2.2 Empirical literature review 11 va 2.3 Measuring herding in financial markets 19 n 2.4 Hypothesis development 25 fu ll CHAPTER 3: RESEARCH METHODOLOGY .27 m oi 3.1 Data collection and sample description 27 nh 3.2 Regression model for testing the hypotheses 28 at z 3.2.1 Regression model for testing the presence of herding bahavior in Vietnamese stock market: 28 z vb ht 3.2.2 Regression model for estimation the degree of herd in rising and falling market: 29 jm k 3.3 Regression methodology 30 gm 3.3.1 Research process 31 om l.c 3.3.2 Quantile regression analysis 31 CHAPTER 4: EMPIRICAL RESULT .34 a Lu 4.1 Decriptive statistics 34 y 4.4 Regression result from Quantile regression analysis 39 te re 4.3.2 Herding behavior in up and down markets 38 n 4.3.1 Evidence on herd presence in Vietnamese stock market 36 va 4.3 Regression result 36 n 4.2 Correlation analysis among variables 35 CHAPTER 5: CONCLUSION AND IMPLICATIONS 45 t to 5.1 Conclusion 45 ng 5.2 Implications of herding behavior in Vietnamese stock market 46 hi ep 5.3 Limitations and further research direction 48 REFERENCES 50 w APPENDICES 55 n lo ad ju y th yi pl n ua al n va ll fu oi m at nh z z ht vb k jm om l.c gm n a Lu n va y te re LIST OF TABLES t to Table 1.1: A summary of empirical evidence on herding behavior 15 ng Table 3.1: Summary of data observations used in the study 27 hi ep Table 4.1: Descriptive statistics for daily market return and cross-sectional absolute deviation (CSAD) for the Vietnamese stock market from 1/2005 to 4/2015 34 w Table 4.2 Correlation among main variables 35 n lo Table 4.3: Regression result of herding behavior in Vietnamese stock market 36 ad Table 4.4: Regression results of herding behavior in rising and declining market 38 y th ju Table 4.5: Analysis of herding behavior in Vietnamese stock market by quantile regression 40 yi pl Table 4.6: A summary of research results 43 n ua al n va ll fu oi m at nh z z ht vb k jm om l.c gm n a Lu n va y te re ABSTRACT t to This study examines the herding behavior of investors in Vietnamese stock ng market using data sample of 299 companies listed on Ho Chi Minh City Stock hi ep Exchange Using a least square method, the author finds evidence of herding presence in rising and falling market when considering over the period of 2005 – 4/2015 as well w n as in the periods of pre-crisis and post-crisis By applying quantile regression analysis lo ad to estimate the herding equation, the author find supporting evidence of herding during y th the period studied as well as when splitting the market into two sub-periods; however, ju the level of this trend is somewhat different conditional on quantile region yi pl Key words: herding behavior, Vietnamese stock market, quantile regression, n ua al asymmetry n va ll fu oi m at nh z z ht vb k jm om l.c gm n a Lu n va y te re CHAPTER 1: INTRODUCTION t to This chapter presents the introduction of the study It contains the research ng background, research gap, research objectives, research methodology and scope and hi ep research structure 1.1 Research background w n lo Traditional financial framework understands financial market by using models ad which meet four foundation conditions: (i) investors are assumed to be rational, (ii) y th ju market is efficient, (iii) investors make a decision on portfolios based on the rules of yi mean-variance portfolio theory, and (iv) the expected returns are a function of risk pl ua al (Statman, 2014) Among them, the condition of rational investor is considered a central assumption in which people make decisions reasonably and no biases in their n n va future prediction However, the world economy has been shaken by the global ll fu financial crisis in 2008, which originated from US and then expanded globally As oi m soon as the crisis began, many economists and financial forecasters were no longer nh able to analyze the bankruptcy of a variety of enterprises or banks in an intensive way at The Vietnamese stock market is not an exception From its foundation in 2000, z z the Vietnamese stock market experiences “hot” growth and drastical fluctuation vb ht without stability causing virtual stock matter The value of VN-Index in 2000 of 100 k jm points increases to 571 points after just one year and