## wpiea2024212-print-pdf

## Source details

**Canonical URL:** [wpiea2024212-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024212-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024212-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024212-print-pdf.pdf.json)

---

### Overview
- Soft power is defined as the ability to obtain preferred outcomes by attraction rather than coercion or payment.
- The paper develops a Global Soft Power Index (GSPI) to measure soft power across countries and over time.
- GSPI covers the period 1990–2021 and is built from 29 indicators along six dimensions.
- Index construction follows a three-step approach:
  - (i) normalization of variables;
  - (ii) aggregation of normalized variables into sub-indices representing a particular dimension; and
  - (iii) aggregation of the sub-indices into the final index.

### Data coverage and scope
- Six GSPI dimensions: Commercial, Culture, Digital, Education, Global Reach, and Institutions.
- Total variables: 29.
- Balanced panel constructed for 66 countries over the period 2007–2021 (data availability constraints).
- For variables released less often than annually (e.g., Olympic medals), latest available information is used as the yearly realization.
- Where data are not available for the latest year (e.g., 2021), values are set equal to the latest available observations (e.g., 2020).
- The authors prioritize constructing indices only from dates where information is available for all variables of interest to yield a balanced dataset.

### Index construction and methodology
- Normalization:
  - z-score standardization: z_{i,t} = (x_{i,t} − x̄_{i, t}) / σ_{i, t}.
  - All variables and sub-indices are standardized (cross-sectional z-scores).
- Aggregation:
  - Sub-indices are constructed for each of the six dimensions using only linear combinations.
  - The headline GSPI is the simple average of the sub-indices.
  - Final step: subtract the median score from the headline GSPI so the overall index has 0 as its median score.
- Weighting:
  - Principal component analysis (PCA) selects weights, following OECD (2008) guidance.
  - PCA groups collinear indicators and extracts factors (eigenvectors) associated with largest eigenvalues to explain variation.
  - Retain all factors that contribute individually to at least 10 percent of the overall variance.
  - Small factor loadings below 0.10 are set to 0.00 (Nicoletti et al., 1999).
  - Final variable weights in sub-indices are computed as weighted sums of selected factor weights using explained variances as weights.
- Aggregation formula:
  - GSPI_i,t = (sum over s of Sub-Index_i,t,s) / n_s, where s = {commercial, culture, digital, education, global reach, and institutions}.

### Key descriptive findings and statistics
- Cross-country variation:
  - GSPI (2021) ranges from a minimum of -0.59 in Dominican Republic to a maximum of 1.68 in South Korea.
  - Advanced economies tend to have higher soft power on average, with exceptions.
- GSPI distribution (2021, sample of 66 countries):
  - Minimum: -0.59
  - Maximum: 1.68
  - Median: 0
  - Standard deviation: 0.49
- Top GSPI scores (2021):
  - South Korea: 1.68
  - Japan: 1.25
  - Germany: 1.18
  - China: 1.17
- Bottom GSPI scores (2021):
  - Dominican Republic: -0.59
  - Algeria: -0.55
  - Philippines: -0.53
  - Azerbaijan: -0.47
- Example country trajectories (2004 → 2021):
  - China: 0.70 in 2004 → 1.17 in 2021.
  - United Kingdom: 1.32 in 2004 → 0.85 in 2021.

