# Forecasting Financial Markets Chaos Theory

Forecasting Financial Markets – Chaos Theory

Slide: 1

Overview
Fractals
Self-similarity

Fractal Market Hypothesis Long Term Memory Processes
Rescale Range Analysis Biased Random walk Hurst Exponent

Cycles Phase Space Chaos & the Capital Markets
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 2

The Chaos Game
Plot point P at random Roll dice Proceed halfway from P to point labeled with rolled number & plot new point Repeat 10,000 times A (1, 2)

P

C (5, 6)

B (3, 4)
Forecasting Financial Markets – Chaos Theory Slide: 3

The Sierpinksi Triangle

Forecasting Financial Markets – Chaos Theory

Slide: 4

Chaos
Local randomness, global determinism
Apparently random process may contain deterministic pattern

Stable, self-similar structure Sierpinksi Triangle
Plot order impossible to predict But odds of plotting each point are not equal • Empty spaces in each triangle have zero probability • Local randomness does not equate to independence
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 5

Characteristics of Fractals
Self-similarity
The part is similar to the whole • Precise similarity in case of Sierpinski triangle

Scale Invariance
Sub-parts not to same scale as parent

Dimension
Euclidean space features integer dimensions Fractals occupy fractional dimension • E.g dimension of Sierpinski triangle is more than a
line but less than a plane (1 < d < 2)
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 6

Fractal Time Series
Dimension measures how “jagged” series is
Straight line has fractal dimension of 1 Random time series has fractal dimension of 1.5 • 50% chance of rising or falling A line can have fractal dimension between 1 and 2 At values ≠ 1.5 series is less or more jagged than a random series • Non-Gaussian

Forecasting Financial Markets – Chaos Theory

Slide: 7

Non-Gaussian Properties of Financial Markets
Distribution of Returns
Higher peak at mean than Normal Fatter tails • Uniformly fatter
– As many observations at 4σ away from mean as at 2σ

Markets tend to stay still or make major moves more often than theory predicts • Reflected in option volatility smiles

Term Structure of Volatility
Scales at faster rate than T1/2
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 8

Example: Returns on the DJIA
Dow Jones Industrials Returns
0.60 0.50

0.40

0.30

Normal 30 Day Returns
0.20

0.10

0.00 -4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Standard Deviations

Forecasting Financial Markets – Chaos Theory

Slide: 9

Volatility Term Structure
DJIA Volatility Term structure 1888-1900
0.0 0.0 -0.5 1000 Days 1.0 2.0 3.0 4.0 5.0

Log(SD)

-1.0 -1.5 -2.0 -2.5

Actual Theoretical

Log(Days)

Forecasting Financial Markets – Chaos Theory

Slide: 10

Regression Analysis of Term Structure
Chart indicates clear breakdown in volatility term structure after n = 1,000 days Regression analysis confirms this:
Days < Days > 1,000 1,000 Intercept -1.95 -1.38 Slope 0.53 0.31 2 R 99.3% 47.2% F 1823.0 5.4 p 0.00% 5.99%
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 11

Conclusion Re term Structure
Risk-Return
Riskier to invest for period < 4 years Increasingly less risk incurred beyond 4 years Tied to business cycle?

Sharpe Ratio
Gets larger for longer time horizons

Forecasting Financial Markets – Chaos Theory

Slide: 12

Sharpe Ratio & Time Horizon
DJIA Sharpe Ratio and Time Horizon
5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0 1000 2000

Sharpe Ratio

y = 9E-08x 2 - 0.0001x + 1.1433 R2 = 0.7465

3000

4000

5000

6000

7000

Days

Forecasting Financial Markets – Chaos Theory

Slide: 13

Fractal Theory of Markets
Stable markets
Investors with many investment horizons • Ensures liquidity

Information set depends on investment horizon
Short term: market sentiment & technical factors Long term: fundamental analysis

Unstable markets

Prices set by combination of ST & LT valuation
ST trends are noise. LT trends tied to economic cycles

Forecasting Financial Markets – Chaos Theory

Slide: 14

Rescaled Range Analysis
Developed by H.E. Hurst 1950’s Brownian Motion
Distance traveled R ∝ T0.5

Hurst Exponent
(R/S)T = cTH • H is the Hurst Exponent • c is a constant • T is # observations • (R/S)T is the rescaled range, a standardized measure

of distance traveled Note for random time series H = 0.5
Forecasting Financial Markets – Chaos Theory Slide: 15

Hurst Exponent & Market Behavior
H measures persistence Correlation C = 2(2H-1) - 1 White Noise: H = 0.5, C = 0 Black Noise: 0.5 < H < 1 , 0 < C < 1
Persistent, trend reinforcing series “Long memory”

Pink Noise: 0 < H < 0.5, C < 0
Antipersistent, mean-reverting Choppier, more volatile than random series
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 16