a half which astonishes economic gm experts; however, this increment does not last long and rush to fall under 140 points in om 2007 as Vietnamese stock market has the highest growth of 1100 points l.c 2003, 150-200 points in 2004 The peak of growing phase is in the period of 2006- n va points on March 12th, 2007 – the highest level in the world This event makes stock n market growth of 135% Particularly, the VN-Index reaches to the record of 1170.67 a Lu (approximately 145%) in Asia – Pacific region, even exceeding the Shanghai stock y bubble formation in the stock market te re experts and market managers difficult to understand, thereby bring out the fear of After a long time of increasing prices, the Vietnamese stock market has signal t to to considerably decrease with the lowest record of 236 points on February 24th, 2009 ng The happening in the market during this period is very complicated to anticipate Once hi ep again, economics experts doubt the precision of the efficient market theory A paradox is present that when the stock price is driven further from the fundamental value of w n 30% investors still trade constantly; whereas, when the stock prices decrease at an lo ad attractive level in declining market investors massively sell stocks instead of buying Is y th it true that the Vietnamese stock market operation does not abide by any rules or there ju are phenomena dominating the market which cause an unusual fluctuation? Failure of yi pl the economists as well as their theories leads to a list of different questions in different ua al context: Are people rational? Or are they influenced by emotion such as fear, greed n which caused wrong decision? va n Then, a new branch of financial research appears beside traditional financial fu ll framework which helps economic experts and finance researchers partial explain m oi unusual fluctuation Behavioral finance is a new strand of finance which investigates at nh the behavior of investors in financial market; in other word, it is a combination between psychology and finance It considers psychological factors as essential input z z to financial analysis Behavioral finance can elucidate several financial reactions that vb ht contrast with standard financial theory and can thus make a contribution to avoidance jm of mistakes as well as advancing investment strategies (Fromlet, 2001) k gm Previous researchers put sustained effort to understand investors’ behaviour in l.c the market as well as its impact on stock price These investment behaviors are om influenced by some factors such as investors’ insight, criterion to measure investment rational and always strive to optimize their actions but the fact that the rationality the actions of others, which engages in herd behavior y on the investment behavior of market participants regarding to their tendency to follow te re optimism, herding, representativeness … and so on In this research, the author focuses n va appears to be inhibited by numerous cognitive biases, such as overconfidence, over- n a Lu efficiency or market instability… In terms of psychology, investors are assumed to be Herding behavior is defined as the trend of investors to imitate the actions of t to others (Luu, 2013) This tendency is considered an inherent psychology of investors ng but it becomes stronger as they have to make decision in a market condition with high hi ep uncertainty and low transparency Over last decades, research regarding this topic receives an attention from scientists and empirical researchers A numerous theories w n are developed and empirical investigations are conducted to examine the presence and lo ad reasons of this phenomenon in financial market Researchers in this field believe that y th the presence of herding behavior has impact on results derived from asset pricing ju model because it influences stock price fluctuation, thus influencing risk and return of yi pl stocks (Tan et al, 2008) Similar to speculation, herding behavior may be rational or ua al irrational If market participants follow market consensus, the fluctuation is more and n more serious that can leads to instability in financial system, particularly in the period va n of global crisis In addition, herding behavior lasting so long