Deal with cases when the agent wants to short #749
Conversation
Yes; but anything will break under those circs. We may need to handle this in the future (esp when trying to have a discrete-continuous problem jointly with the portfolio problem) but that's a problem for another day. |
|
Yeah, the code assumes a single-crossing property / monotone FOC, but
that's consistent with most of our other solvers, which assume that the
(marginal) value function being fed to it from solution_next came from the
solution to the same model and thus is guaranteed to have certain
properties.
A while back I wrote a "robust" ConsIndShock solver that used Giu Fella's
method for handling non-concave value functions. Just like my attempt at
implementing his discrete labor supply choice model with that method, it
put the "bust" in "robust". 18 months ago I began work with Fedor Iskhakov
on an improved method.
Anyway, thanks for doing this so fast, I'll take a look soon.
…On Thu, Jul 2, 2020 at 3:34 PM Christopher Llorracc Carroll < ***@***.***> wrote:
I guess it will break if one considers non-concavities in the value
function?
Yes; but anything will break under those circs. We may need to handle this
in the future (esp when trying to have a discrete-continuous problem
jointly with the portfolio problem) but that's a problem for another day.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#749 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADKRAFNMPCD2QUPWZ3KOJ43RZTONHANCNFSM4OPFZGMQ>
.
|
Yes, exactly. Since I know we will eventually have to deal with this, I just want to keep a mental list of where things might break. |
|
Yes, this is exactly correct. And I guess I had already handled the 100% constraint, now labeled as |
This implements the solution proposed by @mnwhite in #741 for dealing with cases where the agent wants to hold less than a 0% share of the risky asset.
I did some testing and it seems to be well behaved. The approach assumes that as long as the FOC at s=0 is negative, the agent wants to hold less than 0. I guess it will break if one considers non-concavities in the value function?
I got rid of the try-except block as @mnwhite suggested.