Introduction

The Cook Islands entered FY2025/26 on an improving trajectory by several conventional measures: net Crown debt has fallen steadily as a share of GDP, from 54.8% in 2021/22 toward a projected 24.8% by 2026/27, and tourism, the economy's dominant industry, delivered a 16% real GDP rebound in the year to June 2024 and continued growth into 2025. That trajectory was interrupted in December 2025, when New Zealand, the Cook Islands' largest bilateral partner, confirmed a pause of NZD 20 million in core budget support over two years and a further NZD 10.5 million suspension of Public Sector Strengthening funding — NZD 30 million in total, against a total FY2025/26 Budget of roughly NZD 372 million once the resulting Supplementary Budget was accounted for. The Public Accounts Committee's own language is unambiguous: record visitor arrivals and higher-than-forecast VAT collections helped offset part of the shortfall, but fiscal space remains extremely limited, at only NZD 2 million for the year, an amount the Committee itself judged too small to support any major new initiative.

This is a genuinely different dependency structure from anything else examined in this series. Fiji, Solomon Islands and Zambia depend on commodity and resource revenue; Marshall Islands, Kiribati and Nauru depend on trust-fund drawdowns or a single dominant contract; the Cook Islands depends, in this specific instance, on the continuity of bilateral development assistance from a single partner, with its own domestic VAT base — itself driven by an equally concentrated tourism sector reliant on visitor arrivals from New Zealand, Australia and a handful of other source markets — as the only lever available to offset a shortfall when that assistance is paused. A second, structurally separate feature of the Cook Islands' revenue system is its role as an active offshore financial centre: the Financial Services Commission and Financial Services Development Authority regulate an international trust and company sector that has operated for decades and remains internationally significant for asset-protection trust structures, a sector this paper treats as a distinct transparency and integrity question rather than folding it into the domestic tax-compliance models developed for other countries in this series.

The research question this paper takes up has three parts: how much of a bilateral-aid shock of the kind just experienced a domestic VAT windfall can realistically be expected to absorb, and what that implies for how much fiscal buffer the government should hold against the next one; how visitor-arrival data, available well before VAT returns are finalised, can give the Ministry of Finance and Economic Management an earlier read on the government's actual fiscal space than the current quarterly reporting cycle allows; and how the offshore trust and company sector's transparency obligations should be monitored as a standing function, given the sector's international significance and the reputational stakes any lapse would carry.

Current tax issues

2.1 A specific, dated, quantified aid shock

Unlike most fiscal-risk discussion in this series, which addresses a general category of shock (disaster, commodity-price movement, trust-fund return volatility), the Cook Islands case is a single, dated, precisely quantified event: NZD 20 million in paused core budget support and NZD 10.5 million in suspended Public Sector Strengthening funding, confirmed in late 2025 and requiring an early Supplementary Budget tabled in December rather than the usual February timing. This precision is itself an advantage for modelling purposes — the shock's size is known, not estimated.

2.2 A VAT windfall that helps, but does not fully offset

Record visitor arrivals in 2025 drove VAT collections above forecast, and the Public Accounts Committee credited this directly with offsetting part of the shortfall. The operative word is part: the government's own account is that fiscal space after the offset remains at NZD 2 million, described as too small to support any major new initiative, and projected to decline further in subsequent years even as it ticks up slightly in the near term. The relationship between the aid shock and the VAT offset has not, to this paper's knowledge, been formally modelled; sections 3.1 and 4.3 below do so directly.

2.3 A tourism-concentrated revenue base

VAT collections depend on visitor arrivals, which are themselves concentrated in a small number of source markets — principally New Zealand, Australia, Europe and other Asia-Pacific origins. New building approvals for residential and short-term accommodation rose 23% in FY2025 and newly registered vehicles, a discretionary-spending indicator, rose 12.2%, both signs of a tourism sector still expanding briskly in the near term even as the Asian Development Bank projects growth moderating from 15.1% (2024) to 2.7% (2026). A revenue base tied this closely to tourism inherits tourism's own volatility, on top of whatever aid volatility the government separately faces.

