BGP and perfect foresight

Good morning,

I had a rather bizarre comment of a reviewer:
“The procedure to solve and simulate dynamic general equilibrium model with growth requires determining the balanced growth path of the economy and subsequently, to normalize the variables to have the model expressed in stationary terms. In this paper, the economy in per capita terms has three sources of potential growth (labor-augmenting technological change and energy-saving technological progress in the use of electricity and non-electricity energy sectors).
As it is standard in the literature, the authors need to analytically (i) demonstrate the existence of a balance growth path for the economy in per capita terms, (ii) determine the long-run growth rate of the per capita variables and (iii) normalize the model by discounting growth to have the model defined in stationary variables. After doing that, the model can be numerically simulated. The authors do not proceed with any of these steps. Thus, I cannot follow how the authors claim that the growth rate of the economy (provided that a balance growth path exists) is related to the population growth and the labor-saving technical progress, but has no relation whatsoever with the two sources of energy-saving technical progress.”

I chose the perfect foresight setup for my model because I did not want to impose a BGP to my model. Indeed, I did not want some of my variables to grow at these energy saving technical progresses. If I were to have a BGP, I would need them to grow at that technical progress. But this wasn’t what I wanted. Hence, my economy was shocked by these energy saving technical progresses.

In my point of view, this comment from the reviewer is only true to perform stochastic simulations that needs an approximation around a SS or a BGP and hence needs normalization etc. Indeed, for stochastic simulations, perturbations techniques needs a well defined SS or BGP to approximate the model around this SS or BGP. But this does not necessarily apply to perfect foresight set up. Am I getting this wrong ?

The referee indeed seems to have a model solved with perturbation techniques in mind where you are doing an approximation about a well-defined steady state in intensive form.
That’s not necessarily the case for perfect foresight simulations that can be solved via an extended path method where you simply solve a nonlinear equation system with two boundary conditions. But it’s still your job to explain to referee how you exactly solve the model. The challenge is deriving proper boundary conditions.

1 Like