White Noise Process
Fractal Random Walk
20 0

-20

-40

-60

H = 0.5

-80

-100

-120

-140

Forecasting Financial Markets – Chaos Theory

Slide: 17

Black Noise Process
Fractal Random Walk
100

0

-100

-200

-300

-400

H = 0.9 • Smoother series • Trend

-500

-600

Forecasting Financial Markets – Chaos Theory

Slide: 18

Pink Noise Process
Fractal Random Walk
15 10 5 0 -5 -10 -15 -20 -25 -30 -35

H = 0.1 • More volatile • Antipersistent
– Mean reverting
Forecasting Financial Markets – Chaos Theory Slide: 19

Simulating a Fractal Random Walk
Feder (1988):
n(M−1) ⎛ n−H ⎞ ⎧ nt (H−0.5) ⎫ (H−0.5) (H−0.5) ∆yH (t) = ⎜ −i E(1+n(M+1)−i) + ∑ (n+i) E(1+n(M−1+t)−i) ⎬ ⎜ Γ(H +0.5) ⎟×⎨∑i ⎟ i=1 ⎭ ⎝ ⎠ ⎩i=1

[

]

• Ei is a strict white noise process, No(0, 1) • M is the number of periods for which long memory • • •
is generated n is set to 5 t is set to 1 H is Hurst exponent
Forecasting Financial Markets – Chaos Theory Slide: 20

Calculating (R/S)
Form series of returns
rt = Ln(Pt / Pt -1) for t = 1, 2, . . . , T

Divide into A contiguous sub-periods Compute average for each sub-period ra = ∑ rak k =1 Form cumulative series
Length n, such that An = T
n

X ka = ∑ (ria − ra )
i =1

k

Define range Ra = Max(Xk,a) - Min(Xk,a)
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 21

Calculating (R/S)
Compute standard deviation for each subperiod
⎡1 n S a = ⎢ ∑ (rka − ra ) ⎢ n k =1 ⎣
2 1/ 2

⎤ ⎥ ⎥ ⎦

Calculate average R/S for each n
1 A ( R / S ) n = ∑ ( Ra / S a ) A a =1

Use OLS Regression to Estimate H
Ln(R/S)n = Ln(c) + H Ln(n)
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 22

Example: RS Worksheet
Generate random sequence using RAND() fn.
Periods of length n = 10, 20, . . ., 100

Calculate mean and SD for each sub-period Form cumulative series Calculate R, R/S and Ln(R/S) for each sub-period Repeat 10 times Plot Ln(n) against average Ln(R/S) Fit linear trend
OLS slope estimate = H

Forecasting Financial Markets – Chaos Theory

Slide: 23

Example: RS Analysis
Rescaled Range Analysis 3.0 2.5 2.0
Ln(R/S)

1.5 1.0 0.5 0.0 -1.0 -0.5 0.0 1.0 2.0
Ln(N) y = 0.507x + 0.0572 R2 = 0.9619

3.0

4.0

5.0

Forecasting Financial Markets – Chaos Theory

Slide: 24

Hurst Exponent for Random Series 1,000 Simulations
Simulated Values of the Hurst Exponenent
350 300 250 200 150 100 50 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Forecasting Financial Markets – Chaos Theory

Slide: 25

Testing (R/S): E(R/Sn)
Anis & Lloyd (1976)

Γ[0.5(n − 1)] n −1 n − i E (R / Sn ) = ∑ i π Γ(0.5n) i =1
For large n (>350) Use Stirling Function
⎛ n − 0.5 ⎞⎛ nπ ⎞ E(R / Sn ) = ⎜ ⎟⎜ ⎟ ⎝ n ⎠⎝ 2 ⎠
−0.5 n −1 i =1

n−i i

Forecasting Financial Markets – Chaos Theory

Slide: 26

Testing (R/S)
E(H) = Ln(E[R/Sn]) / Ln(n) Var(H) = 1 / T
If underlying process is random Gaussian, H will be Normally distributed

V-Statistic
Recall (R / Sn) = cnH V-Statistic • Divide by √n • V(n) = (R / Sn) / √n = cn(H-0.5)
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 27

V-Statistic
V-Statistic V(n) = (R / Sn) / √n = cn(H-0.5) For Persistent Series H > 0.5
V(n) is increasing fn. of n

For Random Process H = 0.5
V(n) is constant

For Antipersistent process H < 0.5
V(n) is declining fn. of n

Forecasting Financial Markets – Chaos Theory

Slide: 28

V Statistic for E(R/Sn)
V Statistic for E(R/Sn )
1.3 1.3 1.2 1.2 Vn = E(R/Sn) / n0.5 1.1 1.1 1.0 1.0 0.9 0.9 0.8 2.0 3.0 4.0 5.0 Log(n) 6.0 7.0 8.0

Forecasting Financial Markets – Chaos Theory

Slide: 29

V-Statistic for Random Series
V-Statistic
2.0 1.5 1.0 0.5 0.0 2.0 3.0 4.0 5.0

Forecasting Financial Markets – Chaos Theory

Slide: 30

Example: Long memory Process in DJIA (20-day Returns)
R/S Analysis DJIA (20 Day Returns)
1.8 1.6 1.4 1.2 Log(R/S) 1 y = 0.6119x - 0.148 R = 0.9898
2

