Calibration for a growth model with BGP

Dear members,

Does the calibration of parameters for a growth model need to include the growth rates?

For example, in the case of a standard DSGE model one would calibrate the capital depreciation rate from the SS capital accumulation equation (K(t+1)=Ik+K(t)*(1-delta(k))), giving delta(k) = I/K, because at SS K(t+1)=K(t)=K.

But for a model with a BGP, should I take into account the growth rate of the economy such as : delta(k)= I/K - g ? I/K being the ratio of investment on capital stock at SS (long-run average). With my capital accumulation equation : K(t+1)=Ik+K(t)(1-delta(k)). At BGP, K(t+1) becomes K(t)(1+g).

Or should I calibrate my parameters following the steady state of my economy and hence calibrate delta(k)=I/K ?


The correct way would be explicitly account for growth. See for example

1 Like

Thanks, very helpful.

Hi Prof. Pfeifer; this link is not working. May I kindly ask if you have a new one? I am trying to calibrate a long-run endogenous growth model where the growth rate of variables in the model (K, Y, C,…) is driven by the growth rate of human capital (H). For a stationary steady-state, the model is written in terms of K/H, Y/H, C/H,…

What I have seen some people do is to get parameters like population growth rate (n) and depreciation rate (\delta) from data. And parameters like discount factor (\beta) are taken from the literature, for example.

But the calibration will not match the target variables of interest. My target variable here is the steady-state growth rate of the model variables (H, Y, C, K…). So some people use the productivity parameter (A) in the production function or, say, the productivity parameter (E) in the human capital accumulation function to get the desired value or dynamics of their target variable. So A and E here are just scaling parameters. That is fine, yeah?

And is it necessary for K/Y in growth models to match K/Y in data? Like, I can get steady-state values for K/H and Y/H using the calibrated parameters I have. It is necessary, in growth models, for (K/H)/(Y/H) = K/Y to match K/Y data? It seems it is something the researcher can decide whether is necessary or not…per my knowledge, based on the papers I have read so far. Thus, they do not match ratios like K/Y…but sure they match other target variables like output growth rate, income life cycle, and others.

The file is now at Chapter_2_RBC.pdf - Google Drive