## wpiea2025103-print-pdf

## Source details

**Canonical URL:** [wpiea2025103-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025103-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025103-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025103-print-pdf.pdf.json)

---

### Key statistical findings (Table 2: Summary Statistics, 1997-2023)
- Unit of analysis: Tourism-based SDS economies following the definition of International Monetary Fund (2024), covering 15 countries between 1997 and 2023.
- Actual GDP Growth (N=60): Mean 2.64 percent; St. Dev. 5.63; Min −33.40; Median 2.95; Max 41.70.
- Foreigners Visitors Growth (N=60): Mean 1.93 percent; St. Dev. 52.50; Min −380.67; Median 3.83; Max 452.18.
- Production Function Potential GDP Growth (N=42): Mean 2.57 percent; St. Dev. 3.00; Min −4.36; Median 2.01; Max 17.44.
- TFP Growth (N=42): Mean −0.06 percent; St. Dev. 1.61; Min −3.33; Median 0.03; Max 4.46.
- Capital Contribution to Growth (N=42): Mean 1.84 percent; St. Dev. 3.24; Min −7.69; Median 1.42; Max 20.12.
- Labor Contribution to Growth (N=42): Mean 0.87 percent; St. Dev. 0.67; Min −0.23; Median 0.75; Max 2.78.
- Production Function Output gap (N=438): Mean −0.41; St. Dev. 5.40; Min −32.24; Median −0.02; Max 17.23.
- GDP Growth 5Y ahead (WEO) (N=497): Mean 3.20; St. Dev. 1.47; Min −1; Max 38.
- HP Potential GDP Growth (N=550): Mean 4.09; St. Dev. 12.67; Min −148.37; Median 4.14; Max 123.37.
- Number of Rooms Growth (N=572): Mean 2.56; St. Dev. 16.03; Min −161.11; Median 0.78; Max 172.34.
- Number of Beds Growth (N=537): Mean 2.12; St. Dev. 14.97; Min −163.78; Median 0.00; Max 177.50.
- Investment Growth (N=419): Mean 3.27; St. Dev. 9.38; Min −40.59; Median 3.16; Max 95.61.
- Net Emigration Growth (N=565): Mean 3.38; St. Dev. 7.90; Min −30.54; Median 2.56; Max 23.60.
- Current Account Deficit Growth (N=392): Mean 6.63; St. Dev. 84.70; Min −306.81; Median 0.00; Max 309.10.
- Services Balance Growth (N=280): Mean 8.49; St. Dev. 47.13; Min −191.35; Median 5.90; Max 293.02.
- Remittances Growth (N=545): Mean 4.50; St. Dev. 53.05; Min −566.09; Median 2.96; Max 638.27.
- Foreign Direct Investment Growth (N=328): Mean 4.04; St. Dev. 89.74; Min −438.70; Median 3.68; Max 414.72.
- Revenues-to-GDP Growth (N=130): Mean 2.99; St. Dev. 18.01; Min −57.82; Median 2.74; Max 46.72.
- Expenditures-to-GDP Growth (N=130): Mean 1.99; St. Dev. 12.03; Min −20.00; Median 0.14; Max 42.83.
- Debt-to-GDP Growth (N=128): Mean −3.69; St. Dev. 16.34; Min −89.19; Median −6.11; Max 73.92.
- Data sources: Production Function estimates derived from Penn World Table inputs; Foreigners Visitors, Number of Rooms and beds and Net Emigration from UN database; WEO for other variables; Fiscal variables sourced from Fiscal Monitor 2024.

### Interpretation and substantive findings
- Aggregate growth and productivity
  - Actual GDP Growth averages 2.64 percent with high variability (St. Dev. 5.63), indicating exposure to large shocks (examples include global downturns and natural disasters).
  - Production Function Potential GDP Growth mean 2.57 percent closely tracks actual GDP growth, implying on average limited slack.
  - TFP Growth mean −0.06 percent signals stagnant or marginally declining productivity over the sample.
- Tourism activity and infrastructure
  - Foreigners Visitors Growth shows extreme volatility (St. Dev. 52.50) reflecting large swings in international arrivals for tourism-based SDS.
  - Number of Rooms Growth and Number of Beds Growth exhibit substantial dispersion (St. Dev. 16.03 and 14.97), indicating heterogeneous investment cycles in tourism infrastructure.
- External balances and capital flows
  - Current Account Deficit Growth has high dispersion (St. Dev. 84.70) underlining difficulty managing external imbalances; services balance and remittances are important buffers.
  - FDI Growth mean 4.04 with St. Dev. 89.74 indicates positive inflows on average but unstable and episodic investment.
- Labor and migration
  - Net Emigration Growth mean 3.38 with St. Dev. 7.90 underscores outward migration pressures affecting labor supply for tourism sectors.
- Fiscal caveat
  - Debt-to-GDP Growth mean −3.69 (St. Dev. 16.34) suggests a declining debt-to-GDP trajectory in available fiscal samples, but fiscal-variable sample sizes are smaller and interpretation is cautious.

### Econometric approach and identification
- Estimation framework
  - Local Projection (LP) to estimate dynamic responses of capital, labor, Balance of Payments, potential growth, and other variables to recessions and booms over horizons h.
  - Baseline LP: y_{i,t+h} − y_{i,t} = α_h + β_h shock_{i,t} + γ_h ∆y_{i,t−1} + θ naturaldisaster_{i,t} + fixedeffects_i + ε_{i,t}.
  - Natural disaster dummy from EM-DAT included to isolate internally originated shocks; country-fixed effects included; results robust to including time-fixed effects.
- Shock identification
  - Baseline recession definition (following Çelik et al. (2023)): period from a peak to a trough where trough is a year with output growth negative and at least one standard deviation below its long-term mean → baseline yields 44 recession episodes.
  - Alternative recession definitions: HP/detrended GDP variants; tourism-flow-based narrow sample (40 episodes); trough defined by negative growth alone (53 episodes).
  - Baseline expansion: trough to next peak where peak = positive growth and ≥ one standard deviation above long-term mean → 45 expansions; alternative expansion definitions yield 32 and 60 episodes.
  - Tourism-specific boom/bust definitions based on arrivals growth ± one standard deviation thresholds.

