New Zealand has generated roughly the same ~40–44 TWh (terawatt-hours — the standard unit for a country's yearly electricity output) for twenty-five years — almost entirely renewable, yet with no productive surplus in any year. So Copolicy built an agent-based model where supply is not assumed but emerges from the decisions of generators, regulators, banks, factories and households. The honest result: the frictionless prize is real, but the chicken-and-egg investment lock only partly breaks — even under the full policy package, the achievable abundance is about half the frictionless ceiling. We show the number, the method, and the gap.
A national agent-based model with six agent types and endogenous supply (supply that emerges from agents' own decisions, not simply assumed), back-tested to return Band A (no productive surplus) for every year 2000–2024 — the historically-correct answer. The v3 ceiling model shows the prize: an Abundance supply path reaches ~$89.8B of surplus-GDP by 2050 (Band D). But the ABM asks whether that path emerges on its own — and it does not, fully. Anchor contracts (government as first mover, agreeing long-term power deals with a large buyer that absorbs the surplus) break the chicken-and-egg lock partly: the achievable endogenous outcome is Band C / ~$42B by 2050 — about half the frictionless ceiling, delayed five-to-ten years. These are modelled-scenario projections, clearly not bankable (not yet something a bank or investor could finance and build today).
The predecessor model was an accounting identity: hand it a supply number and a population, and it returns how much surplus is left and how much GDP that surplus could create. It assumes the electricity exists. The Abundance ABM does not. Supply is whatever results when simulated power-station investors, regulators, banks, factories and households each make their own year-by-year decisions. If nobody builds, supply does not grow — no matter how desirable abundance would be.
That lets the model ask the one question the accounting identity cannot: if Cabinet declares an Abundance pathway, halves consent times, and underwrites a handful of anchor industrial contracts — does ~100 TWh of generation by 2035 actually emerge, or does the chicken-and-egg investment problem persist? A generator will not build without a creditworthy buyer; an energy-intensive firm will not commit without cheap, certain, abundant power already scheduled. Each waits for the other. Supply stays flat. That is the lock.
This is the honest centre of the case. The frictionless ceiling assumes the supply path simply happens. The achievable figure is what the same model produces when supply has to be built by agents responding to incentives. The gap between them — roughly half — is the cost of the chicken-and-egg lock: anchor contracts break it, but only partly, and the trajectory arrives five-to-ten years later than the frictionless path implies.
The model advances one simulated year at a time, with six classes of agent making real decisions. Generators invest only when projected supply falls short and a project's discounted cashflow clears its margin. The regulator clears a consent queue on policy-set timing and approval rates. The capital market sets the discount rate generators discount at. Industrial firms enter only when there is genuine headroom and their return clears a hurdle. Households carry demand. The grid runs last each year — aggregating supply, clearing a merit-order market, running storage and curtailment (spilling surplus power that can't be used or stored), then pouring usable energy through the tier waterfall.
The waterfall is preserved verbatim from the predecessor: usable energy is poured into an essential floor, then a wellbeing increment, then a resilience reserve — and whatever remains is Tier-3 productive surplus. Surplus-per-capita sets the Band (A–E), and surplus-GDP is that surplus multiplied by the value-weighted mix of industries that consume it.
| Agent | Real decision rule | Why it matters |
|---|---|---|
| Generator | build if NPV clears margin T2 | Won't commit capacity without a creditworthy buyer locked in — the supply half of the deadlock. |
| Industrial firm | enter if headroom + hurdle met T2 | Won't commit without cheap, certain power already scheduled — the demand half of the deadlock. |
| Regulator (RMA) | consent time · approval rate T2 | The consent-reform lever: baseline 7.0 yr / 55%, reform 3.0 yr / 75–85%. Sets build latency. |
| Capital market | discount rate − 50 bp if guaranteed T2 | Cheaper capital raises every NPV — but de-risking a plant nobody has agreed to buy from does not manufacture demand. |
| Grid (Transpower) | merit order → storage → waterfall T2 | Runs last each year so the Band reflects every other agent's decisions; holds the v3 tier allocator verbatim. |
| Anchor contract | pre-committed · hurdle −10 pp the lever | Government (via NZTE) signs the first off-take, manufacturing the demand that gives generators a reason to build. |
The non-negotiable calibration test: run on the baseline scenario from 2000 to 2024 with endogenous supply, the model must classify every one of the 25 years as Band A — no productive surplus. That is the historically-correct answer; NZ had no electricity surplus in any of those years. The model passes, tracking the MBIE (NZ's Ministry of Business, Innovation and Employment) historical band and landing within tolerance of the 2024 anchors (generation, the residential/industrial split, the wholesale spot price). If it had produced a surplus, its investment parameters would be miscalibrated.
With the back-test secured, the model tests the policy ladder — and the ordering of levers is the central result: baseline (flat, Band A forever) < consent reform alone ≈ consent reform plus cheaper capital < the full package with anchor contracts. Faster paperwork builds plant faster but without buyers; a sovereign guarantee de-risks a plant nobody has agreed to buy from. Only the anchor contracts — the government acting as first mover — qualitatively break the lock.
The honest reading: even under the full policy package, NZ cannot reach abundance by 2035 — the build-and-consent physics will not allow it (first new plant only lands ~2031). The achievable outcome is Band B by ~2040 and Band C by 2050: materially below the frictionless ceiling, but a structural break from a status quo that stays flat at ~48 TWh forever.
The model does not recommend a policy position; it reframes the question. If New Zealand intends to be a country where renewable abundance creates new industrial value, the volume needed is roughly twice what current official scenarios deliver — and the composition of new demand matters as much as the volume. Food processing emerges as the single largest sectoral sensitivity in the surplus-GDP decomposition: a ±20% swing on that one figure moves the surplus-GDP ceiling by roughly ±$12.6B. The number is more sensitive to what the surplus is for than to most physical parameters.
These are modelled-scenario projections, not forecasts and not bankable. They are single-seed point estimates from a model with explicit, stated limitations: the household electrification dynamics are still a stub in the documented build, the capital-market logic is simplified, the grid is single-node (no inter-island HVDC constraint — the high-voltage cable link between NZ's two islands), and the time step is annual (no hourly dry-year hydrology — the risk that a low-rainfall year cuts hydropower output). The surplus-GDP figures are scale indicators, calibrated order-of-magnitude — explicitly not investment-grade. We surface the gap between the frictionless ceiling and the achievable outcome precisely so the prize is never mistaken for a guarantee.
The ~$89.8B ceiling and the ~$42B achievable figure are stated openly, each with its provenance (Copolicy NZ-Abundance ABM) and the caveat that they are surplus-GDP scenarios, not bankable returns.
About half the frictionless prize is lost to the chicken-and-egg lock. Anchor contracts break it partly; nothing breaks it fully by 2035. The honest number is the lower one — and the delay.
Food processing dominates the surplus-GDP sensitivity. What the surplus is used for moves the headline more than most physical inputs — so the mix is a first-order policy choice, not a detail.
Single-seed, calibrated order-of-magnitude, with a stub household model and an annual time step. We name the limitations rather than dress a scenario as a forecast or a bankable case.
The NZ case took the first calibration of the framework. Subsequent jurisdictions inherit the universal layer and re-instantiate five cores — now 12–24 weeks to first instance. The full read, including the gated instrument and finance layers, is available under engagement.
Commissioned by national governments, DFIs (development finance institutions), or anchor investors. Fixed-fee scoping precedes pricing.