A New Proxy for Investor Sentiment: Evidence from an Emerging Market

Journal of Business Studies Quarterly 2014, Volume 6, Number 2 ISSN 2152-1034 A New Proxy for Investor Sentiment: Evidence from an Emerging Market Dima Waleed Hanna Alrabadi Associate Professor, Department of Finance and Banking Sciences, Faculty of Economics and Administrative Sciences, Yarmouk University, Irbid- Jordan Abstract This study proposes order imbalance as a proxy for investor (market) sentiment. Using daily data from Amman Stock Exchange over the period (2004-2013), we find that the daily aggregate order imbalance measures and specifically the value of buyer-initiated shares less the value of seller initiated shares serves as an excellent proxy for investor sentiment. Thus, it shows a highly significant effect on the daily aggregate market returns even after controlling for the effects of lagged market returns and liquidity. Order imbalance proxies convey more information about market returns than trading activity measures. The Granger causality tests illustrate a bi-directional causality between investor sentiment and aggregate market returns. Keywords: Investor Sentiment, Order Imbalance, Market Returns, Behavioral Finance, Amman Stock Exchange. 1. Introduction The conventional finance theory states that investment return is purely determined by firm-specific fundamentals assuming that investors are rational and markets are efficient. On the other hand, behavioral finance is a relatively new field in finance that proposes behavioral and cognitive psychological theory to provide explanations for why people make irrational financial decisions. The investor sentiment is a key concept in behavioral finance; it has attracted the interest of many researchers during the last decade. Investor sentiment which is also called market sentiment is defined as the general attitude of investors toward a certain security or the whole financial market. Market sentiment is the feeling or the tone of the market,

or its crowd psychology, as revealed through the activity and price movement of the securities traded in that market. Empirically, investment sentiment is found to affect both the aggregate market and the cross-section of returns. However, such a psychological phenomenon (unobservable variable) is not easy to be measured. Many proxies have been identified in the academic literature with no any consensus on a certain one. These proxies are classified into two major types, direct and indirect measures. The direct measures mainly depend on investors' surveys of their response for the anticipated movement of stock market and the aggregate economy in order to construct an investor sentiment index. On the contrary, the indirect measures employ market related implicit sentiment proxies derived from the selected market statistics, one of the most well known indirect investor sentiment measures is the one developed by Baker and Wurgler (2006), it is the first principal component of six variables and their lags. These variables include the closed-end fund discount, share turnover, the number and average first-day returns on IPOs, the equity share in new issues, and the dividend premium. Examples of other indirect sentiment proxies include change in dollar trading volume, put/call ratio and mutual fund cash position. This study examines the effect of investor (market) sentiment measured by aggregate order imbalance on aggregate market returns. Daily data of Amman Stock Exchange over the period (2004-2013) is used. In fact we argue that order imbalances (buy less sell orders) can provide additional power beyond trading activity measures such as trading volume in reflecting the overall investors' attitude (sentiment) and explaining stock returns. Thus, positive order imbalances can signal excessive investor interest in a stock, and if this interest is auto-correlated, then order imbalances could be related to future returns. On contrast, negative order imbalance reflects contrarian investor expectations of stock prices. What we are trying to say is that the implications and effects of trading 1000 shares generated by 500 buyer initiated trades and 500 seller initiated trades is definitely very different from those of trading 1000 shares generated of either 1000 seller or buyer initiated trades. The remainder of the study is organized as follows: Section 2 reviews the literature, Section 3 describes data and methodology, Section 4 reports the empirical results and Section 5 concludes. 2. Literature Review Numerous researchers investigate investor or market sentiment. Early, Solt and Statman (1988) examine the Bearish Sentiment Index published by Investor Intelligence in USA. It reflects the ratio of bearish to the total number of investment advisors. In specific, they investigate whether this index is useful as an indicator for forthcoming stock price changes and find no evidence. They find that the number of correct forecasts from the index equals the incorrect ones. De Long et al. (1990) model two types of investors on the market, rational and irrational (noise) investors. Irrational investors are subject to the influence of sentiment. They conclude that the unpredictability of noise traders' beliefs (sentiment) creates a risk in the price of the asset that deters rational arbitrageurs from aggressively betting against them. As a result, prices can diverge significantly from fundamental values even in the absence of fundamental risk. De Bondt (1993) shows a significant negative relationship between the sentiment index published by the American Association of Individual Investors (AAII) and future returns. Fisher and Statman (2000) study the sentiment of three 19

