2.1 Sentiment and Market Returns
In the classical financial theory, investor sentiment does not play a role in the cross-section of stock prices, realized returns, or expected returns. For instance, in the traditional CAPM (Capital Asset Pricing Model) theory, the only explanatory factor for asset returns is systematic risk, which is measured by asset beta timing the market risk premium. The more systematic risk investors assume, the more returns investors
obtain. The relationship between asset returns and systematic risk is positive. Baker and Wurgler (2007), however, challenge this point of view. They consider that even if speculative and hard-to-arbitrage securities have higher beta value, according to their theoretical diagram, these securities should have lower returns.
Baker and Wurgler (2006) also find that when beginning-of-period sentiment indicators are low, subsequent returns are relatively high for stocks with the following qualities: small, young, high volatility, unprofitable, non-dividend-paying, extreme growth, and distressed. When sentiment indicators are high, these categories of stocks earn relatively low subsequent returns. This finding is consistent with their prediction that investor sentiment has a large effect on securities whose valuations are highly subjective and difficult to arbitrage. In addition, several firm characteristics display no unconditional forecasting power originally, but those characteristics hide strong conditional patterns that become visible only after conditioning on sentiment.
Although investor sentiment indicators have been widely used, a small ropotion of literature have focused on their efficacy. Clarke and Statman (1998) find that the Bullish Sentiment Index, which is a survey-based measure of the bullishness of newsletter writers, does not have significant predictive power for S&P returns. Brown and Cliff (1999) find that survey measures of sentiment are driven largely by lagged returns. They also find that their composite sentiment measures produced by using the principle component analysis method have predictive power for subsequent returns at 2-year and 3-year horizon and for deviations of stock prices.
Simon and Wiggins (2001) investigate the predictive power of market-based sentiment measures such as VIX (Volatility index), put-call ratio, and TRIN for subsequent returns on the S&P 500 futures contract over 10-day, 20-day, and 30-day horizons from January 1989 through June 1999. They find that these three sentiment measures generally have both statistical and economic forecasting power for
subsequent S&P 500 futures over the sample period of January 1989 through June 1999. They also use stimulation to find evidence that greater returns and risk-adjusted returns will have been earned from buying the S&P 500 futures when the sentiment indicators are flashing a high versus low level of fear.
Unlike previous articles from authors such as Baker and Wurgler (2006 and 2007), and Brown and Cliff (1999), which used principal component analysis to extract composite sentiment indicators, we adopt a method like the one proposed by Simon and Wiggins (2001) to directly examine the forecasting power of sentiment indicators. The method is used because it can preserve the information included in the investor sentiment measures. We also use the market-based sentiment measures such as VIX, VXN and put-call ratio, and survey-based sentiment measures such as AAII in our empirical studies.
2.2 Sentiment Beta and Hard-to-value, Difficult-to-arbitrage Hypothesis (HV-DA)
Glushkov (2006) uses sentiment beta to measure investor sentiment. The definition of sentiment beta is the sensitivity of returns to sentiment. He tests two hypotheses in his paper. The first hypothesis is the so-called HV-DA hypothesis. This postulates that the stocks of some firms are more easily affected by investor sentiment because of the differences in firm characteristics. He finds that more sensitive stocks are smaller, younger, with great short-sales constraints, higher idiosyncratic volatility and lower dividend yields, and this result is consistent with Baker and Wurgler (2006 and 2007). The second hypothesis predicts that stocks which are more sensitive to the movement in investor sentiment are more likely to be held by individual investors.
Evidence supporting the second hypothesis is mixed: institution investors got rid of stocks with high sentiment sensitivity throughout the 1980‘s, but held more of these
stocks throughout the 1990‘s.
