Finance when no one believes the textbooks Roy Batchelor Director, Cass EMBA Dubai Cass Business School, London
What to expect Your fat finance textbook A class test Inside investors heads Something about ants
The Modern Theory of Finance Benefits of diversification Beta: You don t get compensated for taking any old risk Alpha: Outperformance has to be risk-adjusted Efficient Markets Hypothesis: Impossible to outperform, as price changes depend only on unpredictable news Smart Investors will hold Cash + Index Fund only Proportion will depend on risk aversion
Crisis in the Modern Theory The Efficient Markets Hypothesis No excess rewards to active forecasting at the margin. Wrong beliefs will be punished by trading losses No speculative trades, since beliefs of representative agent(s) will converge on true data generating process Random walk with drift: long term trends follow fundamentals, short term price changes unpredictable. Conflict with Reality But market can only be made efficient by efforts of forecasters Noise trader phenomenon, many different beliefs can coexist High trading volume, uncorrelated with deviations of prices from fundamentals Excess volatility, with FX rates and stock prices persistently away from fundamentals for 3-5 years
The Smart Investor in the MTF Expected Utility. Stable attitude to risk, focussed on own future returns. Rational Expectations processes information as if in a statistically optimal way.
Utility and Wealth in the MTF Utility Wealth
Is that right? A mental experiment Suppose investors over each year can put 100m in either A safe 1-year bond which definitely pays 5% A risky stock market investment which might pay +30% or -10% Choices will depend on investor attitudes to risk Risk-averse investor A will buy the bond Risk-loving investor B might buy the stock market investment How will they feel at the end of the year?
Investor A Rational MTF model: Get utility of 105m = U(105), a fixed number no incentive to change behaviour next year But in reality: Utility depends on what happens in the stock market - if investment fell, U(105) is very high (Schadenfreude) -if risky investment rose, U(105) is not so high (Regret) Investor experiencing regret may take more risk next year
Investor B Rational MTF model: Get utility of 90m or 130m, U(90) or U(130) Possible Loss U(90)-U(100) may be equal to possible Gain U(130)-U(100) No incentive to change behaviour next year But in reality: two possible reactions Prospect Theory: if market falls, takes more risk to recoup losses if market rises, takes less risk to preserve gains House Money Theory: if market rises, investor takes more risk since is now playing with the house money
Prospect Theory and the Disposition Effect Evidence favours Prospect Theory over House Money theory Chicago Day Traders take more risks p.m. if they lost a.m. Hedge Funds lock in above-average first-reporting period returns Prospect theory explains actual investor behaviour The Disposition Effect : tendency to hang on to losers, due to increased risk-taking after losses tendency to sell winners too soon, decreased risk-taking after gains
Prospect Theory Value Function Utility Losses Gains
The Smart Investor in the MTF Expected Utility. Stable attitude to risk, focussed on own future returns. Rational Expectations processes information as if in a statistically optimal way.
Class test Exam conditions No open books No talking No copying No writing (until asked) Use the clickers... Which we will now test
An Actual Investor in the Markets Expected Utility. Stable attitude to risk, focussed on own future returns. Rational Expectations processes information as if in a statistically optimal way.
An Actual Investor in the Markets Inconsistent attitudes to risk and return. Bounded rationality: Biases in processing information Heuristics (rules of thumb) used to make decisions
Processing Information the Rational Model Information Acquisition Processing Feedback Forecast
How we actually process information Fog of Data System 1. Automatic Intuitive Judgmental System 2. Effortful Logical Untrained
Information processing biases Acquisition: Availability Processing: Poor intuitive statistical reasoning Anchoring and Adjustment Framing Biases Feedback: Illusion of Control, Need for narrative Overconfidence, Ego defences
Anchoring on chart lines
Conservatism in earnings forecast revisions
Framing - How to Lie with Statistics
The traders and the mouse Traders told to control a dot on the screen using keystrokes, so as to keep the dot in top right segment. Illusion of control: inversely related to profit contribution and remuneration, analytical skills directly related to people skills directly related to religiosity, superstition, belief in ESP, karma,...
Overconfidence and Ego Defence behaviours Natural bias to optimism Performance attribution: If I ve made money, it shows I have skill If I ve lost money, well that s just bad luck Hindsight bias. I knew it all along Self-rated probabilities of an event (e.g. sub-prime collapse) much higher after the event has occurred than before People often deny earlier opinions, rewrite history (Greenspan wrong not to burst asset price bubbles earlier )
But irrational investors can survive Suppose informed investors estimate fundamental value of Ford to be $20, but uninformed (noise, small) investors drive price down to $15. MTF says informed investors can drive price back to fundamentals by buying Ford, selling a similar company like GM (to remove market risk) But this is not riskless - uninformed traders may temporarily drive price even lower, and Ford and GM are not perfect substitutes Hence Irrational investors will not be eliminated from the market, and will coexist with Smart investors
Heterogeneous Agents the new paradigm Heterogeneous Agent Models Investors use two or more forecasting models e.g. regression to mean (fundamentals) extrapolation of trend (technical) Or anything else! Probabilistic switching between models as relative profitability changes Congruence with Reality Assumptions - many theories of exchange rates, share prices, and fundamentalists and chartists do coexist Simulated equilibrium prices display: deviations from fundamentals volatility clustering extreme sensitivity to shocks (nonlinear, chaotic process) Limited short term predictability
What happens the Ants model Two food sources. When ants meet, 1-δ = prob that one ant changes to source of the other, ε = probability of spontaneous change Random meetings lead to apparent regime switches periods when ants are persistently at one food source rather than another
Ants in the stock market Lux and Marchesi (1999) show switching between fundamentals and trend followers Volatility clearly increases with percentage of market following chartist rules, shows typical slow decay after shocks
HAMs and Chaos de Grauwe and Grimaldi (2005) prices away from fundamentals for long periods realistic values of β and γ lead to systems with deeply nonlinear (chaotic) characteristics extreme sensitivity to initial conditions
Simulations of a chaotic stock market Logistic Map.xls
New reasons for unpredictability Why can t we forecast in financial markets? It s all a random walk we have too many structural breaks and now... maybe it s deeply nonlinear Empirical evidence mixed: no sign of low order chaotic processes < dimension 6 But inference difficult when chaos is overlaid with noise Possible payoff to use of: low tech pattern recognition methods high-tech nonlinear modelling - eg neural networks
The Luck-Skill Continuum
Why?
Understanding the Markets, and Ourselves New theories, old predictions: still best not to try to pick winners = cash + index still unpredictable but possible high dimension Experts fund managers, talking heads - show no skill, but play to our need for narrative You probably won t believe this, because you are only human, you know someone who outperformed the market, you want your fund manager to be skilful and active