### PCA illustration and Education sub-index example
- PCA: Education sub-component variance explained (selected factors and shares):
  - Factor 1 — 71% — Education Expenditure: 1% — PISA: Maths: 20% — PISA: Reading: 20% — PISA: Science: 19% — Primary Completion: 0% — Journal Articles: 29% — Years of Schooling: 5% — Tertiary Education: 6%
  - Factor 2 — 12% — Education Expenditure: 11% — PISA: Maths: 13% — PISA: Reading: 10% — PISA: Science: 15% — Primary Completion: 0% — Journal Articles: 33% — Years of Schooling: 3% — Tertiary Education: 14%
  - Factor 3 — 7% — Education Expenditure: 0% — PISA: Maths: 0% — PISA: Reading: 0% — PISA: Science: 0% — Primary Completion: 1% — Journal Articles: 25% — Years of Schooling: 1% — Tertiary Education: 72%
  - Factor 4 — 4% — Education Expenditure: 30% — PISA: Maths: 0% — PISA: Reading: 2% — PISA: Science: 1% — Primary Completion: 1% — Journal Articles: 0% — Years of Schooling: 65% — Tertiary Education: 2%
  - Factor 5 — 3% — Education Expenditure: 53% — PISA: Maths: 0% — PISA: Reading: 0% — PISA: Science: 1% — Primary Completion: 10% — Journal Articles: 11% — Years of Schooling: 19% — Tertiary Education: 6%
  - Factor 6 — 2% — Education Expenditure: 4% — PISA: Maths: 0% — PISA: Reading: 0% — PISA: Science: 0% — Primary Completion: 87% — Journal Articles: 1% — Years of Schooling: 7% — Tertiary Education: 0%
  - Factor 7 — 1% — Education Expenditure: 0% — PISA: Maths: 36% — PISA: Reading: 60% — PISA: Science: 2% — Primary Completion: 0% — Journal Articles: 0% — Years of Schooling: 0% — Tertiary Education: 1%
  - Factor 8 — 0% — Education Expenditure: 0% — PISA: Maths: 30% — PISA: Reading: 8% — PISA: Science: 61% — Primary Completion: 0% — Journal Articles: 0% — Years of Schooling: 0% — Tertiary Education: 0%
- Sub-index weight calculation example (PISA: Maths):
  - w_PISA: Maths = (0.20*0.71)+(0.13*0.12) / 0.62+0.11 = 0.22
- Education sub-index final weights:
  - Education Expenditure: 0.02
  - Journal Articles: 0.33
  - PISA: Maths: 0.22
  - PISA: Reading: 0.19
  - PISA: Science: 0.21
  - Tertiary Education: 0.02
  - Primary Completion and Years of Schooling excluded because factor weights < 0.10.

### Weights for GSPI sub-components (reported percentages)
- Commercial Prowess:
  - Outward Foreign Investment — 28%
  - Patents — 68%
  - Trademarks — 4%
- Culture:
  - Cultural Exports — 26%
  - International Tourists — 0%
  - Olympic Medals — 0%
  - World Heritage Sites — 74%
- Digital Footprint:
  - Internet Access — 49%
  - Mobile Phones Access — 51%
- Education:
  - Education Expenditure — 2%
  - PISA: Maths — 22%
  - PISA: Reading — 19%
  - PISA: Science — 21%
  - Primary Completion — 0%
  - Journal Articles — 33%
  - Years of Schooling — 0%
  - Tertiary Education — 2%
- Global Reach:
  - Aid and Assistance — 0%
  - Embassies — 74%
  - Migrants — 26%
  - Refugees — 0%
  - Diplomatic Events — 0%
- Institutions:
  - Bureaucratic Effectiveness — 18%
  - Corruption — 18%
  - Democratic Accountability — 23%
  - Government Stability — 22%
  - Rule of Law — 19%

### Sample, clustering, and group interpretations
- Sample: 66 countries, yearly observations from 2007 to 2021.
- K-Means clustering (using 2021 sub-index values; number of clusters selected following Pham et al. (2005)) grouped the 66 countries into four groups:
  - Group 1 (low soft power): ALB, ARG, AZE, BGR, BLR, BRA, CHL, COL, CRI, DOM, DZA, IDN, JOR, KAZ, LBN, MAR, MDA, MEX, MYS, PAN, PER, PHL, QAT, ROU, SAU, SRB, THA, TTO, TUR, UKR, URY, VNM.
  - Group 2 (medium soft power): AUS, AUT, CAN, CHE, CYP, CZE, DNK, EST, FIN, HRV, HUN, ISL, ISR, LTU, LVA, MLT, NOR, NZL, POL, PRT, SGP, SVK, SVN, SWE.
  - Group 3 (high soft power): CHN, DEU, ESP, FRA, GBR, ITA, RUS, USA.
  - Group 4 (high soft power, distinct profile): JPN, KOR.
- Clustering interpretation:
  - Group 1: developing countries with low soft power; especially behind in Education and Institutions.
  - Group 2: mostly developed countries with medium soft power; high Education and Institutions but limited Culture and Global Reach impact.
  - Group 3: high soft power countries (advanced economies and powerful developing nations like China, Russia).
  - Group 4: Japan and South Korea — high soft power but with significantly higher Commercial Prowess and meaningfully lower Culture scores relative to Group 3.

