Introduction

Tuvalu closes this research series as its most structurally distinctive case. With a population under 11,000, no central bank and the Australian dollar as legal tender, fiscal policy is, in the Asian Development Bank's own words, Tuvalu's only tool for macroeconomic management — there is no monetary lever to pull. The public sector accounts for roughly two-thirds of GDP, and government revenue depends on a combination of sources with no close parallel elsewhere in this series: fishing-licence fees, which reached 40% of GDP in a single exceptional year (2014); and licensing royalties from the commercial use of the ".tv" country-code internet domain, sold internationally for its branding value to television, streaming and video-related businesses far beyond Tuvalu's own borders. Both revenue streams are episodic, contract-renewal-driven, and largely outside Tuvalu's own control in terms of underlying market demand.

To manage the volatility this creates, Tuvalu established the Tuvalu Trust Fund in 1987 — contemporaneous with, though structurally distinct from, Kiribati's older Revenue Equalization Reserve Fund examined elsewhere in this series. The TTF's statutory design is itself distinctive: rather than a discretionary or return-based withdrawal rule, the TTF automatically distributes any value above its own maintained real level to a secondary buffer, the Consolidated Investment Fund, which in turn provides additional revenue directly to the government budget. This two-tier, rules-based architecture — a value-preservation fund feeding an automatic overflow buffer — has no direct equivalent among the trust-fund structures this series has examined for Kiribati, Marshall Islands, Nauru or FSM, each of which uses a different underlying mechanism.

Tuvalu's expenditure side carries its own distinctive shocks. IMF reporting documents that purchases of goods and services rose from 19% of GDP in 2019 to 31% in 2022, driven primarily by a once-in-50-years land-lease renegotiation with the customary landowners on whose land Tuvalu's public buildings sit — an expenditure category tied to land-tenure arrangements specific to Tuvalu's own customary law, not a recurring or easily forecastable budget line. Over the same period, spending on the Tuvalu Overseas Medical Referral Scheme and scholarships rose from 12.6% to 16.6% of GDP, reflecting Tuvalu's reliance on referring patients and students abroad given its own extremely limited domestic specialist capacity.

The research question this paper takes up has three parts: how the .tv domain and fishing-licence revenue streams, both episodic and renewal-cycle-driven, should be modelled jointly rather than as two unrelated volatile sources; how well the TTF/CIF's specific automatic-distribution rule actually stabilises the budget once the underlying revenue volatility feeding it is made explicit; and how land-lease renegotiation and TOMRS/scholarship spending, both large and lumpy but arguably forecastable given known renegotiation cycles and referral-demand trends, should be anticipated in forward budget planning rather than treated as unpredictable shocks each time they recur. As throughout this series, the organising identity for the domestic tax-side question applies to what remains a comparatively small formal tax base once these larger, non-tax revenue sources are set aside:

R^{pot}_{domestic tax,t} - R_{domestic tax,t} = \underbrace{\left(R^{pot}_{domestic tax,t} - R^{law}_{domestic tax,t}\right)}_{\text{policy gap}} + \underbrace{\left(R^{law}_{domestic tax,t} - R_{domestic tax,t}\right)}_{\text{compliance gap}}

Current tax issues

2.1 A domain name as sovereign revenue

The .tv domain's commercial value derives entirely from its resemblance to the English abbreviation for television, a branding coincidence Tuvalu has monetised through licensing arrangements with international registry operators since the late 1990s. This revenue stream depends on global demand for streaming, broadcast and video-content branding, a market entirely outside Tuvalu's control, and on periodic renegotiation of the underlying registry agreement, whose terms determine Tuvalu's ongoing royalty share.

2.2 Fishing-licence fees at extraordinary, variable scale

Fishing-licence revenue reached 40% of GDP in 2014, among the highest such concentrations documented anywhere in this series, and remains, per the IMF's own recent language, a source of fiscal and external-balance volatility alongside grants. Tuvalu's exclusive economic zone, like Kiribati's, is large relative to its land area, making tuna licensing a natural but inherently variable revenue base.

