Output measure in DGE models

Dear modelers,

I would like to estimate the elasticities of substitution for my production technology, which combines energy and (capital-labor) viewed as GVA. However, I encounter some problems regarding the correspondence between my model and national account data.

If I take the output approach, the GDP is defined in national accounts as the some of gross value added and taxes less subsidies. Should I take out ‘taxes less subsidies’ when I perform estimations on the substitution elasticities. My first guess would be yes. However, the output in my model would not correspond anymore to the output as measured by national accounts as there would be some king of statistical gap.

What should I take then:
a. GVA for (KL) bundle + energy costs for the energy bundle
b. GDP (GVA + taxes less subsidies) for the (KL) bundle + energy costs for the energy bundle?


If I understand your problem correctly, the issue is that your model is not able to distinguish between GDP at factor cost and GDP at producer price. The answer which one to use should depend on your estimation strategy and what you are after. If you are interested in firm choices, then what matters are factor costs

Thank you for the answer.

My issue is rather which the data to take for my estimation procedure (I’m using the MiceconCES tool on R from Henningsen and Henningsen, not Dynare).

What I have done:

  • estimate my elasticity of the production function (Y) combining (KL)-E using GVA data for (KL) bundle, energy cost data for energy, and GVA+Energy Cost data for Y.
  • but in my dynare steady state file, Y (from the resource constraint) does not correspond to GVA+EnergyCost but GDP+EnergyCost (hence integrating taxes less subsidies to GVA+EnergyCost). IS that ok? Or should I leave out taxes less subsidies from the output measure ?

Best regards,

There is an inherent tension in your model. But have you checked whether the difference matters? Often times, the difference is small, so it does not really matter.

The difference is small indeed.