## 7. Selected Dynamic Multipliers

## Source details

**Canonical URL:** [7. Selected Dynamic Multipliers](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021218-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021218-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021218-print-pdf.pdf.json)

---

### Introduction and objective
- Objective: quantify the impact of major external and exogenous shocks on ECCU economies, including shocks from foreign direct investment (FDI) and passport (Citizen-by-Investment, CBI) sales, using Canonical-Correlation Analysis (CCA) and dynamic panel regression analysis (panel VAR-X).
- Data coverage: annual data covering the periods between 1980 and 2017.
- Key conclusion preview: statistically and economically significant effects of external shocks on output; only oil-price fluctuations have significant effects on overall CPI; FDI and passport sales significantly affect the external balance.

### Stylized facts and historical episodes
- Tourism / Global activity:
  - Hotels and restaurants’ real GDP declined by more than ten percent between 2008 and 2009.
  - Transportation real GDP contracted by about seven percent between 2008 and 2009.
  - Wholesale and trade fell by about ten percent between 2008 and 2011.
  - Overall real GDP contracted by about five percent between 2008 and 2011.
- Oil prices:
  - 2014–2015 oil-price decline of approximately fifty percent → large decline in value of fuel imports → about a ten percent decline in ECCU total goods imports and a significant narrowing of the current account balance; overall deflation in consumer prices recorded.
- FDI:
  - Early 2000s FDI windfall: FDI rose from about EC$ 200 million in 2000 to EC$ 1,400 million in 2003; accompanied by almost EC$ 2 billion increase in imports; construction GDP grew more than twenty percent between 2002 and 2004.
- Natural disasters:
  - Hurricane Maria (2017) in Dominica caused damage equivalent to 270 percent of GDP and a real GDP contraction near 10 percent.
- Citizen-by-Investment (CBI) programs:
  - St. Kitts and Nevis: CBI fees rose from about two percent of GDP in 2010 to twelve percent in 2013; FDI related to CBI grew by approximately 70 percent between 2010-2012 and 2013; CBI-boosted real GDP growth reached around 6 percent of GDP in 2013-2014 (compared with a 3 percent annual average in 2000-2008); CBI revenues declined to about 6 percent of GDP in 2016-2017.

### Data, methods, and identification
- Variables:
  - Internal: Real GDP of ECCU, CPI, REER, Reserves as percent of GDP, Current Account as percent of GDP.
  - External: Number of Tourist Arrivals, Real GDP of US/UK/CAN, Oil Prices, FDI inflows, CBI Revenues, Natural Disaster Cost in percent of GDP.
- Techniques:
  - Canonical-Correlation Analysis (CCA): explores contemporaneous multivariate associations between external (X) and internal (Y) groups; variables transformed into stationary expressions where applicable.
  - Dynamic panel regressions (panel VAR-X with individual fixed effects): includes lagged Y, contemporaneous and lagged X, individual fixed effects; provides impulse response functions and dynamic multipliers.
- Econometric choices:
  - Preferred fixed-effects specification for panel VAR-X due to stability; robustness checks include Pooled-OLS and Arellano-Bond.

### Canonical-Correlation Analysis (CCA) — key statistics
- Canonical correlations (five pairs): 0.71, 0.53, 0.44, 0.28, 0.01.
- Interpretations:
  - 0.71 (highest): current account balance fluctuations linked to external flows—oil prices, FDI inflows, and CBI receipts.
  - 0.53 (second): reserves/GDP strongly associated with FDI inflows and CBI receipts.
  - 0.44 (third): local GDP growth positively correlated with trade partners’ GDP (U.S., U.K., Canada), negatively affected by higher oil prices and natural disaster costs.
  - 0.28 (fourth): relationship between oil prices and internal prices (CPI), current account balance, and REER.

