Download presentation
Presentation is loading. Please wait.
Published byDana Simpson Modified over 9 years ago
1
A model of investor sentiment Barberis, Shleifer & Vishny Journal of Financial Economics,1998 Cedric Foucart Dries Heyman
2
Anomalies Underreaction to news: E(r t+1 |z t =G)> E(r t+1 |z t =B) Empirical evidence: Cutler et al (1991): positive autocorrelation in excess returns over 1- 12 months period Bernard (1992): past earnings announcement return predicts future (60 days) return Jegadeesh and Titman (1993) Rouwenhorst (1997)
3
Anomalies (2) Overreaction to a series of news: E(r t+1 |z t =G, z t-1 =G, z t-j =G)< E(r t+1 |z t =B, z t-1 =B, z t-j =B) (j≥1) Empirical evidence: Cutler et al (1991): slightly negative autocorrelation over 3-5 year period De Bondt and Thaler (1985):Losers outperform winners Zarowin (1989)
4
Modelling over- and underreaction Representative, risk neutral investor Earnings follow random walk Investor beliefs: two regime world Regime 1: mean reverting Regime 2: trending Regimes capture two psychological phenomena: conservatism and representativeness
5
Conservatism and representativeness Individuals change beliefs too slow → leads to underreaction Mean reverting regime = conservatism People think they see patterns in truly random sequences → Leads to overreaction Trending regime = representativene ss
6
Regime switching model Underlying switching process follows a Markov process Switches are rare Most likely to be in model 1 (mean reverting) Bloomfield, Hales (2002): Experimental evidence supports regime-shifting beliefs
7
Earnings forecast Determine model that governs earnings Probability of being in model 1(q t ) depends on → previous probability of being in model 1 → new earnings observation → transition probabilities Earnings reversals increase q t Same earnings decrease q t
8
Pricing If random walk: P t =N t /δ If regime switching model: P t =N t /δ + y t (p 1 – p 2 q t ) → Mispricing: y t (p 1 – p 2 q t ) <0: underreaction >0: overreaction Both underreaction and overreaction are possible (depending on parameters)
9
Simulation Earnings generation: random walk 2000 firms over six years 2 portfolios: positive and negative realisations N= 1 to 4 Compute returns in year after formation Difference between returns Results confirm expectations
10
Conclusion Model of investor sentiment Explain both over- and underreaction Makes use of conservatism and representativeness Simulation results confirm expectations
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.