2.3 A two-tier automatic trust-fund rule

The TTF/CIF structure is Tuvalu's own answer to the volatility in sections 2.1-2.2: rather than a discretionary or return-based withdrawal rule of the kind this series has documented for Kiribati's RERF, the TTF automatically distributes value above its own maintained real level to the CIF, a secondary buffer that then supplies additional budget revenue directly. This rule's stabilising power depends on how the underlying .tv and fishing-licence revenue behaves, a dependency this paper's Model B makes explicit for the first time in a formal sense, to this paper's knowledge.

2.4 A once-in-50-years expenditure shock, now realised

The land-lease renegotiation that pushed goods-and-services spending from 19% to 31% of GDP between 2019 and 2022 reflects Tuvalu's customary land-tenure arrangements, under which government buildings sit on land whose lease terms are periodically renegotiated with landowners on a multi-decade cycle. Because this cycle is long but not unknown, it is, in principle, forecastable well in advance, unlike an unpredictable external shock.

2.5 Growing reliance on overseas medical and education referral

TOMRS and scholarship spending rising from 12.6% to 16.6% of GDP reflects Tuvalu's limited domestic specialist healthcare and tertiary education capacity, an inherent feature of its extremely small population rather than a policy failure, but one whose growth trajectory is worth tracking and forecasting rather than absorbing as a recurring surprise.

2.6 A public-sector-dominated economy with no monetary policy lever

With the public sector accounting for roughly two-thirds of GDP and no central bank or independent currency, fiscal policy carries the entire burden of macroeconomic management in Tuvalu, a structural constraint this paper's models are designed to support rather than one any of them can relax directly.

Models

3.1 Model A: .tv domain and fishing-licence renewal-cycle revenue model

Model A treats both revenue streams as renewal-cycle-driven rather than smoothly recurring, since each depends on a periodically renegotiated contract whose terms, not just annual volume, determine the revenue realised:

DomainRevenue_t = \rho_t \cdot N_t, \qquad FishingRevenue_t = \sum_v Q_{v,t} \cdot Price_t

where rho_t is the effective royalty rate under the current registry agreement (a step function that changes only at renegotiation dates), N_t the volume of .tv registrations, Q_{v,t} licensed fishing days by vessel and Price_t the prevailing per-day licence price under the current Vessel Day Scheme terms. The joint volatility this creates:

\text{CombinedRevenue}_t = DomainRevenue_t + FishingRevenue_t, \qquad \sigma^2_{\text{Combined}} = \sigma^2_{Domain} + \sigma^2_{Fishing} + 2\text{Cov}(Domain,Fishing)

is the input this paper's Model B needs to test the TTF/CIF rule against, since a low or negative covariance between the two streams (domain demand driven by global media trends, fishing revenue driven by tuna markets and licensing negotiations) would mean the combined revenue is more stable than either stream alone, a diversification benefit worth quantifying explicitly rather than assuming.

3.2 Model B: TTF/CIF automatic-distribution adequacy

Model B formalises Tuvalu's specific two-tier rule:

TTF_{t+1} = \max\!\left(TTF^{maintained}_t,\; TTF_t(1+r)\right), \qquad Overflow_t = \max\!\left(0,\; TTF_t(1+r) - TTF^{maintained}_t\right)

CIF_{t+1} = CIF_t + Overflow_t - Withdrawal_t, \qquad Withdrawal_t = \min\!\left(CIF_t,\; \text{BudgetSupportNeeded}_t\right)

where TTF-maintained is the fund's statutorily protected real value, Overflow_t the amount automatically passed to the CIF, and Withdrawal_t the amount the CIF actually supplies to the budget in a given year, capped by the CIF's own available balance. Because Overflow_t depends on realised TTF returns, itself only loosely connected to Model A's CombinedRevenue_t volatility (the TTF's own investment portfolio, not Tuvalu's domain/fishing revenue directly, drives Overflow_t), the model's central question is whether the CIF buffer, fed by an overflow rule disconnected from the revenue volatility it is meant to smooth, is actually sized appropriately for that purpose — a structural question this paper's illustrative results in section 4.3 test directly.

