Skip to content

Funded defined contribution pillar, part 1: parameters and account functions - #1283

Merged
rickecon merged 5 commits into
PSLmodels:masterfrom
jdebacker:FDC
Nov 18, 2026
Merged

rickecon merged 5 commits into
PSLmodels:masterfrom
jdebacker:FDC

Conversation

@jdebacker

@jdebacker jdebacker commented Nov 14, 2026 •

Copy link
Copy Markdown
Member

This is the first of four PRs that add a funded defined contribution (FDC) pension pillar to OG-Core. Contributions go into individual accounts that are invested at the household portfolio return r_p. At retirement each balance buys an actuarially fair life annuity. Unlike the existing PAYG systems, the accounts are held by a fund outside the government budget, and its assets add to the domestic capital supply.

This PR adds only the parameters and the account calculations. Nothing reads them yet, so model results are unchanged.

Design

  • Separate pillar. FDC runs alongside whichever pension_system is selected, which allows combinations such as NDC plus a funded premium pension, and PAYG-to-funded reforms. fdc_tau = 0 (the default) turns it off.
  • Contributions. Households pay fdc_tau × (labor income in the FDC band) in period t, at ages below the retirement age, out of period-t resources.
  • Timing. End-of-period holdings earn the next period's r_p, the same timing as private savings b. Each contribution therefore earns the return through the retirement period itself, one period more than the NDC crediting formula.
  • Annuity. At R the balance buys an annuity-due, fixed in real terms. It is priced on the mortality of the household's income group and on the cohort's r_p path, so each group's annuity pool pays out exactly what it holds. A household already retired when the time path begins re-annuitizes its remaining balance.
  • Death before retirement. Pre-retirement holdings are returned so that a later PR can add them to bequests. After retirement, the holdings of those who die stay in the annuity pool.

Parameters

  • fdc_tau (default 0.0). This is a cohort pension parameter, added to COHORT_PENSION_PARAMS, so pension_regimes can set it by birth cohort just like tau_p.
  • Earnings band. There is no FDC-specific cap. With pension_earn_cap true, contributions are paid only on labor income up to dc_earn_cap_factor × avg_labor_inc_t1_model_units. This is the cap shared with the NDC system, and it applies under any pension_system. With pension_earn_cap false, there is no cap.
    • fdc_earn_floor_factor (default [0.0]) sets the lower cutoff, for example to cover only the earnings above a DB band.
    • The floor is a cohort pension parameter, entered and indexed like db_earn_cap_factor and dc_earn_cap_factor. A single value is deflated by e^{-g_y t}, so it is fixed in real terms, and pension_regimes can change it by cohort.
    • Specifications raises an error if, with pension_earn_cap true, the cap is below the floor at any period and age where fdc_tau is positive.

Functions (ogcore/pensions.py)

  • fdc_lifetime and its numba core fdc_lifetime_loop compute one household's account over its life:
    • contributions, balances, benefits, and end-of-period holdings (the fund's assets);
    • d(contribution)/dn, and the benefit at R per unit contributed at each age, which feed the labor FOC accrual term in the next PR.
  • fdc_ss computes the steady-state accounts of every income group.
  • Also added: fdc_annuity_factor_loop, fdc_earn_bands, fdc_mort_rates, and the FDCLifetime named tuple.

Planned follow-up PRs

  1. Steady state: FDC flows in the household budget (kept out of net_taxes), an accrual term in FOC_labor, pre-retirement holdings in BQ, fund assets F in get_K_splits(B + F, …), and outputs.
  2. Transition path: r_p path pricing, initial holdings from the baseline steady state, and re-pricing at t=0 of cohorts already annuitized.
  3. Example and docs: a PAYG-to-FDC transition example and FDC equations in the pensions chapter.

Tests

New tests in tests/test_pensions.py:

  • hand calculations of the annuity factor and of a full lifetime, including the timing of compounding through R;
  • the earnings band (floor and cap);
  • re-annuitization for a household already retired;
  • with the default fdc_tau = 0, every account quantity is zero;
  • the present value of benefits equals the balance at R, and the pool is exhausted at the last age;
  • a finite-difference check of the accrual derivative;
  • fdc_tau set by cohort through pension_regimes;
  • calendar-indexed returns;
  • the earnings band with the shared cap on and off, single-value indexing and a cohort-specific floor;
  • the error raised when the cap is below the floor.

tests/test_pensions.py, tests/test_parameters.py, and tests/test_household.py pass locally (non-local tests).

🤖 Generated with Claude Code

@codecov-commenter

codecov-commenter commented Nov 14, 2026 •

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 68.03279% with 39 lines in your changes missing coverage. Please review.
✅ Project coverage is 79.57%. Comparing base (9b8a3b4) to head (ff1e253).