0.8 0.6 0.4 0.2 0 0.0 0.5 1.0 1.5 Log(n) 2.0 2.5 3.0 Dow E(R/S) Linear (Dow)

Forecasting Financial Markets – Chaos Theory

Slide: 31

Example: Long memory Process in DJIA (20-day Returns)
V-Statistic
1.40

n=52
1.30 1.20 1.10

V
1.00 0.90 Dow 0.80 0.70 0 100 200 300 E(R/S)

n

400

500

600

700

Forecasting Financial Markets – Chaos Theory

Slide: 32

Dow R/S Regression Analysis
Estimated Hurst Exponent
H = 0.62 • OLS estimate of regression slop coefficient Indicates fractal persistent memory process

10 < n < 52
H = 0.71, R2 = 99.9% • Highly persistent Hurst process

52 < n < 650
H = 0.49, R2 = 99.8% • White noise process
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 33

Dow: Conclusions
R/S analysis indicates long memory process Average cycle length approx 4 years
Tied to economic cycle Events occurring today affected by events up to 4 years ago

Long memory effects dissipated after 4 years

Forecasting Financial Markets – Chaos Theory

Slide: 34

R/S Analysis of Stocks
Hurst Exponent 0.72 0.73 0.75 0.70 0.65 0.68 Cycle (months) 18 18 18 42 42 90
Slide: 35

IBM Xerox Apple Coca-Cola McDonald’s Con Edison

Forecasting Financial Markets – Chaos Theory

Innovative, high growth firms
Have high H and short cycles

Stable, low growth firms
Have low H and long cycles

Implications for risk
High H firms are less risky • Less noise in series • Contradicts standard theory What about diversification? • Dow index has one of the highest H exponents

Forecasting Financial Markets – Chaos Theory

Slide: 36

Lab: R/S Analysis of S&P 500 Index
Monthly returns April 75 to Feb 99
Detrended to remove short term memory effects

Calculate
RS, E(R/S), v-statistic (actual and expected)

Plot
Ln(n) vs. Ln(R/S) N vs. V-statistic

Estimate
Cycle length Hurst exponents pre and post cycle

Forecasting Financial Markets – Chaos Theory

Slide: 37

Lab: R/S Analysis of S&P 500 Index
Monthly S&P500 Index Returns Apr 75 - Feb 99
15.00% 10.00% 5.00% 0.00%

Apr-75

Apr-77

Apr-79

Apr-81

Apr-83

Apr-85

Apr-87

Apr-89

Apr-91

Apr-93

Apr-95

-5.00% -10.00% -15.00% -20.00% -25.00%

Forecasting Financial Markets – Chaos Theory

Apr-97

Slide: 38

Solution: R/S Analysis of S&P 500 Index - Ln(R/S) Plot
R/S Analysis S&P500 Monthly Returns

2.8 2.6 2.4 2.2
Ln(R/S)

2.0 1.8 1.6 1.4 1.2 1.0 2.0 3.0
Ln(n)

y = 0.5153x - 0.0637 R2 = 0.9692 SP500 E(R/S) Linear (SP500)

4.0

5.0
Slide: 39

Forecasting Financial Markets – Chaos Theory

Solution: R/S Analysis of S&P 500 Index - V-Statistic
V-Statistic
1.25 1.15

1.05

V

0.95

0.85

0.75

SP500 E(R/S)

0.65 0 20 40 60

n

80

100

120

140

Forecasting Financial Markets – Chaos Theory

Slide: 40

SP500 Regression Analysis

Forecasting Financial Markets – Chaos Theory

Slide: 41

Economic Indicators
Industrial Production Index
H = 0.79 • Based on monthly data, Jan 1946 - Jan 1999 • Strongly persistent Cycle 42 months • Shorter than 4-year cycle accepted by economists Ties in with S&P 500 Index

Forecasting Financial Markets – Chaos Theory

Slide: 42

R/S Analysis: Currencies
True Hurst process: no cycle length
R/S continues scaling at rate H indefinitely with n Infinite memory process Not tied to economic cycles No “fundamental” valuation of currency Less persistent, more volatile than stocks

H Exponents (based on daily data)
• Yen: H = 0.64 • GBP: H = 0.63 • DM: H = 0.62
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 43

R/S Analysis: Other Financial Markets
Treasury Bonds
H = 0.68 • Based on daily yields, Jan 1950-Dec 1989 Cycle length 5 years

Gold
Some evidence of 4-year cycle H = 0.58, but not significant

Volatility
A true pink noise antipersistent process H = 0.31 for S&P 500 Index vol. (realized)
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 44

Chaos Theory & Financial Markets Summary
R/S Analysis confirms aperiodic cycles in many series
Stocks, stock indices, bonds, economic indicators Cycle length related to economic cycle

Currencies are true Hurst process
Scale indefinitely

Volatility is only known antipersistent financial time series (apart from Wheat futures!)
Mean-reverting?
Copyright © 1999 -2006 Investment Analytics Forecasting Financial Markets – Chaos Theory Slide: 45