International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622

Vol.4 (8), AUGUST (2015), pp. 7-13

Online available at indianresearchjournals.com

MODELING LONG RANGE DEPENDENCE USING FRACTIONAL

BROWN MOTION IN BSE SENSEX

DHANU P*; ASHOK KUMAR M**

*PHD (FT) RESEARCH SCHOLAR IN MANAGEMENT

DEPARTMENT OF MANAGEMENT STUDIES AND RESEARCH
KARPAGAM UNIVERSITY
COIMBATORE, INDIA.

**RESEARCH SUPERVISOR
PROF AND HEAD DEPT OF MANAGEMENT STUDIES AND RESEARCH
KARPAGAM UNIVERSITY
COIMBATORE, INDIA.

ABSTRACT

Brownian motion which is used to model stock price returns has the assumption that price
changes are independent and identically distributed (IID), since many studies have shown that
they are dependent. Fractional Brownian motion is a modification of Brownian motion with a
parameter H (Hurst exponent). H determines the autocorrelation of time series. If H=1/2 it is a
Brownian motion, H>1/2 persistent and H<1/2 anti persistent. This study shows that markets are
not efficient.

KEY WORDS - Fractional Brownian motion, Long range dependence, Joseph effect, Hurst
exponent, self similarity.

Introduction
Brownian motion is used to model stock price changes. It assumes that the underlying
distribution is independent and identically distributed. Efficient market hypothesis predicts that
stock market returns are normally distributed. Many researchers have pointed out that the
fluctuations are not independent but there exists trends in the stock price changes called Joseph
effect. The name comes from biblical story of Joseph which says about seven years of plenty and
seven years of famine. Such trends are present in stock market and also in many natural
phenomenon. Fractional Brownian motion is introduced by Benoit Mandelbrot and Van Ness
(1965) to model such events.

Fractional Brownian motion

Pioneering research is done by H E Hurst who was studying about Nile hydrology; he was
studying the yearly variation in the levels of river Nile. He used rescale range (R/S) statistics
where R is range of partial sum of data and S is the sample standard deviation. Under the general
assumption the finite variance and independent and identically distributed observations the R/S
7
International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622
Vol.4 (8), AUGUST (2015), pp. 7-13
Online available at indianresearchjournals.com

statistics should grow like n1/2 where n is the sample size however the data indicated a growth
rate of nH. The Brownian motion has standard deviation at time t is t1/2 and that of fractional
Brownian motion is tH where H ≠ 1/2

The term is fractional because this process can be represented as an integral with respect to
Brownian motion B(t).
0 1 1 𝑡 1
𝐻− 𝐻− 𝐻−
𝐵𝐻 𝑡 = 𝑡−𝑠 2 − −𝑠 2 𝑑𝐵 𝑠 + 𝑡−𝑠 2 𝑑𝐵(𝑠)
−∞ 0

∞ 1 1
= 𝑡 − 𝑠 +𝐻−2 − −𝑠 +𝐻−2 𝑑𝐵(𝑠)
−∞

The fractional Brownian motion (FBM) represented as BH(t) is the random process with
Gaussian increments that satisfies the following diffusion rule

for all t and T, E[BH(t + T) - BH(t)] = 0 and

E[BH(t + T) - BH(t)] = T2H

The value H = 1/2 yields the Wiener Brownian motion, whose diffusion is called "Fickian."
However, the exponent H is only constrained to 0< H < 1. For H =1/2, the diffusion of FBM is
widely called non Fickian." In a different terminology, a mysterious but widely used one, H is
called "strength of singularity" at time t. This process was introduced in Mandelbrot and Van
Ness 1965 and fully described
If H=1/2 then BH(t) is similar to classical Brownian motion denoted by B(t).
If H>1/2 then BH(t) is persistent means that is the stock price increase today then the next day it
has tendency to increase and vice versa.
If H<1/2 then BH(t) is anti persistent means if the stock price increase today then the price tend
to decrease tomorrow and vice versa.

