The Role of Media in the Stock Market. November Paul Tetlock Columbia University

Transcription

1 The Role of Media in the Stock Market November 2013 Paul Tetlock Columbia University

2 Motivating Questions What kind of information moves stock prices? Fundamentals (E(profits)) vs. investor sentiment Mundane vs. sensationalist news Do stock prices move too much or too little? How can we tell? News data shed light on these questions

3 Reaction to News and Non-news In efficient markets (EMH), stock prices should: 1) React quickly and accurately to news about fundamentals 2) Not react to irrelevant facts/rumors (no excess volatility ) We ll systematically examine both predictions First, consider the example of United s stock United Airlines filed for bankruptcy in 2002 The firm exited bankruptcy in 2006 On 9/7/08, United s stock market cap is $1.6B

4 United Stock on Sept 8, 2008 Google News posts a 6-year-old Chicago Tribune story about United s 2002 bankruptcy United s stock falls 76% within minutes United rebounds, but remains down 11% on the day

5 Volatility Should Be Linked to News News can help us test the EMH if news events signify large changes in stocks fundamental values EMH: Prices should react to changes in fundamentals (i.e., news); and prices should not move otherwise Two EMH predictions concerning news and prices: Large price moves should occur on news days Large price moves shouldn t occur on non-news days

6 Cutler, Poterba, and Summers (1989)

7 Cornell (2013): Largest price moves from 1988 to 2012 Rank Date Return News 1 10/13/ % 2 10/28/ % 3 10/15/ % 9 8/8/ % 20 1/8/ % 24 7/29/ % 25 1/3/ % Governments throughout the world announce moves to support troubled banks. Late rally on Wall Street as rebound in stocks defies latest economic news. Falling retail sales and rising wholesale prices spike fears of recession and erase Monday's record rally. Fearful investors reacted to the US losing its coveted AAA credit rating. Stocks were hammered by worries about the economy and possibly stricter market regulation. Investors, growing less worried about earnings and corporate bookkeeping, are ready to buy again. Citing signs of economic slowdown, the Fed unexpectedly cut its target to 6%. 31 8/4/ % Lukewarm jobs report hints that economic deterioration may be stagnation; triggers Fed worries. Source: CRSP Index; Cornell (JPM, 2013)

8 Tests for Overreaction EMH: If the market reacts appropriately to non-news (or uninformative news) events, post-event returns should not be predictable But post-event returns are predictable Specifically, the market reaction to the event negatively predicts post-event returns Chan (2003) and Tetlock (2007)

9 Which News Might Elicit Overreaction? Some financial news is sensationalist May convey or influence traders psychological states Tetlock (2007) investigates market-wide news Uses a popular WSJ column: Abreast of the Market Description of the column from a former WSJ editor: One of the more fascinating sections of the WSJ is on the inside of the back page under the standing headline Abreast of the Market. There you can read each day what the market did yesterday, whether it went up, down or sideways as measured by indexes like the Dow Jones Industrial Average.... In that column, you can also read selected post-mortems from brokerage houses, stock analysts and other professional track watchers explaining why the market yesterday did whatever it did, sometimes with predictive nuggets about what it will do today or tomorrow. This is where the fascination lies. For no matter what the market did up, down or sideways--somebody will have a ready explanation.... -Vermont Royster (January 15, 1986)

10 Staying Abreast of the Market Many traders and journalists claim to know why the market moved at least, in hindsight Fed policy Housing market Oil supply Exchange rates Innovation War

11 Example of Quantifying Content Abreast of the Market on Feb 17, 2009 Headline: Market s `Hope Balloon Loses Air; Tepid Upturns Haven t Stopped the Slide Financial markets are supposedly driven by two competing forces: fear and greed. Fear just made another grab for the steering wheel. Disappointment with the government s planned credit-market bailout and concerns that the $787 billion stimulus plan won t jolt the economy fast enough snuffed out the budding stockmarket rally. Now investors are worried that stocks could fall back to their November lows -- and possibly even farther. Consider a simple methodology for evaluating content Compute daily relative frequency of negative words e.g., 9 negative out of 82 = 11.0%, which is high in this case Adopt a list of negative words from psycho-linguistics (Harvard-IV) E.g., collapse, disappoint, fear, pessimism, worry, ugly, etc.

12 A General Content Measure Tetlock (2007) measures 77 word categories in the Harvard-IV psychosocial dictionary Examples: negative, positive, active, passive, strong, weak, overstated, and painful words For each day s Abreast of the Market column, compute the relative frequency of all 77 categories Extract the first principal component from the 77 by 77 covariance matrix of the word categories Examine the eigenvector s 77 elements (categories) Negative, weak, fail, and fall are the key elements

13 Cumulative Returns on the DJ Index (Basis Points) Negativity Predicts DJ Decline and Reversal Days After the Initial Pessimism Shock

14 News, Information, and Underreaction Financial news is often not sensationalist May be related to firms fundamental values Could signify changes in traders expectations Tetlock et al. (2008) explore firm-specific news Use all DJNS and WSJ stories about S&P 500 firms Compute daily relative frequency of negative words Analyze what happens after negative words appear

15 Example of the Reaction to News DJ News Service story about Microsoft on 1/8/99 The alleged pricing abuse will only get worse if Microsoft is not disciplined sternly by the antitrust court, said Mark Cooper, director of research for Consumer Federal of America. (possible $10 billion class action lawsuit) News = 0 or 1 (or # of wire messages, which was five) Fraction of negative words: 5 of 29 total words = 17.2% Ranks in the 99th percentile of 1998 news data Whole article ranks in the top decile for 1998, too Microsoft s CARs were -42, -141, and -194 basis points on days -1, 0, and +1 surrounding the news event

16 If Content Is Informative, Negative Words Should Predict Firm Earnings

17 Number of News Stories Fundamental Content Is Important Coverage patterns suggest earnings news is important Time Relative to Earnings Announcement (Trading Days)

18 Cumulative Abnormal Returns (in Basis Points) Brief Underreaction to Information 150 Difference WSJ Difference DJNS Positive WSJ Positive DJNS Negative WSJ Negative DJNS Trading Day Relative to Story Release

19 Underreaction to Earnings Information Focus on stories that mention word stem earn*

20 Key Lessons from Research Trading activity and price movements are related to news, but less than we might think Some investors find it hard to distinguish informative and uninformative news Underreact to real info and overreact to non-info

21 Application to Financial Crises News could help us discern whether a stock market crash is justified by fundamentals Prices should be based on macroeconomic news Are prices moving on news or noise trading? Test: Analyze concurrent macroeconomic news E.g., is consumer spending likely to be low? Caveat: We have only a few data points on crises

22 A Tale of Two Crises 1) 1987 stock market crash 2) financial crisis

23 S&P 500 Index (Log Scale) Consumer Sentiment (100 = Normal) Identify Relevant Valuation Data Data: 87 was noise, whereas 08 was news S&P 500 Index Consumer Sentiment Jan-80 Jun-85 Dec-90 Jun-96 Nov-01 May-07 Nov-12

24 Future Work News now arrives on social networks, blogs, RSS feeds, and numerous other channels Investors actively search for Internet content Combining quantitative and qualitative content can improve: Our understanding of market prices and volume Our ability to predict key economic events