## _wp15195 - Appendix A

## Source details

**Canonical URL:** [_wp15195 - Appendix A](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15195.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15195.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15195.pdf.json)

---

### Overview and scope
- Covers the period 1980-2012 for 25 countries and economies in East and South Asia.
- Focus: determinants of growth in Asia—role of investment, exchange rate regime, financial risk, and capital account openness under different fiscal and monetary policy settings.
- Regional environment noted: high private investment, more modest public investment, varying shares of internal funding and significant FDI in some countries.
- Exchange rate regimes: range from fixed to free floating; majority classified as an "intermediate float" (currency not fully flexible).
- Trend: gradual movement towards greater exchange rate flexibility in a number of countries, but persistent foreign exchange intervention and a regional propensity to avoid appreciation more than depreciation.
- Two crisis periods used as dummy variables: the Asian financial crisis (1997-99), and the global financial crisis (2008-11).

### Data, classification, and sample construction
- Sample: 25 Asian economies including Singapore, Hong-Kong SAR and China; excludes small Pacific Island countries, Australia and New Zealand, and countries with missing data for key variables.
- Time coverage: 1980–2012.
- Exchange rate regime measurement:
  - Uses IMF AREAER de jure and de facto classifications.
  - Mapped into three regimes: pegged, intermediate, and floating.
  - Two dummy variables for de facto regimes (floating and intermediate) and two for de jure regimes; pegged regimes are the omitted category.
  - For 3- and 5-year periods, the dummy captures the typical regime during the interval: the regime that occurred most of the time (at least three out of five years or two out of three years).
  - Over the sample period, only four out of 25 countries did not change their exchange regime.
- Capital account openness:
  - Baseline measure: relative openness to OECD countries based on methodology relying on stocks of net foreign assets adjusted for reserves (Lane and Milesi-Ferretti, 2007).
  - Larger value implies a more open capital account.
- Financial risk index:
  - Defined as in the International Country Risk Guide with weights: political risk 50 percent, economic risk 25 percent, financial risk 25 percent.
  - Higher index value implies lower financial risk.
- Real exchange rate misalignment:
  - Captured by annual volatility of the growth of the real effective exchange rate (REER), following Aghion and others (2005).
- Monetary anchor proxy:
  - Inflation below a 10 percent threshold used as a proxy for whether there is an effective monetary anchor.
- "Fear" measures:
  - Fear of floating dummy = 1 when de jure floating regime is associated with de facto less flexible regimes.
  - Fear of pegging dummy = 1 when de jure pegged regime is associated with de facto more flexible regimes.

### Empirical specification and estimation methods
- Core estimating equation:
  - Growth_{i,t} = α X_{i,t} + β R_{i,t} + σ_i + γ_t + ε_{i,t}
  - Growth_{i,t}: growth rate of real per capita GDP in national currency of country i in year t.
  - X_{i,t}: vector of explanatory variables (variously measured at the beginning of periods or as averages over periods).
  - R_{i,t}: vector of exchange rate regime dummies (coefficients represent performance of flexible regimes relative to pegged regime).
  - σ_i: country-specific effects; γ_t: time-specific effects; ε_{i,t}: error term.
- Estimation strategies:
  - Arellano and Bond (1991) generalized method of moments (GMM) to address endogeneity and correlation between country effects and explanatory variables.
  - Feasible generalized least squares (FGLS) to control for group heteroscedasticity.
  - Models estimated on annual data and on non-overlapping 3-year and 5-year averages to reduce business cycle bias and measurement error attenuation.
  - FGLS specifications include initial income (log per capita income); GMM uses lagged dependent variable (lagged real GDP per capita growth).
  - Robustness checks include alternative regime dummy specifications and alternative capital openness definitions.

### Baseline explanatory variables
- Per capita income (natural log) — convergence hypothesis: coefficient expected negative if income convergence present.
- Private investment (share of GDP) — expected positive effect on growth.
- Public investment (share of GDP) — expected positive effect; public investment marginal impact observed to be larger than private investment in the results.
- Public consumption (share of GDP) — ambiguous expected effect on growth; included to account for fiscal considerations affecting exchange rate credibility and sustainability.
- Financial risk index (ICRG) — higher index (lower risk) expected to improve growth (excluding investment channel already in regressions).
- Capital account openness (baseline: relative openness to OECD countries) — effect conditional on supporting policies; somewhat agnostic on sign for Asian sample.
- Real exchange rate volatility (REER volatility) — volatile REER expected to harm growth.
- Crisis dummy(s) for 1997-99 and 2008-11; interaction of crisis dummy and capital account openness also included.
- Additional variations tested: FDI (share of GDP), inflation, financial deepening (credit/GDP), trade openness (exports+imports/GDP), health and education indicators.

### Results — methodological and partial substantive findings
- Models reported in Tables 1a–1c (annual, 3-year averages, 5-year averages) use FGLS and GMM with the baseline variable set, regime dummies, financial risk, crisis dummy, capital account openness index, REER volatility, and crisis × capital openness interaction.
- Instrumentation: GMM specifications maintain number of instruments not exceeding conventional levels.
- Consistent finding across samples and estimators:
  - Private and public investments are consistently positive and statistically significant across all samples.
  - The marginal impact of public investment is consistently larger than that of private investment.
- The source text indicates an intended numerical statement for private investment impact but the main text fragment ended; an appended line provides an implied numeric range for private investment effects: "0.13 to 0.20 percent in the three samples, with the averaged samples having somewhat larger."