### Panel VAR-X baseline results — selected dynamic multipliers and elasticities
- Tourism / source-country GDP channel:
  - A 1-percent increase in GDP of the set {U.S., U.K., Canada} → about a 0.5 percent increase in ECCU GDP.
- Oil prices:
  - A 10 percent increase in oil prices → associated with a 0.15 p.p. reduction in real GDP growth.
  - A 10 percent increase in oil prices → increases inflation by around 0.3 p.p. in the same year.
- Natural disasters:
  - A hurricane causing damage equal to 50 percent of GDP → model predicts a 1.8 percent fall in GDP.
  - Using the model: the 270 percent of GDP damage from Hurricane Maria → predicted an 8 percent economic contraction (10 percent below potential), close to the actual 10 percent contraction.
- FDI shocks:
  - A 5 percentage-point (pp) positive shock in FDI/GDP → could boost GDP by around 0.8 p.p.
  - A 5-pp positive shock in FDI/GDP → translates into a lower current account balance by 2.5 percent of GDP (implying a 50 percent imported component of FDI).
- Model fit and statistical notes:
  - Baseline fixed-effects specification explains volatility well for most endogenous variables (less so for International Reserves as percent of GDP).
  - Cointegration tests cannot reject the null hypothesis of at least one cointegration vector (see Annex I).
  - Comparisons across Pooled-OLS, Fixed-Effects, and Arellano-Bond show no substantial differences; fixed-effects preferred for stability.

### Annex I — selected econometric coefficients and diagnostics
- Representative lagged dependent-variable coefficients (Fixed-Effects Panel VAR-X):
  - Log of GDP in USD (constant prices) = L,0.8577*** (p-value (0.00))
  - CPI (log) = L,-0.1047** (p-value (0.01))
  - REER (log) = L,0.8599*** (p-value (0.00))
  - Reserves/GDP = L,-0.2454** (p-value (0.03))
  - Current Account/GDP = L,0.0012*** (p-value (0.00))
- Selected contemporaneous external-variable coefficients (Fixed-Effects):
  - Real GDP Index (US,UK,CAN; log) 0.6495*** (0.01)
  - Oil price (log) -0.0156*** (0.00)
  - FDI inflows /GDP 0.0012** (0.02)
  - Natural disaster costs/GDP -0.0004*** (0.00)
- Unit-root and cointegration tests (Annex tables):
  - Panel unit-root p-values reported for multiple variables (examples):
    - GDP: 0.1553 / 0.1096 ; 0.0000 / 0.0000
    - CPI: 0.0000 / 0.0000 ; 0.0188 / 0.0068
    - FDI/GDP: 0.0000 / 0.0000 ; 0.0000 / 0.0000
  - Pedroni and Kao cointegration test statistics and p-values reported (examples):
    - Modified Phillips-Perron t 2.05170 (p-value 0.0201)
    - Augmented Dickey-Fuller t -2.74240 (p-value 0.003)
- Robustness checks (selected comparisons):
  - Log of GDP in USD (lagged): Pooled OLS L,1.0042*** (0.00); Fixed-Effects L,0.8577*** (0.00); Arellano-Bond L,0.8664*** (0.00).
  - Oil price (log) contemporaneous negative coefficients across specifications (e.g., -0.0110**, -0.0156***); lagged positive coefficients (e.g., 0.0203***, 0.0160**).
  - Natural disaster costs/GDP: contemporaneous negative and significant across specifications (e.g., -0.0003***, -0.0004***); lagged small positive coefficients (e.g., 0.0003*, 0.0003).
  - Current Account/GDP: contemporaneous FDI inflows /GDP negative and significant (e.g., -0.5483***, -0.5671***, -0.5853***).

### COVID-19 application — scenarios, assumptions, and simulation outcomes
- Rationale: COVID-19 is a tourism-specific demand shock not fully captured by source-country GDP; alternative specification includes Hotels & Restaurants GDP as proxy for tourist arrivals and orthogonalizes tourist arrivals and source-country GDP.
- Considered channels: source-country GDP contraction, near-halt of tourist arrivals, reductions in FDI, oil-price fluctuations; CBI demand uncertain (monthly data indicate no significant change in CBI demand since pandemic start).
- Simulation scenarios:
  - i) Combined shock: halt of tourism + contraction in global economy + reduction of oil prices + reduction of FDI inflows.
  - ii) Tourism halt only (orthogonal to global slowdown).
- Key assumptions:
  - For 2020, tourism effectively shut down for three quarters → decline in tourist arrivals of about three quarters of the 2019 level.
  - Shocks assumed to materialize in one period (t=1) and economy returns to previous state in subsequent periods (model flexible for multi-period shocks).
- Simulation outcomes (summary):
  - GDP collapse broadly in line with international/national projections.
  - Differential country impacts driven by tourism share of GDP:
    - Most affected: Antigua and Barbuda; St. Lucia.
    - Moderately affected: St. Kitts and Nevis; Grenada.
    - Mildly affected: Dominica; St. Vincent and the Grenadines.
  - Current account balances projected to deteriorate in all ECCU countries under assumed shock combinations.
  - Average reduction in number of tourist arrivals to the Caribbean up to October 2020: around 70 percent (y-o-y).
- Offsetting considerations:
  - Historically oil prices and global economic activity have been positively correlated; large oil-price declines could partially offset current-account deterioration, but tourism collapse dominates in the COVID-19 simulations and leads to current-account deterioration.

