No News is News: Do Markets Underreact to Nothing?

Transcription

1 No News is News: Do Markets Underreact to Nothing? Stefano Giglio and Kelly Shue University of Chicago, Booth School of Business April 3, 2013

2 No News is News No news and the passage of time often contain information General contexts Sustained period without terrorist attacks Employee executes task without screwing up Dog that Didn t Bark Financial contexts Firm gets a negative shock: after the shock, the firm does not default, declare bankruptcy, or lay off workers Firm has lack of new investments

3 How Do Markets React to No News? Rational Response Bayesian updating on no news events In financial markets, prices can drift along the path of no news (insider trading: Marin and Olivier, 2008; Gao and Ma, 2012) Behavioral Response: Imperfect updating on no news Limited Attention: attention directed at salient or vivid signals (Tversky and Kahneman, 1983; Hirshleifer and Teoh, 2004; Corwin and Coughenour 2008) Underreaction to no news can cause mispricing and persistent misallocation of resources

4 Empirical Context: Mergers Mergers offer a convenient setting to test for reactions to no news Clear start point: announcement of intent to merge Target price depends on a clearly defined and risky end point: completion (usually 4-5 months) withdrawal (SEC, DoJ, shareholder approval, financing)

5 Passage of Time vs. Explicit News Explicit news (news stories or rumors) may be released Define no news component of each day as the passage of time i.e. what we should learn about the merger even if we are unable to observe any explicit news Measure the information content of the passage of time by estimating event-time variation in empirical hazard rates Explicit news and passage of time may contain the same info Possible that the market underreacts to both explicit news and passage of time At a minimum, the market underreacts to the passage of time

6 Passage of Time vs. Explicit News An Example: Suppose after month 6, the hazard rate of completion falls to 0 Even if we are unable to observe any explicit news, once we reach month 6 the price should have dropped This is the information content of the passage of time Now, at month 6 we might also receive a news report the merger is dead If the price still hasn t dropped, agents are underreacting to both signals

7 Types of Mergers 1. Cash financed target shareholders obtain fixed $X/share 2. Equity financed (stock swap): exchange shares of acquirer per share of the target We exclude mergers with known expiration dates (tender offers) and hybrid deals

8 Outline 1. Are hazard rates non-constant? 2. Do hazard rates predict returns? 3. A simple behavioral underreaction model 4. Alternative rational explanations (e.g. risk)

9 Are Merger Hazard Rates Non-Constant? No news is news if the underlying hazard rates are not constant Measure empirical hazard rates of completion and withdrawal h(t): probability of completion during period t conditional on no completion or withdrawal until t w(t): probability of withdrawal during period t conditional on no completion or withdrawal until t

10 Hazard Rates: Cash hazard Full Sample hazard Early Sample trading weeks after announcement trading weeks after announcement Late Sample hazard trading weeks after announcement Completions 95% CI Withdrawals 95% CI

11 Hazard Rates: Equity Full Sample Early Sample hazard hazard trading weeks after announcement trading weeks after announcement Late Sample hazard trading weeks after announcement Completions 95% CI Withdrawals 95% CI

12 Hazard Rates: Heterogeneity We may be estimating hazard rates with error more information would lead to even more predictability in returns (bias against us) Show that, with standard unobserved heterogeneity (frailty), even more time variation in h(t)

13 Moving to Returns Rational markets should incorporate all available information If the merger is likely to complete tomorrow high price today expected returns only compensate for risk A puzzle: returns are predictable in event time A further puzzle: returns align with hazard rates over event time

14 Hazard Rates and Returns Cash Mergers: Hazard Rates Equity Mergers: Hazard Rates hazard hazard trading weeks after announcement trading weeks after announcement Completions 95% CI Withdrawals 95% CI Completions 95% CI Withdrawals 95% CI Cash Mergers: Returns in Event Time Equity Mergers: Returns in Event Time weekly return weekly return trading weeks after announcement trading weeks after announcement Mean Weekly Return 90% Pointwise CI Mean Weekly Return 90% Pointwise CI

15 Why do Hazard Rates Predict Returns? We usually don t expect returns to vary predictably over time, unless Behavioral biases + limits to arbitrage 2. Changes in risk or frictions

16 Simple Behavioral Model of Underreaction At t = 0 the merger is announced: acquirer offers P C ˆP(t) is price of the target conditional on no completion or withdrawal up to t If the merger completes, the price jumps to P C If the acquirer withdraws, the price reverts to P 0 dp 0 (t) = µp 0 (t)dt + σp 0 (t)dz(t) We solve for ˆP(t): depends on beliefs ĥt Realized returns however depend on both ĥt (beliefs) and h t (true hazards)

17 Model Predictions about Returns Suppose that all risk is idiosyncratic If beliefs are correct, mean excess returns should always be zero If the market underestimates the true hazard rate of completion, mean excess returns will be greater than 0 (positive surprises from completions) and vice versa If agents underreact to no news (ĥ varies less than the true h), they will underestimate h when h is high and overestimate h when h is low This generates positive correlation of hazards and returns These predictions still hold if there is constant systematic risk over event time. Will discuss later about risk changing over event time!