### Key empirical findings on investment and growth
- Public investment:
  - An increase in the ratio of public investment to GDP of one percentage point leads to an increase in growth per capita per year in the range of 0.18 to 0.32 percent in the samples.
  - Public investment coefficients reported across specifications include: 0.322***, 0.276***, 0.218**, 0.184*, 0.321***, 0.286***, 0.275**, 0.239*, 0.320***, 0.258***, 0.269***, 0.204*, 0.337***, 0.304***, 0.386*, 0.249**, 0.282***, 0.221**, 0.309**, 0.257*, 0.320***, 0.284**, 0.317**, 0.292**.
- Private investment:
  - Private investment coefficients reported positive and significant across specifications, with examples: 0.132***, 0.147***, 0.160***, 0.171***, 0.193**, 0.185***, 0.212***.
  - Supplemental fragment: private investment effect reported as "0.13 to 0.20 percent in the three samples, with the averaged samples having somewhat larger."
- Financial risk (frisk):
  - Tends to be positive and statistically significant in many specifications (examples: 0.111***, 0.093**, 0.088**, 0.087**; also 0.098**, 0.094**, 0.109*, 0.074).
- FDI (fdi_ngdp):
  - Positive and significant in some specifications (examples: 0.117**, 0.104**); other specifications report positive but insignificant coefficients.
- Government consumption (govc_ngdp):
  - Generally negative, sometimes significant (examples: -0.120**, -0.127*, -0.164**, -0.139*), implying some weakly negative effect on growth.
- Initial level of real GDP per capita:
  - Not significant in the main sample, implying no convergence of incomes in this sample of countries.

### Exchange rate regimes, volatility, and crises
- Exchange rate regime dummies:
  - Inconsistently significant; when significant they imply that more flexible regimes boost growth (examples: floatdf 1.021*; floatdf 1.423**; floatdf 1.617**; floatdf 0.979; floatdf 1.173).
- REER volatility (reer_sd):
  - Negative and significant in some estimations:
    - Negative and significant in the three-year averaged estimations (examples: reer_sd -0.116***, -0.118***, -0.108*).
    - Negative and significant in the five-year averaged FGLS de facto estimation (examples: reer_sd -0.051**, -0.040).
  - Other specifications show insignificant reer_sd coefficients (examples: -0.008, 0.016, 0.021, 0.030).
- Interaction kaopen*crisis:
  - Negative and generally statistically significant, suggesting that capital account openness combined with financial turmoil amplifies adverse effect on growth (examples: -0.005, -0.006*, -0.003**, -0.003*; -0.007**, -0.007**, -0.012**, -0.009***; -0.010***, -0.009***, -0.011***, -0.010***).
- Crisis dummy (crisis):
  - Negative but inconsistently significant across specifications (examples: -1.868, 0.582, -1.584**, -1.551**; -0.169, -1.347*, -0.288, -0.402; -1.867***, -0.944, -0.735, -0.973).

### Magnitudes and selected regression statistics (examples)
- Annual data (Table 1a examples):
  - lag of real GDP per capita growth: 0.147* (0.057); 0.135* (0.060)
  - privateinv_ngdp: 0.132*** (0.035), 0.147*** (0.033)
  - publicinv_ngdp: 0.322*** (0.036), 0.276*** (0.031)
  - r2: 0.540, 0.542; rmse: 2.329, 2.328
  - Number of observations: 334 (in presented columns)
- 3-year averages (Table 1b examples):
  - privateinv_ngdp: 0.160*** (0.035), 0.178*** (0.038), 0.202*** (0.035)
  - publicinv_ngdp: 0.321*** (0.058), 0.286*** (0.054), 0.275** (0.083)
  - reer_sd: -0.116*** (0.018), -0.118*** (0.019)
  - r2: 0.695, 0.691; rmse: 1.474, 1.519; N: 123
- 5-year averages (Table 1c and Table 2c examples):
  - privateinv_ngdp: 0.171*** (0.038), 0.168** (0.054)
  - publicinv_ngdp: 0.320*** (0.058), 0.258*** (0.058)
  - kaopen*crisis: -0.010*** (0.002), -0.009*** (0.002)
  - r2: 0.755, 0.697; rmse: 1.153, 1.264; N: 76

### Alternative specifications and robustness checks
- Alternative models include inflation (inf), FDI (fdi_ngdp), and dummy variables for fear of floating and fear of pegging.
- Inflation:
  - Significantly negative in most samples. Examples: inf -0.042*** (0.009), -0.041*** (0.010), -0.065* (0.024), -0.047** (0.012); other samples show inf -0.052***, -0.049***, -0.055***, -0.056***.
- FDI:
  - Positive and significant in some alternative specifications (examples: fdi_ngdp 0.117** (0.034), 0.104** (0.031)).
- Fear dummies:
  - Dummy variables for fear of floating and fear of pegging are not significant in the samples examined.
- Other robustness notes:
  - When including inflation and fear dummies, exchange rate regime dummies became somewhat less significant and lagged growth is no longer significant.
  - Results are robust to reducing the number of instruments in GMM; table footnote: "Our results are robust to reducing the number of instruments to 66."