2.4 A debt position improving, which somewhat cushions the shock

Net Crown debt has fallen from 54.8% of GDP in 2021/22 toward a projected 24.8% by 2026/27, giving the government more borrowing headroom than it would have had entering this shock a few years earlier. This is a genuinely favourable structural feature relative to several other countries in this series, and this paper's Model A treats it as part of the available response, not only the VAT offset.

2.5 An offshore financial-services sector with its own transparency stakes

The Cook Islands hosts an internationally significant asset-protection trust and international company sector, regulated domestically by the Financial Services Commission with policy input from the Financial Services Development Authority. This sector is structurally distinct from the domestic VAT and aid-dependency questions in sections 2.1-2.3: its revenue contribution to government is smaller, but its transparency obligations under international information-exchange standards carry reputational and market-access consequences disproportionate to that direct revenue contribution, making it worth a dedicated monitoring model rather than being folded into general domestic compliance work.

Models

3.1 Model A: aid-shock and VAT-offset fiscal space adequacy

Model A defines fiscal space explicitly as what remains after an aid shock and its partial offset, rather than treating the two as separately reported facts:

\text{FiscalSpace}_t = \text{BudgetedSpace}_t - \text{AidShock}_t + \text{VATOffset}_t - \text{DebtHeadroomUsed}_t

where AidShock_t is the confirmed reduction in bilateral support, VATOffset_t is the excess of actual over forecast VAT collections attributable to the tourism windfall, and DebtHeadroomUsed_t is any borrowing drawn to further cushion the gap, made explicit rather than absorbed silently into the debt trajectory. The offset ratio, the share of a given aid shock a VAT windfall of observed scale can cover, is the single most policy-relevant output:

\text{OffsetRatio}_t = \frac{\text{VATOffset}_t}{\text{AidShock}_t}

An OffsetRatio well below 1 (as this paper's illustrative results in section 4.3 show) is a quantified version of the Public Accounts Committee's own qualitative warning, and directly informs how large a discretionary reserve the government should hold against a repeat, sized to the residual, unoffset share of a shock of comparable magnitude:

\text{TargetReserve} = (1 - \overline{\text{OffsetRatio}}) \times \text{TypicalAidShock}

3.2 Model B: visitor-arrival-linked VAT nowcasting

Model B exploits the fact that visitor arrivals, reported at high frequency by source market, are observable well before quarterly VAT returns are finalised, using a standard nowcasting regression linking contemporaneous and lagged arrivals to VAT receipts:

\widehat{VAT}_t = \alpha + \sum_{k=0}^{K} \beta_k \cdot \text{Arrivals}_{t-k} + \gamma \cdot \text{AvgSpend}_t + \delta \cdot \text{SeasonalFactor}_t + u_t

where Arrivals are broken out by source market (New Zealand, Australia, Europe, other Asia-Pacific) since spending patterns and VAT-relevant consumption differ by origin, AvgSpend is a per-visitor spending proxy from tourism survey data, and SeasonalFactor captures the Cook Islands' own visitor-season pattern. The nowcast is re-estimated monthly as new arrivals data lands, giving the Ministry of Finance and Economic Management a revenue signal weeks to months ahead of the formal quarterly cycle — directly relevant given how thin the FiscalSpace_t buffer from Model A already is, and how much earlier warning of a VAT shortfall (or windfall) would be worth in that context.