BH(t) and the phenomenon of long-run statistical dependence. The most striking single
property of Bu(t) concerns the quantities BH 0 − BH (−T) /T, called past average and BH T −
BH(0)/T, called future average. Both are Gaussian random variables, and their correlation is
easily seen to be
1 2T 2H − T 2H − T 2H
C= = 22H−1 − 1
2 (T H )2

Another important property of FBM is Self Affinity. The random function X(t,ω) defined for -
−∞ < t > ∞ are said to be self affine with exponent H ≥0 if for any h >0 and for any t0
𝑋 𝑡0 + 𝜏, 𝜔 − 𝑋 𝑡0 , 𝜔 ≜ {ℎ−𝐻 𝑋 𝑡0 + ℎ𝜏, 𝜔 − 𝑋 𝑡0 , 𝜔 }
Self affinity is the phenomenon where the scaling is at different rate at different directions. In
finance the scaling happens faster horizontally than vertically.

8
International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622
Vol.4 (8), AUGUST (2015), pp. 7-13
Online available at indianresearchjournals.com

H is called Hurst exponent or Hӧlder exponent that is independently developed by H E Hurst

and L O Holder. Hurst who worked as Hydrologist for building a dam across Nile developed H
for modeling Nile water discharge. L O Hӧlder developed the same in the field of pure
mathematics. H is calculated using the following algorithm.
For a (partial) time series of length , , the rescaled range is
calculated as follows:
1. Calculate the mean;

2. Create a mean-adjusted series;

3. Calculate the cumulative deviate series ;

4. Compute the range ;

5. Compute the standard deviation ;

6. Calculate the R/S (rescaled range) and average over all the partial time series
of length The Hurst exponent is estimated by fitting the power law to the data.

This can be done by plotting the logarithm of rescale range as a function of log(n), and
fitting a straight line; the slope of the line gives H.

FBM is simulated by Matlab program and used different H parameter

9
International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622
Vol.4 (8), AUGUST (2015), pp. 7-13
Online available at indianresearchjournals.com

Fig 1. For H=0.2 anti persistent series

Fig 2. Graph for H=0.5 Brown Motion

10
International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622
Vol.4 (8), AUGUST (2015), pp. 7-13
Online available at indianresearchjournals.com

Fig .3Graph of H=0.7 Persistent series.

Fig 4.H.0.615 (Calculated value of BSE sensex)

Fig 5. H=0.9

11
International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622
Vol.4 (8), AUGUST (2015), pp. 7-13
Online available at indianresearchjournals.com

Interpretation
The figures show that when H ranges from 0.2 to 0.9 the graph gets more and more smooth. The
roughness is measured by fractal dimension D =2-H so we can see that as the Hurst exponent
value increases fractal dimension decreases. When H>0.5 (persistent series) then an increase is
followed by an increase and vice versa so it does not make it too rough. But when H<0.5 (
antipersistent series) an increase is followed by a decrease and vice versa so the graph is more
rough.

The calculated value of H=0.615 that means that price changes are persistent. If BSE sensex
increases today then there is more chance that it will increase tomorrow. And if the price drops
today then there is more chance of drop tomorrow. This is a marked difference from independent
and identically distributed (IID) as predicted by efficient market hypothesis. The price changes
are influenced by human behavior if there is an increase investors are becoming more optimistic
and they tend to buy more stock and when the stock price drops they become pessimistic and
they sell so again drop. It means that markets are not efficient. In technical analysis when the
chart breaks the highs high then it starts the bull run and when the lows low is breached it enters
a bear run. This empirical observation is the evidence for the existence of long term dependency.

Conclusion
The fractional brown motion is a modified form the traditional Brownian motion. If the data is
not independent and if they are correlated as in the case of stock returns, then we can use
fractional Brownian motion to model the stock returns. Fractional Brownian motion has the
ability to include non cyclic trends like the persistence and antipersistence that are observed in
many natural phenomenons such as hydrology and signal processing. As the parameter H=0.615
shows that the price changes are persistent. Science doesn’t take us from wrong to right but from
right to more right fractal Brownian motion is more right than Brownian motion.

