New model: Unemployment search

[mnwhite]

Create module for models of consumption-saving-search. Consumption-saving model with persistent unemployment; agent must exert costly effort to increase probability of finding job / receiving offer.

Lots of variations on this to do. Can pick one or two easy ones to start to show models with multiple controls in HARK (or rather: multiple controls in some states).

[mnwhite]

This is a good candidate for an early outside contributor who wants to write a new model.

Consider the utility form: u(c,s) = ((1-s)^alpha*c)^(1-rho))/(1-rho), where c is consumption and s is search effort (0 when employed); alpha and rho are parameters > 0.

[giang-nghg]

@mnwhite I'd like to help, but I only have experience in Python, not scientific modeling. Could you explain to me the general idea of how to implement a new model? Or point me to an example/resource that demonstrate it? Thank you very much.

[mnwhite]

If you're an experienced programmer but you don't have graduate-level training in economics, this isn't a good issue to get your to toes wet with HARK. In the post above, I should have specified "an early outside contributor who is an economist".

We have other issues that are good places to contribute for someone who knows how to program, but doesn't have a PhD in economics. We want various numeric tools that have no economic content, but are useful in a wide array of situations. The most obvious one is #49 to code up the Gauss-Hermite approximation to the normal distribution; I think I even link to Matlab code that can be translated.

I should add other "tool request" issues for non-economists to work on. E.g. optimization routines.

[sbenthall]

I'm an experienced programmer with a PhD that's not in economics; but I think I have enough background to make sense of the consumption shock models and their variations.

As these seem like the canonical models used in HARK so far, I figure I should figure out how to build one before making many other contributions.

If you agree, I'd be happy to take on this ticket as an exercise.

[mnwhite]

Writing code to solve a new model is actually the largest possible contribution, but I definitely won't tell you not to try.

On Sat, Oct 19, 2019 at 5:51 PM Sebastian Benthall notifications@github.com wrote:

I'm an experienced programmer with a PhD that's not in economics; but I think I have enough background to make sense of the consumption shock models and their variations.

As these seem like the canonical models used in HARK so far, I figure I should figure out how to build one before making many other contributions.

If you agree, I'd be happy to take on this ticket as an exercise.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/econ-ark/HARK/issues/21?email_source=notifications&email_token=ADKRAFJE5CVLDGVGYUAFMKDQPN6OJA5CNFSM4B2O6WBKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBX5MBQ#issuecomment-544200198, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADKRAFOOCLWREVW2BKKFNM3QPN6OJANCNFSM4B2O6WBA .

[sbenthall]

I'm looking at this model that already factors in unemployment:

https://github.com/econ-ark/HARK/blob/master/HARK/ConsumptionSaving/ConsMarkovModel.py#L983-L1126

Can you explain how the model you have in mind for this issue differs from this one?

A costly unemployment search could be modeled as a further reduction of "income" when unemployed.

Or is search voluntary (and action with learned policy, not assumed)?

[mnwhite]

Yes, the intensity of employment search is a second choice (or control) variable. A more intense search increases the probability of becoming re-employed, but is increasingly costly in terms of utility, not money. It can be modeled as the agent choosing their probability of being re-employed s_t, with a larger hit to utility from larger values of s_t. For example, the form I've worked with in the past was something like:

u(c,s) = ((1-s)^\alpha * c)^{1-\rho} / (1 - \rho)

Where alpha is a parameter that governs the disutility of search. When s_t = 0, that leading factor is just 1 and the agent loses nothing from searching.

On Sun, Oct 20, 2019 at 4:20 PM Sebastian Benthall notifications@github.com wrote:

I'm looking at this model that already factors in unemployment:

https://github.com/econ-ark/HARK/blob/master/HARK/ConsumptionSaving/ConsMarkovModel.py#L983-L1126

Can you explain how the model you have in mind for this issue differs from this one?

A costly unemployment search could be modeled as a further reduction of "income" when unemployed.

Or is search voluntary (and action with learned policy, not assumed)?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/econ-ark/HARK/issues/21?email_source=notifications&email_token=ADKRAFO74ONJF6VJFSVHUZ3QPS4QVA5CNFSM4B2O6WBKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEBYS7MQ#issuecomment-544288690, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADKRAFITOBT7Y2E4QAMCCULQPS4QVANCNFSM4B2O6WBA .

[mnwhite]

I have multiple PhD students working on variations of this model right now. Submissions are due ~Sept 1, will update.

[mnwhite]

Those students in 2020 did not produce workable code. Endogenous unemployment models remain desired in HARK, but it's fallen by the wayside.