groups of investors, large, small and medium. They find that the sentiment of both small investors and large ones are reliable contrary indicators for future S&P 500 Index returns. The relationship between the sentiment of individual investors and future S&P 500 Index returns is negative and statistically significant and so is the relationship between the sentiment of Wall Street strategists and future S&P 500 returns. Studies in the last decade have started to combine proxies to form sentiment indices that might better measure sentiment. Brown and Cliff (2004) combine twelve direct and indirect measures of sentiment to form an index and investigate its effect on both contemporaneous and near-term stock returns. They find that their index strongly correlates with contemporaneous returns but not with future returns. Once again, Brown and Cliff (2005) use the Kalman filter and the principal components analysis to construct their composite sentiment measures based on survey data, IPO activities and other technical indicators. They examine the relations between the composite sentiment measure and market returns by VAR systems. Brown and Cliff (2005) find that investor sentiment does not predict short-term market returns at weekly and monthly intervals but that investor sentiment predicts long-term market returns at the next two to three years. They attribute these findings to limited arbitrage in the longrun but not in the short term. Baker and Wurgler (2006) combine six previously studied sentiment proxies to form a sentiment index and test it against various portfolios of stock returns. In particular, they find that this index correlate with next period returns of smaller and younger stocks particularly significantly. While the earlier evidence focuses on USA, recent studies provide further evidence on the impact of investor sentiment on stock returns from different stock exchanges around the world. Tas and Akdag (2012) suggest trading volume as a proxy for investor sentiment. Using data from Istanbul Stock Exchange over the period (2005-2009), they find that almost all beta coefficients of volume trend values have positive signs, which reflects the positive contribution of volume changes on the corresponding stock returns. Dash and Mahakud (2013) analyze the role of investor sentiment on stock returns in the Indian stock market over the period (2003-2011). They follow a top-down approach by using various market related implicit sentiment proxies in order to construct an investor sentiment index. Their results indicate that sentiment represents a source of systematic priced risk that holds even after controlling for the well known common risk factors. They explain its negative coefficient by arguing that since positive sentiment results in over valuation of stocks, the expected return for such stocks will be lower in the subsequent period. Consistently, Xu and Green (2013) examine the stock price impact of sentiment in China by incorporating the sentiment risk factor into the Fama and French three factor model. They find that sentiment helps explain the miss-pricing component of returns in the Fama and French model and the time variation in the factors themselves. Moreover, they argue that Fama and French factors become less significant if they are conditioned by sentiment. Sahli and Boubaker (2013) test the impact of investor sentiment on stock returns in the Tunisian stock market over the period (2007-2011). Using VAR and a sentiment index based on indirect indicators, they find no significant relationship between investor sentiment and stock performance, except for very tangible portfolio. 20

To the best of the researcher knowledge, this is the first study that investigates the investor sentiment and its effect on stock returns in the Jordanian stock exchange. More importantly, it is the first study worldwide that empirically suggests order imbalance as a proxy for sentiment. 3. Data and Methodology Data set consists of the daily aggregate seller and buyer initiated number of transactions, number of shares and their values over the period 2004-2013. In addition, the data set includes the daily values of the free float index of Amman Stock Exchange and the daily aggregate executed number of transactions, number of shares and their values over the same period. The trading activity measures are used for comparison purposes. The order imbalance variables are calculated as follows: OIMNUM: the number of buyer-initiated trades less the number of seller-initiated trades on day t OIMSH: the buyer-initiated shares less the seller-initiated shares on day t OIMV: the buyer-initiated value of shares paid less the seller-initiated value of shares on day t The market return is calculated as follows: MKT=Ln( I t / I t 1) (1) Where the I t is the index value on day t while I t 1 is the index value on day t-1. Vector auto regression (VAR) is employed to investigate the dynamic short term effect of investor sentiment measured by its three proxies of order imbalance on aggregate market return as follows: MKT C MKT MKT OIM OIM OIM Control e t 1 t 1 2 t 2 3 t 4 t 1 5 t 2 6 (2) Where MKT is the daily aggregate market return. OIM is the order imbalance (investor sentiment) measured by the three previously mentioned proxies. The equation is estimated three times each using a certain order imbalance measure. Control is the trading activity control variable which changes each time the equation is estimated. Thus, we use the executed number of transactions (Number) as a control variable when OIMNUM is used as a proxy for investor sentiment. The number of shares of executed transactions (Shares) is used as a control variable when OIMSH is used as a proxy for investor sentiment. The value of shares of executed transactions (Value) is used as a control variable of trading activity when OIMV is used as a proxy for investor sentiment. Moreover, Granger causality test is used to investigate the direction of the relationship between the investor sentiment and market return. 4. Empirical Results Table 1 reports the descriptive statistics of the study variables. All order imbalance proxies' means and medians are negative indicating that selling initiated trades are more than buying initiated trades over the study period in such an emerging market. The average daily traded shares over the study period is 15,568,544 with a value of 34, 813,131. The daily average market return is near to zero (0.0001) however, its maximum value was around 5% during the study period. The standard t 21

