G. Mujtaba Mian
The NUS Business School
National University of Singapore
Singapore
E-mail: bizgmm@nus.edu.sg
Srinivasan Sankaraguruswamy
The NUS Business School
National University of Singapore
Singapore
E-mail: bizsrini@nus.edu.sg
November 13, 2007
ABSTRACT
We test the hypothesis that the prevailing market-wide investor sentiment sways the stock market response to good and bad corporate news in the direction of the sentiment. We use the Baker and Wurgler (2006) index of investor sentiment, and investigate stock price response to earnings shocks. Consistent with our hypothesis, we find that the three-day announcement period return for positive (negative) earnings news is greater for the earnings that are announced during high (low) sentiment periods than those announced during low (high) sentiment periods. Furthermore, the effect of sentiment persists in the near term. Over the 60 days following the announcement of earnings, the well-documented stock price drift associated with positive (negative) earnings news is greater for the earnings that are announced during high (low) sentiment periods than those that are announced during low (high) sentiment periods. In the cross-section, the relation between sentiment and the stock price response to news is more pronounced for small stocks, young stocks, volatile stocks, non-dividend paying stocks and distressed stocks.
Keywords: Investor Sentiment, Corporate News, Event Studies, Behavioral Finance
JEL Classification: D14, D21, G24
Introduction
Do waves of market-wide optimism or pessimism, or investor sentiment, influence the stock market response to firm specific news? According to the efficient markets view, the answer is an unequivocal no—stock prices in efficient markets have little to do with non-fundamental factors such as sentiment. Motivated by this view, the voluminous event study literature in finance and other areas of economics typically pool together events that happen during boom times with events that happen during bear periods. Tests are then conducted, for instance, to quantify the impact of various corporate events on the fundamental value of the firm2 under the maintained assumption that stock price reaction to corporate news is independent of the state of the stock market.
In contrast, both the anecdotal evidence3 and several papers in the recent behavioral finance literature suggest that the prevailing sentiment could significantly influence the way investors respond to new information and update their belief. One strand of this literature provides evidence that the optimism reflected in generic non-economic proxies of investor mood is positively correlated with the optimistic beliefs about future economic conditions (Hirshliefer and Shumway (2003), Edmans, Garcia and Norli (2007), and Puri and Robinson (2007)). To the extent that extreme bouts of market-wide investor sentiment and the mood of marginal investor are intertwined, the link between positive mood and optimistic assessments of future prospects implies a similar relation between positive sentiment and optimistic assessment of new information by investors. Furthermore, a related strand of the recent behavioral literature focuses directly on developing measures of sentiment and relating these to expected stock return (Baker and Wurgler (2006, 2007), Lemmon and Portniaguina (2006), Qiu and Welch (2006), Brown and Cliff (2005)). This literature begins with the premise that shocks to speculative demand combine with limits on arbitrage to generate mispricing in stocks. The speculative demand tends to be high (low) during periods of high (low) sentiment, which pushes up (down) the contemporaneous stock prices, and lowers (increases) the future stock returns. Given that investors often trade heavily around significant corporate news announcements4, during high (low) sentiment periods when speculative demand is high (low), investors are more likely to bid up (down) the price around a corporate news announcement. Specifically, the stock price response to good (bad) news that arrives in high sentiment period is likely to be greater (lower) than the stock price response to good (bad) news that arrives during low sentiment period, other things being equal.