### Application: soft power and real exchange rate (REER) volatility
- Annual REER volatility estimation:
  - VOL_i,t = sum_{m=1}^{12} r_{i,m-t}^2
  - where r_{i,m-t} = log(REER_{i,m-t}) − log(REER_{i,(m−1)-t}).
- Empirical framework:
  - Panel regression with country fixed effects and time fixed effects.
  - Dependent variable: annual REER volatility (VOL_i,t).
  - Controls (9 variables from Cevik, Harris, and Yilmaz (2017), alphabetical): credit, current account, export concentration, inflation, stock market capitalization, trade openness, volatility of government consumption, volatility of labor productivity growth, volatility of terms of trade.
  - Include autoregressive lags of VOL_i,t; lag length K selected via BIC (chosen K = 2).
  - Two model specifications:
    - Model 1: VOL_i,t = μ_i + γ_t + sum_{k=1}^{K} α_k VOL_i,t−k + δ' Z_i,t + β GSPI_i,t + ε_i,t
    - Model 2 (expanded): VOL_i,t = μ_i + γ_t + sum_{k=1}^{K} α_k VOL_i,t−k + δ' Z_i,t + β' X_i,t + ε_i,t, where X_i,t is the 6x1 vector of sub-indices.
- Main regression findings:
  - Model 1 (headline GSPI):
    - Overall GSPI coefficient is negative (higher GSPI → lower REER volatility) but not statistically significant; p-value = 0.22.
    - Most meaningful control variables: inflation, volatility of labor productivity growth, volatility of terms of trade; trade openness also relevant to a lesser extent.
  - Model 2 (sub-indices):
    - Culture and Global Reach sub-indices are significant in explaining REER volatility, with negative coefficients (higher Culture or Global Reach → lower REER volatility).
    - Global Reach appears to be the most significant sub-index of the two.
- Interpretation and policy-relevant insights:
  - Composite GSPI is a systematic, multidimensional measure comparable across countries and over time, but the headline GSPI alone does not show a statistically significant relationship with REER volatility once standard macroeconomic controls and fixed effects are included.
  - The Culture and Global Reach dimensions have statistically significant, negative associations with REER volatility, controlling for fundamentals and dynamics.
  - Global Reach—interpreted as the “purest” measure of international soft power reach—is the most impactful sub-index with respect to REER volatility.
  - Practical takeaway: Policy attention to enhancing Global Reach and Culture dimensions of soft power could be associated with lower exchange rate volatility, holding other macro fundamentals constant.

