Time Variation in the Relationship Between News and Returns

How does the stock market respond to news reports about a company? Previous work has found that it takes a few days for prices to fully absorb news, and this underreaction has been attributed to investors’ limited ability to process information. We find evidence for an alternative explanation: Sophisticated institutional investors trade slowly on news to avoid tipping their hands. Once we control for this effect, we find that the market actually overreacts to news.

Academic interest in how markets process information dates back to Eugene Fama’s pioneering work in the 1960’s.  Fama argued that markets are highly informationally efficient, though others, notably Robert Shiller, have pushed back against this conclusion.  It was against this backdrop that Tetlock, Saar-Tsechansky, and Macskassy (2008) (hereafter TSM) showed that stock prices of S&P 500 firms briefly underreact to news.  TSM found that when the Dow Jones News service ran an article about a specific company, that company’s stock price would react on the day of the article’s release (the event day) in the direction of the article’s tone.  Positive articles were associated with positive contemporaneous stock price reactions, and negative articles were associated with negative contemporaneous stock price reactions.  This was not surprising because event day news carried potentially new information about company fundamentals (as measured by future earnings), and these stock price moves were the market’s response to this new information.  What was surprising about TSM’s findings was that stock prices continued to move in the direction of the event day news for the next 10 trading days.  Stock prices did not react only to contemporaneous news, but appeared also to react to lagged news.  In fact, the economic magnitude of this apparent underreaction of stocks to news was quite large.  A trading strategy constructed to take advantage of this underreaction earned paper profits of over 20 percent per year.

The TSM paper was very influential for two reasons.  First, it was one of the first papers to use natural language processing tools to answer an important economic question.  In fact, two of the three authors are computer scientists, not economists.  Second, and perhaps more important, the paper seemed to provide direct evidence of an apparent market inefficiency – evidence that was previously inaccessible to economists because they lacked the tools to measure the information to which market participants were reacting.  Once such tools became available, it turned out that the world did not completely adhere to Fama’s conception of efficient markets. Market participants were apparently leaving money on the table.  If investors had paid a little bit more attention, they would have noticed that, following positive or negative news, stock prices continued to drift in the direction of the news for the next 10 trading days.  Investors should have put on trades to take advantage of this information, thus forcing markets back towards efficiency.  So why didn’t this happen?

In our paper, using news articles from the Thomson Reuters news service from 1996-2018, we shed light on this important question.  First, we show that the degree of stock price underreaction to news fluctuates over time.  The underreaction was most pronounced in the early part of our sample, then decreased in the several years following the financial crisis, but has increased again in the most recent time period (2015-2018).  An interesting facet of our analysis is that, not only does the underreaction of stocks to news vary over time, but so does the contemporaneous stock price reaction to news.  During some time periods, stock prices are more responsive to event day news than during other times.  We exploit this time variation in how stock prices react to contemporaneous and lagged news to identify a potential mechanism for this behavior.

If stock price underreaction to news is due to market inefficiency, then we would expect periods of large underreaction to be associated with relatively sluggish contemporaneous stock price responses to news.  The stock price on the day of the news release does not move enough in the direction of the news, and therefore future stock prices continue to drift in the direction of lagged news to compensate for this lack of event day reaction.  To test for this phenomenon, we need a conditioning variable that is potentially correlated with stock responsiveness to contemporaneous news.   One such variable is the capitalization level of large financial intermediaries (like Goldman Sachs and JPMorgan).  When the Goldmans and JPMorgans of the world are well capitalized and able to bear risk, we would expect stock prices to be very responsive to contemporaneous news.  This is exactly what we find.

If a high contemporaneous price reaction to news means that more of event day information is incorporated into stock prices, then the response of stock prices to lagged news should be relatively low.  To our surprise, however, we found the exact opposite effect.  At times of high intermediary risk-bearing capacity, when stock prices are very responsive to event day news, stock prices are also very responsive to lagged news: The degree of stock price underreaction to news increases when financial intermediaries are well capitalized.  This is inconsistent with the interpretation that stock underreaction to news is indicative of market inefficiency, because such underreaction is most pronounced at times when we expect the market to be most efficient (since financial intermediaries are present to exploit any apparent mispricing).