can drive the stock prices fu ll further fundamental value which causes destabilization If investors are dominated by oi m sentiment such as greedy or fear of loss, they can trade in a “frenzied” way; as a at nh results, economic bubbles are created and may collapse the stock market In sum, herding behavior can lead to bad consequences of reducing the efficiency of market, z z even result in the market instability and financial collapse vb ht Basing on these arguments, doing research about herding behavior can help jm investors have an objective overview and be prudent when making investment k gm decision Therefore, the author decides to a research of “Herding behavior in l.c Vietnamese stock market: an empirical evidence from Quantile regression analysis” om The study applies research model proposed by Chang, Cheng and Khorona (2000) and n stock market a Lu modified by Chiang et al (2010) to investigate the presence of herd in Vietnamese y many region throughout the world, form developed to emerging countries For te re Several empirical studies have examined and detected the herding behavior in n va 1.2 Research gap 58 t to ng Dependent Variable: CSAD T Method: Least Squares Date: 11/05/15 Time: 11:28 Sample: 749 Included observations: 749 hi Coefficient Std Error t-Statistic Prob C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 0.939460 0.593464 -0.987413 -0.063148 -0.201384 0.044725 0.063486 0.085127 0.015280 0.023048 21.00534 9.347882 -11.59930 -4.132617 -8.737610 0.0000 0.0000 0.0000 0.0000 0.0000 ep Variable w n lo ad 0.254873 0.250867 0.698969 363.4864 -792.0226 63.62199 0.000000 Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat ju y th yi pl 1.492718 0.807566 2.128231 2.159064 2.140112 0.962411 n ua al R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) va n The Wald test ll fu 0.0000 0.0000 0.0000 Value Std Err -0.924264 0.082072 ht 744 (1, 744) vb -11.26159 126.8234 126.8234 z Probability z df at Value nh t-statistic F-statistic Chi-square oi Test Statistic m Wald Test: Equation: Untitled jm k Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: om l.c C(3) - C(4) gm Normalized Restriction (= 0) Restrictions are linear in coefficients a Lu n  Period of 2008 – 4/2015 n va Coefficient y Variable te re Dependent Variable: CSAD T Method: Least Squares Date: 11/05/15 Time: 11:31 Sample: 1819 Included observations: 1819 Std Error t-Statistic Prob 59 t to C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 ng hi 1.716533 0.378105 -0.345985 -0.122695 -0.095775 ep 0.118982 0.117039 0.448863 365.4805 -1121.457 61.24541 0.000000 R-squared Adjusted R-squared S.E of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) 0.021273 0.035210 0.033094 0.009821 0.008529 80.69076 10.73869 -10.45448 -12.49372 -11.22918 0.0000 0.0000 0.0000 0.0000 0.0000 w n lo ad Mean dependent var S.D dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter Durbin-Watson stat 1.870446 0.477686 1.238545 1.253680 1.244129 0.800091 ju y th yi pl The Wald test al n ua Wald Test: Equation: Untitled df Probability -7.277658 52.96430 52.96430 1814 (1, 1814) 0.0000 0.0000 0.0000 n ll fu oi m t-statistic F-statistic Chi-square Value va Test Statistic nh -0.223290 0.030682 ht vb Std Err z C(3) - C(4) Value z Normalized Restriction (= 0) at Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: k jm Restrictions are linear in coefficients gm n n va y te re Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.1) Date: 11/05/15 Time: 11:40 Sample: 2568 Included observations: 2568 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.025266 Estimation successfully identifies unique optimal solution a Lu  Period of 2005 – 4/2015 om behavior l.c Appendix 4: Quantile regression analysis to test for the presence of herding 60 t to ng hi Coefficient Std Error t-Statistic Prob C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 0.434347 0.991820 -0.867615 -0.216511 -0.176847 0.027960 0.044010 0.061686 0.010362 0.013204 15.53436 22.53644 -14.06508 -20.89488 -13.39332 0.0000 0.0000 0.0000 0.0000 0.0000 ep Variable 0.141964 0.140625 0.972313 0.873103 2.583978 0.000000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) w n lo ad Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic 1.760276 0.617355 255.7348 298.0466 