Evidence Regarding Market Efficiency From Studies

1
Evidence Regarding Market Efficiency From Studies
2
Background Information

• Early 1970s, Fama MacBeth did a famous study
  testing the CAPM.
• They found weak evidence that portfolios of
  stocks with higher betas had higher returns, and
  found an intercept slightly higher than zero.
  (CAPM Assumes Alpha 0)

3
Beta Return of Portfolios
Return
Beta
4
Early Evidence

• Early evidence basically supported the weak and
  the semi-strong form EMH.

5
Early Weak Form EMH Tests

• () Serial Correlation
• returns follow returns for a given stock or -
  returns follow - returns for a given stock
• Called momentum or inertia

6
Early Weak Form EMH Tests

• (-) Serial Correlation
• returns follow - returns for a given stock or -
  returns follow returns for a given stock.
• Called reversals

7
Tie to a Random Walk

• If we find () or (-) serial correlation, this is
  evidence against the weak-form EMH as it implies
  that past prices can be used to predict future
  prices.
• (Technical analysis)

8
Early Weak Form EMH Tests

• In 1960s Fama showed that
• 1. Stock Prices followed a random walk
• 2. No evidence of serial correlation. The price
  of a stock is just as likely to rise after a
  previous days increase as after a previous days
  decline.

9
Early Semi-Strong Form EMH Tests

• Event studies in the 1960s 1970s looked at
  stock prices around the release of new
  information to the public.
• (Fundamental analysis)

10
Graph of a Typical Study

• Keown and Pinkerton (1981) CARs for target
  firms around takeover attempt.
• See graph on p. 371 in text

11
Challenges to the EMH

• 1980s 1990s
• Empirical evidence began to accumulate that
  provided evidence first against the semi-strong
  EMH and later against the weak form EMH
• Initially any evidence against EMH called an
  anomaly.

12
More Recent Tests of the Semi-Strong Form EMH

• Are abnormal risk-adjusted returns possible if
  you trade after information is made public?
  (fundamental analysts)
• General Equation for Abn. Returns
• Actual Rit Predicted Ri,t

13
Abn. ReturnsUse Historic Data

• Without a risk adjustment
• Actual Rit Actual Rm,t
• With a risk adjustment
• Actual Rit ai BiActual Rm,t
• Or,
• Actual Rit Actual Rmatch,t

14
Challenges to Testing

• Difficult to measure risk-adjusted returns
• a) Is beta the proper measure of risk?
• b) CAPM is forward looking and you are using
  historic data.
• c) Is your matched firm the best match?

15
Quarterly Earnings Surprises

• (Quarterly EPS Released
• Forecasted Quarterly EPS)
• Measure the abnormal risk-adjusted return after
  an earnings surprise.
• Measure CAR Actual Rit Predicted Ri,t
• (Used CAPM)

16
Quarterly Earnings Surprises

• Rank from highest to lowest by magnitude of
  earnings surprises and place stocks into decile
  portfolios.
• See if trading on earnings surprises results in
  subsequent abnormal returns.
• (Cumulative Abnormal Returns (CARs) are the daily
  abnormal returns summed up over time)

17
Evidence Quarterly Earnings Surprises

• For positive earnings surprises
• The larger the earnings surprise the higher the
  positive abnormal return.
• The upward drift in the stock price continues a
  couple of months after the earning announcement!

18
Evidence Quarterly Earnings Surprises

• For negative earnings surprises
• The larger the negative earnings surprise the
  larger the loss as measured by the abnormal
  return.
• The downward drift in the stock price continues a
  couple of months after the earning announcement!

19
Interpretation Mkts Efficient Measurement
Errors

• Markets are efficient. The evidence of abn.
  risk-adjusted returns is due to various
  Measurement Errors when using the CAPM.
• (1) Benchmark Error Beta SML wrong
• (2) CAPM is a forward looking model are
• testing it with historic or ex-post
  data.

20
InterpretationCAPM Not Valid

• Markets are efficient. The evidence of abnormal
  risk-adjusted returns (evidence against market
  inefficiency) is inconclusive as the CAPM may not
  be the proper risk adjustment model.
• Joint or Dual Hypothesis Problem!
• If the CAPM is wrong, then abnormal risk-adjusted
  returns using this model are wrong.