3.3 Model C: offshore trust and financial services transparency

Model C is a compliance-monitoring rather than revenue-costing model, tracking the international trust and company sector's adherence to information-exchange and beneficial-ownership obligations:

\text{TransparencyScore}_e = \theta_1 \cdot \mathbb{1}[\text{beneficial ownership on file and current}] + \theta_2 \cdot \mathbb{1}[\text{timely CRS/AEOI reporting}] + \theta_3 \cdot \mathbb{1}[\text{no adverse FATF/OECD findings in period}]

for each regulated entity e (trust, international company, or the licensed trustee/agent administering it), aggregated to a sector-level dashboard the Financial Services Commission can monitor continuously rather than only at periodic international peer-review intervals. A declining aggregate TransparencyScore over time is an early-warning signal worth escalating before, not after, an international standard-setting body's own review identifies a gap.

Algorithms and illustrative results

4.1 Algorithm 1: fiscal-space adequacy tracker

QUARTERLY:
  aid_shock       <- confirmed reduction in bilateral support vs prior commitment
  vat_actual      <- actual VAT collections for the period
  vat_forecast    <- pre-shock forecast VAT collections
  vat_offset      <- max(0, vat_actual - vat_forecast)
  debt_headroom   <- any additional borrowing drawn this period, explicitly logged
  fiscal_space    <- budgeted_space - aid_shock + vat_offset - debt_headroom
  offset_ratio    <- vat_offset / aid_shock if aid_shock > 0 else null
  PUBLISH fiscal_space, offset_ratio to Cabinet and the Public Accounts Committee
ANNUALLY:
  target_reserve  <- (1 - trailing_avg(offset_ratio)) * typical_aid_shock_size
  COMPARE current reserve balance against target_reserve; flag shortfall

4.2 Algorithm 2: visitor-arrival VAT nowcast

MONTHLY:
  arrivals_by_market <- latest visitor arrival counts, by source market
  FIT (or update) regression: VAT_hat = alpha + sum(beta_k * arrivals_{t-k}) +
      gamma * avg_spend_t + delta * seasonal_factor_t
  PUBLISH VAT_hat for the current and next quarter, with a confidence interval
  WHEN actual VAT returns land: compare VAT_hat to VAT_actual, log forecast error
  QUARTERLY: recalibrate beta, gamma, delta using the accumulated forecast-error
    history (simple online regression update)

4.3 Illustrative results: the offset ratio

Using the confirmed aid shock (NZD 30 million) and an illustrative VAT windfall consistent with the order of magnitude implied by the Public Accounts Committee's account (VAT collections above forecast, though the Committee did not publish a precise figure this paper could cite directly), an illustrative VAT offset of NZD 6.5 million against the NZD 30 million shock implies an OffsetRatio of about 0.22 — the tourism windfall covering roughly a fifth of the aid shortfall. The resulting target reserve, sized to the unoffset 78% share, comes to roughly NZD 23.5 million for a future shock of similar scale. This is the concrete number behind the Public Accounts Committee's own qualitative language that remaining fiscal space is too thin for new initiatives: even a favourable tourism year did not come close to fully absorbing the shock, and a less favourable tourism year, arriving at the same time as a future aid pause, would leave materially less cushion still.

4.4 Illustrative arithmetic: nowcast value

If the visitor-arrival nowcast in Model B gives the Ministry of Finance a VAT signal even six to eight weeks ahead of the current quarterly reporting cycle, that lead time is worth more to the Cook Islands specifically than to a country with a larger fiscal buffer, precisely because the NZD 2 million remaining space documented in section 2.2 leaves almost no room to absorb a forecasting surprise discovered late. Early warning does not create fiscal space, but it converts a fixed amount of space into a more usable one, since decisions can be made before commitments are locked in rather than after a shortfall is already realised.

Solutions, software, data science and AI models

5.1 Issue-to-solution map

5.2 Model catalogue

5.3 Reference architecture

flowchart TD
  A[Sources: MFEM budget and revenue records, Cook Islands Statistics Office arrivals data, aid-partner commitment records, Financial Services Commission trust/company register] --> B[Ingestion and data quality]
  B --> C[Warehouse: bitemporal]
  C --> D[Fiscal-space adequacy tracker]
  D --> E[MFEM and Public Accounts Committee dashboard]
  C --> F[Visitor-arrival VAT nowcast]
  F --> D
  C --> G[Trust-sector transparency scoring]
  G --> H[Financial Services Commission dashboard]

The nowcast feeds directly into the fiscal-space tracker, since VATOffset_t in Model A is exactly the quantity Model B is designed to forecast ahead of confirmation.