### Variables, data sources, and descriptive statistics (selected)
- Variables and Sources (Variable Name — Dimension — Source — Definition) include:
  - Outward Foreign Investment — Commercial Prowess — UNCTAD — Outward Foreign Investment
  - Patents — Commercial Prowess — World Bank — Number of international
  - Trademarks — Commercial Prowess — World Bank — Number of trademarks
  - Cultural Exports — Culture — UNCTAD — Exports of cultural goods
  - International Tourists — Culture — World Bank — Number of international
  - Olympic Medals — Culture — Olympic Committee — Total number of medals in
  - World Heritage Sites — Culture — UNESCO — Number of UNESCO World
  - Internet Access — Digital Footprint — World Bank — Number of internet users
  - Mobile Phones Access — Digital Footprint — World Bank — Number of mobile phones
  - Education Expenditure — Education — World Bank — Government expenditure on
  - Journal Articles — Education — World Bank — Number of journal articles
  - PISA: Maths — Education — World Bank — Mean performance on the
  - PISA: Reading — Education — World Bank — Mean performance on the
  - PISA: Science — Education — World Bank — Mean performance on the
  - Primary Completion — Education — World Bank — Primary completion rate
  - Tertiary Education — Education — World Bank — Gross tertiary educational
  - Years of Schooling — Education — World Bank — Barro-Lee: Average years of
  - Aid and Assistance — Global Reach — World Bank — Official development
  - Diplomatic Events — Global Reach — GDELT — Share of diplomatic cooperation
  - Embassies — Global Reach — Lowy Institute — Lowy Institute number of
  - Migrants — Global Reach — World Bank — Number of migrants
  - Refugees — Global Reach — World Bank — Number of refugees
  - Bureaucratic Effectiveness — Institutions — ICRG — The PRS Group International Country
  - Corruption — Institutions — ICRG — The PRS Group International Country
  - Democratic Accountability — Institutions — ICRG — The PRS Group International Country
  - Government Stability — Institutions — ICRG — The PRS Group International Country
  - Rule of Law — Institutions — ICRG — The PRS Group International Country
- Descriptive statistics (Variable Name — Start Date — Mean — Standard Deviation — 5th Percentile — 95th Percentile), selected:
  - Outward Foreign Investment — 1/1/1970 — 2.2825 — 5.87 — -0.25 — 5.27
  - Patents — 1/1/1980 — 0.01 — 0.03 — 0 — 0.04
  - Trademarks — 1/1/1980 — 0.29 — 1.4 — 0.53 — 0.53
  - Cultural Exports — 1/1/2007 — 0.17 — 0.38 — 0 — 0.63
  - International Tourists — 1/1/1995 — 123.63 — 37.74 — 0.53 — 644.25
  - Olympic Medals — 1/1/1960 — 9.45 — 22.6 — 0.19 — 33.24
  - World Heritage Sites — 1/1/1978 — 3.78 — 6.64 — 0 — 16
  - Internet Access — 1/1/1990 — 28.03 — 29.87 — 0.03 — 86.53
  - Mobile Phones Access — 1/1/1980 — 57.29 — 54.1 — 0.03 — 146.92
  - Education Expenditure — 1/1/1970 — 4.34 — 2.68 — 1.54 — 7.38
  - Journal Articles — 1/1/2000 — 0.03 — 0.05 — 0 — 0.15
  - PISA: Maths — 1/1/2000 — 461.39 — 59.83 — 361.53 — 543.8
  - PISA: Reading — 1/1/2000 — 458.29 — 55.26 — 358.31 — 526.89
  - PISA: Science — 1/1/2000 — 464.82 — 54.74 — 375.37 — 539.47
  - Primary Completion — 1/1/1970 — 78.41 — 24.78 — 28.31 — 105.67
  - Tertiary Education — 1/1/1970 — 22.16 — 23.21 — 0.68 — 72.45
  - Years of Schooling — 1/1/1970 — 6.18 — 3.29 — 1.09 — 11.53
  - Aid and Assistance — 1/1/1960 — 6.35 — 10.96 — 0.01 — 24.69
  - Diplomatic Events — 1/1/1979 — 9.5 — 17.13 — 0.97 — 18.18
  - Embassies — 1/1/2006 — 78.57 — 40.54 — 24 — 152
  - Migrants — 1/1/1960 — 11.13 — 15.87 — 0.26 — 50.33
  - Refugees — 1/1/1960 — 1.24 — 5.16 — 0 — 4.51
  - Bureaucratic Effectiveness — 1/1/1984 — 2.17 — 1.15 — 0 — 4
  - Corruption — 1/1/1984 — 2.92 — 1.32 — 0 — 5.5
  - Democratic Accountability — 1/1/1984 — 3.81 — 1.65 — 0 — 6
  - Government Stability — 1/1/1984 — 7.45 — 2.11 — 4 — 11
  - Rule of Law — 1/1/1984 — 3.65 — 1.42 — 0 — 6

*Source: IMF Working Paper — Measuring Soft Power: A New Global Index (Executive Summary and Appendix, wpiea2024212-print-pdf).*

### Executive Summary ......................................................................................................

### Executive Summary

### Overview
- Soft power is defined as the ability to obtain preferred outcomes by attraction rather than coercion or payment.  
- The paper develops a Global Soft Power Index (GSPI) to measure soft power across countries and over time.  
- GSPI covers the period 1990–2021 and is built from 29 indicators along six dimensions.  
- The index construction follows a standard three-step approach: (i) normalization of variables; (ii) aggregation of normalized variables into sub-indices representing a particular dimension; and (iii) aggregation of the sub-indices into the final index.

### Data coverage and scope
- Six GSPI dimensions: Commercial, Culture, Digital, Education, Global Reach, and Institutions.  
- Total variables: 29.  
- Balanced panel constructed for 66 countries over the period 2007–2021 (data availability constraints).  
- For variables released less often than annually (e.g., Olympic medals), latest available information is used as the yearly realization.  
- Where data are not available for the latest year (e.g., 2021), values are set equal to the latest available observations (e.g., 2020).  
- The authors prioritize constructing indices only from dates where information is available for all variables of interest to yield a balanced dataset.