### Summary of robustness across models (Table 3 indicators)
- Private investment: positive, strongly significant and robust across specifications (marked pos/+++/yes/yes/yes).
- Public investment: positive, strongly significant and robust across specifications (pos/+++/yes/yes/yes).
- Volatility of the REER: generally negative and significant in some specifications (neg/+++/no/no/yes or neg/+/no/no/yes depending on model).
- Financial risk: positive and robust in several specifications (pos/+++/no/no/yes).
- Capital account openness: negative in interaction with crisis and robust in some checks (neg/+++/yes/yes/yes for baseline).
- Inflation: negative and significant in many specifications (neg/+++/no/no/no in some summaries).
- FDI: positive and significant in alternative model specifications (pos/++/yes/no/yes).
- Fear of floating and fear of pegging: generally not significant across samples.

### Conclusions and policy implications
- Most robust empirical result: private and public investments are significant determinants of growth in Asian countries.
- Lower financial risk and higher FDI also appear to support growth.
- Public consumption may exert some drag on growth.
- Some evidence suggests growth benefits from more flexible exchange rate regimes.
- Real exchange rate volatility and fears of floating do not consistently limit growth in the sample.
- A more open capital account during financial turbulence may reduce growth, but the effect is marginal.
- Inflation has a statistically significant negative effect on growth in most specifications.

### Appendix A — Sample and country income classification
- Sample includes:
  - six high income economies
  - two upper-middle-income economies
  - eleven lower-middle-income economies
  - six low income economies
- Table A1. Asian Countries in the Full Sample (income level as listed):
  1. Bangladesh Low-income
  2. Bhutan Lower-middle-income
  3. Brunei Darussalam High-income
  4. Cambodia Low-income
  5. China Lower-middle-income
  6. Fiji Upper-middle-income
  7. Hong Kong SAR, China High-income
  8. India Lower-middle-income
  9. Indonesia Lower-middle-income
  10. Japan High-income
  11. Korea High-income
  12. Lao P.D.R. Low-income
  13. Malaysia Upper-middle-income
  14. Maldives Lower-middle-income
  15. Mongolia Lower-middle-income
  16. Myanmar Low-income
  17. Nepal Low-income
  18. Pakistan Lower-middle-income
  19. Papua New Guinea Lower-middle-income
  20. Philippines Lower-middle-income
  21. Singapore High-income
  22. Sri Lanka Lower-middle-income
  23. Taiwan Province of China High-income
  24. Thailand Lower-middle-income
  25. Vietnam Low-income
- Note on income grouping: Economies are divided according to 2013 GNI per capita using World Bank Atlas method with groups: low income, $1,045 or less; lower middle income, $1,046–4,125; upper middle income, $4,126–12,745; and high income, $12,746 or more.

### Appendix A — Variable definitions and data sources (selected)
- Dependent variable:
  - rgdppcg: Real GDP growth per capita (percent) — IMF, WEO
- Selected independent variables and sources:
  - govc_ngdp: Government consumption (percent of GDP) — IMF, WEO
  - privateinv_ngdp: Private investment (percent of GDP) — IMF, WEO
  - publicinv_ngdp: Public investment (percent of GDP) — IMF, WEO
  - fdi_ngdp: FDI (percent of GDP) — IMF, WEO
  - frisk: Financial risk (index) — International Country Risk Guide (ICRG)
  - Volatility of REER: Standard deviation of REER change — IMF, INS
  - openness: Sum of imports and exports (percent of GDP) — IMF, WEO
  - healthexp: Health expenditure (percent of GDP) — World Bank, WDI
  - pri_enroll: Primary school enrollment rate (percent of the population of relevant age) — World Bank, WDI
  - inf: Inflation (percent) — IMF, WEO
  - floatdj / intdj: de jure regime dummies — IMF AREAER
  - floatdf / intdf: de facto regime dummies — IMF AREAER, Bubula and Ötker-Robe (2002), CUHK
  - float_10b, int_10b, float_10a, int_10a: regime dummies with below/above 10% inflation — Staff estimates
  - fear of floating / fear of pegging: Staff estimates
  - IT: Inflation targeting — IMF, INS
  - kaopen: Relative openness to OECD countries (Lane and Milesi-Ferretti, 2007)