363.8822 ju y th yi df Probability 2563 (1, 2563) 0.0000 0.0000 0.0000 n n va -10.94419 119.7753 119.7753 ua t-statistic F-statistic Chi-square Value al Test Statistic pl Wald Test: Equation: Untitled ll fu m oi Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: -0.651104 0.059493 z Std Err at C(3) - C(4) Value nh Normalized Restriction (= 0) z ht vb Restrictions are linear in coefficients jm k Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.25) Date: 11/05/15 Time: 11:43 Sample: 2568 Included observations: 2568 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.049137 Estimation successfully identifies unique optimal solution om l.c gm C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 1.046455 0.776987 -0.754474 -0.194648 -0.177619 0.052128 0.054572 0.058406 0.010951 0.011571 20.07473 14.23787 -12.91770 -17.77435 -15.35033 0.0000 0.0000 0.0000 0.0000 0.0000 Pseudo R-squared 0.103571 Mean dependent var 1.760276 y Prob te re t-Statistic n Std Error va Coefficient n a Lu Variable 61 t to Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) ng 0.102172 0.676439 1.461366 1.671628 0.000000 S.D dependent var Objective Restr objective Quasi-LR statistic 0.617355 454.0348 506.4926 334.7335 hi ep Wald Test: Equation: Untitled w n Test Statistic df Probability -10.93351 119.5416 119.5416 2563 (1, 2563) 0.0000 0.0000 0.0000 Value Std Err lo Value ad ju y th t-statistic F-statistic Chi-square yi pl Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: n ua al Normalized Restriction (= 0) C(3) - C(4) -0.559826 n va Restrictions are linear in coefficients 0.051203 ll fu m oi Dependent Variable: CSAD T Method: Quantile Regression (Median) Date: 11/05/15 Time: 11:47 Sample: 2568 Included observations: 2568 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.070948 Estimation successfully identifies unique optimal solution at nh z z ht vb C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 1.477117 0.575438 -0.602372 -0.154175 -0.151820 0.022724 0.043788 0.036133 0.014345 0.010275 65.00365 13.14154 -16.67100 -10.74749 -14.77555 0.0000 0.0000 0.0000 0.0000 0.0000 n 1.760276 0.617355 534.2016 584.2147 402.9448 va Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic n y te re 0.085607 0.084180 0.601376 1.794159 0.992950 0.000000 a Lu Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) om Prob l.c t-Statistic Std Error gm Coefficient k jm Variable Wald Test: Equation: Untitled 62 Test Statistic t to ng hi t-statistic F-statistic Chi-square df Probability -13.92476 193.8989 193.8989 2563 (1, 2563) 0.0000 0.0000 0.0000 Value Std Err -0.448196 0.032187 ep Value Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: w n Normalized Restriction (= 0) lo ad C(3) - C(4) y th Restrictions are linear in coefficients ju yi pl Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.75) Date: 11/05/15 Time: 11:48 Sample: 2568 Included observations: 2568 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.049137 Estimation successfully identifies unique optimal solution n ua al n va k l.c gm om 2563 (1, 2563) 0.0000 0.0000 0.0000 Value Std Err te re -8.745658 76.48654 76.48654 y Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: Normalized Restriction (= 0) n Probability va df n Value a Lu t-statistic F-statistic Chi-square jm Wald Test: Equation: Untitled Test Statistic 1.760276 0.617355 444.5464 468.6433 180.2981 ht Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic vb 0.051418 0.049938 0.660173 2.095829 1.425604 0.000000 0.0000 0.0000 0.0000 0.0000 0.0000 z Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 63.59565 7.000529 -9.896518 -4.304771 -7.668512 z 0.028105 0.065074 0.052372 0.024822 0.016677 at 1.787334 0.455551 -0.518298 -0.106855 -0.127885 Prob nh C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 t-Statistic oi Std Error m Coefficient ll fu Variable 63 C(3) - C(4) -0.411443 0.047045 t to Restrictions are linear in coefficients ng hi ep Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.9) Date: 11/05/15 Time: 11:50 Sample: 2568 Included observations: 2568 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.025266 Estimation successfully identifies unique optimal solution w n lo ad ju y th Variable Coefficient t-Statistic Prob 0.054455 0.115812 0.114200 0.037119 0.039652 