Quasi hyperbolic pf by MridulS · Pull Request #833 · econ-ark/HARK

MridulS · 2020-09-03T13:12:55Z

implements #215

replaces #815

- Hyperbolic beta modifies current solution discount factor. - Presolve processes generate geometric solution if needed (Hyperbolic_beta !=1)

github-actions · 2020-09-26T03:47:21Z

👈 Launch a binder notebook on this branch

llorracc · 2020-10-01T16:23:05Z

/binder

github-actions · 2020-10-01T16:23:20Z

👈 Launch a binder notebook on this branch

sbenthall · 2020-12-10T16:51:13Z

@Mv77 to look at this as CDC's RA

Mv77 · 2021-06-15T13:28:46Z

I did not get to this in my time as Chris's RA this winter and I don't anticipate working on it for now.

sbenthall · 2021-06-15T18:38:07Z

@llorracc @MridulS I'm wondering if there is somebody else who could review this.

sbenthall · 2022-04-06T16:12:39Z

@MridulS to fix conflict, then @mnwhite to review?

Copilot

Pull request overview

This PR implements quasi-hyperbolic discounting for perfect foresight consumer agents in HARK, addressing issue #215 and replacing PR #815. The implementation uses a "beta-delta" approach where naive agents have present-biased preferences.

Changes:

Modified solveOneCycle in core.py to support geometric solutions for backward induction
Added HyperbolicBeta parameter to PerfForesightConsumerType and its solver
Updated discount factor calculation to incorporate hyperbolic discounting
Added example configurations and a demonstration notebook

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 9 comments.

File	Description
HARK/core.py	Modified `solveOneCycle` to conditionally use `geometric_solution` attribute for backward induction instead of current solution
HARK/ConsumptionSaving/ConsIndShockModel.py	Added `HyperbolicBeta` parameter to solver and agent type; updated effective discount factor calculation; reformatted init dictionary
examples/Journeys/Quickstart_tutorial/Jounery_1_param.py	Added `Hyperbolic_beta` parameter to example configuration
HARK/ConsumptionSaving/ConsIndShockModelDemos/testing_hyperbolic_discounting.ipynb	New Jupyter notebook demonstrating and validating the hyperbolic discounting implementation with algebraic verification

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-28T19:19:13Z

 PermGroFacAgg = 1.0                 # Aggregate permanent income growth factor (only matters for simulation)
 T_age = None                        # Age after which simulated agents are automatically killed
 T_cycle = 1                         # Number of periods in the cycle for this agent type
+Hyperbolic_beta = 1                 # Naive hyperbolic discount factor


The parameter name "Hyperbolic_beta" in the example file is inconsistent with "HyperbolicBeta" used in the main implementation. Parameter names should be consistent across the codebase. Consider using "HyperbolicBeta" (without underscore) to match the implementation.

Copilot · 2026-01-28T19:19:13Z

+
        Returns:
        ----------


The docstring is missing a blank line before the "Returns:" section and has inconsistent formatting with an empty line after the description. The formatting should match the project's documentation standards.

Suggested change

Returns:

----------

Returns

-------

Copilot · 2026-01-28T19:19:14Z

+        "MaxKinks",
+        "BoroCnstArt",
+        "HyperbolicBeta",
+        "geometric_solution",


Adding "geometric_solution" to time_inv_ makes it a required parameter that must be provided during initialization. This could break existing code that creates PerfForesightConsumerType instances without this parameter. Consider making this parameter optional by checking for its existence with hasattr() (as already done in core.py line 923) rather than requiring it as a time-invariant parameter.

Suggested change

"geometric_solution",

Copilot · 2026-01-28T19:19:14Z

+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "I implement hyperbolic discounting by modyfing the `solveOneCycle` function in `core.py`. If a `geometric_solution` attribute is specified, the `solveOneCycle` function uses the solutions specified in this attribute list in its backward induction loop, instead of the current solution. The exponential discount factor is also resized by a factor of the `hyperbolic` discount variable. "


The notebook documentation contains a typo: "modyfing" should be "modifying".