The HV-DA hypothesis states that some stocks are more affected by irrational investor sentiment than others because of differences in their characteristics. For some younger growth stocks with short earning history and no dividends, it is hard to use the discount cash flow (DCF) model to evaluate their present value. This means that investor individual judgment plays a vital role when deciding the present value of those stocks. Therefore, hard-to-value stocks may be more sensitive to the fluctuation of investor sentiment. On the other hand, small stocks may be more sensitive to sentiment because they are difficult to short sell (Jones and Lamont (2002), D‘Avolio (2002)). Even if short selling is allowed, it is still difficult and costly for arbitragers to maintain a short position for a period of time. Because hard-to-value stocks are more sensitive to the fluctuation of sentiment, astute investors will lose their interest in the arbitrage of these kinds of stocks. This noise trader risk (De Long et al. (1990)) makes hard-to-value stocks also difficult-to-arbitrage. Thus, given the arbitrage limits and risks that arbitragers encounter, sentiment investors may exert their significant influence over the prices of stocks which are smaller, younger and volatile, and make them more vulnerable to sentiment change.
There are several new findings not documented in Baker and Wurgler (2006).
First, evidence shows that age, the firm‘s dividend policy and growth potential have explanatory power on relative sentiment sensitivities. Second, after controlling size and volatility, growth stocks are more sensitive to sentiment than distressed stocks.
As mentioned above in the two subsections, those articles discuss cross-section stock market returns. Our paper, however, is about time-series market returns. That is because our paper maintains the focus on the relationship among market liquidity, market returns and sentiment indicators. This is another big difference between our paper and previous literature concerning investor sentiment.
2.3 Market Liquidity and Market Returns
Order imbalance is one indicator to represent market liquidity. Most existing literature analyzing order imbalance is about specific events. For example, Lauterbach and Ben-Zion (1993) examines the behavior of the small market during the October 1987 crash. Blume et al. (1989) analyze order imbalances and stock price movements around the October 1987. Fung (2007) discovers the interactions between the arbitrage spread and order imbalances. Hasbrouck and Seppi (2001) find out common factors in returns, order flow, and market liquidity for thirty Dow Jones stocks during 1994. Brown et al. (1997) study the relationship between order imbalances and stock returns in the Australian stock market over one and two years, respectively.
There are also a growing number of studies suggesting that market liquidity predicts stock returns, both at the firm level and in the time series of the aggregate market. Amihud and Mendelson (1986) find that bid-ask spread is a factor to explain expected returns. Brennan and Subrahmanyam (1996) investigate that the relation between required rate of returns and the measure of liquidity. Jones (2002) suggests that time-series variation in aggregate liquidity is an important determinant of conditional expected stock market returns.
Furthermore, Chordia et al. (2001) argue that equity market returns and recent market volatility affect liquidity and trading activity. And they also find some factors which influence liquidity and trading activity. Those factors include short- and long- term interest rates, default spreads, market volatility, recent market movements, and indicator variables for the day of the week, for holiday effects, and for major macroeconomic announcements.
Backer and Stein (2004) establish a model to explain why increase in liquidity forecasts lower subsequent returns. They also discuss the relationship between market
liquidity and investor sentiment in their model. The most important conclusion in their paper is that market liquidity can be an investor sentiment indicator in the market with short-sale constraints.
Chordia et al. (2002) try to discover the tripartite association among trading activity, liquidity, and stock market returns. They have several important empirical results.
Firstly, order imbalances are strongly related to past market returns. Investors behave like contrarians, namely they buy after market declines and sell after market advances. This behavior is more apparent when the market declines.
Secondly, strong contemporaneous association exists between changes in the absolute level of market-wide order imbalance and market-wide liquidity. The relationship between order imbalances and market liquidity may arise when specialists cannot adjust the quotes on both sides of the market during the period of large order imbalance. Although order imbalance appears to have no forecasting ability, evidences reveal that both the number of trades and the market returns can forecast future changes in liquidity. A bear market predicts lower market liquidity the next day, and a bull market predicts higher market liquidity the next day. This consequence is consistent with inventory models of liquidity proposed by Stoll (1978a).
Thirdly, stock returns are strongly and contemporaneously related to order imbalances. Evidence shows that market prices tend to reverse following declines and continue to follow previous up moves. There is also evidence that returns are predictable using past imbalances and past returns following large-negative-imbalance, large-negative-return days, but there is no forecasting power following high-positive-imbalance, high-positive-return days. Therefore, as mentioned above, the results reveal that order imbalances influence market liquidity and market returns.
This result is consistent with Kraus and Stoll (1972a, b), in which large sales are followed by reversals but large buys are not.
Fourthly, there is strong relationship between order imbalances and contemporaneous absolute returns after controlling for market volume and market liquidity.