L1: Behavioral Finance

1
L1 Behavioral Finance

• Discussions on Barberis and Thaler (2003) A
  Survey of Behavioral Finance
• Discussions on Other Papers

2
Market Efficiency and Limit to Arbitrage

• In a world where agents are rational and there
  are no frictions, a securitys price equals its
  fundamental value. Friedman (1953) rational
  traders will quickly undo any dislocations caused
  by irrational traders
• Limit to arbitrage
• Fundamental Risk
• Risk that a surprise is related to a specific
  company
• Noise Trader risk
• They trade irrationally
• Implementation costs

3
Evidence on Irrationality

• Twin shares
• E.g., Royal Dutch and Shell Transport
• Index inclusion
• Internet Carve-outs
• 3Com and Plam Inc.
• A case where there is no fundamental risk and no
  noise trader risk
• The key is the barrier to short selling ?
  arbitrade was limited and mispricing persists

4
Belief-based Behavioral Explanations

• Overconfidence
• Optimism
• People are overconfident of their judgements
• biased parameters
• Representativeness
• People tend to draw a conclusion after observing
  few data points
• Conservatism
• Opposite to representativeness

5
Belief-based Behavioral Explanations

• Belief perseverance
• Once people have an opinion, they stick to it too
  long
• Reluctant to search for evidence against their
  belief
• Treat such evience with excessive skepticism
• Anchoring
• Anchoring too much on the initial number
• Availability Biases

6
Preference-based Explanations

• Prospect Theory
• About investor preferences
• Risk aversion to gains (loss aversion) risk
  loving to losses

7
Preference-based Explanations (2)

• Ambiguity Aversion
• People do not like situations where they are
  uncertain about the probability distribution of a
  gamble.
• Prefer certainty

8
Aggregate Stock Applications

• Equity premium
• - using annual data from 1871-1993,
  Campbell and Cochrane (1999) report that the
  average log return on the SP 500 index is 3.9
  higher than the average log return on short-term
  commercial paper.
• Volatility
• Stock returns and price-dividend ratios are
  highly volable. Annual standard deviation of
  excess log returns on the SP is 18, while that
  of log price-dividend ratio is 0.27
• Predictability
• Stock returns are forecastable. Using monthly,
  real, equal-weighted NYSE returns from 1941-1986,
  FF (1988) show that dividend-price ratio is able
  to explain 27 of the variation of cumulative
  stock return over the subsequent four years.

9
Cross-Sectional Predictions

• Size premium
• Long-term reversals
• The predictive power of scaled-price ratios
• Momentum
• Earnings announcement effect
• Dividend initiations and omissions
• Stock repurchases
• IPOs and SEOs

10
Explanations

• Representativeness
• Overconfident
• others

11
Applications in Investor Behavior

• Insufficient diversifications (home bias)
• Ambiguity aversion
• Naïve diversification
• Excessive trading
• Disposition effect
• Buying decision is attention driven

12
Hong and Stein (JF 1999) -- tong

• Main Idea present a unified framework for
  underreaction, momentum trading and overreaction
  in asset market. It assumes there are two types
  of investors (1) newswatchers who observe some
  private invormation, but dont extract
  information from prices, and (2) momentum
  traders. If information diffuses gradually across
  the population, prices underract in the short
  run, thus momentum traders can profit from trend
  chasing. Simple implementation of momentum
  trading leads to over-reaction at long horizons.

13
(No Transcript)
14
Testable Implications

• Stocks with most information asymmetry enjoy the
  biggest momentum effect
• Small stocks
• Stocks having few analysts to follow
• Stocks most momentum-prone are most
  reversal-prone.

15

• Mental accounting, loss aversion and individual
  stock returns
• by Barberis, N., and M. Huang (2001, JoF) -- Fu
• Improving the way we model investor preferences
• Loss aversion People are more sensitive to
  losses than to gain
• Dynamic loss aversion the degree of loss
  aversion depends on Ri,t-1
• Mental accounting over which people think about
  and evaluate?
• Narrow framing People pay attention to narrowly
  defined gains and losses (firm-level stock
  returns) when making decision
• 1.Individual stock accounting U(Ct, Ri,t-1),
  high mean, more volatile, large value premium
  (P/D effect) aggregate stock returns are
  predictable in the TS.
• 2.Portfolio accounting U(Ct, Pt-1), mean value
  falls, less volatile, value premium in CS
  disappears, more correlated with each other. Less
  successful
• WHY? Change discount rate ?ri,t f(Ri,t-1)

16
Model A Individual Stock Accounting
17
Model B Portfolio Accounting
18
Style InvestingBarberis and Shleifer (JFE, 2003)
-- Tina

• Purpose study asset prices in an economy where
  some investors categorize risky assets into
  different styles and move funds among these
  styles depending on their relative performance.