Suggested change

"I implement hyperbolic discounting by modyfing the `solveOneCycle` function in `core.py`. If a `geometric_solution` attribute is specified, the `solveOneCycle` function uses the solutions specified in this attribute list in its backward induction loop, instead of the current solution. The exponential discount factor is also resized by a factor of the `hyperbolic` discount variable. "

"I implement hyperbolic discounting by modifying the `solveOneCycle` function in `core.py`. If a `geometric_solution` attribute is specified, the `solveOneCycle` function uses the solutions specified in this attribute list in its backward induction loop, instead of the current solution. The exponential discount factor is also resized by a factor of the `hyperbolic` discount variable. "

Copilot · 2026-01-28T19:19:14Z

        solution_t = solveOnePeriod(**temp_dict)
        solution_cycle.insert(0, solution_t)
-        solution_next = solution_t
+        # solution_next = s/olution_t


There is a commented-out line with a typo ("s/olution_t" instead of "solution_t"). This commented code should be removed entirely as it serves no purpose.

Suggested change

# solution_next = s/olution_t

Copilot · 2026-01-28T19:19:14Z

                           'T_age' : T_age,
-                           'T_cycle' : T_cycle
+                           'T_cycle' : T_cycle,
+                           'Hyperbolic_beta' : Hyperbolic_beta


The parameter name "Hyperbolic_beta" (with underscore) is inconsistent with "HyperbolicBeta" (without underscore) used elsewhere in the code. This should be "HyperbolicBeta" to match the implementation in ConsIndShockModel.py.

Copilot · 2026-01-28T19:19:15Z

+    "PermGroFacAgg": 1.0,  # Aggregate permanent income growth factor (only matters for simulation)
+    "T_age": None,  # Age after which simulated agents are automatically killed
+    "T_cycle": 1,  # Number of periods in the cycle for this agent type
+    "HyperbolicBeta": 1,


The HyperbolicBeta parameter is added to init_perfect_foresight but lacks a descriptive comment like all other parameters in the dictionary. Add a comment such as "# Quasi-hyperbolic discount factor (naive agents)" to maintain consistency.

Suggested change

"HyperbolicBeta": 1,

"HyperbolicBeta": 1, # Quasi-hyperbolic discount factor (naive agents)

Copilot · 2026-01-28T19:19:15Z

        """
        self.defUtilityFuncs()
-        self.DiscFacEff = self.DiscFac * self.LivPrb
+        self.DiscFacEff = self.DiscFac * self.LivPrb * self.HyperbolicBeta


The new HyperbolicBeta functionality lacks automated test coverage. While there is a Jupyter notebook demonstrating the feature, there are no unit tests in the test_PerfForesightConsumerType.py file that verify the quasi-hyperbolic discounting behavior. Consider adding tests that verify the consumption functions match the expected algebraic solutions when HyperbolicBeta is not equal to 1.

Copilot · 2026-01-28T19:19:15Z

+    "geom_solution = deepcopy(PFexample.solution)\n",
+    "PFexample_hyperbolic = deepcopy(PFexample)\n",
+    "PFexample_hyperbolic.geometric_solution = geom_solution\n",
+    "PFexample_hyperbolic.Hyperbolic_beta=0.7\n",


The notebook uses "Hyperbolic_beta" (line 223) which is inconsistent with the parameter name "HyperbolicBeta" used in the implementation. This should be "HyperbolicBeta" (without underscore) to match the code.

Suggested change

"PFexample_hyperbolic.Hyperbolic_beta=0.7\n",

"PFexample_hyperbolic.HyperbolicBeta=0.7\n",

MridulS · 2026-03-08T15:45:37Z

The issue was turned into a discussion. Should I still try to clean this up and sync up with new changes? @mnwhite

mnwhite · 2026-03-08T16:00:01Z

No, you don't need to do it. This is still a task for me. I left the PR open because it's still a new model that I don't want to throw away. It will probably be moved to a totally new PR when I get around to that, because it's so out of date.

