Investor sentiment and the cross-section ... Clea

1
Investor sentiment and the cross-section of stock
returns

• Malcolm Baker HBS
• Jeffrey Wurgler NYU Stern

2
Introduction

• Classical finance theory
• Investor sentiment doesnt affect prices,
  because the demands of any sentimental investors
  are neutralized by arbs
• Challenges to classical theory
• Clear violations of market efficiency (momentum,
  post-earnings announcement drift, index inclusion
  effects, negative stub values, etc.)
• This paper
• Theory and evidence that investor sentiment is
  real, measurable time-series phenomenon that and
  that it has pervasive cross-sectional effects

3
Theory

• What is investor sentiment? Does it affect
  different stocks in different ways?
• Observation
• Mispricings are invariably caused by two factors
• 1. An uninformed (e.g. sentimental) demand
  shock
• 2. A binding constraint on arbitrage
• Implication
• For a wave of sentiment to have cross-sectional
  effectsnot just cause equal mispricings across
  all stocksfactor 1, 2, or both, must vary across
  stocks

4
Cross-sectional variation in sentiment

• One potential definition of sentiment the
  marginal investors propensity to speculate
• Then sentiment is the relative demand for
  intrinsically speculative stocks, and thus causes
  cross-sectional effects even when arbitrage is
  equally difficult across stocks.
• What is an intrinsically speculative stock? A
  stock with a highly subjective/uncertain
  valuation
• Prediction stocks whose valuations are most
  subjective canonical young, unprofitable,
  extreme-growth potential stock, or a distressed
  stock will be especially sensitive to
  fluctuations in propensity to speculate

5
Cross-sectional variation in arbitrage

• Another potential definition of sentiment
  marginal investors (over-) optimism or
  (over-)pessimism about stocks in general.
• By this definition, indiscriminate waves of
  sentiment will still affect the cross-section to
  the extent that arbitrage forces are weaker in
  certain subsets of stocks.
• Arbitrage limits that vary across stocks
  fundamental risks, transaction costs/liquidity,
  short-selling costs, predatory trading risks,
  noise-trader risks, etc.
• Prediction time-varying optimism or pessimism
  has biggest effects on stocks that are hardest to
  arbitrage

6
Main hypothesis

• Observation Roughly speaking, the same stocks
  that are the hardest to arbitrage are also the
  most speculative /hardest to value
• Robust prediction Young, small, unprofitable,
  extreme-growth and distressed stocks are most
  sensitive to fluctuations in investor sentiment

7
Anecdotal history of investor sentiment,
1961-2002

• high sentiment period ? demand for speculative
  stocks
• low sentiment period ? demand for safety,
  quality stocks
• 1960-61 tronics small, growth stocks bubble
• 1967-69 small, growth stocks bubble
• early 1970s nifty fifty bubble
• late 1970s through mid-1980s small, sometimes
  industry-concentrated bubbles, e.g. biotech, oil
• late 1990s Internet bubble

8
Empirical approach

• Mispricing is hard to identify directly. Our
  approach is to look for systematic patterns of
  correction of mispricings.
• E.g., if returns on young and unprofitable firms
  are low when beginning-of-period sentiment is
  estimated to be high may represent the
  correction of a bubble in growth stocks. Ex post
  evidence of ex ante mispricing.

9
Measuring investor sentiment

• We consider six proxies the average discount on
  closed-end equity funds, NYSE share turnover, the
  number of and average first-day returns on IPOs,
  the equity share in new issues, and the dividend
  premium
• To smooth out noise, we also form a composite
  index based on their first principal component
• Sentiment proxies are annual, 1962 through 2001

10
Sentiment Index
11
Conditional predictability Size portfolios
12
Conditional predictability Volatility portfolios
13
Conditional predictability Sales growth
portfolios
14
Et cetera

• Patterns are not due to time-varying betas or
  plausible patterns of compensation for systematic
  risk
• (The EMH explanation would require that older,
  profitable, dividend-paying, and less-volatile
  firms are (when sentiment is high) actually
  require higher returns than younger,
  unprofitable, nonpaying, highly-volatile firms.
  Very counterintuitive.)

15
Conclusion

• Investor sentiment is a real, measurable
  phenomenon. It has large effects on the
  cross-section of stocks.
• Several novel findings emerge characteristics
  that have no unconditional predictive power have
  much power once one conditions on sentiment!
• Approach embraces, rather than ignores, evidence
  of bubbles and crashes