19
Style InvestingBarberis and Shleifer (JFE, 2003)

• Findings
• 1. assets in the same style comove too much
• 2. assets in different styles comove too little
• 3. reclassifying an asset into a new style
  raises its correlation with that style
• 4. style returns exhibit a rich pattern of own-
  and cross-autocorrelations
• 5. style-level momentum and value strategies are
  even more profitable than those of asset-level

20

• Returns to Buying Winners and Selling Losers
  Implications for Stock Market Efficiency by
  Jegadeesh and Titman (1993) -- Jeff
• Buying past winners and shorting past losers
  generates
• significant positive stock returns over 3 to 12
  months holding
• periods. For example, 6-month/6-month strategy
  can realize a
• compounded excess return of 12.01 per year.
• Profitability is not persistent Part of abnormal
  returns
• generated in the first year after portfolio
  formation dissipates
• in the following two years.

21
Trading Strategies

• J-month/K-month strategy select stocks on the
  basis of returns over the past J months and holds
  them for K months.
• At the beginning of each month t, the stocks are
  ranked in ascending order based on their returns
  in the past J months.
• Based on these rankings, ten decile portfolios
  are formed that equally weight the stocks
  contained in each decile.
• Top is losers and bottom is winners.
• In each month t, buy winners and sell losers and
  hold this position for K months.

22
Source of Profitability

• 3 sources of excess returns cross-sectional
  dispersion in expected returns market factor
  and firm-specific (idiosyncratic) components
• Profitability is not related to systematic risk
  not related to delayed stock price reactions to
  common factors.
• But consistent with delayed price reactions to
  firm-specific information.
• Other tests
• Size and beta based subsamples
• Subperiod January effect
• Event time
• Back-testing

23

• Investor Psychology and Security Market
  Under-and Overreaction by Daniel, Hirshleifer,
  and Subrahmanyam (1998) -- Jeff
• Propose a theory of stock market under- and
  overreaction
• based on two psychological biases
• Overconfidence Overestimate the precision of
  privation
• information, but not public information
• Biased self-attribution Attribute events that
  confirm the
• validity of actions to high ability and events
  that disconfirm the actions to noise.

24
Model 1 Constant Confidence Level

• 2 investors and 4 days
• I (informed) those who receive the signal
• U (uninformed) those who do not receive the
  signal
• Day 0 endowment
• Day 1 I receives the signal and trades with U
• Day 2 Noisy public signal comes trade further
• Day 3 conclusive public info arrives.
• The risky security has a terminal value of ? the
  private information signal received by I at day 1
  is
• s1 ? e U correctly assesses the e but I
  underestimate it to be ?c2 lt ?e2 (key
  overreaction assumption)

25
Model 2 Outcome Dependent Confidence

• No longer require, initial overconfidence, ?c2 lt
  ?e2
• Assume public signal is discrete, with s2 1 or
  -1 at day 2.
• If sign (? e) sign (s2), confidence
  increases, so investors assessment of noise
  variance decreases to ?c2 k, 0 lt k lt ?c2
• If sign (? e) ? sign (s2), confidence remains
  constant ?c2
• Model 1 or Overconfidence implies negative
  long-lag autocorrelations, excess volatility,
  and, when managerial actions are correlated with
  stock mispricing, public-event-based return
  predictability.
• Model 2 or attribution implies short-lag
  autocorrelations (momentum), short-run earnings
  drift.

26
Evidence on the Characteristics of Cross
Sectional Variation in Stock Returns (Daniel and
Titman JF 1997) -- Liem

• Firm sizes and B/M ratios are both highly
  correlated with average returns of common stocks.
  DT find that return premia on small cap and high
  B/M does not arise because of the co-movements of
  these stocks with pervasive factors. It is the
  characteristics rather than the covariance
  structure of returns that appear to explain the
  cross-sectional variation in stock returns.
• Model 1 The Null Hypothesis
• Returns are generated by the following factor
  structure

27
Evidence on the Characteristics of Cross
Sectional Variation in Stock Returns (Daniel and
Titman JF 1997)

• Model 2 A Model with Time Varying Factor Risk
  Premia
• Factor loadings do not change as firms become
  distressed. A factors risk premium increases
  following a string of negative factor
  realizations.
• There is no separate distress factor fD. The
  remaining ßs in this model are constant over
  time.

