When Arrivals Recover
but Paid Nights Do Not

01 / The observed recovery

Three measures.
Three different answers.

Saint Lucia’s 2025 arrival count was above its pre-pandemic level. Published paid nights were lower. Yet real production in accommodation and food services was higher.

Read the boundaries with the numbers. The 2025 results are preliminary, and night levels are rounded. Accommodation and food services include activity serving residents and exclude tourism activity in other industries. This is not total tourism value added or income reaching residents. Source: the paper, §§3–4 and Appendix A; Economic and Social Review 2025, Tables 2 and 5.

The comparison year matters. Arrivals, paid nights and real industry value added all declined between 2024 and 2025. Recovery from the pre-pandemic level and a recent slowdown can coexist.

02 / The accounting question

A visitor count is the start
of an income question.

03 / What the counts cannot settle

The same published path
can hide opposite income outcomes.

In a constructed example, arrivals rise 10%, published nights fall 10% and household prices rise 10%. Spending per true night and income per visitor dollar stay unchanged. Only the reporting relationship differs between the two cases.

These are synthetic examples from §6.2, not estimates of Saint Lucia. They use a 10% night decline, distinct from the observed annual decline of about 13.6% above. They show why the direction of resident-income change cannot be read from arrivals and published nights alone.

The paper derives transparent conditional bounds when coverage, spending yield, prices and resident-income shares are uncertain. The bounds depend on stated assumptions. They are not automatically confidence intervals, and they are not a measured national income range.

04 / A result that changes measurement priorities

Sometimes the spending total
already answers the nights question.

If reliable spending is known for exactly the same visitor group, valuation and period, its total already combines true nights with spending per true night.

Once that product is known, treating coverage and spending per night as independent unknowns can create avoidable uncertainty in the income comparison. The next useful measurement may instead concern production costs, resident earnings, ownership income or prices.

The conditions are strict. A spending total for all stay-over visitors cannot simply replace spending for the paid-stayer group. Different timing or booking valuations can also break the link. Information on the accommodation frame may remain useful through shared survey or accounting constraints. Saint Lucia’s published total expenditure is therefore not automatically treated as a matched paid-segment measure. See §6.3.

Collect the missing observation that can change the answer.

The executed survey experiments also distinguish precision from representativeness. More responses can reduce random error while still missing the same visitor groups. The paper’s comparison of measurement options illustrates how to target a stated decision; its costs are hypothetical, not a Saint Lucia budget estimate.

05 / Contribution and limits

An audited case.
An executable measurement protocol.

The contribution combines a reconciled Saint Lucia record with established decomposition methods, conditional income bounds, executed coverage and survey-selection experiments, and a finite comparison of measurement choices. It identifies both defensible conclusions and the observations that could make the next decision more reliable.

The paper does not claim to invent the arrivals–nights identity, Shapley allocation or interval arithmetic. It does not estimate a current national retention rate or the causal effect of a Prime Minister, marketing partnership or policy. Unverified 2026 endpoints remain outside the principal empirical analysis. The evidence cutoff is 6 October 2026.