Test random Walk

[diego45]

Hi everybody

I would like to test that stock market price follow a random walk.

So i did stochastic calculus and i saw that a random walk with a drift could be written like that :

Yt = Y0 + b.dt + o.dWt ...

Where b.dt is the drift
and dWt is wienner or brownien process

So in eviews if i generate x = nrnd

and i regress

DYt = c + x

And that coefficient aren't significant and my residual are white noise is it a random walk or i did a mystake in the procedure ?

thanks

[startz]

You probably want to learn about testing for unit roots.

[diego45]

no i have done the dick fuhler test and their is a unit root

Sont i cant regress Pt = Pt-1 + ut

due to the presence of this unit root ?

[startz]

Your test shows that stock prices have a unit root. That's not quite the same as saying stock prices follow a random walk, but it's close.
Maybe you could tell us more about what you're trying to do.

[diego45]

Thanks for answer.

I want to test if use technical analysis to forecast price is a good way.

In fact ii i find that their is a random walk i'm sure that technical analysis bring nothing the analyst.
If i found that their is not random walk and if i find that their is serial-correlation between past and day price that i could use technical analysis.
I have found a study that show their is correlation between past and day price but using 1 price per each quarter during 22 years .

[startz]

You might define the left hand side variable as the return, dlog(p), and see if you can find varibles that help predict that.