deviation values are relatively high taken into consideration the volatility of the stock market returns and trading activity. Table 2 shows the correlation matrix between the study variables. Interestingly, the results show that OIMV is highly related to stock returns which could strongly propose it as the main indicator of market sentiment. Thus, it shows a correlation value of 59.13% with the daily aggregate market return (MKT). The other investor sentiment proxies namely OIMNUM and OIMSH show a correlation with MKT of 37.5% and 29.8%, respectively. Moreover, Table 2 shows that order imbalance proxies are more correlated with MKT than trading activity measures which in turn highlights the information content that these proxies reveal about the overall market. On the other hand, the correlation values between order imbalance and trading activity measures are negative and low indicating the different informational signals that the two groups of variables contain. Figure 1 shows the trend of the market return and market sentiment proxies over the study period. The charts show that OMIV and market return exhibit very similar trends which consecutively support our argument that OIMV could be a vital proxy for market (investor) sentiment. Table 1: Descriptive statistics of the variables of the study. OIMV Value OIMSH Shares OIMNUM Number MKT Mean -11155003 34813131-4148396 15568544-2179 9109 0.0001 Median -9365961 25350162-3290061 13692575-2125 7963 0.0003 Maximum 105000000 447000000 51022158 99395426 9221 29883 0.0469 Minimum -212000000 2044401-82299542 1104356-20697 1683-0.0453 Std. Dev. 16125370 32056582 11657727 9650467 2298 5121 0.0101 Table 2: Correlation Matrix OIMV Value OIMSH Shares OIMNUM Number MKT OIMV 1.0000-0.1249 0.4620 0.0567 0.4276-0.0910 0.5913 Value -0.1249 1.0000 0.1582 0.4759 0.1489 0.8107 0.0793 OIMSH 0.4620 0.1582 1.0000 0.1869 0.4737 0.2029 0.2981 Shares 0.0567 0.4759 0.1869 1.0000 0.0283 0.5926 0.0708 OIMNUM 0.4276 0.1489 0.4737 0.0283 1.0000-0.0357 0.3752 NUMBER -0.0910 0.8107 0.2029 0.5926-0.0357 1.0000 0.0651 MKT 0.5913 0.0793 0.2981 0.0708 0.3752 0.0651 1.0000 22