21
Interpretation Mkts Not Efficient

• Behavioral Finance Psychological and behavioral
  elements lead to predictable biases.
• Arbitrage
• Not always possible to execute arbitrage trades.
• Arbitrage is risky and therefore limited

22
Evidence of Abn Risk Adj. Returns .

• After share repurchase announcements
• (Ikenberry (1995))
• After dividend initiations and omissions
• (Michaely (1995))
• After stock splits
• (Ikenberry (1995))
• After seasoned equity offerings after IPOs
• (Loughran and Ritter (1995))

23
Size Effect

• Portfolios of small cap stocks earn positive
  abnormal risk-adjusted returns ( alphas)

24
Size Effect

• January Anomaly Most of the abnormal returns
  occur in January! (tax loss selling??)
• Grossman/Stiglitz Professionals move prices to
  efficiency. Dont buy at the small cap end of the
  market much due to limits on portfolio positions.

25
Problem With CAPM?

• Possible sources of risk for small caps
• Neglected by analysts and institutional
  investors, so is less information, which implies
  higher risk.
• Less Liquidity Higher trading costs as bid-ask
  spreads are wider, and broker commissions are
  larger.

26
Background Information

• Back to Early 1970s, Fama MacBeth test of
  CAPM.

27
Fama MacBeth CAPM Test Early 1970s
Return
Beta
28
Relationship Between Beta and Returns

• Fama French re-examined the earlier tests of
  the CAPM forming size decile portfolios.

29
Fama-French 1992
30
Beta Return of Portfolios
Small cap stocks
Return
Large cap stocks
Beta
31
Fama-French Interpretation

• See that small cap stocks have higher betas than
  large cap stocks. Fama and French concluded that
  size is driving the relationship between beta and
  return not beta!

32
Previous Slide (cont)

• Also see that within the small cap groupings,
  portfolios of stocks with lower betas have higher
  returns! The same is true within the large cap
  groupings.

33
Interesting Fact

• Fama, once a strong proponent of the CAPM now
  claimed that beta was dead. Beta was a rough
  proxy for size in his earlier tests!!

34
The Cross Section of Expected Stock Returns
Table 1 Panel A

35
Interesting Result

• Within each size group, the higher the beta the
  lower the return.

36
The Cross Section of Expected Stock ReturnsTable
5
37
Value Puzzle

• It is not evident why value stocks should
• be riskier than growth stocks. Value stocks
  have lower standard deviations than growth stocks
  after controlling for size.

38
Value Puzzle

• Value Puzzle
• Value stocks have lower standard deviations and
  higher returns!

39
Fama-French Findings

• Beta does not explain returns.
• Small cap stocks have higher returns. Small cap
  stocks have higher betas, but it is size not beta
  driving higher returns.
• Low P/E or high Book-to-Market of equity stocks
  have higher returns.

40
Explanations for Fama-French Results

• Alternative Explanations for their results?
• Market Semi-Strong Efficient
• Small cap stocks and low P/E (high B/M) stocks
  generate higher returns because they are riskier.
  However, this risk is not captured by Beta!

41
Problem

• Lack of a theoretical model to explain why size
  and style (value vs growth) are important risk
  factors. The CAPM had an elegant, logical theory
  underlying it, this has none!

42
Explanations for Fama-French Results

• Market Semi-Strong Efficient
• Abnormal risk-adjusted returns for small cap
  stocks or for stocks with low P/E (or high B/M)
  are due to various measurement errors when using
  the CAPM.
• (1) Benchmark Error Beta SML wrong
• (2) CAPM is a forward looking model we are
• testing it with a historic or ex-post
  data.

43
Explanations for Fama-French Results

• Market Semi-Strong Efficient. Abnormal
  risk-adjusted returns (evidence against market
  inefficiency) are inconclusive as the CAPM may
  not be the proper risk adjustment model.
• Joint or Dual Hypothesis Problem!
• If the CAPM is wrong, then abnormal risk-adjusted
  returns using this model are wrong.