Equally importantly, some of the most cited cognitive biases on the part of individuals, observed in the experimental psychology and noted in the behavioral finance literature, provide micro foundations for a hypothesized relation between sentiment and stock marker response to corporate news. Individuals tend to suffer from confirmatory bias whereby they interpret new evidence in a fashion consistent with their prior beliefs. They are reluctant to accept the inconsistent facts, attributing these to luck or faulty data gathering (Barberis and Thaler (2003), and Hirshliefer (2001)). The confirmatory bias would, therefore, lead investors to respond more strongly to good (bad) news during periods of high (low) sentiment. Moreover, as Shiller (2005) argues, investors have a tendency to form their expectations about future price changes by anchoring on recent price changes. The representativeness heuristic reinforces such extrapolation of the recent price trends. The problem gets especially severe if, as Statman, Thorley and Vorkink (2006) note, investors ignore that “rising water lifts all boats” and anchor on the absolute value of recent price changes rather than on changes relative to the market. Consequently, in time of high (low) sentiment, which tend to be preceded by stock price increases (drops), investors underreact to bad (good) news as they believe the stock price would continue its recent rise (drop). Overconfidence, social interaction5 and media also play a critical role in reinforcing the representativeness heuristic and in persuading investors to underreact to evidence that contradicts the prevailing sentiment (Shiller (2005)).
Understanding the impact of sentiment on stock market resposne to news could yield new perspectives on key debates in financial economics. First, despite a number of recent studies, the debate on the importance of sentiment is far from settled, and much remains to be learnt about the role sentiment plays in financial markets. The evidence in recent studies that stocks are mispriced in the direction of the sentiment raises the possibility that a significant amount of this mispricing occurs on the days of the arrival of news. Specifically, overreaction (underreaction) of investors to good news or undereraction (overreaction) to bad news during periods of high (low) sentiment could be an important channel through which stocks become overpriced (underpriced) during high (low) sentiment periods. A confirmation of this would not only help identify a channel through which sentiment causes mispricing but would also suggest that the importance of sentiment documented in the recent literature is not spurious. Second, by focusing on the effect of sentiment on stock price response in the short window around the announcement of news, one could garner fresh evidence on the efficiency of the instantaneous response of the stock market to new information. Many researchers interpret the evidence accumulated in the voluminous event studies literature as an indication that human psychology plays no role in how markets respond to news, and treat the evidence as a key pillar in support of the efficient market hypothesis (see, for example, Fama (1991) and Ross (2005)).6 Any evidence that proxies of sentiment influence the stock market response to news would call into question the efficiency of the stock market’s instantaneous reaction to news. Finally, the event-study methodology has been widely used by researchers in finance and other areas of economics, to quantify the impact of various corporate events on the fundamental value of the firm.7 In a world where the stock market reaction to news varies significantly across periods of high and low sentiment, it is not clear how useful it is to rely solely on the market reaction as a measure of the valuation impact of the news event.
In this paper, we examine empirically how the prevailing sentiment influences stock market response to news. Specifically, we test the hypothesis that prevailing sentiment sways the stock market response to news in the direction of the sentiment. That is, stock market response to good news is greater during high sentiment periods than that during low sentiment periods. Similarly, stock market response to bad news is greater during low sentiment periods than that during high sentiment periods.
To test our hypothesis, we rely on the proxies of the sentiment developed by Baker and Wurgler (2006, 2007). Baker and Wurgler (2007) sentiment index is based on six proxies: trading volume as measured by NYSE turnover, the dividend premium, the closed-end fund discount, the number and first day returns on IPOs, and the equity share in new issues. This index is available at monthly frequency. The corporate news event we primarily focus on is earnings surprises, although in the latter part of our paper, we confirm the robustness of our results for several other corporate events including dividend changes, stock splits and stock repurchases. Earnings news is perhaps the most prominent news event, which unlike many other corporate events, is not voluntary and is a regular feature of corporate calendars. We measure the news content of an earnings announcement by comparing the actual earnings with the analyst consensus forecast in the month prior to the earnings announcement date. We use the standard event study methodology and examine the 3-day abnormal returns around the announcement of the news events. Our analyses incorporate standard control variables that have been identified by prior research to explain the stock price reaction to earnings surprises. Our results indicate that market reacts more to good news during high sentiment periods than during low sentiment periods. Similarly, market reacts more to bad news during low sentiment periods than during high sentiment periods. We confirm that our results also hold for other corporate events, namely dividend changes, stock splits and stock repurchases.