40.53411 1.463886 -3.286124 0.031327 -1.949662 0.0000 0.1433 0.0010 0.9750 0.0513 yi Std Error 2.207301 0.169536 -0.375276 0.001163 -0.077309 pl C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 n ua al Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic 1.760276 0.617355 274.3341 283.8624 52.88753 n ll fu oi m at nh 0.033567 0.032058 0.898029 2.433319 4.003594 0.000000 va Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) z z ht vb Wald Test: Equation: Untitled 2563 (1, 2563) 0.0003 0.0003 0.0003 Value Std Err -0.376439 0.103005 om l.c -3.654565 13.35584 13.35584 Probability gm df k t-statistic F-statistic Chi-square Value jm Test Statistic Normalized Restriction (= 0) n a Lu Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: n va C(3) - C(4) y te re Restrictions are linear in coefficients  Period of 2005 – 2007 64 t to ng hi Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.1) Date: 11/05/15 Time: 11:56 Sample: 749 Included observations: 749 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.038099 Estimation successfully identifies unique optimal solution ep w n Variable lo ad C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 Std Error t-Statistic Prob 0.271696 0.960971 -0.916017 -0.199567 -0.180967 0.026013 0.065505 0.054277 0.017676 0.012001 10.44473 14.67008 -16.87686 -11.28999 -15.07892 0.0000 0.0000 0.0000 0.0000 0.0000 ju y th Coefficient yi 0.300338 0.296576 1.000545 0.555308 1.425830 0.000000 pl Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 1.492718 0.807566 56.63094 80.94040 378.8741 n ua al Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic n va ll fu oi df Probability -14.11402 199.2057 199.2057 744 (1, 744) 0.0000 0.0000 0.0000 Value Std Err -0.716450 0.050762 at Value nh Test Statistic m Wald Test: Equation: Untitled z z ht vb t-statistic F-statistic Chi-square jm om l.c Restrictions are linear in coefficients C(3) - C(4) gm Normalized Restriction (= 0) k Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: n a Lu n va y te re Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.25) Date: 11/05/15 Time: 11:57 Sample: 749 Included observations: 749 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.074094 Estimation successfully identifies unique optimal solution 65 t to ng hi ep Variable Coefficient Std Error t-Statistic Prob C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 0.376316 1.037830 -1.017762 -0.210179 -0.201663 0.032397 0.081857 0.066547 0.023459 0.015235 11.61563 12.67859 -15.29395 -8.959387 -13.23696 0.0000 0.0000 0.0000 0.0000 0.0000 0.281481 0.277618 0.898598 0.866259 1.154455 0.000000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) w n lo ad Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic 1.492718 0.807566 120.5742 167.8093 436.4316 ju y th al Value df Probability 744 (1, 744) 0.0000 0.0000 0.0000 n va -13.15809 173.1354 173.1354 n ua t-statistic F-statistic Chi-square pl Test Statistic yi Wald Test: Equation: Untitled ll fu oi m Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: nh Normalized Restriction (= 0) Std Err -0.807583 0.061375 at Value z C(3) - C(4) z ht vb Restrictions are linear in coefficients jm k Dependent Variable: CSAD T Method: Quantile Regression (Median) Date: 11/05/15 Time: 11:58 Sample: 749 Included observations: 749 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.10698 Estimation successfully identifies unique optimal solution om l.c gm C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 0.609556 1.094023 -1.119272 -0.211234 -0.218877 0.047779 0.133225 0.098342 0.044499 0.025989 12.75769 8.211876 -11.38140 -4.746984 -8.421889 0.0000 0.0000 0.0000 0.0000 0.0000 Pseudo R-squared Adjusted R-squared 0.217135 0.212926 Mean dependent var S.D dependent var 1.492718 0.807566 y Prob te re t-Statistic n Std Error va Coefficient n a Lu Variable 66 t to S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.769742 1.366989 1.367792 0.000000 Objective Restr objective Quasi-LR statistic 186.2682 237.9314 302.1702 ng hi ep Wald Test: Equation: Untitled w Value df Probability -10.03523 100.7059 100.7059 744 (1, 744) 0.0000 0.0000 0.0000 n Test Statistic lo ad ju y th t-statistic F-statistic Chi-square yi pl Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: Value Std Err -0.908037 