### Appendix A — Descriptive statistics (selected)
- Table A3. Descriptive Statistics, 1980-2012 (Source: IMF staff estimates)
  - Real GDP per capita growth (percent change): Mean 4.2, Standard Deviation 3.6, Min -11.5, Max 16.2
  - Private investment (percent of GDP): Mean 19.0, Standard Deviation 6.4, Min 3.0, Max 41.0
  - Public investment (percent of GDP): Mean 7.3, Standard Deviation 5.6, Min 2.0, Max 32.8
  - Government consumption (percent of GDP): Mean 10.3, Standard Deviation 2.9, Min 3.5, Max 17.6
  - Standard deviation of REER growth (units): Mean 6.4, Standard Deviation 5.1, Min 0.6, Max 41.0
  - Financial risk (index): Mean 37.5, Standard Deviation 7.4, Min 14.0, Max 48.5
  - FDI (percent of GDP): Mean 4.0, Standard Deviation 6.6, Min -2.8, Max 53.1
  - Capital account openness (index): Mean 88.9, Standard Deviation 172.5, Min 5.5, Max 1,024.6
- Table A4. Selected Variable Means by Type of Exchange Rate Regime, 1980-2012 (de Jure classification) — Regime counts: Floating (109), Intermediate (153), Pegged (74)
  - Floating: Real GDP per capita growth 3.7; Private investment 20.3; Public investment 5.1; Government consumption 11.1; Standard deviation of REER growth 8.2; Financial risk 37.0; FDI 2.0; Capital account openness 22.6
  - Intermediate: Real GDP per capita growth 4.8; Private investment 18.2; Public investment 8.9; Government consumption 11.0; Standard deviation of REER growth 5.4; Financial risk 38.1; FDI 4.9; Capital account openness 67.3
  - Pegged: Real GDP per capita growth 3.5; Private investment 19.0; Public investment 7.1; Government consumption 8.9; Standard deviation of REER growth 6.1; Financial risk 37.0; FDI 5.2; Capital account openness 229.7
- Table A5. Selected Variable Means by Type of Exchange Rate Regime, 1980-2012 (de Facto classification) — Regime counts: Floating (48), Intermediate (181), Pegged (107)
  - Floating: Real GDP per capita growth 3.0; Private investment 17.2; Public investment 4.7; Government consumption 10.7; Standard deviation of REER growth 6.5; Financial Risk 34.9; FDI 3.0; Capital account openness 22.7
  - Intermediate: Real GDP per capita growth 4.4; Private investment 20.6; Public investment 6.9; Government consumption 10.6; Standard deviation of REER growth 6.7; Financial Risk 38.3; FDI 4.1; Capital account openness 62.5
  - Pegged: Real GDP per capita growth 4.3; Private investment 17.4; Public investment 9.1; Government consumption 9.7; Standard deviation of REER growth 5.9; Financial Risk 37.4; FDI 4.4; Capital account openness 162.9

### Appendix A — Correlations (selected)
- Correlation matrix excerpts:
  - rdpppc with inincome0: -0.05351
  - privateinv_ngdp with rdpppc: 0.21960
  - publicinv_ngdp with rdpppc: 0.331
  - floatdf with rdpppc: -0.13670
  - intdf with rdpppc: 0.07230
  - govc_ngdp with rdpppc: 0.1169
  - frisk with rdpppc: 0.3133
  - reer_sd with rdpppc: -0.165
  - crisis with rdpppc: -0.1923
  - kaopen with rdpppc: -0.0357

### Appendix B — De Jure and De Facto Classification (key points and regime distributions)
- Two classification schemes:
  - IMF de jure regime classification (governments' claimed regimes)
  - IMF de facto classification (observed policies and exchange rate behavior), supplemented by CUHK data where missing.
- IMF moved from a purely de jure classification to a hybrid de facto methodology from 1999, applied retrospectively as well.
- Table B1. Distribution of Regimes (1980-2012, totals)
  - De Jure: Floating 203 (24.6 percent), Intermediate 293 (35.5 percent), Pegged 329 (39.9 percent); Total 825 (100)
  - De Facto: Floating 102 (12.4 percent), Intermediate 434 (52.6 percent), Pegged 289 (35.0 percent); Total 825 (100)
- Interpretation:
  - Under de facto categorization, there are only half as many floating regimes compared with de jure, indicative of a "fear of floating" in Asian countries.
  - Pegged regimes slightly prevail in the de facto sample.
- Country-level notes:
  - Never floating (de facto): Bangladesh, Brunei Darussalam, China, Fiji, Hong Kong SAR, Lao P.D.R, Maldives, Myanmar, Nepal, Singapore, Vietnam.
  - Never pegged (de facto): Japan, Korea, Taiwan, Province of China.
  - No intermediate regime at any time: Nepal and Pakistan.
  - Countries with intermediate regimes most of the time: Sri Lanka, Singapore, Korea, Indonesia, Cambodia.
- Table B2 totals (selected): Floating 102, Intermediate 347, Pegged 376, Total 825; "Floats in Percent of Total" column example ending with 12.

*Source: _wp15195 - Appendix A (content from the provided PDF excerpt).*

### Appendix A .............................................................................................................

### _wp15195 - Appendix A

### Overview and scope
- Covers the period 1980-2012 for 25 countries and economies in East and South Asia.
- Focus: determinants of growth in Asia—role of investment, exchange rate regime, financial risk, and capital account openness under different fiscal and monetary policy settings.
- Notes a regional environment of high private investment, more modest public investment, varying shares of internal funding and significant FDI in some countries.
- Observes a range of exchange rate regimes from fixed to free floating; majority classified as an "intermediate float" (currency not fully flexible).
- Identifies gradual movement towards greater exchange rate flexibility in a number of countries, but with persistent foreign exchange intervention and a regional propensity to avoid appreciation more than depreciation.
- Two crisis periods identified and used as dummy variables: the Asian financial crisis (1997-99), and the global financial crisis (2008-11).