### Index construction and methodology
- Normalization: z-score standardization is used to convert indicators to a common scale with mean 0 and standard deviation 1. The z-score transformation is described as z_{i,t} = (x_{i,t} − x̄_{i, t}) / σ_{i, t}.  
- Aggregation: sub-indices are constructed for each of the six dimensions; the GSPI is an aggregation of those sub-indices.  
- Weighting: principal component analysis (PCA) is used to select weights, following OECD (2008) guidance.  
  - PCA groups collinear indicators and extracts factors (eigenvectors) associated with largest eigenvalues to explain variation.  
  - The number of PCA factors retained follows the Handbook on Constructing Composite Indicators (OECD, 2008): keep all factors that contribute individually to at least 10 percent of the overall variance.  
  - Small factor loadings below 0.10 are set to 0.00 (Nicoletti et al., 1999), and final variable weights in sub-indices are computed as weighted sums of selected factor weights using explained variances as weights.

### Key descriptive findings
- Substantial cross-country variation in soft power.  
- GSPI (as of 2021) ranges from a minimum of -0.59 in Dominican Republic to a maximum of 1.68 in South Korea.  
- Advanced economies tend to have higher soft power on average, but there are important cross-country and temporal exceptions.  
- Example country trajectories:  
  - China: GSPI increased from 0.70 in 2004 to 1.17 in 2021.  
  - United Kingdom: GSPI declined from 1.32 in 2004 to 0.85 in 2021.

### Macro-financial application: soft power and exchange rates
- The GSPI is tested for macro-financial relevance via its impact on real effective exchange rate (REER) volatility.  
- Main empirical findings:  
  - The headline GSPI does not appear to be statistically significant in explaining REER volatility.  
  - The Culture and Global Reach sub-indices are clearly relevant at almost all levels of statistical significance in explaining REER volatility.  
  - Low soft-power countries present a dynamic between soft power and REER volatility that is significantly different from that for medium and high soft-power countries.  
  - Global Reach is the most significant sub-index and is responsible for most of the differentiation between medium and high soft-power countries.

### Advantages and potential uses
- GSPI is a systematic, transparent framework that captures the multidimensional matrix of soft power characteristics.  
- The aggregation into sub-indices allows granular comparisons across countries and over time at both headline and component levels.  
- The GSPI can be used in multiple econometric applications beyond the presented macro-financial example to explore soft power’s influence on economic and political developments.

### Summary points on robustness and limitations
- The construction balances comprehensiveness and data availability; some dimensions (e.g., culture) have weaker historical coverage.  
- Normalization and PCA weighting are used to address differing measurement units, dispersion, and collinearity across variables.  
- The sample and balanced-panel constraints mean that the most comprehensive GSPI is available for 66 countries over 2007–2021.

*Source: IMF Working Paper — Measuring Soft Power: A New Global Index (Executive Summary).*

### appendix provides an illustration of the steps to obtain the aforementioned subindex weights. We begin by

### wpiea2024212-print-pdf — appendix provides an illustration of the steps to obtain the aforementioned subindex weights. We begin by

### Index construction methodology
- Three-step approach to construct the GSPI:
  - (i) normalization of variables to standard z-scores.
  - (ii) aggregation of normalized variables into sub-indices representing each functional dimension using only linear combinations.
  - (iii) aggregation of sub-indices into the final GSPI using an additive aggregation method (Fagerberg (2002) and (OECD, 2008)).
- All variables and sub-indices are standardized (cross-sectional z-scores); the headline GSPI is the simple average of the sub-indices.
- Final step: subtract the median score from the headline GSPI so the overall index has 0 as its median score.

- Sub-index weight calculation (illustration for PISA: Maths):
  - w_PISA: Maths = (0.20*0.71)+(0.13*0.12) / 0.62+0.11 = 0.22
  - Education sub-index final weights:
    - Education Expenditure: 0.02
    - Journal Articles: 0.33
    - PISA: Maths: 0.22
    - PISA: Reading: 0.19
    - PISA: Science: 0.21
    - Tertiary Education: 0.02
  - Primary Completion and Years of Schooling are excluded because their factor weights in the most relevant factors are lower than 0.10.
- Same methodology applied to commercial, culture, digital, education, global reach, and institutions sub-indices (complete set of weights reported in appendix Table A6).

- Aggregation formula for each country i at year t:
  - GSPI_i,t = (sum over s of Sub-Index_i,t,s) / n_s
  - where s = {commercial, culture, digital, education, global reach, and institutions}.