5.4 Software stack

5.5 Governance of AI use

• The nowcast is advisory. Model B's VAT signal informs budget planning; it does not replace the formal quarterly VAT return as the basis for actual fiscal accounting.
• Aid-shock accounting is transparent to the Public Accounts Committee by design. Model A's outputs are built to be published in the same reporting channel the Committee already uses, not held internally within MFEM.
• Trust-sector scoring supports, not replaces, international peer review. Model C is a domestic early-warning tool; it does not substitute for the Financial Action Task Force, OECD Global Forum, or other international assessment processes the sector is independently subject to.
• No automated funding or licensing decisions. A low TransparencyScore for a given entity triggers Financial Services Commission review, not an automatic licence action.

5.6 Team and indicative effort

A small team within the Ministry of Finance and Economic Management — one analyst covering the fiscal-space and nowcasting models, working with existing MFEM and Statistics Office data — plus a separate, smaller compliance-monitoring function within the Financial Services Commission for Model C, given its distinct regulatory audience, is proportionate to the Cook Islands' scale. The nowcasting model in particular could be built relatively quickly given the Cook Islands' existing high-quality tourism-statistics infrastructure. Cost and staffing figures are estimates.

Limitations

• The VAT offset figure used in section 4.3 is illustrative; this paper does not have access to the precise higher-than-forecast VAT collection amount the Public Accounts Committee referenced only in qualitative terms, and MFEM's own budget documents should supply the actual figure before the target-reserve calculation informs a real decision.
• The visitor-arrival nowcast (Model B) assumes a reasonably stable relationship between arrivals and VAT receipts; a material change in visitor spending patterns, exchange rates, or the VAT base itself would require recalibration beyond the routine quarterly update Algorithm 2 describes.
• This paper does not have visibility into the specific terms of New Zealand's funding pause beyond public reporting, including whether or when the paused funding might resume; Model A treats the shock as realised and past, not as an ongoing negotiation this paper is positioned to comment on.
• The trust-sector transparency model (Model C) is necessarily limited by this paper's lack of access to the Financial Services Commission's actual register and compliance data; the model specifies a monitoring structure, not a current assessment of the sector's actual transparency performance.
• This paper treats aid dependency and tourism-VAT dependency as the central fiscal-risk story for the Cook Islands at this specific moment; a longer-run treatment would also need to address climate vulnerability and the outer-islands/Rarotonga economic divide, which are outside this paper's scope.

References

• Public Accounts Committee, Cook Islands Parliament. Report on the Review of the Supplementary Budget 2025/26, as reported in "New Zealand funding pause forces Cook Islands into early budget overhaul", Pacific Islands News Association, 10 December 2025.
• Cook Islands Government, Ministry of Finance and Economic Management. Appropriation Bill 2025 / Budget Estimates 2025/26; 2024/25 Half-Year Economic and Fiscal Update; 2024/25 Supplementary Budget Speech.
• Asian Development Bank. Pacific Economic Monitor / South Pacific Economies, Asian Development Outlook, April 2026.
• Asian Development Bank and Government of New Zealand. Supporting Sustainable Recovery Program, Cook Islands, December 2021.
• Cook Islands Financial Services Commission and Financial Services Development Authority. Public guidance on the international trust and company sector.
• Methods: as elsewhere in this series — Hutton, E. (2017), The Revenue Administration Gap Analysis Program, IMF Technical Notes and Manuals 17/04; standard nowcasting methodology as applied in central-bank and treasury GDP/revenue nowcasting practice. Method citations are from memory; verify before publication.