For instance, if one is running a regression with deflated prices for a given year and one of the independent variables is a price index for a given sector, does it make any sense to 'deflate' this variable or somehow extract the content that might be related to the general economy (assuming I am deflating with a more agregated variable, like a consumer price index)? One might think of demand estimation or supply estimation, if that helps.

My first thought was that I shouldn't bother, but now I'm thinking of running an auxiliary regression with the agreggated price index and the sectoral price index as a dependent variable and save the residuals -- does that make sense?

Edit: Added some more intuition, as follows:

Think of an equation of a residual demand such as $q_t$$=$$\beta p_t$+$\epsilon_t$, where $q_t$ are the quantities, $p_t$ is the price and $\epsilon_t$ is an estochastic error. But since $Cov(p_t,\epsilon_t)≠0$ -- the price $p_i$ is inherently endogenous in a demand function because of the existence of another equation defining price and quantities, the supply equation --, one should estimate demand adequately through instrumental variables estimation. Eligible instruments for demand are supply shifters, such as the company cost. But one could also use some sort price index of a sector as an instrument, a sector that sells inputs to the company out of which you are estimating the residual demand. The thing is that this price index might be related to the price index that you are using for deflating your variables. In that situation, how do you extract these fluctuations related to the general economy out of your price index being used as an instrument?

  • $\begingroup$ what are you trying to accomplish with your model? $\endgroup$
    – snoram
    May 11 '15 at 22:45
  • $\begingroup$ Think of demand estimation. A sectorial price index, if related to a cost of an industry, would be an eligible supply shifter -- so you could include it as an instrument in a two-stage regression. But if all variables are deflated for a given period with a price index such as the consumer price index, would it make sense to extract the portion of the sectorial price index that is related to the consumer price index? Or in some way 'deflate' it? $\endgroup$
    – John Doe
    May 11 '15 at 22:56
  • $\begingroup$ It always helps to write the model explicitly in the post. It is not at all clear what the model is. For example, the independent variable is apparently "deflated prices" -"price-s?" Do you have a system of equations? And deflated by what ? Etc. $\endgroup$ May 12 '15 at 22:37

In that situation, how do you extract these fluctuations related to the general economy out of your price index being used as an instrument?

Regress the sectorial price index on the general price index, and use as instrument in the demand equation the "residuals plus estimated constant" from the previous regression.

These residuals will be by construction orthogonal to the general price index which indeed could be also a demand shifter, something that would invalidate the sectorial price index as an instrument.

As is the case with any time-series data, one should also deal with possible issues of non-stationarity, co-integration etc.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.