to deflate the serie

[efes]

Hi all,
I am working on an arima model. Although i have almost completed the processes, i have a question from the begining steps. The variable is the monthly amount of silk import in terms of dollar. The serie is nonstationary. The variance is not stable and there is an increasing trend. First i log the serie to stabilize the variance. Then there was still a trend, so i take the first differences of log serie. At last it was stationary. I set the model and forecasted with arima. The forecast values seems ok for the moment.

I wrote the above explanation because thinking it's concerned to my below question. As i heard it is not needed to deflate the serie if we take the first difference of ln(y) (which i did the same). I mean working with the serie d(lny). Is that true or not. Should i deflate my serie with real exchange rate or anything else ?

I searched in forum but couldnt find what i really interest. I hope someone can help.

Thanks
B.R.

[trubador]

We had similar discussions in the forum before (please see Here and Here).

Deflating a series is one thing and making it stationary is something else. In other words, there is no guarantee that your series will become stationary after you deflate it. You may still need to make further transformations to achieve it. If the variables under your consideration are subject to significant price effects, then for the sake of the validity (and reliability) of your results you should better deflate your series either beforehand or inside the equation (EViews allows this type of normalization). For instance, if the macroeconomic environment is suffering from high inflation problem and/or the sample period is too long, then nominal values might blur the actual relationships.

Please keep in mind that you'll have to estimate future values of each component if you decide to decompose the imports variable (e.g. price, volume, exchange rate, etc.) from the forecasting point of view.