44
Explanations for Fama-French Results

• Market Not Semi-Strong Form Efficient
• Can make abnormal returns using public
  information regarding market capitalization and
  P/E or B/M ratio.
• How can this persist?

45
Behavioral Finance

• Decisions people make deviate from the maxims of
  economic rationality in predictable ways
• 1. Attitudes towards Risk
• 2. Non Bayesian Expectation
• Formation
• 3. Framing Effects of Decisions

46
Attitudes Toward Risk Example

• 90 chance of 1 million 10 chance of 0. I
  offer to buy you out for 900,000. Will you take
  my offer?

47
Attitudes Toward Risk Example

• 90 chance to lose 1 million 10 chance of 0.
  I will take the bet if you pay me 900,000. Will
  you take my offer?

48
Behavioral Finance

• Attitudes Towards Risk
• People look at gains and losses relative to some
  reference point rather than the levels of final
  wealth.
• Display Loss Aversion! Outcome Typically Doesnt
  follow standard von Neumann-Morgenstern
  rationality.

49
Behavioral Finance

• Non-Bayesian Expectation Formation
• Representativeness Predict the future taking a
  short history of data and determine the model
  driving the data. (Too small a weight on
  chance.)
• Conservatism Slow updating to new information
  as have extrapolated a short earnings history too
  far into the future.

50
Non-Bayesian Expectations

• 1st 2 winters here mild. Assumed they were
  always like that.
• Investors may extrapolate short histories of
  rapid earnings growth too far in the future and
  may overprice glamour stocks.

51
Behavioral Finance

• Framing Effects
• How data is presented can affect the decisions
  people make.

52
Framing Effects Example

• Investors will allocate more money to stocks
  rather than bonds when they see long-term
  cumulative wealth graphs than they will if you
  only show them volatile short-term stock returns.

53
Behavioral Finance Explanation for Quarterly
Earnings Surprise

• In this case, would argue that initially there is
  slow updating or conservatism as a reaction to
  the news released by the earnings surprise.
  Short run under-reaction
• Eventually keep seeing good news so
  representativeness sets in then get
  over-reaction.

54
Mkt Not Efficient?

• (Lakonishok, Shleifer and Vishney)
• These professors offer a different
  interpretation. Markets are inefficient. People
  overreact with a lag. Overprice firms with good
  recent returns (growth) and underprice firms with
  poor recent returns (value).

55
Long-Term Horizons Test of Weak-Form EMH

• DeBondt and Thaler (1985)
• Create Loser and Winner portfolios based on past
  36 months of CARs. Top decile are Winners, bottom
  decile are Losers.
• Examine CARs for next 36 months.
• Losers outperform winners Is an overreaction
  followed by a correction.

56
Efficient Market Believers Say....

• Evidence is due to market risk premiums varying
  over time. Is not overshooting correction but
  instead a rational response to changes in the
  discount rate.

57
Short Horizons(Tests of Weak Form EMH)

• Lo and MacKinlay (1988) test to see if there is
  serial correlation of weekly stock returns for
  NYSE stocks.

58
Lo MacKinlay
Stock Price
momentum
reversal
reversal
- momentum
1
2
Period
59
Lo MacKinlay

• If momentum is present, the variance of returns
  should increase as the number of periods used is
  increased.
• If there is no momentum, gains or losses will
  tend to reverse, keeping the variance of returns
  from becoming wider.

60
Evidence Lo MacKinlay

• Lo and Mackinlay (1988) find serial correlation
  of weekly stock returns for NYSE stocks as the
  variance of returns increases as the return
  interval is lengthened. Implies there is inertia
  in the short run.

61
Evidence Lo MacKinlay

• The effect is the strongest in the small cap
  stocks.
• Not clear if abnormal returns are possible by
  exploiting this information.

62
Intermediate HorizonsTest of Weak-Form EMH

• Study by Jegadeesh and Titman.

63
Intermediate Horizons

• 1.Measure stock rates of return over the past 6
  months.
• 2.Rank the stocks from highest to lowest past 6
  month return and then divide the sample into
  deciles. Losers are the bottom decile and
  winners are the top decile

64
Jegadeesh and Titman

• 3. For the next 36 months, every time one of the
  winners or losers reports quarterly earnings,
  record 3-day returns starting 2 days before the
  earnings announcement and ending the day of the
  announcement.
• 4.Observe the difference in 3-day returns between
  the winners and losers reporting earnings in each
  month.

