Model Specification

[Mahdi Jafari]

Hi,
I want to regress natural gas consumption on the price to estimate price elasticity. The data of natural gas consumption is at the bill level for a period consisting of multiple days and multiple units of households. For example, one bill is for 30 days and 5 units of households (these 5 HH have a common gasometer so their consumption is not measured separately), and the other bill is for 40 days and 1 unit of household.

I wonder if which of following specifications are correct (or better, in case both of them are true)?

1. Q = beta.Price + delta.NumberofUnits + lambda.NumberofDays
or
2. Q / (NumberofUnits * NumberofDays) = beta.Price

Also I use average price in the above equation. In fact :
Price = Cost / Consumption .
As it is obvious the dependent variable consumption (Q) appears on the denominator of the right hand side of the model.
Is there any concern about endogeneity because of inverse mechanical relationship of Price and Q ? (note that cost also increases with consumption and is not constant)

[Carlo Lazzaro]

Mahdi:
see -reg3-.

[Mahdi Jafari]

Mahdi:
see -reg3-.

Thank you Carlo, But my question is not about Stata command. I asked about the econometrics side of the specification.

[Carlo Lazzaro]

Mahdi:
my reply expressed the concern that you might have an endogeneity issue due to simultaneous equations.
Provided that I do not know the literature of your research field, I would go with your #1 code (I think you have other predictors, too).