llorracc
commented
4 years ago On the [image: R \beta < 1] point, much of my BufferStockTheory http://econ.jhu.edu/people/ccarroll/papers/BufferStockTheory paper is about deriving the most general results I can that define when the problem has a solution. The last line of the BufferStockTheory REMARK https://mybinder.org/v2/gh/llorracc/BufferStockTheory/master?filepath=Code%2FPython%2FBufferStockTheory.ipynb is:
Two tables in the paper https://econ.jhu.edu/people/ccarroll/papers/BufferStockTheory/#Factors-Defined-And-Compared summarize the various definitions, and then articulate conditions required for the problem to have a nondegenerate solution. Among the nondegenerate cases, the most interesting result is that if the Growth Impatience Condition holds there will be a target level of wealth.
And, there is a fairly comprehensive optional reporting system. If you set the “verbose=true” flag in the “.checkConditions()” module, you get all sorts of info about whether the parameter values you have fed in are ones where there is a solution.
In other papers where I have estimated the time preference rate, I’ve used the conditions to put bounds on the parameter search, which yielded a major improvement in the robustness and speed of finding the stable parameter values. Basically, you can compute the “edge” value of [image: \bar{\beta}] such that the problem diverges if [image: \beta \geq \bar{\beta}]. The you can define, say, [image: \hat{\beta} = exp(-\nu) + \bar{\beta}] and search for [image: \nu], knowing that even as [image: \nu \uparrow infty] you’ll never actually propose to the algorithm a [image: \hat{\beta}] for which the model is degenerate.
On Tue, Oct 29, 2019 at 7:14 AM Pablo Winant notifications@github.com wrote:
Merged #6 https://github.com/EconForge/dolark/pull/6 into master.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/EconForge/dolark/pull/6?email_source=notifications&email_token=AAKCK77KSGTJNXNVQLYT2EDQRALJLA5CNFSM4JFUXXN2YY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOUQCBQRI#event-2751731781, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAKCK7Y7Y4HZOBEGSIM5L33QRALJLANCNFSM4JFUXXNQ .
--
- Chris Carroll
The most patient Aiyigari consumers are not diverging; they just have very large target wealth. Also, there's a bug (now commented out) in equilibrium.py