### Long-term relations: recessions and expansions — core empirics and thematic takeaways
- Broad patterns (Annex B):
  - Potential Growth responds negatively to recessions across baseline and multiple alternative specifications (example: baseline t=0: −1.199*** (0.004); narrow identification t=0: −1.433*** (0.003)).
  - Investment Growth falls sharply at recession onset (baseline t=0: −15.178*** (0.025); narrow: t=0: −15.524*** (0.027)) and rises during expansions (expansion t+2: 21.791*** (0.076)).
  - Stock of Capital Growth contracts after recessions (baseline t+2: −6.787*** (0.021)) and expands after booms (expansion t+3: 18.735** (0.094)).
  - Services Balance and hotel-industry metrics exhibit large negative swings in recessions (Services Balance t=0: −62.31*** (0.09); Hotel Rooms t=0: −15.651** (0.071)) and positive spikes during expansions (Hotel Rooms expansion t+2: 13.227*** (0.043)).
  - Public Debt Growth rises at recessions (t=0: 16.858*** (0.027); t+1: 13.574*** (0.032)) and falls or improves during expansions (expansion t+1: −12.978*** (0.044)).
- Thematic interpretation
  - Tourism-dependent SDS experience deeper recessions on average than EMDEs but faster recoveries; recessions temporarily reduce potential growth mainly via capital contractions.
  - Expansions generate initial surges in hotel capacity and investment that tend to fade, producing limited spillovers to broader investment and limited sustained improvement in external stability.
  - Labor outcomes: recessions cause sharp employment contractions and reduced labor efficiency; expansions see employment rebound and accelerated net emigration.

### Short-term influence of tourism, scenarios, and quantified impacts (Sections 5.2 and 6.2)
- Short-term elasticities and differential island responses
  - Estimated short-run elasticity of tourism to economic growth in Cabo Verde ≈ 0.4 over twelve months (baseline, lower bound for other SDS).
  - Decomposition: Θ = Ψ·ω_centric + η·ω_noncentric with calibration (pre-pandemic 2019 INE data excluding Santiago): ω_noncentric ≈ 0.6; ω_centric ≈ 0.4; estimated Ψ ≈ 0.4 and η ≈ 0 → Θ ≈ 0.16.
  - Alternative national point elasticity (1996–2024 averages): Θ ≈ 0.18 with ω_noncentric ≈ 0.54 and ω_centric ≈ 0.46.
- Baseline and scenario magnitudes
  - Baseline: Ψ = 0.4, η = 0, ω_centric = 0.4, ω_noncentric = 0.6 → Θ = 0.16. A 20 percent increase in tourist arrivals → national GDP rises ≈ 3.2 percent.
  - Downside scenario: 20 percent decline in arrivals concentrated in tourism-centric islands → national GDP contraction ≈ 3.2 percent (given Θ = 0.16); using Growth effect = − g_t / (1 + g_t) · (D_{t−1} / Y_{t−1}) with g_t = −0.032 and D_{t−1}/Y_{t−1} = 1.1 implies a rise in the debt ratio of about 3.6 percentage points of GDP.
  - Upside (integration) scenario: if η increases from 0 to 0.4 → Θ_up = 0.4·0.4 + 0.4·0.6 = 0.4; a 20 percent increase in arrivals → national GDP increases 8.0 percent.
- Short-term empirical tables (Annex C highlights)
  - Baseline impact on Tourism-centric Islands (Table 15) at horizon 0: 0.402*** (0.134); horizon 12: 0.379** (0.189). Other Islands Differential shows multiple negative and significant coefficients (e.g., horizon 3: −0.354*** (0.07); horizon 12: −0.55*** (0.174)).
  - After the pandemic (Table 17) Tourism-centric impact at horizon 0: 0.571*** (0.121); horizon 12: 0.241*** (0.059). Other Islands Differential includes significant negative entries (e.g., horizon 5: −0.403*** (0.117); horizon 12: −0.42*** (0.133)).
  - IV regressions (Table 19) show statistically significant positive impacts at selected horizons (e.g., horizon 4: 0.693*** (0.265); horizon 8: 1.158*** (0.302)) and substantial negative differentials for Other Islands (e.g., horizon 8: −1.054*** (0.348)).
  - IV first-stage F-statistics (Table 20) are strong across horizons (Horizon 0: 28.968 and 24.003; Horizon 12: 21.165 and 19.533).

### Robustness and subgroup evidence
- Two robustness strategies yield broadly consistent results with baseline:
  1. Subgroup analysis for Sub-Saharan Africa (SSA), the Caribbean, and Asia:
     - SSA: net emigration significant in expansions and recessions; higher remittances during recessions; FDI inflows during expansions.
     - Caribbean: recessions associated with persistently higher debt levels.
     - Asian SDS: persistent decline in potential growth following recessions.
  2. Alternative recession/expansion definitions (detrended GDP, tourism-gap narrow/broad) generally align with baseline for potential growth, capital, labor, BoP, and fiscal indicators; narrow definitions often produce muted expansionary gains and persistent fiscal weakness.
- Robustness charts and tables referenced in Annex B and Annex C support these conclusions.

### Policy recommendations and implications
- Build resilience to tourism-driven volatility
  - Stabilize external financing and diversify foreign income sources to reduce Current Account Deficit and FDI volatility.
  - Maintain prudent indebtedness and use expansions to implement fiscal consolidation to create buffers for downturns.
- Enhance productivity and inclusive growth
  - Prioritize policies to raise TFP (mean −0.06 percent historically) to sustain potential growth.
  - Foster linkages between tourism and non-tourism sectors to increase η (elasticity in non-tourism-centric islands) and multiply aggregate gains from tourism shocks.
- Labor and social policy measures
  - Implement temporary supports during downturns: unemployment security schemes, access to credit (including public guarantee designs), and targeted social transfers.
- External sector and energy policy
  - Promote diversification and transition to renewable energy to reduce Balance of Payments exposure to food and energy price shocks.
  - Identify and maintain adequate international reserves to respond to temporary external shocks.
- Infrastructure and connectivity
  - Improve inter-island connectivity and integrate non-tourism-centric regions into tourism value chains to convert baseline Θ ≈ 0.16 into higher aggregate elasticities (example: Θ_up = 0.4).
- Research and data priorities
  - Address data limitations on potential growth metrics and sectoral granularity; future work to explore applicability to other small states, cross-country best practices, and heterogeneity by income/fragility status.

