TANK models with share of each agent being dynamic

Hi. I wonder if it’s possible to build a model with two different types of households, like in Galí et al. (2007) where a share \lambda are Ricardian and a share 1–\lambda are non-Ricardian households. My question, is it possible to build a model alike where the share of a determined type of household, say \lambda_t\in(0,1) is a function of time, and presumably other variables in the model and even subject to a uncertainty (e.g. \lambda_t=f(\boldsymbol {\Omega_t})+\varepsilon_t^\lambda )? Given that, at the end this parameter only plays a role as a weight in aggregate variables. Is there some work with a similar approach?

Thanks.

It’s hard to provide a judgment on this. I think it’s less a matter of writing down the time-varying weights than obtaining the actual solution. After all, the resulting problem will be inherently nonlinear and it’s not clear whether perturbation solutions will be able to handle this.

Yes, I think that this kind of very ad-hoc modification could be mathematically problematic for the solution. Thanks!