We also examine the stock returns in the 60 days following the announcement of earnings news to see if the effect of sentiment persists or reverses during this period. The results indicate that sentiment continues to impact the stock price behavior in the period subsequent to the earnings announcement. That is, the upward stock price drift following positive earnings news, commonly documented by prior literature, is greater when earnings are announced in high sentiment periods than those that are announced during low sentiment periods. Similarly, the downward stock price drift following negative earnings news is greater for earnings that are announced during low sentiment periods. In fact, we find that there is no drift associated with negative earnings shocks that are announced during high sentiment periods. This is probably not surprising given that the sentiment measure we use has very high persistence—the autocorrelation coefficient for the monthly Baker-Wurgler index is 0.95 during our sample period. This is also consistent with the recent sentiment literature that argues that the effect of sentiment persists over weekly and monthly frequency, and reverses only at yearly horizons (Brown and Cliff (2004, 2006) and Baker and Wurgler (2006)).
The behavioral finance literature suggests that firms that are more difficult to value and harder to arbitrage are more susceptible to sentiment impacting their stock prices (see, for example, Shleifer and Vishny (1997)). More specifically, Baker and Wurgler (2006) argue that firm size, age, stock price volatility, and growth prospects are the characteristics that delineate firms that are more or less susceptible to market sentiment. Therefore, we examine whether the influence of sentiment is stronger on investor response to news for small firms, young firms, volatile firms, and growth firms. We find that sentiment indeed plays a greater role in determining the stock price reaction to corporate earnings news for these firms than for other firms.
While we document that stock market response to good (bad) news is stronger during high (low) than low (high) sentiment periods, our tests, per se, cannot discern whether stock market overreacts to good (bad) news during high (low) sentiment period or underreacts to good (bad) news during low (high) sentiment periods. For this, we rely on recent studies, example, Baker and Wurgler (2006). These studies find that both high and low sentiment periods contributes to mispricing—stocks that are more prone to sentiment such as small stocks, earn unusually high returns following periods of low returns and earn unusually low returns following periods of high sentiment.
We also consider alternative explanations for our results, but rule them out for being inconsistent with some of the evidence. One possibility is that that investors’ risk aversion changes across periods of high and low sentiment, and that explains market’s differential reaction to news over time. However, we note that for this explanation to hold, stock price response to both good and bad news should be muted during periods of low sentiment. This is because during these times, investors’ increased risk aversion pushes up the discount rate; and any news about the future cash flows is, therefore, worth less in present value terms. Our results are not consistent with this explanation as we find that stock price reaction to bad news is greater during periods of low sentiment. It is also possible that good (bad) news released during high (low) sentiment period has greater information content, and market’s greater response to good (bad) news during high (low) sentiment period simply reflects this differential strength of the news signal. To control for this possibility, we introduce the ex-post earnings change associated with each earnings surprise as an additional variable in our regression analyses. We find that our conclusion remains robust to the inclusion of this variable.
At least two recent papers have documented instances, albeit in narrow settings, where the stock market ignored the firm-level fundamentals apparently under the influence of the broad market sentiment. Lamont and Thaler (2003) document how in the case of equity carve-out of technology companies at the peak of the Internet bubble, investors valued non-technology parent and technology-oriented subsidiary companies at prices that violated the fundamental law of one price. Cooper, Dimitrov and Rau (2001) document that a mere alignment of a company’ name to what is considered fashionable in the market can enhance stock price, even when the operations of the company are little changed. They document that at the peak of the Internet bubble, companies that added dotcom to their names experienced a cumulative abnormal return of 74% in the ten days surrounding the announcement of the name change. Our paper generalizes the findings in these studies by showing that the influence of sentiment on stock market reaction to news extends beyond the narrow settings of these studies.
The rest of the paper is organized as follows. In Section II, we discuss our data selection process and research methodology. In Section III, we report our results on the effect of sentiment on stock price response to earnings news. We then examine the effect of sentiment on stock price response for three other corporate events, namely, dividend changes, stock splits and stock repurchases in Section IV. We conclude the paper in Section V.
