Is it a Must to work with deflated data?

[whitenoise]

or could we work with raw data if so would it suit to the theory of econometrics. For examp: I would like to estimate an Index model. Dependent variable was going to be Stock prices (for an exact corporation) and the ındependent variable was going to be oil prices.
Should I deflate those variable or what?

[trubador]

Deflating is necessary when you want to remove the effect of prices and work with real terms. Therefore, it is usually an important concept in economics. However, from the point of econometrics, what matters is whether your variables are stationary. You either choose an appropriate method to make your variables stationary or try another econometric technique to estimate the model at hand.

In your case, you are already using price data and therefore you should deal with the stationarity. For instance, you may use the first differences of level or log-level of the data before carrying out the estimation. If, however, you find out that the series are cointegrated, then you can build a vector error correction (VEC) without worrying about the stationarity problem.

[whitenoise]

Thank you for your explanation it is very satisfactory...

[riza]

i still don't get it. If i'm using the variable which has wide-range variance such as accounting earning (independet variable) in 17 company (panel data) which the value is ranging from million usd to billion usd, should i deflate the data ? cos the plot will look so scattered. how can i apply the deflation if i want to regress the accounting earning variable to abnormal return (dependent var) for example??
please help. thanks

[startz]

i still don't get it. If i'm using the variable which has wide-range variance such as accounting earning (independet variable) in 17 company (panel data) which the value is ranging from million usd to billion usd, should i deflate the data ? cos the plot will look so scattered. how can i apply the deflation if i want to regress the accounting earning variable to abnormal return (dependent var) for example??
please help. thanks

In general, "deflated" refers to adjusting for changes in the overall price level. Most of the time that's a good idea. In your case, if most of the variation is cross-sectional then deflating won't make any difference. You may want to look at using a weighted regression or using Huber-White robust standard errors to adjust for heteroskedasticity.

[riza]

did u mean by choosing this option?


so 'deflating' is similar to 'transforming' the data, isn't it? cmiiw
btw, what is the Cross-section SUR for?

[startz]

did u mean by choosing this option?


so 'deflating' is similar to 'transforming' the data, isn't it? cmiiw
btw, what is the Cross-section SUR for?

Yes, that could be a good approach.
Here's what the manual says

If you select Cross section weights, EViews will estimate a feasible GLS specification assuming the presence of cross-section heteroskedasticity. If you select Cross-section SUR, EViews estimates a feasible GLS specification correcting for both cross-section heteroskedasticity and contemporaneous correlation.

[EViews Gareth]

did u mean by choosing this option?

Interesting screenshot. An XP theme to look like a Mac, or running on a Mac virtualised?

[riza]

thanks for the infos. don't understand what the manuals says..lol. but dont worry i'll try to figure it out by myself from the econometrics books.
many thanks.

Interesting screenshot. An XP theme to look like a Mac, or running on a Mac virtualised?

.
it just an MacOS theme on windows xp :). otherwise, i had tried the crossovr mac (based-on darwine) running on leopard..but my eviews runs slower so i prefer to get back to windows.