*Source: wpiea2025103-print-pdf — Sections 3.1, 5.1, 6.2, Annex B, and Annex C.*

### 3.1    Long-term assessment of recessions and expansions

### 3.1    Long-term assessment of recessions and expansions

### Key statistical findings (Table 2: Summary Statistics, 1997-2023)
- Unit of analysis: Tourism-based SDS economies following the definition of International Monetary Fund (2024), covering 15 countries between 1997 and 2023.
- Actual GDP Growth (N=60): Mean 2.64 percent; St. Dev. 5.63; Min −33.40; Median 2.95; Max 41.70.
- Foreigners Visitors Growth (N=60): Mean 1.93 percent; St. Dev. 52.50; Min −380.67; Median 3.83; Max 452.18.
- Production Function Potential GDP Growth (N=42): Mean 2.57 percent; St. Dev. 3.00; Min −4.36; Median 2.01; Max 17.44.
- TFP Growth (N=42): Mean −0.06 percent; St. Dev. 1.61; Min −3.33; Median 0.03; Max 4.46.
- Capital Contribution to Growth (N=42): Mean 1.84 percent; St. Dev. 3.24; Min −7.69; Median 1.42; Max 20.12.
- Labor Contribution to Growth (N=42): Mean 0.87 percent; St. Dev. 0.67; Min −0.23; Median 0.75; Max 2.78.
- Production Function Output gap (N=438): Mean −0.41; St. Dev. 5.40; Min −32.24; Median −0.02; Max 17.23.
- GDP Growth 5Y ahead (WEO) (N=497): Mean 3.20; St. Dev. 1.47; Min −1; Max 38.
- HP Potential GDP Growth (N=550): Mean 4.09; St. Dev. 12.67; Min −148.37; Median 4.14; Max 123.37.
- Number of Rooms Growth (N=572): Mean 2.56; St. Dev. 16.03; Min −161.11; Median 0.78; Max 172.34.
- Number of Beds Growth (N=537): Mean 2.12; St. Dev. 14.97; Min −163.78; Median 0.00; Max 177.50.
- Investment Growth (N=419): Mean 3.27; St. Dev. 9.38; Min −40.59; Median 3.16; Max 95.61.
- Net Emigration Growth (N=565): Mean 3.38; St. Dev. 7.90; Min −30.54; Median 2.56; Max 23.60.
- Current Account Deficit Growth (N=392): Mean 6.63; St. Dev. 84.70; Min −306.81; Median 0.00; Max 309.10.
- Services Balance Growth (N=280): Mean 8.49; St. Dev. 47.13; Min −191.35; Median 5.90; Max 293.02.
- Remittances Growth (N=545): Mean 4.50; St. Dev. 53.05; Min −566.09; Median 2.96; Max 638.27.
- Foreign Direct Investment Growth (N=328): Mean 4.04; St. Dev. 89.74; Min −438.70; Median 3.68; Max 414.72.
- Revenues-to-GDP Growth (N=130): Mean 2.99; St. Dev. 18.01; Min −57.82; Median 2.74; Max 46.72.
- Expenditures-to-GDP Growth (N=130): Mean 1.99; St. Dev. 12.03; Min −20.00; Median 0.14; Max 42.83.
- Debt-to-GDP Growth (N=128): Mean −3.69; St. Dev. 16.34; Min −89.19; Median −6.11; Max 73.92.
- Data sources: Production Function estimates derived from Penn World Table inputs; Foreigners Visitors, Number of Rooms and beds and Net Emigration from UN database; WEO for other variables; Fiscal variables sourced from Fiscal Monitor 2024.

### Interpretation of the statistics and substantive findings
- Aggregate growth and variability:
  - Actual GDP Growth averages 2.64 percent but with high variability (St. Dev. 5.63), indicating vulnerability to large shocks (e.g., global downturns, natural disasters).
  - Production Function Potential GDP Growth (2.57 percent) closely tracks actual GDP growth, implying on average limited economic slack.
  - TFP Growth mean is negative (−0.06 percent), suggesting stagnant or declining productivity improvements over the sample period.
- Tourism activity and infrastructure:
  - Foreigners Visitors Growth shows large swings (Mean 1.93 percent; St. Dev. 52.50), capturing significant fluctuation in international tourist arrivals for tourism-based SDS.
  - Number of Rooms Growth and Number of Beds Growth display substantial variability (Means 2.56 and 2.12; St. Dev. 16.03 and 14.97 respectively), reflecting differing investment/expansion cycles in tourism infrastructure.
- External balances and capital flows:
  - High variability in Current Account Deficit Growth (Mean 6.63; St. Dev. 84.70) underscores difficulties managing external imbalances and the role of tourism revenues as a buffer.
  - Foreign Direct Investment Growth is positive at 4.04 percent on average but with large dispersion (St. Dev. 89.74), indicating challenges in securing stable long-term capital inflows.
- Labor and migration:
  - Net Emigration Growth mean 3.38 with St. Dev. 7.90 highlights outward migration pressures that may exacerbate labor shortages in tourism sectors.
- Fiscal metrics caveat:
  - Debt-to-GDP Growth mean is −3.69 (St. Dev. 16.34), reflecting a declining debt-to-GDP trajectory attributed to a tax base growing faster than expenditures and GDP growth; interpretation limited because fiscal-variable sample size is significantly smaller than for other metrics.

### Potential growth comparisons and recent period patterns
- Potential growth metrics (HP Filter, Production Function, WEO 5-year ahead) yield relatively high estimates for 2000–2010 and 2011–2021, approximately in the range of 3 percent to nearly 8 percent depending on method, with notably stronger performance in the lead-up to the global financial crisis.
- Comparisons across groups (World, AEs, EMDEs, tourism-based SDS) show that the gap between actual and Production Function Potential GDP Growth widened significantly for tourism-based SDS during 2011–2021, largely explained by the 2021 cut-off as tourism-dependent economies were in the initial phase of COVID-19 recovery.

### Econometric approach: Local Projection (LP) framework and identification
- Estimation strategy:
  - Use Local Projection Method (LP) to estimate dynamic responses of capital, labor, Balance of Payments, potential growth, and other variables to recessions and booms over horizons h.
  - Baseline LP specification: y_{i,t+h} − y_{i,t} = α_h + β_h shock_{i,t} + γ_h ∆y_{i,t−1} + θ naturaldisaster_{i,t} + fixedeffects_i + ε_{i,t}.
  - Natural disaster dummy (naturaldisaster_{i,t}) included to isolate internally originated shocks (source: EM-DAT).
  - Country-fixed effects included; time-fixed effects unnecessary in baseline because trepresents duration since the event; results robust to inclusion of time-fixed effects.
- Identification of shocks:
  - Recession definition (baseline, following Çelik et al. (2023)): period from a peak in output preceding a business cycle trough to the trough, where trough is a year with output growth negative and at least one standard deviation below its long-term mean. Baseline yields 44 recession episodes.
  - Alternative recession identifications:
    - Detrended GDP via HP filter to compute mean and standard deviation conditions (modified baseline).
    - Recessions defined when tourism flows fall below historical trend (narrower sample of 40 episodes).
    - Trough defined as a year with negative actual GDP growth (no standard deviation condition) yields wider sample of 53 episodes.
  - Expansion definitions:
    - Baseline expansion: period from trough to next peak, with peak defined as year when output growth positive and at least one standard deviation above long-term mean; baseline identifies 45 expansions.
    - Alternative expansion definitions parallel those for recessions: detrended-GDP baseline; expansions where tourism flows rise above historical trend (n=32); expansions above trend without output growth threshold (n=60).
  - Tourism-specific definitions:
    - Tourism burst/boom defined by peaks/troughs in tourism arrivals where a peak is arrivals growth positive and ≥ one standard deviation above long-term mean and trough is arrivals growth negative and ≤ one standard deviation below long-term mean.

### Implications for tourism-dependent small developing states (SDS)
- Tourism-dependency amplifies macroeconomic vulnerability:
  - Large swings in foreign visitor flows and tourism infrastructure growth rates produce sizable short- and long-run effects on GDP, external balances, and labor markets.
- Productivity concerns:
  - Negative mean TFP Growth (−0.06 percent) signals a need to prioritize policies that boost productivity to sustain potential growth.
- External and fiscal resilience:
  - High dispersion in Current Account Deficit Growth and FDI Growth suggests priorities include stabilizing external financing and diversifying revenue sources.
  - Fiscal indicators point to declining debt-to-GDP but smaller sample size advises cautious interpretation; strengthening tax bases while managing expenditure remains important.
- Heterogeneity and policy targeting:
  - Substantial cross-country (and within-country island) heterogeneity in tourism infrastructure and recovery paths implies tailored policies—investment, disaster risk management, and labor-market measures—are needed to enhance resilience.

*Source: wpiea2025103-print-pdf — Section 3.1, "Long-term assessment of recessions and expansions".*

### 5.1    Long-term relation of recessions and expansions

### 5.1    Long-term relation of recessions and expansions

### Resilience of SDS to Recessions
- Tourism-dependent SDS experience temporary reductions in potential growth during recessions, largely attributed to a contraction in capital.
- Emerging Market and Developing Economies (EMDEs) face more permanent declines, driven by sustained reductions in investment, employment, and potential growth in TFP.
- Recessions in tourism-dependent SDS tend to be deeper on average than in EMDEs, yet tourism-dependent SDS recover more quickly after recessions; research by Mu et al. (2023) suggests the milder economic cycles in the AEs (the main source of tourism for most SDS) play a critical role.
- These observations remain robust when the COVID-19 period is excluded from the analysis.
- Expansions in tourism-dependent SDS lead to a delayed and muted response to potential growth, primarily driven by gains in employment and investment growth.

### Capital
- During recessions:
  - Hotel supply metrics (number of rooms and beds available) decline sharply but begin to recover over time.
  - Broader investment metrics show a more prolonged contraction, reflecting a delayed recovery in capital accumulation outside the tourism sector.
  - Divergence suggests the tourism sector adapts relatively quickly to demand changes while broader investment faces persistent constraints.
- During expansions:
  - Investment growth in tourism-dependent SDS initially surges, driven by rapid expansion of hotel capacity.
  - This momentum tends to fade, suggesting limited spillover effects into broader investment opportunities.

### Labor
- During recessions:
  - Employment contracts sharply, particularly in tourism-related sectors such as hospitality and services.
  - Labor efficiency is reduced due to lower activity and higher uncertainty.
  - Net emigration growth slows significantly, likely due to fewer opportunities abroad, intensifying local labor market competition.
- During expansions:
  - Employment growth rebounds alongside a resurgence in tourism activity, particularly in service-oriented sectors.
  - Net emigration growth accelerates during expansions, potentially reflecting improved opportunities globally as economic conditions stabilize.

### Balance of Payments (BoP)
- Structural features:
  - Tourism-dependent SDS often face persistent trade deficits due to reliance on imports for consumption and investment goods; this is offset to some extent by a positive services balance from tourism receipts.
  - Remittances are a critical foreign income source for SDS.
- During recessions:
  - Current account deficit (CAD) increases significantly, driven by sharp declines in tourism inflows, deteriorating the balance of services and amplifying the trade deficit.
  - Remittances decline, further straining the balance of payments.
- During expansions:
  - Expansions do not produce significant improvements in CAD; tourism inflow gains are offset by rising imports needed to support the expanding tourism sector.
  - Foreign direct investment (FDI) shows only limited recovery during expansions, reflecting persistent structural challenges in attracting sustained external capital.
- Implication: Expansions in SDS may not substantially enhance external stability due to high import dependence and limited diversification of foreign income sources.

### Fiscal
- During recessions:
  - Government spending increases relative to GDP as a stabilizing measure.
  - The sharp GDP contraction leads to a significant rise in the public debt-to-GDP ratio.
- During expansions:
  - Expenditure growth slows relative to GDP, but revenue-to-GDP ratios do not improve sufficiently to rebuild fiscal buffers.
  - Result: missed opportunity to improve fiscal resilience and reduce public debt during growth periods.

*Source: IMF and Authors’ own calculations; figures note the shaded area represents a 90 percent confidence interval.*

### Short-term Influence of Tourism (Section 5.2)
- Differential impact between tourism-centric and non-tourism-centric islands:
  - Tourism-centric islands (example: Sal and Boa Vista) experience a strong and immediate increase in VAT collections; reported mid-point elasticity over 12 months around 0.4.
  - Non-tourism-centric islands show a muted response, with economic activity sometimes declining.
  - The differential influence is statistically significant: tourism-dependent islands experience robust growth while non-tourism-centric regions are more insulated or negatively affected.
- Direct and indirect links of tourism to non-tourism-centric islands (Equation (3) results):
  - Direct impact of local tourism on VAT in non-tourism-centric islands: modest, short-term increase that becomes statistically insignificant over time—local tourism alone does not drive long-term VAT revenues in these regions.
  - Indirect impact of tourism-centric islands on non-centric regions: similarly insignificant—economic activity in Sal and Boa Vista remains largely disconnected from other islands in the short term.
- Decomposition of national elasticity of economic activity to tourist arrivals (Θ):
  - Θ = Ψ·ω_centric + η·ω_noncentric, where:
    - Ψ denotes elasticity of GDP to arrivals in tourism-centric islands.
    - η denotes elasticity in non-tourism-centric islands.
    - ω_centric and ω_noncentric are shares of national GDP from each group.
  - Calibration (pre-pandemic 2019 island-level GDP data from INE, excluding Santiago):
    - ω_noncentric ≈ 0.6
    - ω_centric ≈ 0.4
    - Estimated Ψ ≈ 0.4 and η ≈ 0 imply Θ ≈ 0.16.
  - Alternative national point elasticity using average annual growth rates of real GDP and international arrivals over 1996–2024 yields Θ ≈ 0.18, which implies ω_noncentric ≈ 0.54 and ω_centric ≈ 0.46.
- Baseline and scenario quantitative implications:
  - Baseline: Ψ = 0.4, η = 0, ω_centric = 0.4, ω_noncentric = 0.6 → Θ = 0.16. A 20 percent increase in tourist arrivals would raise national GDP by approximately 3.2 percent.
  - Downside scenario: A 20 percent decline in tourist arrivals to tourism-centric islands → national GDP contraction of approximately 3.2 percent (given Θ = 0.16); using the growth effect expression below with g_t = −0.032 and D_{t−1}/Y_{t−1} = 1.1 implies a rise in the debt ratio of about 3.6 percentage points of GDP.
    - Growth effect expression provided:
      - Growth effect = − g_t / (1 + g_t) · (D_{t−1} / Y_{t−1}),
      - where g_t = −0.032 corresponds to the GDP contraction from a 20% drop in arrivals (given Θ = 0.16), and D_{t−1}/Y_{t−1} = 1.1.
  - Upside scenario: If η increases from 0 to 0.4 (matching Ψ), then
    - Θ_up = Ψ·ω_centric + η·ω_noncentric = 0.4·0.4 + 0.4·0.6 = 0.4,
    - A 20 percent increase in arrivals would lead to an 8.0 percent increase in national GDP.
  - Policy implication: Improving inter-island connectivity and integrating non-tourism-centric regions into the tourism value chain can multiply aggregate gains from tourism shocks.

### Robustness (Section 6.1)
- Rationale: Recessions and expansions in tourism-dependent SDS are intrinsically tied to fluctuations in tourism activity given tourism’s central role.
- Two alternative robustness analyses:
  1. Apply methodology to three subgroups: Sub-Saharan Africa (SSA), the Caribbean, and Asia.
     - Results broadly align with the expanded sample for potential growth, capital, labor, external, and fiscal metrics.
     - SSA: net emigration significant during both expansions and recessions; higher remittances during recessions; additional FDI inflows during expansions.
     - Caribbean: persistent higher level of debt associated with occurrence of recessions.
     - Asian SDS: notable and persistent decline in potential growth following a recession.
  2. Modify recession/expansion specifications:
     - Baseline recession identification from Çelik et al. (2023) yields 44 recession episodes.
     - Narrower recession sample (tourism flows below historical trend) yields 40 episodes; results using detrended GDP align with baseline.
     - Alternative trough definition (drop requirement of standard deviation below mean; focus solely on negative output growth) expands sample to 53 recession episodes.
     - Baseline expansions: periods from trough to next peak where peak = positive output growth at least one standard deviation above long-term mean → 45 expansion episodes.
     - Narrower expansion (tourism flows above historical trend) reduces to 32 cases.
     - Broader expansion (tourism flows exceed trend without requiring output growth threshold) expands to 60 episodes.
- Robustness outcomes:
  - Under narrow and broad definitions, results broadly in line with baseline for potential growth, capital, labor, BoP, and fiscal indicators.
  - Narrow definition: expansions produce muted gains in potential growth and capital accumulation; net emigration decreases after an expansion (diverging from baseline); limited rebound in fiscal metrics persists.
- Full set of robustness charts and tables are presented in Annex B (as noted in the source). 

*Source: IMF and Authors’ own calculations.*

### 6.2    Short-term Influence of Tourism

### 6.2    Short-term Influence of Tourism

### Empirical findings
- Before the pandemic, results align with the baseline: a strong positive response in tourism-centric islands and a negative differential in non-tourism-centric islands.
- After the pandemic, the response in tourism-centric islands is more subdued compared to the baseline, resembling the negative differential in non-tourism-centric regions.
- The recovery from the pandemic may have amplified asymmetry between regions, with non-tourism-centric islands experiencing greater disruptions and reduced benefits from tourism.
- The baseline specification focuses on the dynamic relationship between tourist arrivals and local economic activity rather than an event-study framework.
- Estimated short-run elasticity of tourism to economic growth in Cabo Verde is around 0.4 over twelve months, considered a lower bound for other SDS due to limited spillover effects from tourism to other economic sectors.
- Negligible impact of tourism shocks on non-tourism-centric islands highlights constrained benefits of tourism growth across different regions and sectors within an SDS.
- Observed resilience in Cabo Verde suggests tourism can serve as a robust recovery mechanism.

### Robustness checks and identification
- A difference-in-differences specification uses Sal and Boa Vista as treated units and the onset of COVID (March 2020) as the event; without violating the parallel trend condition, results confirm tourism-centric islands experienced a significantly sharper contraction.
- Analysis of Santiago Island (where the capital Praia is located) shows government-related economic activity does not substantially alter estimated results, suggesting findings are not overly sensitive to inclusion of regions with distinct economic performance.
- An instrumental variable (IV) approach is employed; with high first-stageF-statistics across all regression horizons, the IV estimation confirms validity of the instrument and provides evidence that estimated relations are not driven by endogeneity or omitted variable bias.
- Full set of supporting charts and tables for this section is presented in Annex C.

### Scenarios and quantified impacts
- Asymmetric scenario: a 20 percent decline in arrivals concentrated in tourism-centric islands could lead to:
  - a contraction in national GDP of over 3 percent
  - a mechanical increase in public debt of approximately 3.6 percentage points of GDP
- Integration scenario: under improved inter-island connectivity where non-centric regions become more integrated into the tourism economy, the same increase in arrivals could:
  - raise GDP by up to 8 percent, more than double the baseline effect

### Policy recommendations
- Recognize tourism’s potential as a robust recovery mechanism in the aftermath of recessions, with relatively faster recoveries compared to larger emerging markets.
- Address limited spillover effects and regional disparities by fostering diversification and sectorial integration; strengthen linkages between tourism and non-tourism sectors to ensure broader, more equitable growth.
- Avoid overreliance on tourism to reduce exposure to volatility and external shocks.
- Implement temporary measures during downturns to mitigate adverse effects on economic activity, with emphasis on supporting other sectors. Potential measures include:
  - establishing an effective unemployment security scheme
  - ensuring access to credit for businesses, potentially through the design of a public guarantee policy if necessary
  - providing targeted social transfers to the most vulnerable populations
- Fiscal policy guidance:
  - maintain a prudent level of indebtedness
  - implement fiscal consolidation measures during periods of economic expansion to create fiscal buffers for increased spending during downturns
- External sector guidance:
  - reduce Balance of Payments exposure to shocks from food and energy prices by promoting economic diversification and facilitating the transition to renewable energy
  - identify and maintain an adequate level of international reserves to respond to temporary external shocks

### Limitations and avenues for future research
- Analysis constrained by availability and quality of data, particularly concerning potential growth metrics.
- Future research could benefit from more granular data, especially at the sectoral level, to refine estimates.
- Results may not be directly generalizable to other small states or non-tourism-dependent economies due to the unique characteristics of SDS (reliance on tourism and external aid).
- Future work could explore applicability to other small states, cross-country comparisons to identify best practices in managing economic shocks, and heterogeneity by income level or fragility status within tourism-dependent SDS.

### Annex content (supporting materials)
- Annex A lists the country groups used in the empirical analysis.
- Annex B presents robustness checks on long-term relations of tourism booms and busts across regions (Sub-Saharan Africa, the Caribbean, and Asia) and economic channels (growth, capital, labor, BoP, and fiscal metrics), including results under alternative identification strategies.
- Annex C focuses on the short-term effects of tourism shocks in Cabo Verde, using baseline and IV specifications to show how local fiscal outcomes vary with tourism dependency before and after the pandemic (Local projections and Diff-and-Diff).

*Source: wpiea2025103-print-pdf - 6.2    Short-term Influence of Tourism*

### Annex B: Long-term relation of tourism on growth

### Annex B: Long-term relation of tourism on growth

### Recessions — baseline impacts (Table 5)
- Potential Growth: Production Function
  - t=0: -1.199*** (0.004)
  - t+1: -2.582*** (0.005)
  - t+2: -3.011 (0.022)
  - t+3: -1.949 (0.051)
- Investment Growth
  - t=0: -15.178*** (0.025)
  - t+1: -14.247*** (0.025)
  - t+2: -12.646** (0.057)
  - t+3: -11.897 (0.096)
- Stock of Capital Growth
  - t=0: -2.089* (0.011)
  - t+1: -4.591*** (0.009)
  - t+2: -6.787*** (0.021)
  - t+3: -5.484 (0.048)
- Services Balance
  - t=0: -62.31*** (0.09)
  - t+1: -54.225*** (0.147)
  - t+2: -10.628* (0.063)
  - t+3: -41.449** (0.169)
- Hotel Industry Supply: Rooms Growth
  - t=0: -15.651** (0.071)
  - t+1: -4.693* (0.024)
- Public Debt Growth
  - t=0: 16.858*** (0.027)
  - t+1: 13.574*** (0.032)

### Recessions — robustness checks and alternative specifications
- Detrended GDP baseline (Table 6) — Potential Growth: Production Function
  - t=0: -0.79 (0.007)
  - t+1: -2.275*** (0.005)
- Major AEs growth identification (Table 7) — Potential Growth: Production Function
  - t=0: -0.483 (0.006)
  - t+1: -1.881*** (0.006)
- Narrow identification (recession and negative tourism gap) (Table 8) — Potential Growth: Production Function
  - t=0: -1.433*** (0.003)
  - t+1: -2.659*** (0.009)
- Broad identification (negative tourism gap) (Table 9) — Potential Growth: Production Function
  - t=0: -0.719* (0.004)
  - t+1: -1.466* (0.008)
- Selected other robust patterns under alternative identifications:
  - Investment Growth (narrow): t=0: -15.524*** (0.027); t+1: -14.836*** (0.026)
  - Services Balance (narrow): t=0: -68.048*** (0.132); t+1: -55.546*** (0.086)
  - Public Debt Growth (narrow): t=0: 16.025*** (0.03); t+1: 11.951*** (0.029)
  - Investment Growth (broad): t=0: -9.954*** (0.02); t+1: -9.712*** (0.024)
  - Services Balance (broad): t=0: -38.257*** (0.088); t+2: 19.798*** (0.07)

### Expansions — baseline impacts (Table 10)
- Potential Growth: Production Function
  - t=0: 1.172 (0.018)
  - t+1: 2.801 (0.021)
  - t+2: 12.001** (0.058)
  - t+3: 14.793* (0.078)
- Investment Growth
  - t=0: 6.988** (0.034)
  - t+1: 9.984** (0.046)
  - t+2: 21.791*** (0.076)
- Stock of Capital Growth
  - t=0: 0.83 (0.026)
  - t+1: 4.289 (0.028)
  - t+2: 14.267** (0.071)
  - t+3: 18.735** (0.094)
- Hotel Industry Supply: Rooms Growth
  - t=0: 6.236* (0.038)
  - t+1: 6.321* (0.035)
  - t+2: 13.227*** (0.043)
  - t+3: 11.498** (0.058)
- Public Debt Growth
  - t=0: -7.922 (0.055)
  - t+1: -12.978*** (0.044)
  - t+2: -9.711** (0.045)
  - t+3: -11.338** (0.054)

### Expansions — robustness checks and alternative specifications
- Detrended GDP baseline (Table 11) — Potential Growth: Production Function
  - t=0: 0.247 (0.02)
  - t+1: 1.382 (0.022)
  - t+2: 10.459* (0.06)
  - t+3: 13.323* (0.078)
- Major AEs growth identification (Table 12) — Potential Growth: Production Function
  - t=0: 0.357 (0.003)
  - t+1: 0.825* (0.004)
- Narrow identification (expansion and positive tourism gap) (Table 13) — Potential Growth: Production Function
  - t=0: 0.546 (0.006)
  - t+1: 2.153* (0.012)
- Broad identification (positive tourism gap) (Table 14) — Potential Growth: Production Function
  - t=0: -2.334 (0.081)
  - t+1: 1.008 (0.038)
- Selected other patterns under alternative identifications:
  - Investment Growth (expansion & positive tourism gap): t=0: 6.551*** (0.015); t+1: 8.101* (0.044)
  - Hotel Rooms Growth (expansion & positive tourism gap): t=0: 8.553*** (0.032); t+1: 8.709** (0.034); t+2: 10.119*** (0.037)
  - Services Balance (expansion, major AEs id): t=0: -34.959 (0.478); t+1: 47.722* (0.245); t+2: -91.607*** (0.345)

### Thematic takeaways evident from figures and tables
- Potential Growth responses to recessions are consistently negative across baseline and several alternative specifications (examples: Table 5 t=0: -1.199*** (0.004); Table 8 t=0: -1.433*** (0.003)).
- Investment and stock-of-capital growth show notable negative coefficients following recessions and positive coefficients during expansions (examples: Investment t=0 recession -15.178*** (0.025); expansion t+2 21.791*** (0.076)).
- Services balance and tourism-linked metrics (hotel rooms, beds) register large negative swings in recessions and positive or elevated growth during expansions under several specifications (examples: Services Balancet=0 recession -62.31*** (0.09); expansion Hotel Rooms t+2 13.227*** (0.043)).
- Public debt tends to rise following recessions and to fall or increase less following expansions across multiple specifications (examples: recession Public Debt Growth t=0 16.858*** (0.027); expansion Public Debt Growth t+1 -12.978*** (0.044)).

*Source: IMF and Authors’ own calculations (Annex B: Long-term relation of tourism on growth).*

### Annex C: Short-term influence of tourism on growth

### Annex C: Short-term influence of tourism on growth

### Figures: Impact of 1 Percent increase in Tourism flows on Domestic VAT
- Figure 36: Impact by Tourism Dependency Before the Pandemic. Note: The shaded area represents a 90 percent confidence interval.
- Figure 37: Impact by Tourism Dependency After the Pandemic. Note: The shaded area represents a 90 percent confidence interval.
- Figure 38: Impact by Tourism Dependency Including Santiago Island. Note: The shaded area represents a 90 percent confidence interval.
- Figure 39: Impact by Tourism Dependency IV. Note: The shaded area represents a 90 percent confidence interval.

### Baseline Results (Table 15)
- Impact on Tourism-centric Islands at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.402*** (0.134)
  - 1: -0.012 (0.074)
  - 2: 0.055 (0.08)
  - 3: 0.135* (0.072)
  - 4: 0.394* (0.212)
  - 5: 0.34*** (0.096)
  - 6: 0.553*** (0.181)
  - 7: 0.452*** (0.145)
  - 8: 0.486*** (0.129)
  - 9: 0.365*** (0.129)
  - 10: 0.107 (0.292)
  - 11: 0.449*** (0.133)
  - 12: 0.379** (0.189)
- Other Islands Differential with Tourism-centric at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: -0.175 (0.233)
  - 1: -0.089 (0.125)
  - 2: 0.012 (0.091)
  - 3: -0.354*** (0.07)
  - 4: -0.459** (0.221)
  - 5: -0.562*** (0.125)
  - 6: -0.336* (0.185)
  - 7: -0.165 (0.169)
  - 8: -0.421*** (0.09)
  - 9: -0.65*** (0.104)
  - 10: -0.243 (0.249)
  - 11: -0.513*** (0.122)
  - 12: -0.55*** (0.174)

### Before the Pandemic (Table 16)
- Impact on Tourism-centric Islands at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.175 (0.246)
  - 1: 0.265 (0.376)
  - 2: 1.652* (0.948)
  - 3: 1.787 (1.423)
  - 4: 1.984 (1.819)
  - 5: 2.369 (1.641)
  - 6: 2.406 (1.624)
  - 7: 1.973 (1.34)
  - 8: 1.768 (1.544)
  - 9: 1.866 (1.543)
  - 10: 1.674 (1.42)
  - 11: 1.411 (1.046)
  - 12: 1.442 (1.01)
- Other Islands Differential with Tourism-centric at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.21 (0.499)
  - 1: 0.787 (1.266)
  - 2: -0.815 (1.261)
  - 3: -2.103 (1.772)
  - 4: -1.834 (1.942)
  - 5: -2.373 (1.85)
  - 6: -1.674 (1.483)
  - 7: -0.477 (1.889)
  - 8: -1.426 (1.568)
  - 9: -1.418 (1.43)
  - 10: -1.488 (1.369)
  - 11: -1.137 (1.286)
  - 12: -1.301 (1.35)