### Policy-relevant implications and recommendations
- External shocks (tourism, source-country GDP, oil prices, FDI, CBI receipts, natural disasters) are central determinants of volatility in ECCU economies and explain much of short-run output variability.
- COVID-19 exceptionalism: abrupt halt in tourist arrivals explains most of the projected collapse in ECCU output; travel restrictions and external source-country policies are largely exogenous to ECCU policymakers.
- Forecasting and preparedness:
  - Project realistic combinations of simultaneous external shocks for policy planning.
  - Use econometric models that incorporate historically observed interrelatedness among global growth, tourism demand, and oil prices.
  - Improve data frequency and methodological homogeneity of external data to increase precision of estimated elasticities and better capture dynamics across shocks.

*Source: 7. Selected Dynamic Multipliers, wpiea2021218-print-pdf (annual data 1980–2017) — IMF chapter text provided.*

### REFERENCES .............................................................................................................

### REFERENCES .................................................................................................................25

### TABLES
- 1. Data Deninition and Sources .........................................................................................14
- 2. Canonical Correlation Analysis-Standardized Coeff icients .............................................16

### FIGURES
- 1. Impact of 2009 Tourism Shock ........................................................................................8
- 2. Impact of  Oil Price Slump in 2014-15 ..............................................................................9
- 3. Impact of FDI Windfall in Early 2000s ..........................................................................10
- 4. Impact of Storm Erika (2015) and Hurricane Maria (2017) in Dominica .........................12
- 5.  Impact of CBI Windf all in St. Kitts and Nevis...............................................................13
- 6. Problem Setup ..............................................................................................................16

*Canonical source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021218-print-pdf.pdf*

### 7. Selected Dynamic Multipliers........................................................................................1

### 7. Selected Dynamic Multipliers

### Introduction
- Objective: quantify the impact of major external and exogenous shocks on ECCU economies, including shocks from foreign direct investment (FDI) and passport (Citizen-by-Investment, CBI) sales, using Canonical-Correlation Analysis (CCA) and dynamic panel regression analysis (panel VAR-X).
- Data coverage: annual data covering the periods between 1980 and 2017.
- Key conclusion preview: statistically and economically significant effects of external shocks on output; only oil-price fluctuations have significant effects on overall CPI; FDI and passport sales significantly affect the external balance.

### Background and literature
- Small-state features increasing volatility: high trade dependence, granular production, frequent natural disasters, no fossil-fuel endowments, high reliance on tourism.
- Prior empirical findings highlighted:
  - Hnatkovska and Loayza (2005): a one standard deviation rise in output volatility → 1.3 percentage point drop in real GDP growth.
  - McIntyre and others (2016): real oil price changes explain, on average, 7 percent of real GDP growth variation in the Caribbean (variation across countries: 15 percent in Dominica to less than 1 percent in Guyana).
  - Thacker and others (2012): a 10 percent increase in tourist arrivals per capita → about 0.2 percentage point rise in economic growth for Caribbean countries.
- Motivation: need to quantify shocks’ impact across multiple macro variables and to include understudied shocks (FDI, passport sales).