18 Backing Out Beliefs Cash Mergers 0.06 Completion Hazard Beliefs True Hazard Trading Weeks after Announcement Equity Mergers 0.08 Completion Hazard Beliefs True Hazard Trading Weeks after Announcement

19 Other Information in the Passage of Time Hazard rate is ONE important reason why the passage of time contains info The passage of time may contain other information Value of target in absence of merger or probability of receiving a competing bid may vary in event time with hazard rates We can t distinguish between: 1. Markets underreact to changes in hazard rates 2. Markets have correct beliefs about hazard rates but fail to update on other changes in merger value that move with hazard rates in event time Both represent underreaction to the passage of time!

20 Risk as an Alternative Explanation The results cannot be driven by calendar-time variation in risk or risk premia because our results hold in event time What if risk covaries with hazard rates in event time? Use time-series portfolio returns to explore event-time variation in: 1. Systematic risk 2. Downside risk 3. Idiosyncratic risk

21 Trading Strategies Cash Merger: buy 1 share of the target Equity Merger: buy 1 share of the target, short shares of acquirer Each month, invest equally in all deals that are active within certain event windows (see next slide) Modification of standard merger arbitrage (Mitchell and Pulvino, 2001): enter after announcement, buy and hold until completion/withdrawal

22 Trading strategies hazard low haz 1 high hazard low haz trading weeks after announcement Only use information in hazard rates, not returns 3 strategies: low-hazard 1, high-hazard, low-hazard 2 Each calendar month, invest in deals active in the relevant event window After controlling for risk, rational updating implies α high = α low 1 = α low 2

23 Strategy Fama-French Alphas Cash Equity Individual Strategies Tests: P-values Alpha Stderr High > Low Low hazard High hazard *** Low hazard Buy and hold *** Low hazard *** High hazard *** Low hazard Buy and hold *** Obs 3190 Joint test R Results High>BH robust to computing hazards in earlier period ( ), High>BH>Low trading in later period ( )

24 Event Time Variation in Betas Cash Mergers: Betas in Event Time portfolio strategy betas trading weeks after announcement Equity Mergers: Betas in Event Time portfolio strategy betas trading weeks after announcement Rm Rf SMB HML

25 Downside Risk Strategy Return Minus Risk Free Rate Market Return Minus Risk Free Rate Low Haz 1 Low Haz 2 High Haz Buy and Hold Exposure to option factors also do not vary in event time

26 Downside Risk Idiosyncratic risk does not seem hump shaped Standard deviations of returns are: 0.023, 0.040, Most difference due to diversification and number of deals We have on average 20 deals per month Real arbitrageurs even more diversified

27 Portfolio Returns Comparisons Annual Return Calendar Year High Haz Strategy Market Buy and Hold Strategy Annual Return Calendar Year High Haz Strategy Market Buy and Hold Strategy

28 Other Rational Explanations Buying or selling pressures from liquidity traders Right after announcement, mutual funds sell targets to lock in gains and because M&A doesn t fit their investment focus Arbitrageurs may not be able to buy fast enough to relieve the selling pressure Market liquidity should move negatively with returns in event time Information asymmetry may vary in event time Bid-ask spread should move with returns in event time

29 Giglio and Volume Shue (Chicago doesbooth) not vary with returns in event time (same for turnover Introduction Hazards Returns Model Risk Limits to Arb Conclusion Appendix Volume weekly return Cash Mergers: Returns in Event Time weekly return Equity Mergers: Returns in Event Time trading weeks after announcement trading weeks after announcement Mean Weekly Return 90% Pointwise CI Mean Weekly Return 90% Pointwise CI Cash Mergers: Volume Equity Mergers: Volume median (volume / volume in event week 1) low$haz$1 high$haz low$haz$2 median (volume / volume in event week 1) low$haz$1 high$haz low$haz$ trading weeks after announcement trading weeks after announcement

30 Limits to Arbitrage A behavioral underreaction model can generate observed returns However, sophisticated merger arbitrageurs are likely to exist Behavioral market participants + limits to arbitrage Lack of liquidity, transactions costs, price impact

31 Trading Frictions Cash High hazard alpha Equity High hazard alpha Size Volume Bid-ask spread Time period Small Large Low High High Low Early Late *** *** *** *** ** (0.0024) (0.0021) (0.0022) (0.0021) (0.0028) (0.0017) (0.0021) (0.0024) *** *** *** *** *** *** *** *** (0.0037) (0.0021) (0.0035) (0.0026) (0.0054) (0.0021) (0.0032) (0.0030) P-values Illiquid > Liquid High haz > Low Haz Alphas are higher for smaller and less liquid deals

