12-FinEcon-Model Sel Page 326 Wednesday, February 4, 2004 12:59 PM
326 The Mathematics of Financial Modeling and Investment Management
pt = pt – 1 ++μεt
A time series of this form is called an arithmetic random walk. It is a
generalization of the simple random walk that was introduced in Chap-
ter 6. The arithmetic random walk is the simplest example of an inte-
grated process.
Let’s go back to simple net returns. From the above definition, it is
clear that we can write
με
1 +Rt =e
+ t
If the white noise is normally distributed, then the returns Rt are lognor-
mally distributed. Recall that we found a simple correspondence
between a geometric Brownian motion with drift and an arithmetic
Brownian motion with drift. In fact, using Itô’s Lemma, we found that,
if the process St follows a geometric Brownian motion with drift
dS
------ =μdt +σdB
S
its logarithm st = log St then follows the arithmetic Brownian motion
with drift:
1
ds =
μ–---σ
2
dt +σdB
2
In discrete time, there is no equivalent simple formula as we have to
integrate over a finite time step. If the logarithms of prices follow a discrete-
time arithmetic random walk with normal increments, the prices them-
selves follow a time series with lognormal multiplicative increments
written as
+ t
Pt =( 1 +Rt)Pt – 1 =e
με
Pt – 1
The arithmetic random walk model of log price processes is sug-
gested by theoretical considerations of market efficiency. As we have seen
in Chapter 3, it was Bachelier who first suggested Brownian motion as a
model of stock prices. Recall that the Brownian motion is the continu-
ous-time version of the random walk. Fama and Samuelson formally