### Stylized facts (selected channels and historical episodes)
- Global economic activity (tourism channel)
  - Global Financial Crisis (GFC) impact: hotels and restaurants’ real GDP declined by more than ten percent between 2008 and 2009; transportation real GDP contracted by about seven percent between those years; wholesale and trade fell by about ten percent between 2008 and 2011; overall real GDP contracted by about five percent between 2008 and 2011.
  - Tourism shocks can be income-driven or demand-driven (e.g., 9/11, COVID-19).
- Oil prices
  - 2014–2015 oil-price decline: approximately fifty percent decline in international oil prices translated into a similarly large decline in the value of fuel imports, contributing to a ten percent decline in ECCU total goods imports and a significant narrowing of the current account balance; overall deflation in consumer prices recorded.
  - Note: 2014 drop largely related to rapidly growing oil supply (U.S.) and shifting OPEC policies.
- Foreign Direct Investment (FDI)
  - Early 2000s FDI windfall: FDI rose from about EC$ 200 million in 2000 to EC$ 1,400 million in 2003; FDI increase accompanied by almost EC$ 2 billion increase in imports; construction GDP grew more than twenty percent between 2002 and 2004.
- Natural disasters
  - Example: Hurricane Maria (2017) in Dominica caused damage equivalent to 270 percent of GDP and a real GDP contraction near 10 percent.
  - Natural disasters can trigger reconstruction-related rebounds; severe disasters always show negative growth impacts.
- Citizen-by-Investment (CBI) programs
  - St. Kitts and Nevis example: CBI fees rose from about two percent of GDP in 2010 to twelve percent in 2013; FDI related to CBI grew by approximately 70 percent between 2010-2012 and 2013; CBI-boosted real GDP growth reached around 6 percent of GDP in 2013-2014 (higher than the 3 percent annual average in 2000-2008); CBI revenues declined to about 6 percent of GDP in 2016-2017.

### Data and methodology
- Data sources: WEO-IMF database and additional sources; variables split into internal (Real GDP of ECCU, CPI, REER, Reserves as percent of GDP, Current Account as percent of GDP) and external (Number of Tourist Arrivals, Real GDP of US/UK/CAN, Oil Prices, FDI inflows, CBI Revenues, Natural Disaster Cost in percent of GDP).
- Two complementary techniques:
  - Canonical-Correlation Analysis (CCA): explores contemporaneous multivariate associations between external (X) and internal (Y) groups; variables transformed into stationary expressions where applicable.
  - Dynamic panel regressions (panel VAR-X with individual fixed effects): specification includes lagged Y, contemporaneous and lagged X, individual fixed effects; provides impulse response functions and dynamic multipliers.

### Main results

#### Canonical-Correlation Analysis (CCA) — key statistics and interpretations
- Canonical correlations (five pairs): 0.71, 0.53, 0.44, 0.28, 0.01.
- Interpretations of the top canonical correlations:
  - 0.71 (highest): current account balance fluctuations linked to external flows—oil prices, FDI inflows, and CBI receipts.
  - 0.53 (second): reserves/GDP strongly associated with FDI inflows and CBI receipts.
  - 0.44 (third): local GDP growth positively correlated with trade partners’ GDP (U.S., U.K., Canada), negatively affected by higher oil prices and natural disaster costs.
  - 0.28 (fourth): relationship between oil prices and internal prices (CPI), current account balance, and REER.

#### Baseline econometric (panel VAR-X) results — selected dynamic multipliers and elasticities
- Tourism/source-country GDP channel:
  - A 1-percent increase in GDP of the set {U.S., U.K., Canada} → about a 0.5 percent increase in ECCU GDP.
- Oil prices:
  - A 10 percent increase in oil prices → associated with a 0.15 p.p. reduction in real GDP growth.
  - A 10 percent increase in oil prices → increases inflation by around 0.3 p.p. in the same year.
- Natural disasters:
  - A hurricane causing damage equal to 50 percent of GDP → model predicts a 1.8 percent fall in GDP.
  - Using the model: the 270 percent of GDP damage from Hurricane Maria → predicted an 8 percent economic contraction (10 percent below potential), close to the actual 10 percent contraction.
- FDI shocks:
  - A 5 percentage-point (pp) positive shock in FDI/GDP → could boost GDP by around 0.8 p.p.
  - A 5-pp positive shock in FDI/GDP → translates into a lower current account balance by 2.5 percent of GDP (implying a 50 percent imported component of FDI).
- Model fit and robustness:
  - Baseline fixed-effects specification explains volatility well for most endogenous variables (less so for International Reserves as percent of GDP).
  - Cointegration tests cannot reject the null hypothesis of at least one cointegration vector (see Annex references).
  - Comparisons across Pooled-OLS, Fixed-Effects, and Arellano-Bond show no substantial differences; preferred fixed-effects due to stability.