### Data, classification, and sample construction
- Sample: 25 Asian economies including Singapore, Hong-Kong SAR and China; excludes small Pacific Island countries, Australia and New Zealand, and countries with missing data for key variables of interest.
- Time coverage: 1980–2012.
- Exchange rate regime measurement:
  - Uses IMF AREAER de jure and de facto classifications.
  - Mapped into three regimes: pegged, intermediate, and floating.
  - Two dummy variables for de facto regimes (floating and intermediate) and two for de jure regimes; pegged regimes are the omitted category.
  - For 3- and 5-year periods, the dummy captures the typical regime during the interval: the regime that occurred most of the time (at least three out of five years or two out of three years).
  - Over the sample period, only four out of 25 countries did not change their exchange regime.
- Capital account openness:
  - Baseline measure: relative openness to OECD countries based on methodology relying on stocks of net foreign assets adjusted for reserves (Lane and Milesi-Ferretti, 2007).
  - Larger value implies a more open capital account.
- Financial risk index:
  - Defined as in the International Country Risk Guide with weights: political risk 50 percent, economic risk 25 percent, financial risk 25 percent.
  - Higher index value implies lower financial risk.
- Real exchange rate misalignment:
  - Captured by annual volatility of the growth of the real effective exchange rate (REER), following Aghion and others (2005).
- Monetary anchor proxy:
  - Inflation below a 10 percent threshold used as a proxy for whether there is an effective monetary anchor.
- "Fear" measures:
  - Fear of floating dummy = 1 when de jure floating regime is associated with de facto less flexible regimes.
  - Fear of pegging dummy = 1 when de jure pegged regime is associated with de facto more flexible regimes.

### Empirical specification and estimation methods
- Core estimating equation:
  - Growth_{i,t} = α X_{i,t} + β R_{i,t} + σ_i + γ_t + ε_{i,t}
  - Growth_{i,t}: growth rate of real per capita GDP in national currency of country i in year t.
  - X_{i,t}: vector of explanatory variables (variously measured at the beginning of periods or as averages over periods).
  - R_{i,t}: vector of exchange rate regime dummies (coefficients represent performance of flexible regimes relative to pegged regime).
  - σ_i: country-specific effects; γ_t: time-specific effects; ε_{i,t}: error term.
- Estimation strategies:
  - Arellano and Bond (1991) generalized method of moments (GMM) to address endogeneity and correlation between country effects and explanatory variables.
  - Feasible generalized least squares (FGLS) to control for group heteroscedasticity.
  - Models estimated on annual data and on non-overlapping 3-year and 5-year averages to reduce business cycle bias and measurement error attenuation.
  - FGLS specifications include initial income (log per capita income); GMM uses lagged dependent variable (lagged real GDP per capita growth).
  - Robustness checks include alternative regime dummy specifications and alternative capital openness definitions.

### Baseline explanatory variables
- Per capita income (natural log) — convergence hypothesis: coefficient expected negative if income convergence present.
- Private investment (share of GDP) — expected positive effect on growth.
- Public investment (share of GDP) — expected positive effect; public investment marginal impact observed to be larger than private investment in the results.
- Public consumption (share of GDP) — ambiguous expected effect on growth; included to account for fiscal considerations affecting exchange rate credibility and sustainability.
- Financial risk index (ICRG) — higher index (lower risk) expected to improve growth (excluding investment channel already in regressions).
- Capital account openness (baseline: relative openness to OECD countries) — effect conditional on supporting policies; somewhat agnostic on sign for Asian sample.
- Real exchange rate volatility (REER volatility) — volatile REER expected to harm growth.
- Crisis dummy(s) for 1997-99 and 2008-11; interaction of crisis dummy and capital account openness also included.
- Additional variations tested: FDI (share of GDP), inflation, financial deepening (credit/GDP), trade openness (exports+imports/GDP), health and education indicators.

### Results — methodological and partial substantive findings
- Models reported in Tables 1a–1c (annual, 3-year averages, 5-year averages) use FGLS and GMM with the baseline variable set, regime dummies, financial risk, crisis dummy, capital account openness index, REER volatility, and crisis × capital openness interaction.
- Instrumentation: GMM specifications maintain number of instruments not exceeding conventional levels.
- Consistent finding across samples and estimators:
  - Private and public investments are consistently positive and statistically significant across all samples.
  - The marginal impact of public investment is consistently larger than that of private investment.
- The text indicates an intended numerical statement: "An increase in the ratio of private investment to GDP of one percentage point leads to an increase in growth per capita per period in the range of" — the source content ends at this point and does not provide the completed numeric range.