3.3 Model C: land-lease and TOMRS/scholarship lumpy-expenditure forecasting

Model C treats both expenditure categories as forecastable, not merely large, given their known underlying drivers:

\text{LandLeaseCost}_t = \mathbb{1}[\text{renegotiation cycle year}] \times \text{EstimatedRenegotiatedValue}

with the renegotiation cycle's timing estimated from the historical multi-decade interval (illustratively, the documented "once-in-50-years" cycle) and the estimated renegotiated value updated as the next cycle approaches and preliminary landowner discussions provide better information. TOMRS and scholarship spending is modelled as a trend with demographic and health-referral demand drivers:

\text{TOMRS}_t = \text{TOMRS}_{t-1}(1+g_t), \qquad g_t = f(\text{referral demand growth}, \text{overseas care cost inflation})

allowing the Ministry of Finance to build both categories into medium-term expenditure ceilings as anticipated, planned-for cost growth rather than as recurring budget surprises each time they materialise.

Algorithms and illustrative results

4.1 Algorithm 1: joint domain/fishing revenue and TTF/CIF simulation

ANNUALLY:
  domain_revenue_t <- rho_t * registrations_t   # rho_t updates only at renegotiation dates
  fishing_revenue_t <- sum over vessels of (licensed_days * price_t)
  combined_revenue_t <- domain_revenue_t + fishing_revenue_t
  RECORD combined_revenue_t, its volatility, and the covariance between the
    two component series for use in the TTF/CIF stress test below

FOR the TTF/CIF simulation, each simulated year:
  ttf_maintained_t <- ttf_maintained_{t-1} * (1 + inflation)
  r_t <- drawn TTF investment return for the year
  grown_t <- ttf_{t-1} * (1 + r_t)
  overflow_t <- max(0, grown_t - ttf_maintained_t)
  ttf_t <- max(ttf_maintained_t, grown_t) - overflow_t
  cif_t <- cif_{t-1} + overflow_t
  withdrawal_t <- min(cif_t, budget_support_target_t)
  cif_t <- cif_t - withdrawal_t
  RECORD cif_t; FLAG if cif_t reaches zero (buffer exhausted for the year)
RUN across many simulated return paths; REPORT the distribution of years in
  which the CIF buffer is exhausted, not just a single central-case path

4.2 Algorithm 2: lumpy-expenditure forecast

FOR the land-lease cycle:
  years_since_last_renegotiation <- current_year - last_renegotiation_year
  IF years_since_last_renegotiation approaches the historical cycle length:
    BEGIN preliminary landowner engagement and cost estimation
    INCORPORATE an estimated LandLeaseCost into the medium-term expenditure
      ceiling for the projected renegotiation year, well before it arrives
FOR TOMRS and scholarships:
  FIT a trend model on historical spending as a share of GDP
  PROJECT forward using estimated referral-demand growth and overseas-cost
    inflation
  INCORPORATE the projection into medium-term expenditure ceilings alongside
    the land-lease estimate

4.3 Illustrative results: TTF/CIF buffer adequacy under stress

Using illustrative inputs (a TTF starting value on the order of USD 130 million, a CIF starting value of USD 20 million, a maintained-value growth rate tied to 3% inflation, an expected TTF real return of 4.5% with 10% volatility, and an illustrative annual budget-support draw of USD 6 million), simulated paths vary considerably depending on the realised sequence of TTF returns. In a favourable return sequence, the CIF buffer grows steadily and never approaches exhaustion. In a less favourable but entirely plausible sequence, the CIF buffer is fully exhausted by around year 10 and remains constrained for a further year or two before a run of stronger TTF returns allows the overflow mechanism to rebuild it. This is the concrete evidence behind this paper's central methodological point: because Overflow_t depends on TTF investment returns rather than directly on the .tv/fishing revenue volatility Model A characterises, the CIF buffer's adequacy is not automatically matched to the volatility it is meant to smooth, and a run of poor TTF returns can leave the buffer empty at precisely the time volatile domain or fishing revenue might most need supplementing.

4.4 Illustrative arithmetic: the land-lease and TOMRS combined burden

If goods-and-services spending's rise from 19% to 31% of GDP (2019-2022) is attributed substantially to the land-lease renegotiation, and TOMRS/scholarship spending's rise from 12.6% to 16.6% of GDP over the same period is added, the two categories together account for a combined increase on the order of 16 percentage points of GDP in current-expenditure share over three years — a scale of lumpy, quasi-structural expenditure growth this paper argues should be forecast using Model C's approach well ahead of the next land-lease cycle, rather than absorbed as a repeat surprise when it recurs decades from now.