### Extensions, robustness checks, and interpretation caveats
- Addressed potential Nickell bias and stationarity concerns; mixed set of stationary and integrated variables present.
- Chosen fixed-effects specification balances bias correction and coefficient stability (GMM instrumentation led to less stable estimates given limited variability in some instruments).
- CCA useful for exploratory contemporaneous links; panel VAR-X adds dynamics needed for projections.

### Application to the COVID-19 shock
- Rationale: COVID-19 is a tourism-specific demand shock (travel as virus transmission vector) that is not fully captured by source-country GDP; alternative specification includes Hotels & Restaurants GDP as proxy for tourist arrivals and orthogonalizes tourist arrivals and source-country GDP.
- Considered channels affected by COVID-19: source-country GDP contraction, near-halt of tourist arrivals, reductions in FDI, oil-price fluctuations; CBI demand uncertain (monthly data indicate no significant change in CBI demand since pandemic start).
- Simulation scenarios:
  - i) Combined shock: halt of tourism + contraction in global economy + reduction of oil prices + reduction of FDI inflows.
  - ii) Tourism halt only (orthogonal to global slowdown).
- Assumptions used in simulations:
  - For 2020, tourism effectively shut down for three quarters → decline in tourist arrivals of about three quarters of the 2019 level.
  - Shocks are assumed to materialize in one period (t=1) and economy returns to previous state in subsequent periods (model flexible for multi-period shocks).
- COVID-19 simulation outcomes (summary):
  - GDP collapse broadly in line with international/national projections.
  - Differential country impacts driven by tourism share of GDP:
    - Most affected: Antigua and Barbuda; St. Lucia.
    - Moderately affected: St. Kitts and Nevis; Grenada.
    - Mildly affected: Dominica; St. Vincent and the Grenadines.
  - Current account balances projected to deteriorate in all ECCU countries under assumed shock combinations.
  - Average reduction in number of tourist arrivals to the Caribbean up to October 2020: around 70 percent (y-o-y).
- Notes on potential offsetting effects:
  - Historically oil prices and global economic activity have been positively correlated; large oil-price declines could partially offset CA deterioration, but tourism collapse in COVID-19 simulations dominates and leads to CA deterioration.

### Concluding remarks and policy-relevant implications
- External shocks (tourism, source-country GDP, oil prices, FDI, CBI receipts, natural disasters) are central determinants of volatility in ECCU economies and explain much of short-run output variability.
- COVID-19 exceptionalism: abrupt halt in tourist arrivals explains most of the projected collapse in ECCU output; travel restrictions and external source-country policies are largely exogenous to ECCU policymakers.
- Forecasting suggestions:
  - Projecting realistic combinations of simultaneous external shocks is valuable for policy planning; econometric models that incorporate historically observed interrelatedness among global growth, tourism demand, and oil prices are useful.
- Data and methodological improvements (higher frequency data, methodologically homogeneous external data) would increase precision of estimated elasticities and better capture dynamics across shocks.

*Source: 7. Selected Dynamic Multipliers, wpiea2021218-print-pdf (annual data 1980–2017) — IMF chapter text provided.*

