First difference

[jakeace]

Dear all,
I am trying to run an OLS regression with time series data. My data turns out to be non-stationary at level; however it is stationary at the first difference (ADF). My issue is that when I run the regression with first difference variables the effect they have becomes insignificant. Whereas, if I run the regression normally, the variables have the expected impact and are significant. Is this normal? Or is there anything I can do to resolve this issue?
Thanks in advance.

[startz]

This could be exactly the Newbold-Granger point that two series with unit roots appear to be spuriusly correlated when they aren't. Odds are there's really nothing there.

[obicna89]

Dear all,
I am trying to run an OLS regression with time series data. My data turns out to be non-stationary at level; however it is stationary at the first difference (ADF). My issue is that when I run the regression with first difference variables the effect they have becomes insignificant. Whereas, if I run the regression normally, the variables have the expected impact and are significant. Is this normal? Or is there anything I can do to resolve this issue?
Thanks in advance.

it's probably spurious regression - due to the common trend.
run a johansen test procedure in order to see if they are cointegrated or not- and the differences will be used in the VEC representation of a VAR model if there is a cointegration

[nabilah]

Dear all
i have done ADF test and my variables are stationary in the first difference but some are stationary at 1%, some at 5% and some at 10%..they are stationary in the first difference but not at same percentage..but also a little further problem, at the first difference, there are 2 variables which are not stationary at all. is it possible to conduct the johansen test? help asap please
thanks

[Basyvava]

Dear all
i have done ADF test and my variables are stationary in the first difference but some are stationary at 1%, some at 5% and some at 10%..they are stationary in the first difference but not at same percentage..but also a little further problem, at the first difference, there are 2 variables which are not stationary at all. is it possible to conduct the johansen test? help asap please
thanks

Hum it's not possible. It looks like your stationary series are either I(1) or probably I(2) as you haven't found out the integration order of the 2 remaining series. You've got to run the Johansen cointegration test on variables that are stationary and integrated of the same order, otherwise there is no cointegration risk at all.

[khadija]

Dear all
I need a clarification. If I used log data to two variables ‘x’ and 'y'. And I found both variables are stationary at the first difference. So for analysing Johansen's test, Engle- Granger test, VECM test, which data should I use? Is it first difference data or log data?

thanks

[Shuyilim0619]

Obj: to study the impact of (money supply, interest rate) on exchange rate volatility and stock return volatility.
The result of cointegration test state that all null hypothesis is not rejected. Then what is next step for me to carry out the analysis.. Can tel the way to do the analysis of var in first difference?