Table 3 reports the VAR results of MKT and investor sentiment proxies while incorporating trading activity measures to control the liquidity effect. The results show that the three order imbalance proxies and their lags (OIMV, OIMSH, OIMNUM) have a positive and highly significant effect on market return (MKT) even after controlling for the lagged market returns and liquidity (Number, value, share) effects. Indeed, these results support the notion that order imbalance proxies serve as good proxies of market sentiment that affect aggregate market return. The adjusted R- squared of the estimated VARs ranges from 28.5% to 45.2%. This indicates once again that the OIMV is the best among the three proxies of order imbalance to reflect market sentiment. These results are consistent with Chordia et al. (2002) for NYSE who find that market-wide returns are strongly affected by contemporaneous and lagged order imbalances. Bailey et al. (2009) also find that Order imbalances explain about 31.2% of daily fluctuations in open-to-close excess returns of the stocks listed in the Chinese stock market (Shanghai Stock Exchange). Consistently, Chen and Lin (2012) find that contemporaneous order imbalance exerts an extremely significant impact on market returns and volatility in the same market. Li et al. (2010) show that order imbalance explains more than 90% of intraday returns of the Nikkei 225 Futures in the Osaka Stock Exchange in Japan. Smales (2012) finds that Contemporaneous order imbalance shows a significant impact on market returns in the expected direction thus excess buy (sell) orders drive up (down) prices for the Australian interest rates futures market. Huang (2011) examines the intraday dynamic returnorder imbalance relation to stealth trading in the NASDAQ-100 component stocks. He finds that the contemporaneous order imbalance-return relation is positively significant. Moreover, he argues that the impact of order imbalance on return is stronger than that of trading volume, implying that order imbalance convey more information than trading volume does. Table 4 reports the Granger causality tests between investor sentiment proxies and MKT. The results in Panels A, B, and C show a highly significant bi-directional causal relationship between market returns and the three investor sentiment proxies. Thus, lagged (past) investor sentiment forecasts future market returns as well as lagged market return forecasts future investor sentiment. Choe and Yoon (2007) also apply Granger causality tests and find that the causal relation between order imbalances and the Korean index (KOSPI200) returns is bidirectional. Sua et al. (2012) also examine the causal relationship between return and order imbalance and find that order imbalance is a good indicator for price discovery. Table 3: Vector Autoregression results of investor sentiment proxies and market return. Panel A Panel B Panel C MKT MKT MKT C 0.0011 C -0.0002 C 0.0002-0.0003-0.0004-0.0004 [ 4.27799] [-0.45690] [ 0.39214] MKT(-1) 0.2814 MKT(-1) 0.3176 MKT(-1) 0.3061-0.0201-0.0201-0.0202 [ 14.0264] [ 15.7744] [ 15.1767] MKT(-2) -0.0102 MKT(-2) -0.0270 MKT(-2) -0.0022 23

-0.0200-0.0199-0.0203 [-0.50947] [-1.36186] [-0.10923] OIMV(-1) 0.0000 OIMSH(-1) 0.0000 OIMNUM(-1) 0.0000 0.0000 0.0000 0.0000 [-11.3813] [-14.7299] [-13.0464] OIMV(-2) 0.0000 OIMSH(-2) 0.0000 OIMNUM(-2) 0.0000 0.0000 0.0000 0.0000 [-5.34715] [-5.30720] [-4.96573] OIMV 0.0095 OIMSH 0.0070 OIMNUM 0.0003 0.0000 0.0000 0.0000 [ 41.8203] [ 27.8603] [ 28.2452] Value 0.0000 Shares 0.0000 Number 0.0000 0.0000 0.0000 0.0000 [ 7.29989] [ 1.93022] [ 4.47718] Adj. R-squared 0.4512 Adj. R-squared 0.2848 Adj. R-squared 0.2890 Table 4: Granger Causality Tests Panel A: MKT and OIMV Null Hypothesis: Obs F-Statistic Probability OIMV does not Granger Cause MKT 2459 32.1662 0.0000 MKT does not Granger Cause OIMV 28.2793 0.0000 Panel B: MKT and OIMSH Null Hypothesis: Obs F-Statistic Probability OIMSH does not Granger Cause MKT 2459 21.6317 0.0000 MKT does not Granger Cause OIMSH 43.9147 0.0000 Panel C: MKT and OIMNUM Null Hypothesis: Obs F-Statistic Probability OIMNUM does not Granger Cause MKT 2459 31.3335 0.0000 MKT does not Granger Cause OIMNUM 46.1828 0.0000 Figure 1: Investor Sentiment Proxies and Market Return over the Study Period 24

25-1E+08-80000000 -60000000-40000000 -20000000 0 20000000 40000000 60000000 04/01/2004 04/01/2005 04/01/2006 04/01/2007 04/01/2008 04/01/2009 04/01/2010 04/01/2011 04/01/2012 04/01/2013 OIMBSH OIMBSH -0.06-0.04-0.02 0 0.02 0.04 0.06 04/01/2004 04/01/2005 04/01/2006 04/01/2007 04/01/2008 04/01/2009 04/01/2010 04/01/2011 04/01/2012 04/01/2013 MKT MKT -25000-20000 -15000-10000 -5000 0 5000 10000 15000 04/01/2004 04/01/2005 04/01/2006 04/01/2007 04/01/2008 04/01/2009 04/01/2010 04/01/2011 04/01/2012 04/01/2013 OIMNUM OIMNUM -2.5E+08-2E+08-1.5E+08-1E+08-50000000 0 50000000 100000000 150000000 04/01/2004 04/01/2005 04/01/2006 04/01/2007 04/01/2008 04/01/2009 04/01/2010 04/01/2011 04/01/2012 04/01/2013 OIMBV OIMBV