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: .tv registry agreement terms, fishing licence/VDS records, TTF and CIF financial statements, land-lease and TOMRS/scholarship spending records] --> B[Ingestion and data quality]
  B --> C[Warehouse: bitemporal]
  C --> D[Joint domain/fishing revenue model]
  D --> E[TTF/CIF stress simulation]
  E --> F[Ministry of Finance fiscal-risk dashboard]
  C --> G[Lumpy-expenditure forecast]
  G --> F

The TTF/CIF simulation is run periodically (e.g. annually, or ahead of major fund-policy decisions) rather than continuously, given its reliance on longer-horizon return assumptions.

5.4 Software stack

5.5 Governance of AI use

• The TTF/CIF stress test informs Trust Fund Board and Cabinet decisions; it does not automate them. The Fund's statutory rule remains as legislated; Model B's role is to show how that rule performs under a realistic range of return outcomes, not to change it unilaterally.
• Land-lease cost estimates are prepared collaboratively with landowners, not imposed. Model C's forecasting role is to inform the government's own budget planning; actual renegotiation remains a matter for direct engagement with customary landowners.
• Domain-revenue forecasts respect commercial confidentiality. Registry agreement terms may be commercially sensitive; Model A's outputs should be handled with the same confidentiality as the underlying agreements themselves.
• All three models' outputs are reported together, not separately. Given how interconnected Tuvalu's revenue and buffer mechanisms are, this paper recommends a single consolidated fiscal-risk report rather than three independent ones that obscure the connections Model B in particular depends on.

5.6 Team and indicative effort

Given Tuvalu's very small scale, a single analyst within the Ministry of Finance, working with the Trust Fund Board secretariat for the TTF/CIF-specific work, is proportionate to build and maintain all three models. The domain-revenue component of Model A may benefit from external actuarial or financial-modelling support given the specialised nature of internet-domain licensing agreements. Cost and staffing figures are estimates.

Limitations

• The TTF/CIF simulation parameters in section 4.3 (fund sizes, return assumptions, budget-draw target) are illustrative; Tuvalu's Trust Fund Board holds the actual financial statements and statutory rule parameters that should replace these inputs before the model informs a real Fund-policy decision.
• The .tv domain revenue model (Model A) treats the royalty rate as a step function changing only at renegotiation dates; this paper does not have access to the actual registry agreement terms, which are likely commercially confidential, and the model's structure should be validated against those terms directly.
• The land-lease renegotiation cycle length (illustratively, "once in 50 years") is drawn from a single documented instance; a genuinely robust forecast would need a longer historical record of renegotiation cycles than this paper has been able to establish from public sources.
• This paper's combined-burden arithmetic in section 4.4 attributes expenditure growth substantially to the two named categories based on IMF reporting; other contributing factors to the 2019-2022 expenditure rise are possible and this paper does not claim to have fully decomposed the change.
• As the final country paper in this series, this paper's scope is necessarily bounded by the same synthetic-demonstration, non-empirical-validation caveat that applies throughout: it establishes the models' logical structure and illustrates their mechanics, not validated real-world performance, which only Tuvalu's own Ministry of Finance, Trust Fund Board and Customs/Revenue functions can establish through actual deployment.

References

• International Monetary Fund. IMF Executive Board Concludes 2025 Article IV Consultation with Tuvalu. Press Release 25/293, 12 September 2025.
• International Monetary Fund. Tuvalu: Staff Concluding Statement of the 2025 Article IV Mission. 27 May 2025.
• International Monetary Fund. Tuvalu: Staff Concluding Statement of the 2023 Article IV Mission. 25 April 2023.
• Asian Development Bank. Tuvalu: Fiscal Sustainability and the Tuvalu Trust Fund, ADB project documentation.
• Asian Development Bank. Country Operations Business Plan: Tuvalu, 2016-2018, Linked Document.
• Government of Tuvalu. Tuvalu Trust Fund and Consolidated Investment Fund, statutory framework.
• Parties to the Nauru Agreement (PNA). Vessel Day Scheme documentation.
• Methods: as elsewhere in this series — Hutton, E. (2017), The Revenue Administration Gap Analysis Program, IMF Technical Notes and Manuals 17/04. Method citations are from memory; verify before publication.