Files with missing lines Patch % Lines
ogcore/pensions.py 63.88% 39 Missing ⚠️
Additional details and impacted files

Impacted file tree graph

@@            Coverage Diff             @@
##           master    #1283      +/-   ##
==========================================
- Coverage   79.81%   79.57%   -0.24%     
==========================================
  Files          24       24              
  Lines        6811     6924     +113     
==========================================
+ Hits         5436     5510      +74     
- Misses       1375     1414      +39     
Flag Coverage Δ
unittests 79.57% <68.03%> (-0.24%) ⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines Coverage Δ
ogcore/parameters.py 90.77% <100.00%> (+0.13%) ⬆️
ogcore/pensions.py 71.89% <63.88%> (-1.32%) ⬇️
🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Adds fdc_tau (a cohort pension parameter, default 0), fdc_earn_cap,
fdc_earn_cap_factor, and fdc_earn_floor_factor, and the functions that
compute a household's funded defined contribution account over its
life: contributions, balances earning r_p through the retirement
period, a fair annuity-due priced on income-group mortality and the
r_p path, end-of-period holdings, and the accrual derivatives. Not yet
wired into the household problem or aggregates.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
jdebacker and others added 3 commits November 16, 2026 17:12
… accounts

Remove fdc_earn_cap and fdc_earn_cap_factor. With pension_earn_cap true,
FDC contributions are paid only on labor income up to dc_earn_cap_factor
times avg_labor_inc_t1_model_units, the cap shared with the notional
defined contribution system, under any pension_system.

fdc_earn_floor_factor is now a cohort pension parameter entered and
indexed like db_earn_cap_factor and dc_earn_cap_factor (a single value
is value / e^(g_y t); regimes may change it by cohort), so the floor and
the cap move together. fdc_earn_bands returns arrays by period and age,
read down each cohort's diagonal. Specifications checks that the cap is
at least the floor wherever fdc_tau is positive.

Also resolves the leftover conflict markers in default_parameters.json
from the upstream merge and moves the FDC changelog entries to an
Unreleased section.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@rickecon

Copy link
Copy Markdown
Member

@jdebacker. I have the following suggestions and questions for this PR.

  • I submitted a pull request to your branch adding the parameter r_fdc_wedge that characterizes the return to workers on their FDC contributions such that r_fdc = r_p - r_fdc_wedge. This will allow the pension authority to payout a rate of return that need not correspond to the internal investment rate of return of the pension authority. Any differences will have be received or paid out to the pension authority.
  • Why have the FDC earnings have different interest accrual timing than the NDC. It should be the same. And I like the idea of a one-period delay. But whichever way we choose, it should be the same for both NDC and FDC.
  • [For future parts of this set of PRs] We need to make some assumptions for people already working and already retired at period t=1. What happens to FDC balances of people who have been working and paying into the DB system when the FDC starts. My guess is we have to fund their accounts with what their DB contributions would equal at the starting points. And for retired workers in period t=1, we need to provide and fund an actuarially fair annuity.
  • [For future parts of this set of PRs] For deaths before retirement, should we return the FDC balances to the population of workers through bequests, or should we return the FDC balances to the pension fund? Seems like bequests.

Updated FDC growth rate to r_fdc=r_p-r_fdc wedge
@jdebacker

Copy link
Copy Markdown
Member Author

@rickecon Thanks for the review.

I merged in your PR to this branch -- nice additions!

Re the other points:

Why have the FDC earnings have different interest accrual timing than the NDC. It should be the same. And I like the idea of a one-period delay. But whichever way we choose, it should be the same for both NDC and FDC.

Good catch. Opening a separate PR now to make this consistent. The timing wasn't as important for a pay as you go system, but I agree that it makes sense to have consistency across the two.

[For future parts of this set of PRs] We need to make some assumptions for people already working and already retired at period t=1. What happens to FDC balances of people who have been working and paying into the DB system when the FDC starts. My guess is we have to fund their accounts with what their DB contributions would equal at the starting points. And for retired workers in period t=1, we need to provide and fund an actuarially fair annuity.

Yes, let's revisit.

[For future parts of this set of PRs] For deaths before retirement, should we return the FDC balances to the population of workers through bequests, or should we return the FDC balances to the pension fund? Seems like bequests.

PR #1284 does addresses this -- treating remaining balances like bequests.

@rickecon

Copy link
Copy Markdown
Member

@jdebacker. Looks good. Merging now.

@rickecon
rickecon merged commit 998032d into PSLmodels:master Nov 18, 2026
8 checks passed
john-p-ryan added a commit to john-p-ryan/OG-Core that referenced this pull request Nov 19, 2026
…hocks

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants