-
-
Notifications
You must be signed in to change notification settings - Fork 199
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Remove initialization of TranShkNow #327
Conversation
The problem goes the other direction, in my view. Lifecycle models in HARK
have no input for what the distribution of income shocks at t=0 should be.
The object ThisType.IncomeDstn[0] is what the *solver* needs to know for
t=0, which is the distribution of income shocks at the start of t=1.
Permanent income shocks at t=0 can be handled by the distribution specified
with the pLvlBlahInit variables-- there is no difference between initial
pLvl distribution and "shocks" that occur between birth and the first time
they could be "observed", so they're lumped together.
Transitory income shocks are more difficult, because we have no alternate
way to specify it. Moreover, there are some papers (including by us) that
assume no transitory shock in the very first period. Among other things,
this simplifies ex ante welfare analysis because you only consider the
value function at exactly one state.
In the big overhaul of HARK, this is one of the timing quirks that needs to
be worked out. We would like to be able to specify things so that shock
distributions at the beginning of t=0 are put into an instance like any
other period, but our solver framework knows that it's not actually useful
for solving the problem.
…On Fri, Jun 21, 2019, 4:42 PM Patrick Kofod Mogensen < ***@***.***> wrote:
fixes #326 <#326>
I think it makes no sense to hard-code the transitory income shock. If it
is super important from a practical point of view, we should simply make
the very first distribution a discrete distribution with one point (1.0
ofc). Otherwise the simulations don't really match the model specification.
------------------------------
You can view, comment on, or merge this pull request online at:
#327
Commit Summary
- Remove initialization of TranShkNow
File Changes
- *M* HARK/ConsumptionSaving/ConsIndShockModel.py
<https://github.com/econ-ark/HARK/pull/327/files#diff-0> (2)
Patch Links:
- https://github.com/econ-ark/HARK/pull/327.patch
- https://github.com/econ-ark/HARK/pull/327.diff
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#327?email_source=notifications&email_token=ADKRAFLE4EL5FSADBDOI3W3P3U4MTA5CNFSM4H2UWW72YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G3BLRXA>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/ADKRAFIGY6WFVOXXS7ZO6QLP3U4MTANCNFSM4H2UWW7Q>
.
|
I see your point. The model is solved conditional on the initial condition. Then why don't we introduce a distribution of initial (t_age = 0) income shocks that is only used in these lines - that is it is only used for simulation. It's not necessary for I think an explicit "initial shock distribution for simulation" would be both useful and easy to implement. |
In the talk today, we agree on a solution as sketched above, so I will move forward with that. I'll keep the default as a degenerate "draw" of 1, but allow users to more flexibly control it with a distribution if they want. |
@pkofod what is the status of this? |
fixes #326
I think it makes no sense to hard-code the initial transitory income shock. If it is super important from a practical point of view, we should simply make the very first distribution a discrete distribution with one point (
1.0
ofc). Otherwise the simulations don't really match the model specification.