32 Trading Costs Pooled Strategy (Cash and Equity Mergers) No Trading Costs Alpha Trading Costs Alpha (Method 1) Trading Costs Alpha (Method 2) Low hazard ** (0.0012) *** (0.0010) *** (0.0009) High hazard *** (0.0018) *** (0.0012) (0.0012) Low hazard (0.0029) (0.0009) ** (0.0008) Buy and hold *** (0.0011) ** (0.0010) (0.0009) P-value: High > Low Obs R Simulated trading costs: Impose maximum portfolio weight of 10% per deal, conservative estimates of indirect (price impact) and direct transaction costs (broker commissions)

33 Strategy Returns: Cash Mergers 45 x Strategy starting event week Strategy ending event week 6

34 Strategy Returns: Equity Mergers 45 x Strategy starting event week Strategy ending event week 6

35 Conclusion We test whether markets underreact to no news, using the natural experiment of mergers Merger returns track hazard rates of completion in event time explained by a model of flat beliefs about the hazard rates not explained by event time variation in risk or frictions Consistent with behavioral underreaction and limits to arbitrage Underreaction to no news may cause persistent misallocation of resources and mispricing in many other contexts

36 Descriptive Statistics Cash Mergers Equity Mergers Mean Median Stdev Mean Median Stdev Number of deals Time to completion (trading days) Time to withdrawal (trading days) % Completed within one year % Withdrawn within one year % Pending within one year Premium Size ($mil) Size 1980s ($mil) Size 1990s ($mil) Size 2000s ($mil)

37 Hazard Rates and Returns Dep Var: (1) (2) (3) Weekly Return Cash Equity Cash Equity Cash Equity Weekly hazard *** *** *** *** ** ** (0.011) (0.009) (0.011) (0.011) (0.013) (0.017) Calendar year x month FE N N Y Y Y Y Stderr clustered by deal & Y Y Y Y Y Y (calendar year x month) Split sample N N N N Y Y (early hazards, late returns) Obs R

38 Typical Timeline CASH MERGER Estimated Timeline Month June July August September October Week Due Diligence Negotiate Merger Agreement Sign Merger Agreement and Announce Transaction Draft Proxy Statement File 8-K containing Merger Agreement - Other SEC filings relating to investor communications likely Hart-Scott Rodino Antitrust Filing Hart-Scott-Rodino Waiting Period (30 days) SEC Review/No Review Decision Expected Respond to SEC Comments Mail Proxy (Shaded = No SEC Review) Shareholder Meeting and Vote 1 Closing 1 If the Merger Agreement contains a force the vote provision, the board of directors of target would be unable to terminate the Merger Agreement prior to the shareholder vote. This timing advantage would reduce the likelihood of an interloper succeeding in a topping bid, as the shareholders would have to vote down the transaction before the interloper s transaction could be approved by the board.

39 A Simple Model Price of target if no completion or withdrawal until t: ˆP(t) = E t { T t + T t e r(z t) e z t [ĥ(k)+ŵ(k)]dkĥ(z)p C dz e r(z t) e z t [ĥ(k)+ŵ(k)]dk ŵ(z)p 0 (z)dz + e r(t t) e T t [ĥ(k)+ŵ(k)]dk P 0 (T )} is the expectation under the risk-neutral probability measure, and all hazards here are risk-neutral E t For simplicity, we assume all risk is idiosyncratic (can use objective hazard rates)

40 Model Predictions E [ret t ] = rdt + + [ ] P ] C ˆP(t) 1 [h(t) ĥ(t) dt }{{} >0 [ ] P 0 (t) ˆP(t) 1 [w(t) ŵ(t)]dt }{{} <0

41 Why are Merger Returns High in General? Simple buy-and-hold has a positive alpha Mitchell and Pulvino (2001): 8 basis points of the monthly CAPM alpha for the buy-and-hold is explained by downside risk Most of the alpha is compensation for transaction costs THIS PAPER: split the buy-and-hold into three regions: the high haz region has a larger alpha the two low haz regions and buy-and-hold High haz strategy does not have larger transaction costs than low haz strategies or buy-and-hold

42 Overestimating Hazard Rates Right After Announcement? Fitting the behavioral model: agents overestimate hazard rate of completion right after announcement May seem surprising given that most mergers cannot legally complete in the first month But, if explicit news is released that the merger will complete for certain, target price will jump to P C just as if the merger had completed Agents overestimate probability of receiving good explicit news set prices high today, so as to earn a fair rate of return disappointed by lack of good explicit news tomorrow low returns