- Annual REER volatility estimation (realized volatility from monthly log returns):
  - VOL_i,t = sum_{m=1}^{12} r_{i,m-t}^2
  - where r_{i,m-t} = log(REER_{i,m-t}) − log(REER_{i,(m−1)-t}).

### Sample, coverage, and grouping
- Sample: 66 countries, yearly observations from 2007 to 2021 (latest year with most variable coverage).
- K-Means clustering (using 2021 sub-index values, number of clusters selected following Pham et al. (2005)) grouped the 66 countries into four groups:
  - Group 1 (low soft power): ALB, ARG, AZE, BGR, BLR, BRA, CHL, COL, CRI, DOM, DZA, IDN, JOR, KAZ, LBN, MAR, MDA, MEX, MYS, PAN, PER, PHL, QAT, ROU, SAU, SRB, THA, TTO, TUR, UKR, URY, VNM.
  - Group 2 (medium soft power): AUS, AUT, CAN, CHE, CYP, CZE, DNK, EST, FIN, HRV, HUN, ISL, ISR, LTU, LVA, MLT, NOR, NZL, POL, PRT, SGP, SVK, SVN, SWE.
  - Group 3 (high soft power): CHN, DEU, ESP, FRA, GBR, ITA, RUS, USA.
  - Group 4 (high soft power, distinct profile): JPN, KOR.

### Key descriptive results (cross-section 2021)
- GSPI distribution (2021, sample of 66 countries):
  - Minimum: -0.59
  - Maximum: 1.68
  - Median: 0
  - Standard deviation: 0.49
- Top GSPI scores:
  - South Korea: 1.68
  - Japan: 1.25
  - Germany: 1.18
  - China: 1.17
- Bottom GSPI scores:
  - Dominican Republic: -0.59
  - Algeria: -0.55
  - Philippines: -0.53
  - Azerbaijan: -0.47
- Time-path examples (2004 → 2021):
  - China: 0.70 in 2004 → 1.17 in 2021.
  - United Kingdom: 1.32 in 2004 → 0.85 in 2021.

### Clustering interpretation and sub-index roles
- Group 1: developing countries with low soft power; especially behind in Education and Institutions.
- Group 2: mostly developed countries with medium soft power; high Education and Institutions but limited Culture and Global Reach impact.
- Group 3: high soft power countries (advanced economies and powerful developing nations like China, Russia).
- Group 4: Japan and South Korea — high soft power but with significantly higher Commercial Prowess and meaningfully lower Culture scores relative to Group 3.

### Application: soft power and real exchange rate (REER) volatility
- Empirical framework:
  - Panel regression with country fixed effects and time fixed effects.
  - Dependent variable: annual REER volatility (VOL_i,t).
  - Controls (9 variables from Cevik, Harris, and Yilmaz (2017), alphabetical): credit, current account, export concentration, inflation, stock market capitalization, trade openness, volatility of government consumption, volatility of labor productivity growth, volatility of terms of trade.
  - Include autoregressive lags of VOL_i,t; lag length K selected via BIC (chosen K = 2).
  - Two model specifications:
    - Model 1: VOL_i,t = μ_i + γ_t + sum_{k=1}^{K} α_k VOL_i,t−k + δ' Z_i,t + β GSPI_i,t + ε_i,t
    - Model 2 (expanded): VOL_i,t = μ_i + γ_t + sum_{k=1}^{K} α_k VOL_i,t−k + δ' Z_i,t + β' X_i,t + ε_i,t
      - where X_i,t is the 6x1 vector of sub-indices and β' is 6x1 coefficients vector.
- Main regression findings:
  - Model 1 (headline GSPI):
    - Overall GSPI coefficient is negative (higher GSPI → lower REER volatility) but not statistically significant; p-value = 0.22.
    - Most meaningful control variables: inflation, volatility of labor productivity growth, volatility of terms of trade; trade openness also relevant to a lesser extent.
  - Model 2 (sub-indices instead of composite GSPI):
    - Culture and Global Reach sub-indices are significant in explaining REER volatility, with negative coefficients (higher Culture or Global Reach → lower REER volatility).
    - Global Reach appears to be the most significant sub-index of the two.