0.090485 n ua al Normalized Restriction (= 0) C(3) - C(4) n va Restrictions are linear in coefficients ll fu oi m at nh z z ht vb Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.75) Date: 11/05/15 Time: 11:59 Sample: 749 Included observations: 749 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.074094 Estimation successfully identifies unique optimal solution 1.175330 0.839159 -1.152727 -0.115215 -0.253403 0.111808 0.238357 0.164406 0.074829 0.036752 10.51206 3.520591 -7.011484 -1.539705 -6.894953 0.0000 0.0005 0.0000 0.1241 0.0000 n 1.492718 0.807566 179.4033 206.4208 103.1629 va Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic n y te re 0.130885 0.126213 0.785466 1.945318 2.793505 0.000000 a Lu Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) om C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 l.c Prob t-Statistic gm Std Error k Coefficient jm Variable Wald Test: Equation: Untitled 67 Test Statistic t to ng t-statistic F-statistic Chi-square hi ep Value df Probability -7.389879 54.61031 54.61031 744 (1, 744) 0.0000 0.0000 0.0000 Value Std Err -1.037511 0.140396 Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: w n Normalized Restriction (= 0) lo ad C(3) - C(4) y th Restrictions are linear in coefficients ju yi pl Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.9) Date: 11/05/15 Time: 12:00 Sample: 749 Included observations: 749 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.038099 Estimation successfully identifies unique optimal solution n ua al n va ll fu 1.492718 0.807566 110.9516 122.7930 45.84421 k om l.c gm Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic jm 0.096434 0.091576 1.218471 2.617441 5.739929 0.000000 0.0000 0.0209 0.2600 0.1150 0.6649 ht Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 10.37301 2.315377 -1.127363 1.577929 -0.433274 vb 0.183386 0.147998 1.001092 0.016752 0.537223 z 1.902265 0.342670 -1.128594 0.026434 -0.232765 Prob z C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 t-Statistic at Std Error nh Coefficient oi m Variable a Lu Wald Test: Equation: Untitled n df Probability -1.167432 1.362897 1.362897 744 (1, 744) 0.2434 0.2434 0.2430 n Value va Test Statistic y Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: te re t-statistic F-statistic Chi-square 68 Normalized Restriction (= 0) t to C(3) - C(4) Value Std Err -1.155028 0.989375 ng hi Restrictions are linear in coefficients ep  Period of 2008 – 4/2015 w n lo Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.1) Date: 11/05/15 Time: 12:01 Sample: 1819 Included observations: 1819 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.028344 Estimation successfully identifies unique optimal solution ad ju y th yi pl C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 1.380765 0.284422 -0.208675 -0.103835 -0.075821 Std Error t-Statistic Prob 55.15396 4.981310 -2.744244 -9.003446 -4.782013 0.0000 0.0000 0.0061 0.0000 0.0000 n Coefficient ua al Variable n ll fu oi m 1.870446 0.477686 136.3845 159.5750 298.2854 at z z Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic nh ht vb 0.145327 0.143442 0.620296 1.397002 1.727691 0.000000 va Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) 0.025035 0.057098 0.076041 0.011533 0.015855 k jm 1814 (1, 1814) 0.1381 0.1381 0.1379 Value Std Err -0.104840 0.070668 n -1.483557 2.200941 2.200941 a Lu Probability om df l.c t-statistic F-statistic Chi-square Value Test Statistic gm Wald Test: Equation: Untitled y Restrictions are linear in coefficients te re C(3) - C(4) n Normalized Restriction (= 0) va Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: 69 t to ng hi ep Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.25) Date: 11/05/15 Time: 12:02 Sample: 1819 Included observations: 1819 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.055123 Estimation successfully identifies unique optimal solution w n Variable lo ad C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 ju y th Coefficient Std Error t-Statistic Prob 1.440740 0.455537 -0.455989 -0.142075 -0.127934 0.017582 0.027583 0.035027 0.006002 0.008005 81.94500 16.51527 -13.01806 -23.67010 -15.98173 0.0000 0.0000 0.0000 0.0000 0.0000 yi 0.131758 0.129844 0.506878 1.616798 0.861683 0.000000 pl n ua al Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic 1.870446 0.477686 221.1246 254.6809 415.3896 n va Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) ll fu oi nh Test Statistic m Wald Test: Equation: Untitled df Probability -9.589841 91.96505 91.96505 1814 (1, 1814) 0.0000 0.0000 0.0000 Value Std Err -0.313914 0.032734 at Value z z ht vb t-statistic F-statistic Chi-square jm om l.c Restrictions are linear in coefficients C(3) - C(4) gm Normalized Restriction (= 0) k Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: n a Lu n va y te re Dependent Variable: CSAD T Method: Quantile Regression (Median) Date: 11/05/15 Time: 12:03 Sample: 1819 Included observations: 1819 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.07959 Estimation successfully identifies unique optimal solution 70 t to ng hi Coefficient Std Error t-Statistic Prob C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 1.610446 0.502064 -0.514584 -0.155817 -0.142199 0.017704 0.027269 0.030259 0.006596 0.008268 90.96537 18.41180 -17.00625 -23.62119 -17.19805 0.0000 0.0000 0.0000 0.0000 0.0000 ep Variable 0.104237 0.102262 0.454335 1.860376 0.776175 0.000000 Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) w n lo ad Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic 1.870446 0.477686 279.8111 312.3717 335.6005 ju y th yi pl Wald Test: Equation: Untitled df Probability 1814 (1, 1814) 0.0000 0.0000 0.0000 n n va -12.62066 159.2810 159.2810 ua t-statistic F-statistic Chi-square Value al Test Statistic ll fu oi m Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: Std Err -0.358768 0.028427 at Value nh Normalized Restriction (= 0) z C(3) - C(4) z ht vb Restrictions are linear in coefficients Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.75) Date: 11/05/15 Time: 12:04 Sample: 1819 Included observations: 1819 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.055123 Estimation successfully identifies unique optimal solution k jm 1.850534 0.459811 -0.443785 -0.137079 -0.111889 0.024866 0.043582 0.044073 0.012669 0.013867 74.42045 10.55054 -10.06925 -10.82045 -8.068438 0.0000 0.0000 0.0000 0.0000 0.0000 0.058128 0.056051 0.493676 2.118398 Mean dependent var S.D dependent var Objective Restr objective 1.870446 0.477686 244.6209 259.7178 y Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var te re C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 n Prob va t-Statistic n Std Error a Lu Coefficient om l.c gm Variable 71 Sparsity Prob(Quasi-LR stat) 1.175340 0.000000 Quasi-LR statistic 137.0106 t to ng hi ep Wald Test: Equation: Untitled Test Statistic w n t-statistic F-statistic Chi-square lo ad df Probability -7.533914 56.75986 56.75986 1814 (1, 1814) 0.0000 0.0000 0.0000 Value Std Err -0.306706 0.040710 y th Value ju Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: yi pl Normalized Restriction (= 0) Restrictions are linear in coefficients n ua al C(3) - C(4) n va fu ll Dependent Variable: CSAD T Method: Quantile Regression (tau = 0.9) Date: 11/05/15 Time: 12:05 Sample: 1819 Included observations: 1819 Huber Sandwich Standard Errors & Covariance Sparsity method: Kernel (Epanechnikov) using residuals Bandwidth method: Hall-Sheather, bw=0.028344 Estimation successfully identifies unique optimal solution oi m at nh z z ht vb Coefficient Std Error t-Statistic Prob C TRU_D*RM D01*RM TRU_D*RM2 D01*RM2 2.230599 0.273430 -0.279396 -0.080383 -0.054256 0.047262 0.131146 0.054016 0.046309 0.009021 47.19599 2.084928 -5.172449 -1.735771 -6.014653 0.0000 0.0372 0.0000 0.0828 0.0000 om l.c n a Lu 1.870446 0.477686 154.9891 158.9543 30.22524 Mean dependent var S.D dependent var Objective Restr objective Quasi-LR statistic gm n va 0.024946 0.022796 0.694855 2.396891 2.915309 0.000004 k Pseudo R-squared Adjusted R-squared S.E of regression Quantile dependent var Sparsity Prob(Quasi-LR stat) jm Variable y te re Wald Test: Equation: Untitled Test Statistic Value df Probability 72 t to t-statistic F-statistic Chi-square -4.053626 16.43188 16.43188 0.0001 0.0001 0.0001 Value Std Err -0.199013 0.049095 ng 1814 (1, 1814) hi ep Null Hypothesis: C(3)-C(4)=0 Null Hypothesis Summary: Normalized Restriction (= 0) w n C(3) - C(4) lo ad Restrictions are linear in coefficients ju y th yi pl n ua al n va ll fu oi m at nh z z ht vb k jm om l.c gm n a Lu n va y te re