## stationarity of CPI(price index)

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

Mansur
Posts: 1
Joined: Sun Mar 20, 2016 9:56 am

### stationarity of CPI(price index)

I am trying to determine whether CPI is DS or TS.On the graph it looks like the series is DS.However dfuller test in stata shows no evidence to support DS.I tried ADF using different number of lags.It seems that optimal number of lags is 2 since most of the lags after second one show no significance.With two added lags there is still strong evidence against DS(p=0.001).With 25 lags I get evidence in favour of DS,but there is no reason to add 25 lags,right?with every added lag p-value behaves in an unclear manner,it goes up and down. I also regressed it on time to see if it is TS: the coefficient on t is obviously significant, but residuals are not stationary.Could anyone suggest some help?

startz
Non-normality and collinearity are NOT problems!
Posts: 3399
Joined: Wed Sep 17, 2008 2:25 pm

### Re: stationarity of CPI(price index)

I don't know what data you are using or what Stata is doing, but in EViews using the full set of CPI data it is clear that there is a unit root.

rommellh
Posts: 1
Joined: Sun Apr 10, 2016 5:36 am

### Re: stationarity of CPI(price index)

If at levels a variable with an intercept alone is non-stationary but with intercept and trend it is stationary (I would think this suggests the variable is trend stationary), do you conclude stationarity and proceed modelling? Or must you first detrend the variable, run the unit root test once more to ensure stationarity and then proceed to model.

Also, I saw a video on Youtube that suggested that the results of all three ADF models (intercept alone, intercept and trend, and no intercept or trend) must agree before making a conclusion on stationarity. Is this true? I was not taught this.