### Interpretation and policy-relevant insights
- Composite GSPI:
  - Useful as a systematic, multidimensional measure of soft power comparable across countries and over time.
  - The headline GSPI alone does not show a statistically significant relationship with REER volatility once standard macroeconomic controls and fixed effects are included.
- Sub-index implications:
  - The Culture and Global Reach dimensions of soft power have statistically significant, negative associations with REER volatility, controlling for fundamentals and dynamics.
  - Global Reach—interpreted as the “purest” measure of international soft power reach—is the most impactful sub-index with respect to REER volatility.
- Practical takeaway:
  - Policy attention to enhancing Global Reach and Culture dimensions of soft power could be associated with lower exchange rate volatility, holding other macro fundamentals constant. This suggests possible macro-financial relevance of targeted soft-power-enhancing policies even if aggregate GSPI is not significant in the composite form.

*Source: Authors’ estimations and analysis in wpiea2024212-print-pdf — appendix provides an illustration of the steps to obtain the aforementioned subindex weights. We begin by (IMF Working Paper).*

### Appendix Table A1. List of Countries

### Appendix Table A1. List of Countries

### Country List
- ALBKAZ
- ARGKOR
- AUSLBN
- AUTLTU
- AZELVA
- BGRMAR
- BLRMDA
- BRAMEX
- CANMLT
- CHEMYS
- CHLNOR
- CHNNZL
- COLPAN
- CRIPER
- CYPPHL
- CZEPOL
- DEUPRT
- DNKQAT
- DOMROU
- DZARUS
- ESPSAU
- ESTSGP
- FINSRB
- FRASVK
- GBRSVN
- HRVSWE
- HUNTHA
- IDNTTO
- ISLTUR
- ISRUKR
- ITAURY
- JORUSA
- JPNVNM

### Variables and Data Sources (Variable Name — Dimension — Source — Definition)
- Outward Foreign Investment — Commercial Prowess — UNCTAD — Outward Foreign Investment
- Patents — Commercial Prowess — World Bank — Number of international
- Trademarks — Commercial Prowess — World Bank — Number of trademarks
- Cultural Exports — Culture — UNCTAD — Exports of cultural goods
- International Tourists — Culture — World Bank — Number of international
- Olympic Medals — Culture — Olympic Committee — Total number of medals in
- World Heritage Sites — Culture — UNESCO — Number of UNESCO World
- Internet Access — Digital Footprint — World Bank — Number of internet users
- Mobile Phones Access — Digital Footprint — World Bank — Number of mobile phones
- Education Expenditure — Education — World Bank — Government expenditure on
- Journal Articles — Education — World Bank — Number of journal articles
- PISA: Maths — Education — World Bank — Mean performance on the
- PISA: Reading — Education — World Bank — Mean performance on the
- PISA: Science — Education — World Bank — Mean performance on the
- Primary Completion — Education — World Bank — Primary completion rate
- Tertiary Education — Education — World Bank — Gross tertiary educational
- Years of Schooling — Education — World Bank — Barro-Lee: Average years of
- Aid and Assistance — Global Reach — World Bank — Official development
- Diplomatic Events — Global Reach — GDELT — Share of diplomatic cooperation
- Embassies — Global Reach — Lowy Institute — Lowy Institute number of
- Migrants — Global Reach — World Bank — Number of migrants
- Refugees — Global Reach — World Bank — Number of refugees
- Bureaucratic Effectiveness — Institutions — ICRG — The PRS Group International Country
- Corruption — Institutions — ICRG — The PRS Group International Country
- Democratic Accountability — Institutions — ICRG — The PRS Group International Country
- Government Stability — Institutions — ICRG — The PRS Group International Country
- Rule of Law — Institutions — ICRG — The PRS Group International Country