65
Evidence Jegadeesh and Titman

• For the 1st 7 months, the market is pleasantly
  surprised by the earnings announcements of the
  winners and disappointed by the earnings
  announcements of the losers.
• (momentum in the short run)

66
Evidence Jegadeesh and Titman

• From months 9 - 36, the market is pleasantly
  surprised by the earnings announcements of the
  losers and disappointed by the earnings
  announcements of the winners.
• (Reversals in the intermediate term)
• If the stock market is efficient, it should be
  able to anticipate the good or bad reports in
  advance.

67
Evidence Jegadeesh and Titman

• Abnormal profit opportunities.
• Reversion to the mean.
• The market overreacts with a lag. Consistent with
  Representativeness and Conservatism.
• Short Run Inertia
• Intermediate Run Reversals

68
Technical Analysts

• Technical analysts claim to exploit these trends
  or patterns.

69
Mutual Fund Performance

• If the stock market is not weak or semi-strong
  form efficient, then professional portfolio
  managers should be able to achieve abnormal
  risk-adjusted returns!

70
Evidence Mutual Funds

• Malkiel (1995) examined the alphas of mutual
  funds.
• Recall Regression Model
• (Ri,t RFRt) ?i ?i(Rm,t - RFRt) ei,t
• If market is efficient what should we find
  regarding the multiple-period alpha?

71
Evidence Mutual Funds

• WSJ Article, Stock Funds Just Dont Measure Up.
  Oct. 5, 1999
• After adjusting for size and survivorship bias,
  funds trailed the SP 500 by 1.4 per year which
  is on average what they charge for annual
  expenses.

72
Evidence Mutual Funds

• Other studies 1970s 1990s After expenses
  commissions, only 1/3 beat the market on a
  risk-adjusted basis.

73
STRONG FORM EMH TESTS

• Are abnormal risk-adjusted returns possible if
  you trade using private information?

74
Evidence on Insiders

• Corporate insiders are required to report their
  transactions to the SEC.
• They are not supposed to trade when in the
  possession of material information.
• Even with regulation, they achieve positive
  risk-adjusted abnormal returns.

75
Market Crash of Oct. 1987

• 23 Drop in One Day??
• No large release of news
• Efficient Market explanation Due to chance. Are
  outliers in the distribution. Just an outlier
  observation in a random process.
• Panic Crowd Psychology (behavioral finance
  explanation)

76
Internet Bubble

• Some companies saw their stock price go up just
  by adding dotcom to their names
• When 3-Com spun off Palm Pilot, but kept 95 of
  the shares, The 95 of Palm owned by 3-Com were
  worth more than the market cap of 3-Com. Implies
  negative value for the rest of 3-Com!

77
Internet Bubble

• It is obvious now that the 1998-March 2000 tech
  run-up was a bubble, but was this market
  inefficiency, or merely poor valuations?
• How do you know a bubble when you are in it?
• Should you try to short a bubble if you dont
  know when it will burst?

78
Limits of Arbitrage

• Just because you know something is overvalued or
  undervalued, doesnt necessarily mean you can
  make money off it
• Classic Example We know that someday the sun
  will explode, but you cant short the Earth

79
Shleifer and Vishny (1997) Paper

• Most arbitrage is not carried out by small
  investors, but by large money managers.
• They usually manage OPM (other peoples money)
• Most arbitrage in the real world is actually
  risk arbitrage and requires capital

80

• If money managers observe a price discrepancy and
  commit capital to an arbitrage position based on
  convergence, the initial movement may be away
  from convergence, but that merely means there is
  a greater opportunity for profit, and more
  capital should be committed.

81

• But that is exactly when investors are most
  likely to pull out.
• Investors invest based on PBA (Performance Based
  Arbitrage) rather than expected returns
• This lack of capital prevents arbitrage from
  taking place

82

• This is often given as an explanation for the
  collapse of LTCM (Long-Term Capital Management).
• Amazingly, the Shleifer and Vishny paper came out
  about a year prior to the LTCM collapse