Scope of further research

FBM doesn’t incorporate the large discontinuous price changes in stock market that is called as
Nora effect. Normal distribution is not capable of incorporating theses large fluctuations. Levi
stable distribution which is a general form of distribution, by adjusting its parameter we can
generate distribution ranging from Cauchy to Gaussian. Further research has to be done to
combine long range dependency and discontinuity that are observed in stock market that means
combining FBM and Levi stable distribution.

REFERENCE
1. L. Bachelier - The Theory Of Speculation 1900 -Translated By D. May From Annales
Scientifiques De I´École Normale Supérieure, Sér. 3, 17 (1900), P. 21-86.
2. Eugene F. Fama – Behaviour Of Stock Market Prices - The Journal Of Business, Vol. 38,
No. 1. (Jan., 1965), Pp. 34-10
3. Burton G. Malkiel - The Eficient Market Hypothesis And Its Critics -Journal Of
Economic Perspectives—Volume 17, Number 1—Winter 2003—Pages 59 – 82
4. Benoid B Mandelbort - Fractals Scaling In Finance Discontinuity Concentration And
Risk (1997) Springer.
5. Benoid B Mandelbort & J. Van Ness, Fractional Brownian Motions, Fractional Noises
And Applications, Siam Review, 10, 1968, 422-437

12
International Journal Of Marketing, Financial Services & Management Research ____________________ ISSN 2277-3622
Vol.4 (8), AUGUST (2015), pp. 7-13
Online available at indianresearchjournals.com

6. Benoid B Mandelbort, The Fractal Geometry Of Nature (New York: W. H. Freeman,

1982).
7. Benoit B Mandelbrot- Scaling In Financial Prices: Ii. Multifractals And The Star
Equation- Quantitative Finance Volume 1 (2001) 124–130
8. Benoit B Mandelbrot- Scaling In Financial Prices: Iv. Multifractal Concentration (2001) -
Quantitative Finance Volume 1 (2001) 641–649
9. Murad S Taqqu – Benoid B Mandelbrot And Fractional Brownian Motion (2013)
Statistical Science 2013, Vol. 28, No.1 1, 131-134.
10. Brent Oksendal- Fractional Brownian Motion In Finance (2004) ISBN 82–553–1398–2
No 28, ISSN 0806–2439 June 2003
11. Rossitsa Yalamova; Liu Qi, Yudong Chen; Chenwei Wang, Lin Wang; And Qiao Wang
-Power Laws In Financial Time Series (2007)
12. Dhanu P And M Ashok Kumar-Hurst Exponent Of Bse Sensex (2014)-International
Journal Of Advanced Research In Computer Science And Management Studies.
13. Dhanu P And M Ashok Kumar – Modeling Heavy Tails In Bse Sensex(2015)-
International Journal Of Advanced Research In Computer Science And Management
Studies.
14. Bo Qian Khaled Rasheed- Hurst Exponent And Financial Market Predictability
15. H.E. Hurst, Long-Term Storage Of Reservoirs: An Experimental Study, Transactions Of
The American Society Of Civil Engineers, 116, 1951, 770-79
16. Marcus Davidsson-Serial Dependence And Rescaled Range Analysis(2011)-
International Research Journal Of Finance And Economics- Issn 1450-2887 Issue 64
17. Subrata Kumar Mitra- Trends In Stock Prices And Range To Standard Deviation
Ratio(2011)- International Journal Of Business And Management Vol. 6, No. 1; January
2011
18. Swetadri Samadder, Koushik Ghosh And Tapasendra Basu- Phase Wise Scaling And
Trend Pattern Analysis Of Prime Indian Stock Market Indices During Last Decade -
Universal Journal Of Marketing And Business Research Vol. 1(2) Pp. 044-055, June,
2012
19. José A.O. Matos Sílvio M.A, Gama Heather J. Ruskin, Adel Al Sharkasi, Martin Crane
D-Time And Scale Hurst Exponent Analysis For Financial Markets

13