### REFERENCES

### REFERENCES

### Key references
- Acevedo, Sebastian, 2014, “Debt, Growth and Natural Disasters: A Caribbean Trilogy" IMF Working Paper 14/125.
- Aghion, Philippe, Philippe Bacchetta, Romain Rancière, and Kenneth Rogoff, 2009, "Exchange rate volatility and productivity growth: The role of financial development," Journal of Monetary Economics, vol. 56(4), pages 494-513, May.
- Arellano M. and Stephen Bond, 1991, "Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations," Review of Economic Studies, vol. 58(2), pages 277-297.
- Beaton, Kimberly, Alla Myrvoda and Shernnel Thompson, 2016, “Non-Performing Loans in the ECCU: Determinants and Macroeconomic Impact”, IMF Working Paper 16/229.
- Briguglio, L., 1995, "Small island developing states and their economic vulnerabilities," World Development, 23(9), 1615–1632.
- Cantelmo, Alessandro, Leo Bonato, Giovanni Melina, and Gonzalo Salinas, 2019, "Policy Trade-Offs in Building Resilience to Natural Disasters: The Case of St. Lucia," IMF Working Paper 2019/054.
- Endegnanew, Y., C. Amo-Yartey, and T. Turner-Jones, 2012, “Fiscal Policy and the Current Account: Are Micro-States Different?” IMF Working Paper 12/51.
- Fomby, Thomas B., Ikeda, Yuki and Loayza, Norman, 2009, "The Growth Aftermath of Natural Disasters," World Bank Policy Research Working Paper No. 5002.
- Guerson, Alejandro, 2020, “Government Insurance Against Natural Disasters: An Application to the ECCU,” IMF Working Paper No. 2020/266.
- Komatsuzaki, Takuji and Steve Brito, 2019, “Financial Conditions and Growth at Risk in the ECCU”, IMF Working Paper No. 19/247.
- McIntyre, Arnold, et al., 2016, "Caribbean Energy; Macro-Related Challenges," IMF Working Paper 2016/053.
- Narita, Machiko, 2014, “Fiscal Multipliers in the Caribbean,” chapter in Caribbean Renewal: Tackling Fiscal and Debt Challenges.
- Thacker, N., S. Acevedo, and R. Perrelli, 2012, “Caribbean Growth in an International Perspective: The Role of Tourism and Size,” IMF Working Paper 12/235.
- Worrell, D., 2012, “Policies for Stabilization and Growth in Small Very Open Economies,” (Washington: Group of Thirty).

### Annex I. Econometric Results — Tables and key statistics
- Table A1. Panel VAR-X (Estimated using fixed effects)
  - Dependent-variable coefficient examples (lagged):
    - Log of GDP in USD (constant prices) = L,0.8577*** (p-value (0.00))
    - CPI (log) = L,-0.1047** (p-value (0.01))
    - REER (log) = L,0.8599*** (p-value (0.00))
    - Reserves/GDP = L,-0.2454** (p-value (0.03))
    - Current Account/GDP = L,0.0012*** (p-value (0.00))
  - Other reported coefficients (contemporaneous and lags) include:
    - Real GDP Index (US,UK,CAN; log) 0.6495*** (0.01); = L,-0.4180** (0.02)
    - Oil price (log) -0.0156*** (0.00); = L,0.0160** (0.03)
    - FDI inflows /GDP 0.0012** (0.02)
    - Natural disaster costs/GDP -0.0004*** (0.00); = L,0.0003 (0.11)
  - Observations: 210 (columns 1–4), 204 (column 5)
  - R-squared: 0.9919, 0.9960, 0.8544, 0.3085, 0.6461
  - Number of ifs_code: 6
  - Robust p-values in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1

- Table A2. Fixed Effects Panel VAR-X (With Tourism Sector GDP)
  - Selected coefficients (lagged and contemporaneous):
    - Log of GDP in USD (constant prices) = L,0.9179*** (0.00)
    - CPI (log) = L,0.8362*** (0.00)
    - REER (log) = L,0.7783*** (0.00)
    - Current Account/GDP = L,0.0014** (0.02)
    - Hotels & Rest. real GDP 0.0009** (0.04); = L,-0.0007* (0.07)
    - Oil price (log) 0.0222* (0.06); = L,-0.0236* (0.08)
    - Natural disaster costs/GDP -0.0002* (0.06); = L,0.0003* (0.07)
  - Observations: 168 (columns 1–4), 162 (column 5)
  - R-squared: 0.9801, 0.9937, 0.8145, 0.3456, 0.7023
  - Number of ifs_code: 6

- Table A3. Unit Root Tests of Panel-Data Variables
  - (Table heading present; no numeric entries beyond table numbering in this section.)