5. Conclusions There is no consensus yet on a certain proxy of investor sentiment. Such a psychological phenomenon is hardly observable. This study introduces order imbalance proxies as measures for investor sentiment. It uses daily data from the Jordanian stock Exchange over the period 2004-2013. The findings indicate that the value of buyer-initiated shares minus the value of seller initiated shares is an important proxy of investor sentiment. Daily aggregate market returns are strongly affected by contemporaneous and lagged investor sentiment variables measured by order imbalances. This effect persists even after controlling for past returns and contemporaneous trading activity. Investor sentiment Granger causes market returns and vice versa. Finally, order imbalances are more correlated to market returns than trading activity measures. Future research could replicate our study to other developed and emerging markets. Further evidence would probably highlight the importance of our findings. References Bailey, W., Cai, J., Cheung, YL. & Wang. F. (2009). "Stock Returns, Order Imbalances, and Commonality: Evidence on Individual, Institutional, and Proprietary Investors in China". Journal of Banking & Finance. 33(1): 9 19. Baker, M. & Wurgler, J. (2006). "Investor Sentiment and the Cross-section of Stock Returns", Journal of Finance, (61)4: 1645 1680. Brown, G. W. & Cliff, M. T. (2004). "Investor Sentiment and the Near-term Stock Market". Journal of Empirical Finance, (11)1:1 27. Brown, G. W. & Cliff, M. T. (2005). "Investor Sentiment and Asset Valuation". Journal of Business, 78(2):405 40. Chen, M. & Lin. H. (2012). "Order Imbalance, Individual Stock Returns and Volatility: Evidence from China". Journal of Convergence Information Technology, 7(2):101. Choe, H. & Yoon, S. (2007). "The Impact of Program Trading on Stock Returns". Asia-Pacific Journal of Financial Studies,36(2):281-320. Chordia, T., Roll, R. & Subrahmanyam, A. (2002). "Order Imbalance, Liquidity and Market Returns". Journal of Financial Economics, 65:111 130. Dash,S. R. & Mahakud, J. (2013). "Impact of Investor Sentiment on Stock Return: Evidence from India". Journal of Management Research, 13(3): 131 144. De Bondt, W. (1993)."Betting on Trends: Intuitive Forecasts of financial Risk and Return". International Journal of Forecasting, (9)3:355-71. De Long, J. B., Shleifer, A., Summers, L. H. &Waldmann, R. J. (1990). "Noise Trader Risk in Financial Markets". Journal of Political Economy, 98(4): 703 738. 26

Fisher, K. & Statman, M. (2000). "Investor Sentiment and Stock Returns". The Financial Analysts Journal, March/April: 16-23. Huang, H. (2011). "An Analysis of Intraday Return-Order Imbalance Relation to Stealth Trading". Investment Management and Financial Innovations, 8(1):234-242. Li, M., Endo, M., Zuo, S. & Kishimoto, K. (2010). "Order Imbalances Explain 90% of Returns of Nikkei 225 Futures". Applied Economics Letters, 17(13): 1241-1245. Sahli, L. & Boubaker, A. (2013). "The Impact of Investor Sentiment on the Tunisian Stock Market". Journal of Business Studies Quarterly, 5(2): 91-113. Smales, L.(2012)."Order Imbalance, Market Returns and Macroeconomic News Evidence from the Australian Interest Rate Futures Market". Research in International Business and Finance, 26:410 427. Solt, ME. & Statman, M. (1988). "How useful is the sentiment index?" Financial Analysts Journal, 44(5):45-55. Sua, YC., Huang, HC. & Lina, SF. (2012) "Dynamic Relations between Order Imbalance, Volatility and Return of Top Gainers". Applied Economics, 44(12):1509-1519. Tas, O. & Akdag, O. (2012). "Trading Volume Trend as an Investor s Sentiment Indicator in Istanbul Stock Exchange". DoğuşÜniversitesi Dergisi, 13(2):290 300. Xu, Y. & Green, C. (2013). "Asset Pricing With Investor Sentiment: Evidence from Chinese Stock Markets". The Manchester School, 81(1): 1-32. **This article is sponsored by the Dean of Scientific Research and Graduate Studies-Yarmouk University-Jordan. 27