OG-UK overlapping generations
OG-UK
Calibration targets, the nearest external benchmark, the OBR’s own OLG model, and the limits.
No ground truth. Calibration targets instead
Calibrated; no outcome benchmark: an imposed target cannot disagree.
| Dimension | Level | What would change it |
|---|---|---|
| Implementation fidelity | moderate | UK calibration cross-checked against OG-Core |
| Predictive validation | not applicable | None: a long-run comparative static |
| Identification | not applicable | None: behaviour comes from the calibration |
| Policy counterfactuals | weak | A published UK OLG result to score against |
| Uncertainty | weak | Sensitivity sweep over the key elasticities |
| Vintage reproducibility | weak | Solves reproduce from committed inputs |
Targets, the nearest benchmark, the OBR’s model
Calibration targets, not outcomes
No published replication target exists for this model class.
What the table comparesCalibration inputs, not model outputs
The table compares calibration inputs with official UK aggregates, not model outputs.
What would count as over-identifying evidence, and the dating and structural caveats
The real over-identifying checks are what the steady state must deliver.
- Capital–output ratio and interest rate implied by the calibrated (γ, δ, β) triple.
- Both emerge in an economically sensible range.
- No systematic published reconciliation of achieved-versus-target moments exists for OG-UK 0.3.2.
- Producing one from logged solves is the most valuable next step for this member.
Two dating caveats and two structural caveats from OG-Core apply.
- The official column reports mid-2026 figures.
- The deployed calibration was frozen against slightly earlier vintages.
- The saving ratio is volatile quarter to quarter.
- The fixed β was set against a smoothed level, not the latest print.
- Ability types and lifetime earnings profiles are still the US tax-microdata estimates.
- UK re-estimation of those profiles is open.
About this steptable, notes
| Quantity | Official UK value | Source | Model treatment | Deviation |
|---|---|---|---|---|
| Government debt-to-GDP | 94.9% (June 2026) | ONS · HF6X, vintage 2026-07-26 | imposed via closure target | +0.1pp vs latest ONS (target 95.0%) |
| Household saving ratio | 8.9% (2026Q1, down from 9.6%) | ONS quarterly sector accounts | not targeted — β = 0.965 fixed, not estimated | no model-produced value to compare |
| Potential growth | ≈1.1% / yr | OBR EFO | imposed, gy = 0.011 | ≈0 |
| Depreciation / capital | ≈6–7% (CFC / net stock) | ONS capital stocks bulletin | imposed, δ = 0.065 | within range |
| Labour share of income | ≈0.59–0.60, rising in 2024 | ONS labour-share series / Blue Book 2025 | imposed, γ = 0.35 (OG-Core default) | ≈5pp — the capital share is not yet re-anchored to ONS factor shares |
| Net capital stock | £5.6tn (2024) whole-economy net stock; the model's K maps to the narrower business-capital concept | ONS capital stocks bulletin | emergent — checked, not imposed | concept-adjusted; no published reconciliation |
An imposed target cannot disagree, so a zero deviation above is an anchor, not evidence. Only the emergent row is a check.
The nearest external benchmark
OBR Working Paper 22, Table 5.1: the figures a future run will be judged against, not a match claimed today.
About this stepnotes
1pp basic rate, long run
- GDP per person
- −0.1%
- Labour supply
- −0.2%
- Productivity
- +0.1%
- Status
- future benchmark
Table 5.1 reports a 1pp basic-rate rise in the long run. No OG-UK run has been scored against it yet.
Against the OBR’s own OLG model
Two UK overlapping-generations models, dimension by dimension: OG-UK and the OBR’s (Working Paper No. 22, April 2025).
More detailtable
| Dimension | OG-UK strength | OBR UK OLG strength |
|---|---|---|
| Transition dynamics | Full year-by-year paths | Steady-state only |
| UK institutional detail | Smooth approximations | Explicit bands and rates |
| Within-cohort inequality | Permanent types | Stochastic earnings |
| Production structure | Multi-sector CES | Single-sector Cobb-Douglas |
| Tax instruments | 5 instruments | 3 instruments (more detailed) |
| Fiscal policy levers | G, TR, UBI, Ig | RBI, age-dependent transfers |
| Openness | Open-source, Python | MATLAB, closed toolkit |
| Solve time | Many minutes (not timed) | Under two minutes |
About this stepnotes
OG-UK — the UK calibration of OG-Core, maintained by the Policy Simulation Library — against the OBR UK OLG model of OBR Working Paper No. 22 (Brzezinski, Hantzsche & Watson, April 2025), the closest available external check for a model of this kind.
Limits
| Limit | Detail |
|---|---|
| The worked example’s debt change depends on its denominator and baseline | The worked example divides reform-run aggregates by the reform run’s own GDP (og-model-dashboard #4), so debt in 2030 is 95.38% against a 95.32% baseline: +0.05pp, higher under the tax rise. With baseline GDP as the denominator, as the page used before, the sign flipped. The baseline debt path is now public sector net debt excluding the Bank of England at end-March, from ONS CPOA outturns and the OBR’s November 2025 table 6.13 (#7); it used to be HF6X, which includes the Bank (#6). The calibration target in step 1 is still HF6X. The model’s own numbers come from the 19 March 2026 run and predate OG-UK’s July 2026 labour-tax-rate fix; a rerun is planned (#5). |
| Smoothed tax functions, not statutory bands | A single Gouveia–Strauss function fitted to PolicyEngine-UK output folds income tax and National Insurance together. Reforms changing average liability flow through cleanly; reforms whose mechanism is the kink itself (a new threshold, an allowance taper change) register only as far as the smoothed function shifts. |
| Permanent ability types, no earnings risk | With J = 7 deterministic ability types and no idiosyncratic earnings shock within a type, precautionary savings and earnings-risk-driven inequality are not channels here; the OBR’s own UK OLG model (above) makes the opposite choice. |
| UK as a single entity | Calibration is to UK-wide aggregates, with no England / Scotland / Wales / Northern Ireland breakdown; devolved tax differentials and explicitly regional reforms sit below the model’s granularity. |
| Reforms are step changes | A PolicyEngine reform sets parameter values from a start date, holding thereafter; phased introductions, sunset clauses or year-on-year indexation changes need explicit scripting. |
| Truncated horizon | The default 60-period transition assumes the steady state arrives by year 60; longer-run effects collapse into the steady-state anchor. |
| The dynamic score is a flat overlay | og+microsim applies the steady-state factor flat to one policy year, so no transition path is represented; the distributional incidence of effective-labour changes is reported, not allocated; and the model is real, so it has no price-level effects. |
| Version pin | oguk 0.3.2 pins policyengine-uk==2.88.0. Versions ≥ 2.89 renamed the microdata dataset keys, so mixing them fails with a KeyError at calibration time; the adapter traps it and raises an actionable error. Until the pin lifts, the OLG member scores against an older statute vintage than the microsimulation member — a declared inconsistency, not a hidden one. |
| Gated data | Calibration needs the private enhanced-FRS microdata, so a HUGGING_FACE_TOKEN with access is a hard prerequisite. |
| Thin CI | CI cannot exercise a seventeen-minute solve on every commit, so regressions in solve behaviour are caught by scheduled runs and by hand, not continuously. |
| Steady-state-only in the CLI | The CLI exposes steady-state solves only (Code, step 7): the long-run destination is reachable through the suite, the path there is not. |
| Ability profiles not UK-estimated | Lifetime-ability profiles are inherited rather than re-estimated on UK microdata. On the roadmap. |