- Table A4. Unit Root Tests of Time-Series Variables (P-values reported)
  - Im-Pesaran-Shin and Fisher test p-values — Levels / First differences:
    - GDP: 0.1553 / 0.1096 ; 0.0000 / 0.0000
    - GDP Tourism Contribution: 0.6338 / 0.8559 ; 0.0000 / 0.0000
    - CPI: 0.0000 / 0.0000 ; 0.0188 / 0.0068
    - REER: 0.1020 / 0.1734 ; 0.0000 / 0.0000
    - Reserves/GDP: 0.0000 / 0.0000 ; 0.0000 / 0.0000
    - Current Account/GDP: 0.0000 / 0.0000 ; 0.0000 / 0.0000
    - FDI/GDP: 0.0000 / 0.0000 ; 0.0000 / 0.0000
    - CBI/GDPa/: 0.1129 / 0.0000 ; 0.0000
    - Nat. Disaster Costs/GDP: 0.0000 / 0.0000 ; 0.0000 / 0.0000
  - Time-series unit-root p-values:
    - GDP (US-UK-CAN): Levels 0.8802 / 0.7634 ; First differences 0.0001 / 0.0004
    - Oil Price: Levels 0.0640 / 0.0878 ; First differences 0.0100 / 0.0000
  - Note: Reporting P-values, H0: All panels contain unit roots; HA: Some panels are stationary. a/ Insufficient number of time periods to compute.

- Table A5. Cointegration Tests
  - Pedroni test for cointegration (Number of panels = 6; Avg. number of periods = 35; Lags: 3.00 (Newey-West))
    - Modified Phillips-Perron t 2.05170 (p-value 0.0201)
    - Phillips-Perron t 0.93280 (p-value 0.1755)
    - Augmented Dickey-Fuller t 1.25730 (p-value 0.1043)
  - Kao test for cointegration (Number of panels = 6; Avg. number of periods = 34; Lags: 2.50 (Newey-West))
    - Modified Dickey-Fuller t -1.25530 (p-value 0.1047)
    - Dickey-Fuller t -2.08110 (p-value 0.0187)
    - Augmented Dickey-Fuller t -2.74240 (p-value 0.003)
    - Unadjusted modified Dickey -3.08360 (p-value 0.001)
    - Unadjusted Dickey-Fuller t -2.90780 (p-value 0.0018)

- Table A6. Robustness Checks (Various specifications: Pooled OLS, Fixed-Effects, Arellano-Bond)
  - Panel highlights (selected coefficient values and statistics):
    - Log of GDP in USD (constant prices) — lagged coefficients:
      - Pooled OLS L,1.0042*** (0.00)
      - Fixed-Effects L,0.8577*** (0.00)
      - Arellano-Bond L,0.8664*** (0.00)
    - CPI (log) — lagged:
      - Pooled OLS L,-0.0855** (0.01)
      - Fixed-Effects L,-0.1047** (0.01)
      - Arellano-Bond L,-0.0933*** (0.00)
    - Oil price (log): pooled and fixed negative contemporaneous coefficients (e.g., -0.0110**, -0.0156***), lagged positive coefficients (e.g., 0.0203***, 0.0160**).
    - Natural disaster costs/GDP: contemporaneous negative and significant across specifications (e.g., -0.0003***, -0.0004***); lagged small positive coefficients (e.g., 0.0003*, 0.0003).
    - Observations: 210 (most specifications), 206 for Arellano-Bond in some columns.
    - R-squared reported up to 0.9960 and 0.9919 for different dependent-variable regressions.
  - Additional dependent-variable panels (CPI (log), REER (log), Reserves/GDP, Current Account/GDP) each reported with pooled OLS, fixed-effects, and Arellano-Bond columns and associated coefficient estimates and robust p-values.
  - Examples from Reserves/GDP specification:
    - Reserves/GDP = L,-0.2454** (0.03) in Fixed-Effects; Arellano-Bond L,-0.2420*** (0.00)
    - CBI revenue/GDP 1.5049*** (0.00)
    - Observations: 210; R-squared: 0.2963, 0.3085 in different specifications.
  - Examples from Current Account/GDP specification:
    - Current Account/GDP = L,0.5414*** (Pooled OLS) and L,0.5001*** (Fixed-Effects); Arellano-Bond L,0.5011***.
    - FDI inflows /GDP contemporaneous negative and significant (e.g., -0.5483***, -0.5671***, -0.5853***).
    - Natural disaster costs/GDP = L,-0.0631** (Pooled OLS), L,-0.0589** (Fixed-Effects), L,-0.0541*** (Arellano-Bond).
    - Observations: 210 (Pooled OLS and Fixed-Effects), 206 (Arellano-Bond); R-squared up to 0.6669 and 0.6517.

*Source: wpiea2021218-print-pdf — REFERENCES and Annex I tables (authors’ calculations).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021218-print-pdf.pdf_