*Source: _wp15195 - Appendix A (content from the provided PDF excerpt).*

### 0.13 to 0.20 percent in the three samples, with the averaged samples having somewhat larger

### _wp15195 - 0.13 to 0.20 percent in the three samples, with the averaged samples having somewhat larger

### Key empirical findings on investment and growth
- An increase in the ratio of public investment to GDP of one percentage point leads to an increase in growth per capita per year in the range of 0.18 to 0.32 percent in the samples.
- Private investment coefficients: reported positive and significant across specifications (examples in tables: 0.132***, 0.147***, 0.160***, 0.171***, 0.193**, 0.185***, 0.212***).
- Public investment coefficients: reported positive and significant across specifications (examples in tables: 0.322***, 0.276***, 0.218**, 0.184*, 0.321***, 0.286***, 0.275**, 0.239*, 0.320***, 0.258***, 0.269***, 0.204*, 0.337***, 0.304***, 0.386*, 0.249**, 0.282***, 0.221**, 0.309**, 0.257*, 0.320***, 0.284**, 0.317**, 0.292**).
- Financial risk (frisk) tends to be positive and statistically significant in many specifications (examples: 0.111***, 0.093**, 0.088**, 0.087**; also 0.098**, 0.094**, 0.109*, 0.074; other samples show smaller or insignificant values).
- FDI (fdi_ngdp) is positive and significant in some specifications (examples: 0.117**, 0.104**, and other positive but insignificant coefficients in different specifications).
- Government consumption (govc_ngdp) is generally negative, sometimes significant (examples: -0.120**, -0.127*, -0.164**, -0.139*), implying some weakly negative effect of spending on growth.
- Initial level of real GDP per capita is not significant in the main sample, implying no convergence of incomes in this sample of countries.

### Exchange rate regimes, volatility, and crises
- Exchange rate regime dummies are inconsistently significant; when significant they imply that more flexible regimes boost growth (examples: floatdf 1.021*; floatdf 1.423**; floatdf 1.617**; floatdf 0.979; floatdf 1.173).
- Volatility of the real exchange rate (reer_sd) is negative and significant in some estimations:
  - Negative and significant in the three-year averaged estimations (examples: re e r_sd -0.116***, -0.118***, -0.108*).
  - Negative and significant in the five-year averaged FGLS de facto estimation (examples: re e r_sd -0.051**, -0.040).
  - Other specifications show insignificant reer_sd coefficients (examples: -0.008, 0.016, 0.021, 0.030).
- The interaction term kaopen*crisis is negative and generally statistically significant, suggesting that capital account openness combined with financial turmoil amplifies the adverse effect on growth (examples: -0.005, -0.006*, -0.003**, -0.003*; -0.007**, -0.007**, -0.012**, -0.009***; -0.010***, -0.009***, -0.011***, -0.010***).
- Crisis dummy (crisis) is negative but inconsistently significant across specifications (examples: -1.868, 0.582, -1.584**, -1.551**; -0.169, -1.347*, -0.288, -0.402; -1.867***, -0.944, -0.735, -0.973).

### Magnitudes and selected regression statistics (examples from tables)
- Annual data examples (Table 1a):
  - lag of real GDP per capita growth: 0.147* (0.057); 0.135* (0.060)
  - privateinv_ngdp: 0.132*** (0.035), 0.147*** (0.033)
  - publicinv_ngdp: 0.322*** (0.036), 0.276*** (0.031)
  - r2: 0.540, 0.542; rmse: 2.329, 2.328
  - Number of observations: 334 (in presented columns)
- 3-year averages examples (Table 1b):
  - privateinv_ngdp: 0.160*** (0.035), 0.178*** (0.038), 0.202*** (0.035)
  - publicinv_ngdp: 0.321*** (0.058), 0.286*** (0.054), 0.275** (0.083)
  - reer_sd: -0.116*** (0.018), -0.118*** (0.019)
  - r2: 0.695, 0.691; rmse: 1.474, 1.519; N: 123
- 5-year averages examples (Table 1c and Table 2c):
  - privateinv_ngdp: 0.171*** (0.038), 0.168** (0.054)
  - publicinv_ngdp: 0.320*** (0.058), 0.258*** (0.058)
  - kaopen*crisis: -0.010*** (0.002), -0.009*** (0.002)
  - r2: 0.755, 0.697; rmse: 1.153, 1.264; N: 76

### Alternative specifications and robustness checks
- Alternative models include inflation (inf), FDI (fdi_ngdp), and dummy variables for fear of floating and fear of pegging.
- Inflation is significantly negative in most samples:
  - Examples: inf -0.042*** (0.009), -0.041*** (0.010), -0.065* (0.024), -0.047** (0.012); other samples show inf -0.052***, -0.049***, -0.055***, -0.056***.
- FDI is positive and significant in some alternative specifications (examples: fdi_ngdp 0.117** (0.034), 0.104** (0.031)).
- Dummy variables for fear of floating and fear of pegging are not significant in the samples examined.
- When including inflation and fear dummies, the exchange rate regime dummies became somewhat less significant and lagged growth is no longer significant.
- Results are robust to reducing the number of instruments in GMM (note: table footnote: "Our results are robust to reducing the number of instruments to 66.").

### Summary of robustness across models (Table 3 summary indicators)
- Private investment: positive, strongly significant and robust across specifications (noted as pos/+++/yes/yes/yes in summary).
- Public investment: positive, strongly significant and robust across specifications (pos/+++/yes/yes/yes).
- Volatility of the REER: generally negative and significant in some specifications (neg/+++/no/no/yes or neg/+/no/no/yes depending on model).
- Financial risk: positive and robust in several specifications (pos/+++/no/no/yes).
- Capital account openness: negative in interaction with crisis and robust in some checks (neg/+++/yes/yes/yes for baseline).
- Inflation: negative and significant in many specifications (neg/+++/no/no/no in some summaries).
- FDI: positive and significant in alternative model specifications (pos/++/yes/no/yes).
- Fear of floating and fear of pegging: generally not significant across samples.