### After the Pandemic (Table 17)
- Impact on Tourism-centric Islands at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.571*** (0.121)
  - 1: -0.041 (0.067)
  - 2: -0.069 (0.112)
  - 3: 0.235*** (0.046)
  - 4: 0.346 (0.213)
  - 5: 0.284*** (0.095)
  - 6: 0.522** (0.235)
  - 7: 0.287*** (0.105)
  - 8: 0.042 (0.088)
  - 9: 0.157* (0.08)
  - 10: -0.08 (0.167)
  - 11: 0.215*** (0.037)
  - 12: 0.241*** (0.059)
- Other Islands Differential with Tourism-centric at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: -0.211 (0.187)
  - 1: 0.08 (0.168)
  - 2: 0.007 (0.142)
  - 3: -0.168 (0.13)
  - 4: -0.171 (0.192)
  - 5: -0.403*** (0.117)
  - 6: -0.397** (0.194)
  - 7: 0.028 (0.093)
  - 8: -0.006 (0.106)
  - 9: -0.636*** (0.185)
  - 10: 0.051 (0.113)
  - 11: -0.401*** (0.051)
  - 12: -0.42*** (0.133)

### Including Santiago island (Table 18)
- Impact on Tourism-centric Islands at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.41*** (0.122)
  - 1: -0.005 (0.075)
  - 2: 0.072 (0.076)
  - 3: 0.17*** (0.064)
  - 4: 0.432** (0.208)
  - 5: 0.36*** (0.085)
  - 6: 0.557*** (0.162)
  - 7: 0.474*** (0.132)
  - 8: 0.519*** (0.115)
  - 9: 0.401*** (0.112)
  - 10: 0.144 (0.276)
  - 11: 0.49*** (0.125)
  - 12: 0.423** (0.183)
- Other Islands Differential with Tourism-centric at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: -0.193 (0.171)
  - 1: -0.042 (0.12)
  - 2: -0.015 (0.088)
  - 3: -0.316*** (0.057)
  - 4: -0.447** (0.201)
  - 5: -0.503*** (0.107)
  - 6: -0.397** (0.159)
  - 7: -0.234 (0.152)
  - 8: -0.502*** (0.117)
  - 9: -0.609*** (0.1)
  - 10: -0.29 (0.248)
  - 11: -0.544*** (0.118)
  - 12: -0.537*** (0.175)

### IV regressions Results (Table 19)
- Impact on Tourism-centric Islands at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: -0.01 (0.266)
  - 1: 0.026 (0.271)
  - 2: 0.2 (0.273)
  - 3: 0.523* (0.28)
  - 4: 0.693*** (0.265)
  - 5: 0.585** (0.274)
  - 6: 0.569** (0.274)
  - 7: 0.41 (0.285)
  - 8: 1.158*** (0.302)
  - 9: 0.287 (0.315)
  - 10: 0.359 (0.308)
  - 11: 0.707** (0.317)
  - 12: 0.456 (0.315)
- Other Islands Differential with Tourism-centric at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: -0.004 (0.319)
  - 1: 0.256 (0.326)
  - 2: -0.164 (0.323)
  - 3: -0.44 (0.325)
  - 4: -0.632** (0.307)
  - 5: -0.618* (0.317)
  - 6: -0.561* (0.316)
  - 7: -0.501 (0.329)
  - 8: -1.054*** (0.348)
  - 9: -0.928** (0.363)
  - 10: -0.426 (0.357)
  - 11: -0.923** (0.363)
  - 12: -0.744** (0.359)

### IV First Stage F-statistics Across Horizons (Table 20)
- Horizons 0 through 12 (F-statistics):
  - Horizon 0: Impact on Tourism-centric Islands = 28.968; Other Islands Differential with Tourism-centric = 24.003
  - Horizon 1: Impact on Tourism-centric Islands = 28.645; Other Islands Differential with Tourism-centric = 23.679
  - Horizon 2: Impact on Tourism-centric Islands = 27.508; Other Islands Differential with Tourism-centric = 23.375
  - Horizon 3: Impact on Tourism-centric Islands = 26.011; Other Islands Differential with Tourism-centric = 23.040
  - Horizon 4: Impact on Tourism-centric Islands = 25.376; Other Islands Differential with Tourism-centric = 22.613
  - Horizon 5: Impact on Tourism-centric Islands = 24.872; Other Islands Differential with Tourism-centric = 22.162
  - Horizon 6: Impact on Tourism-centric Islands = 24.322; Other Islands Differential with Tourism-centric = 21.791
  - Horizon 7: Impact on Tourism-centric Islands = 23.900; Other Islands Differential with Tourism-centric = 21.390
  - Horizon 8: Impact on Tourism-centric Islands = 23.412; Other Islands Differential with Tourism-centric = 21.026
  - Horizon 9: Impact on Tourism-centric Islands = 22.998; Other Islands Differential with Tourism-centric = 20.599
  - Horizon 10: Impact on Tourism-centric Islands = 22.801; Other Islands Differential with Tourism-centric = 20.249
  - Horizon 11: Impact on Tourism-centric Islands = 21.981; Other Islands Differential with Tourism-centric = 19.984
  - Horizon 12: Impact on Tourism-centric Islands = 21.165; Other Islands Differential with Tourism-centric = 19.533

### Non Tourism-centric Islands Specification (Table 21)
- Direct impact from Local Tourism at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.103 (0.221)
  - 1: 0.072 (0.102)
  - 2: 0.356*** (0.114)
  - 3: 0.003 (0.24)
  - 4: 0.032 (0.202)
  - 5: -0.184 (0.248)
  - 6: 0.248 (0.352)
  - 7: 0.219 (0.576)
  - 8: -0.048 (0.208)
  - 9: -0.164 (0.374)
  - 10: -0.107 (0.269)
  - 11: -0.097 (0.239)
  - 12: -0.18 (0.346)
- Indirect Impact from Tourism-centric Islands at horizons 0 through 12 (Coefficient (Standard Error)):
  - 0: 0.202 (0.387)
  - 1: 0.11 (0.392)
  - 2: -0.337* (0.189)
  - 3: -0.097 (0.44)
  - 4: -0.013 (0.249)
  - 5: 0.167 (0.236)
  - 6: -0.301* (0.175)
  - 7: 0.032 (0.413)
  - 8: 0.457** (0.213)
  - 9: -0.475 (0.315)
  - 10: 0.243 (0.335)
  - 11: 0.354 (0.337)
  - 12: 0.087 (0.563)

### Event Study
- Figure 40: Event Study: Response of local economy Around March 2020 (treated units: Sal and Boa Vista). Source: IMF and authors’ own calculation.

*Long and Short-term Impact of Tourism on Growth in Small Developing States — Working Paper No. WP/2025/103*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025103-print-pdf.pdf_