28
Evidence on the Characteristics of Cross
Sectional Variation in Stock Returns (Daniel and
Titman JF 1997)

• Model 3 A Characteristic-based Pricing Model
• Firms exist that load on the distressed factors
  but which are not themselves distressed, and
  therefore have a low theta and commensurately
  low return.
• There is no separate distress factor fD. The
  remaining ßs in this model are constant over
  time.

29
What is the Intrinsic Value of the Dow? (Lee,
Myers and Swaminathan JF 2000) -- Liem

• They model the time-series relation between
  price and intrinsic value as a co-integrated
  system so that price and value are long-term
  convergent. They compare the performance of
  alternative estimates of intrinsic value for the
  Dow 30 stocks.
• Traditional market multiples such as B/P, E/P,
  and D/P ratios had little predictive power.
• However, a V/P ratio, where V is based on a
  residual income valuation model, has
  statistically reliable predictive power. Further
  analysis shows time-varying interest rates and
  analyst forecasts are important to the success of
  V. Alternative forecast horizons and risk premia
  are less important.

30
What is the Intrinsic Value of the Dow? (Lee,
Myers and Swaminathan JF 2000)

• The Residual-Income Valuation Model
• Returns are generated by the following factor
  structure
• Model Implementation Issues
• Forecast horizons and terminal values
• Cost of equity capital
• Explicit earnings forecasts
• Matching book value to I/B/E/S forecasts
• Dividend payout ratios

31
What is the Intrinsic Value of the Dow? (Lee,
Myers and Swaminathan JF 2000)

• Intrinsic Value Measures
• DJDP
• DJEP
• DJBM
• VP
• Tracking the Dow Index
• Without time trend (eq 10)
• With time trend (eq 11)

• Business Cycle Variables
• Default spread
• Term spread
• Returns prediction
• Forecast regression methodology (eq 12)
• Forecasting regression results
• Univariate regressions (eq 13)
• Multivariate regressions involving DJDP, DJEP,
  DJBM, and VP

32
Does the stock market overreact?Werner F. M. De
and Richrd Thaler Leon

• The paper tests that whether overreaction
  affect stock prices.
• overreaction is an implicit comparison to
  appropriate reaction, which tells us Bayes
  rule prescribes the correct reaction to new
  information.
• Individuals tend to overweight recent information
  and underweight prior information.
• Early researchs J. M. Keynes, Williams, Arrow,
  Shiller, Kleidons, Reinganum, Basu, Graham,
  Russell and Thaler.

33
The methodology

• Two hypothesis 1. extreme movements in stock
  will be followed by subsequent price movement in
  the opposite direction, 2. the more extreme the
  initial price movement , the greater will be the
  subsequent adjustment. ------ To test whether the
  overreaction hypothesis is predictive.
  and .
• is market-adjusted excess return
  , and if it is a efficient market, then
• Use , ,
  to test and find overreaction.

34
Findings

• loser and winner are both overreacting, and loser
  overreact more (asymmetric )
• most of the excess return realized in January
  (January effect)
• the overreaction mostly occurs during the second
  and third year of the test period.

35

• A model of investor sentiment
• N. Barberis, A. Shleifer, R. Vishny. JFE (1998)
  -- Daryl
• Earnings streams follow a random walk process
• Investors form expectations based on one of two
  non-random walk models mean-reverting or a
  trend.
• Investors exhibit representativeness, the
  tendency to view events as typical and ignore
  statistical probabilities.
• Investors make forecasts based on (i) the
  strength of evidence (ii) the statistical
  weight of evidence
• Model predicts that stocks
• Underreact to low strength evidence high weight
• Corporate announcements
• Overreact to high strength and low statistical
  weight of evidence.
• Consistent patterns of good or bad news

36

• Underreaction E(rt1ztG) gt E(rt1ztB)
• over-confidence about prior information
• Overreaction E(rt1ztG,, zt-jG) lt
  E(rt1ztB,,zt-jB)
• seeing order among chaos
• Model
• Investors believe earnings follow one of two
  regimes according to a specific regime switching
  process.
• Model 1 Mean Reverting or Model 2 Trend
  (Markov)
• Investor is convinced that he knows both pH pL

37

• Regime switching between models based on
  probability parameters,?i , which are assumed low.

• To value a security, an investor needs to
  forecast earnings.
• Investor task is then to understand which of the
  two regimes is currently governing earnings.
• At time t, after observing shock yt, investor
  estimates the probability qt that yt was generate
  by Model 1. Formally,
• qt Pr (st1yt, yt-1, qt-1) or
• qt1
• If earnings are generated by regime-switching
  process, then prices may be decomposed to a
  random-walk component and and a deviation
  component from fundamental value. (Prop 1)