What might be going on?  Starting with Kyle (1985), economists have understood that trading by large institutions conveys information to the market.  When Fidelity’s Magellan fund starts buying a stock, that stock’s price begins to increase as other market participants sense that there is a large buyer in their midst.  So, when Magellan buys, it does so carefully.  Rather than executing the entirety of a large order over the course of an hour or two, Magellan is likely to trade slowly, so as not to spook the market; economists refer to this as strategic trading.  There is ample evidence that large trade execution often takes place over several days (e.g. Keim and Madhavan 1995 and Huang, Tan, and Wermers 2019).  Perhaps the underreaction first documented by TSM is more indicative of such strategic trading by large institutions than it is of market participants underreacting to new information in news stories.  When Magellan is slowly buying a large new stock position, we may expect that stock’s price to increase over the several days that it takes Magellan to execute the trade.  This would be consistent with both of our findings with regard to the capitalization level of financial intermediaries: (1) When financial intermediaries are better capitalized, stock prices are more responsive to contemporaneous news, and (2) when financial intermediaries are better capitalized, stock prices are more responsive to lagged news because institutional investors trade actively but strategically in response to such news.

To test whether this mechanism is responsible for the stock-news underreaction, we try to isolate particular situations where institutional investors might have the most incentive to trade in response to news.  We do so in two ways.  First, we examine the relationship between short-interest in a stock and the sentiment of event day news.  Second, we examine the relationship between the degree of a stock’s institutional ownership and event day news sentiment.  We find, first, that stock price underreaction to news is most pronounced for heavily shorted stocks in companies subject to good news.  In such cases, stock prices have a pronounced tendency to go up over the 10 days following the news release.  We interpret this as short-covering by large institutional traders, where executing short-covering trades takes several days.  Second,  stocks that institutions invest heavily in and that experience bad news show a pronounced tendency to fall in price over the ensuing 10 days.   We interpret this as window-dressing by large institutional investors who are averse to owning a stock that is the subject of bad news.  Again, such repositioning might take large investors several days.  What remains after we control for cases where institutions are likely to be active traders?  Stock prices overreact to news.  Indeed, stocks that are not heavily shorted and not owned largely by institutions tend to move in the direction opposite to event day news over the ensuing 10 days.  We interpret this as strong evidence that strategic trading by large institutions is responsible for stock-news underreaction.  However, that stock prices sometimes overreact to news is itself potentially a sign of irrational behavior by investors.

In our paper we also show that passive institutional ownership, like intermediary capitalization, is a good variable for capturing time variation in price responses to contemporaneous and lagged news.  The results from this interaction are qualitatively similar to those we’ve already discussed.  Furthermore, we analyze whether the time variation in the news-return relationship may be related to time variation in the information content of the news itself, and we find evidence supportive of this idea.


Huang, A., H. Tan, and R. Wermers, 2019, “Institutional trading around corporate news: Evidence from textual analysis,” working paper.

Keim, D., and A. Madhavan, 1995, “Anatomy of the trading process Empirical evidence on the behavior of institutional traders,” Journal of Financial Economics, 37, 371–398.

Kyle, A., 1985, “Continuous auctions and insider trading,” Econometrica, 53, 1315–1335.

Tetlock, P., M. Saar-Tsechansky, and S. Macskassy, 2008, “More than words: Quantifying language to measure firms’ fundamentals,” Journal of Finance, 63 (3), 1437–1467.

This post comes to us from Paul Glasserman, the Jack R. Anderson Professor of Business at Columbia Business School; Fulin Li, a PhD candidate at the University of Chicago’s Booth School of Business; and Professor Harry Mamaysky at Columbia Business School. It is based on their recent paper, “Time Variatino in the News-Returns Relationship,” available here.