…

On Sun, Mar 8, 2026 at 11:45 AM Mridul Seth ***@***.***> wrote: *MridulS* left a comment (econ-ark/HARK#833) <#833 (comment)> The issue was turned into a discussion. Should I still try to clean this up and sync up with new changes? @mnwhite <https://github.com/mnwhite> — Reply to this email directly, view it on GitHub <#833 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADKRAFLK2C6UMMMWAOJYD2T4PWITPAVCNFSM6AAAAACTG5K2OGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHM2DAMJZGI4TMNRVHA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

g-carranza and others added 7 commits December 20, 2018 17:54

Implementation of geometric_solution

93c573b

Default Hyperbolic_beta value

612e032

Introduction of hyperbolic_beta

6baa7c6

- Hyperbolic beta modifies current solution discount factor. - Presolve processes generate geometric solution if needed (Hyperbolic_beta !=1)

Delete presolve

ef0c95d

Example of hyperbolic discounting

a1e3eb1

fix merge conflicts

5c11375

fix up example

24a967d

mnwhite self-assigned this Sep 3, 2020

MridulS added the Status: Review Needed label Sep 17, 2020

llorracc assigned MridulS Sep 24, 2020

Merge branch 'master' into quasi-hyperbolic-pf

465768d

econ-ark deleted a comment from github-actions Bot Sep 26, 2020

MridulS removed their assignment Oct 1, 2020

sbenthall unassigned mnwhite Dec 10, 2020

sbenthall added this to the 1.x.y milestone Dec 10, 2020

sbenthall assigned Mv77 Jun 14, 2021

sbenthall unassigned Mv77 Jun 15, 2021

sbenthall requested a review from mnwhite April 6, 2022 16:12

akshayshanker self-assigned this May 22, 2025

mnwhite moved this to Stale in Issues & PRs Jan 3, 2026

mnwhite added this to Issues & PRs Jan 3, 2026

alanlujan91 requested a review from Copilot January 28, 2026 19:15

Copilot started reviewing on behalf of alanlujan91 January 28, 2026 19:16 View session

Copilot AI reviewed Jan 28, 2026

View reviewed changes

	"I implement hyperbolic discounting by modyfing the `solveOneCycle` function in `core.py`. If a `geometric_solution` attribute is specified, the `solveOneCycle` function uses the solutions specified in this attribute list in its backward induction loop, instead of the current solution. The exponential discount factor is also resized by a factor of the `hyperbolic` discount variable. "
	"I implement hyperbolic discounting by modifying the `solveOneCycle` function in `core.py`. If a `geometric_solution` attribute is specified, the `solveOneCycle` function uses the solutions specified in this attribute list in its backward induction loop, instead of the current solution. The exponential discount factor is also resized by a factor of the `hyperbolic` discount variable. "

	"HyperbolicBeta": 1,
	"HyperbolicBeta": 1, # Quasi-hyperbolic discount factor (naive agents)

	"PFexample_hyperbolic.Hyperbolic_beta=0.7\n",
	"PFexample_hyperbolic.HyperbolicBeta=0.7\n",

Uh oh!

Uh oh!

Conversation

MridulS commented Sep 3, 2020

Uh oh!

github-actions Bot commented Sep 26, 2020

Uh oh!

llorracc commented Oct 1, 2020

Uh oh!

github-actions Bot commented Oct 1, 2020

Uh oh!

sbenthall commented Dec 10, 2020

Uh oh!

Mv77 commented Jun 15, 2021

Uh oh!

sbenthall commented Jun 15, 2021

Uh oh!

sbenthall commented Apr 6, 2022

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 28, 2026

Choose a reason for hiding this comment

Uh oh!

MridulS commented Mar 8, 2026

Uh oh!

mnwhite commented Mar 8, 2026 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants