search

uchicago-computation-workshop / steven_durlauf

Repository for Steven Durlauf's presentation at the CSS Workshop (2/28/2019)
0 stars 0 forks source link

Possibility of more complex non-linear systems? #28

Open KenChenCompEcon opened 5 years ago

KenChenCompEcon commented 5 years ago

Looking forward to your presentation today!

I am just curious about the possibility of more complex nonlinear systems that generate social interactions and structures endogenously. What is the pros and cons of modeling in a nonlinear manner?

Thanks sincerely.

sdurlauf commented 5 years ago

I think that nonlinearities naturally arise in these environments and so it is a natural feature of many social interactions models. You ask a complicated question in that you are interested in a dynamic environment in which social structure coevolves with behaviors given structure at a point in time. There are a number of such models. My own work on neighborhood formation and intergenerational mobility has this feature. The linear models are restrictive in terms of the type of phenomena they are characterize. (There are senses in which the micrfoundations that produce linearity are nongeneric.) The JPE paper focuses on them because identification problems are the "worst" in such cases and because they are standard in empirical work.