### Conclusions and policy implications
- The most robust result: private and public investments are significant determinants of growth in Asian countries.
- Lower financial risk and higher FDI also appear to support growth.
- Public consumption may exert some drag on growth.
- Some evidence suggests that growth benefits from more flexible exchange rate regimes.
- Real exchange rate volatility and fears of floating do not consistently limit growth in the sample.
- A more open capital account during financial turbulence may reduce growth, but the effect is marginal.
- Inflation has a statistically significant negative effect on growth in most specifications.

*Source: IMF staff estimates (extracted from the provided content of the IMF PDF)._

### References

### References

### Bibliographic sources
- Extensive list of references on exchange rate regimes, growth, and related macroeconomic topics (authors include Aghion; Arellano and Bond; Bailliu, Lafrance, and Perrault; Berg and Miao; Bleaney and Francisco; Bubula and Otker-Robe; Calvo and Reinhart; Cavoli and Rajan; Chinn and Ito; Coudert and Dubert; Detken and Gaspar; De Vita and Kyaw; Eichengreen; Filardo and Grenville; Frankel; Ghazanchyan and Stotsky; Ghosh et al.; Goyal; Harms and Kreschmann; Henning; Huang and Malhotra; Husain, Mody, and Rogoff; IMF AREAER; Iwata, Khan, and Murao; Johnson, Ostry, and Subramanian; Klein and Shambaugh; Kumhof, Li, and Yan; Lane and Milesi-Ferretti; Levine and Renelt; Levy-Yeyati and Sturzenegger; Mankiw, Romer, and Weil; McKibbin and Lee; Miles; Moreno; PRS Group; Park and Park; Patnaik et al.; Petreski; Rajan; Reinhart and Rogoff; Rodrik; Rogoff et al.; Rose; Schnabl; Sokolov, Lee, and Mark; Stotsky et al.; Weber.)
- Data sources cited include IMF WEO, IMF AREAER, International Country Risk Guide (ICRG), World Bank WDI, Lane and Milesi-Ferretti (2007) database, Chinese University of Hong-Kong (CUHK) database, IMF staff estimates, and PRS Group International Country Risk Guide.

### Appendix A — Sample and country income classification
- Sample includes:
  - six high income economies
  - two upper-middle-income economies
  - eleven lower-middle-income economies
  - six low income economies
- Noted histories:
  - "Singapore, Hong-Kong SAR and Taiwan Province of China have had an extensive history of exchange rate regimes."
  - "Indonesia, Korea, the Philippines, and Thailand have, at different times, adopted inflation targeting regimes."
- Table A1. Asian Countries in the Full Sample (income level as listed):
  1. Bangladesh Low-income
  2. Bhutan Lower-middle-income
  3. Brunei Darussalam High-income
  4. Cambodia Low-income
  5. China Lower-middle-income
  6. Fiji Upper-middle-income
  7. Hong Kong SAR, China High-income
  8. India Lower-middle-income
  9. Indonesia Lower-middle-income
  10. Japan High-income
  11. Korea High-income
  12. Lao P.D.R. Low-income
  13. Malaysia Upper-middle-income
  14. Maldives Lower-middle-income
  15. Mongolia Lower-middle-income
  16. Myanmar Low-income
  17. Nepal Low-income
  18. Pakistan Lower-middle-income
  19. Papua New Guinea Lower-middle-income
  20. Philippines Lower-middle-income
  21. Singapore High-income
  22. Sri Lanka Lower-middle-income
  23. Taiwan Province of China High-income
  24. Thailand Lower-middle-income
  25. Vietnam Low-income
- Source: World Bank
- Note on income grouping: Economies are divided according to 2013 GNI per capita using World Bank Atlas method with groups: low income, $1,045 or less; lower middle income, $1,046–4,125; upper middle income, $4,126–12,745; and high income, $12,746 or more.

### Appendix A — Variable definitions and data sources (Table A2)
- Dependent variable:
  - rgdppcg: Real GDP growth per capita (percent) — IMF, WEO
- Independent variables and sources (selected):
  - govc_ngdp: Government consumption (percent of GDP) — IMF, WEO
  - privateinv_ngdp: Private investment (percent of GDP) — IMF, WEO
  - publicinv_ngdp: Public investment (percent of GDP) — IMF, WEO
  - fdi_ngdp: FDI (percent of GDP) — IMF, WEO
  - frisk: Financial risk (index) — International Country Risk Guide (ICRG)
  - Volatility of REER: Standard deviation of REER change — IMF, INS
  - openness: Sum of imports and exports (percent of GDP) — IMF, WEO
  - healthexp: Health expenditure (percent of GDP) — World Bank, WDI
  - pri_enroll: Primary school enrollment rate (percent of the population of relevant age) — World Bank, WDI
  - inf: Inflation (percent) — IMF, WEO
  - floatdj: Floating exchange rate regime dummy, de jure regime classification — IMF AREAER publication, various years
  - intdj: Intermediate exchange rate regime dummy, de jure regime classification — IMF AREAER publication, various years
  - floatdf: Floating exchange regime dummy, de facto regime classification — IMF AREAER, Bubula and Ötker-Robe (2002), and Chinese University of Hong-Kong, http://intl.econ.cuhk.edu.hk/index.php
  - intdf: Intermediate exchange rate regime dummy, de facto regime classification — same sources as floatdf
  - float_10b, int_10b, float_10a, int_10a: Floating/intermediate dummies with below/above 10% inflation — Staff estimates based on earlier classification
  - fear of floating: The country announces a de jure floating exchange regime but de facto deviates from it — Staff estimates
  - fear of pegging: The country announces a de jure pegged exchange regime but de facto deviates from it — Staff estimates
  - IT: Inflation targeting — IMF, INS
  - kaopen: Variable measuring capital account openness — Relative openness to OECD countries, based on Lane and Milesi-Ferretti (2007) database