### Descriptive Statistics (Variable Name — Start Date — Mean — Standard Deviation — 5th Percentile — 95th Percentile)
- Outward Foreign Investment — 1/1/1970 — 2.2825 — 5.87 — -0.25 — 5.27
- Patents — 1/1/1980 — 0.01 — 0.03 — 0 — 0.04
- Trademarks — 1/1/1980 — 0.29 — 1.4 — 0.53 — 0.53
- Cultural Exports — 1/1/2007 — 0.17 — 0.38 — 0 — 0.63
- International Tourists — 1/1/1995 — 123.63 — 37.74 — 0.53 — 644.25
- Olympic Medals — 1/1/1960 — 9.45 — 22.6 — 0.19 — 33.24
- World Heritage Sites — 1/1/1978 — 3.78 — 6.64 — 0 — 16
- Internet Access — 1/1/1990 — 28.03 — 29.87 — 0.03 — 86.53
- Mobile Phones Access — 1/1/1980 — 57.29 — 54.1 — 0.03 — 146.92
- Education Expenditure — 1/1/1970 — 4.34 — 2.68 — 1.54 — 7.38
- Journal Articles — 1/1/2000 — 0.03 — 0.05 — 0 — 0.15
- PISA: Maths — 1/1/2000 — 461.39 — 59.83 — 361.53 — 543.8
- PISA: Reading — 1/1/2000 — 458.29 — 55.26 — 358.31 — 526.89
- PISA: Science — 1/1/2000 — 464.82 — 54.74 — 375.37 — 539.47
- Primary Completion — 1/1/1970 — 78.41 — 24.78 — 28.31 — 105.67
- Tertiary Education — 1/1/1970 — 22.16 — 23.21 — 0.68 — 72.45
- Years of Schooling — 1/1/1970 — 6.18 — 3.29 — 1.09 — 11.53
- Aid and Assistance — 1/1/1960 — 6.35 — 10.96 — 0.01 — 24.69
- Diplomatic Events — 1/1/1979 — 9.5 — 17.13 — 0.97 — 18.18
- Embassies — 1/1/2006 — 78.57 — 40.54 — 24 — 152
- Migrants — 1/1/1960 — 11.13 — 15.87 — 0.26 — 50.33
- Refugees — 1/1/1960 — 1.24 — 5.16 — 0 — 4.51
- Bureaucratic Effectiveness — 1/1/1984 — 2.17 — 1.15 — 0 — 4
- Corruption — 1/1/1984 — 2.92 — 1.32 — 0 — 5.5
- Democratic Accountability — 1/1/1984 — 3.81 — 1.65 — 0 — 6
- Government Stability — 1/1/1984 — 7.45 — 2.11 — 4 — 11
- Rule of Law — 1/1/1984 — 3.65 — 1.42 — 0 — 6

### PCA: Education Sub-Component — Variance Explained (Factor — Education Expenditure — PISA: Maths — PISA: Reading — PISA: Science — Primary Completion — Journal Articles — Years of Schooling — Tertiary Education)
- Factor 1 — 71% — 1% — 20% — 20% — 19% — 0% — 29% — 5% — 6%
- Factor 2 — 12% — 11% — 13% — 10% — 15% — 0% — 33% — 3% — 14%
- Factor 3 — 7% — 0% — 0% — 0% — 0% — 1% — 25% — 1% — 72%
- Factor 4 — 4% — 30% — 0% — 2% — 1% — 1% — 0% — 65% — 2%
- Factor 5 — 3% — 53% — 0% — 0% — 1% — 10% — 11% — 19% — 6%
- Factor 6 — 2% — 4% — 0% — 0% — 0% — 87% — 1% — 7% — 0%
- Factor 7 — 1% — 0% — 36% — 60% — 2% — 0% — 0% — 1% — 1%
- Factor 8 — 0% — 0% — 30% — 8% — 61% — 0% — 0% — 0% — 0%

### Weights for GSPI Sub-Components (Sub-Index — Variable Name — Weight)
- Commercial Prowess — Outward Foreign Investment — 28%
- Commercial Prowess — Patents — 68%
- Commercial Prowess — Trademarks — 4%
- Culture — Cultural Exports — 26%
- Culture — International Tourists — 0%
- Culture — Olympic Medals — 0%
- Culture — World Heritage Sites — 74%
- Digital Footprint — Internet Access — 49%
- Digital Footprint — Mobile Phones Access — 51%
- Education — Education Expenditure — 2%
- Education — PISA: Maths — 22%
- Education — PISA: Reading — 19%
- Education — PISA: Science — 21%
- Education — Primary Completion — 0%
- Education — Journal Articles — 33%
- Education — Years of Schooling — 0%
- Education — Tertiary Education — 2%
- Global Reach — Aid and Assistance — 0%
- Global Reach — Embassies — 74%
- Global Reach — Migrants — 26%
- Global Reach — Refugees — 0%
- Global Reach — Diplomatic Events — 0%
- Institutions — Bureaucratic Effectiveness — 18%
- Institutions — Corruption — 18%
- Institutions — Democratic Accountability — 23%
- Institutions — Government Stability — 22%
- Institutions — Rule of Law — 19%

*Appendix Table A1. List of Countries — IMF Working Papers: Measuring Soft Power: A New Global Index*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024212-print-pdf.pdf_