### Appendix A — Descriptive statistics (Table A3 and Table A4/A5 excerpts)
- Table A3. Descriptive Statistics, 1980-2012 (Source: IMF staff estimates)
  - Real GDP per capita growth (percent change): Mean 4.2, Standard Deviation 3.6, Min -11.5, Max 16.2
  - Private investment (percent of GDP): Mean 19.0, Standard Deviation 6.4, Min 3.0, Max 41.0
  - Public investment (percent of GDP): Mean 7.3, Standard Deviation 5.6, Min 2.0, Max 32.8
  - Government consumption (percent of GDP): Mean 10.3, Standard Deviation 2.9, Min 3.5, Max 17.6
  - Standard deviation of REER growth (units): Mean 6.4, Standard Deviation 5.1, Min 0.6, Max 41.0
  - Financial risk (index): Mean 37.5, Standard Deviation 7.4, Min 14.0, Max 48.5
  - FDI (percent of GDP): Mean 4.0, Standard Deviation 6.6, Min -2.8, Max 53.1
  - Capital account openness (index): Mean 88.9, Standard Deviation 172.5, Min 5.5, Max 1,024.6
- Table A4. Selected Variable Means by Type of Exchange Rate Regime, 1980-2012 (de Jure classification) — Regime counts: Floating (109), Intermediate (153), Pegged (74)
  - Floating: Real GDP per capita growth 3.7; Private investment 20.3; Public investment 5.1; Government consumption 11.1; Standard deviation of REER growth 8.2; Financial risk 37.0; FDI 2.0; Capital account openness 22.6
  - Intermediate: Real GDP per capita growth 4.8; Private investment 18.2; Public investment 8.9; Government consumption 11.0; Standard deviation of REER growth 5.4; Financial risk 38.1; FDI 4.9; Capital account openness 67.3
  - Pegged: Real GDP per capita growth 3.5; Private investment 19.0; Public investment 7.1; Government consumption 8.9; Standard deviation of REER growth 6.1; Financial risk 37.0; FDI 5.2; Capital account openness 229.7
- Table A5. Selected Variable Means by Type of Exchange Rate Regime, 1980-2012 (de Facto classification) — Regime counts: Floating (48), Intermediate (181), Pegged (107)
  - Floating: Real GDP per capita growth 3.0; Private investment 17.2; Public investment 4.7; Government consumption 10.7; Standard deviation of REER growth 6.5; Financial Risk 34.9; FDI 3.0; Capital account openness 22.7
  - Intermediate: Real GDP per capita growth 4.4; Private investment 20.6; Public investment 6.9; Government consumption 10.6; Standard deviation of REER growth 6.7; Financial Risk 38.3; FDI 4.1; Capital account openness 62.5
  - Pegged: Real GDP per capita growth 4.3; Private investment 17.4; Public investment 9.1; Government consumption 9.7; Standard deviation of REER growth 5.9; Financial Risk 37.4; FDI 4.4; Capital account openness 162.9

### Appendix A — Correlations (Table A6, selected)
- Correlation matrix headings include: rdpppc, inincome0, privateinv_ngdp, publicinv_ngdp, floatdf, intdf, govc_ngdp, frisk, reer_sd, crisis, kaopen.
- Example entries (as provided):
  - rdpppc with rdpppc: 1
  - rdpppc with inincome0: -0.05351
  - privateinv_ngdp with rdpppc: 0.21960
  - publicinv_ngdp with rdpppc: 0.331
  - floatdf with rdpppc: -0.13670
  - intdf with rdpppc: 0.07230
  - govc_ngdp with rdpppc: 0.1169
  - frisk with rdpppc: 0.3133
  - reer_sd with rdpppc: -0.165
  - crisis with rdpppc: -0.1923
  - kaopen with rdpppc: -0.0357
- Sources: IMF World Economic Outlook and staff estimates.

### Appendix B — De Jure and De Facto Classification (key points and regime distributions)
- Two classification schemes adopted:
  - IMF de jure regime classification (governments' claimed regimes)
  - IMF de facto classification (observed policies and exchange rate behavior), with data available through AREAER and supplemented by CUHK data where missing.
- IMF moved from a purely de jure classification to a hybrid de facto methodology from 1999, applied retrospectively as well.
- Table B1. Distribution of Regimes (IMF AREAER and staff estimates)
  - De Jure (in total): Floating 203 (24.6 percent), Intermediate 293 (35.5 percent), Pegged 329 (39.9 percent); Total 825 (100)
  - De Facto (in total): Floating 102 (12.4 percent), Intermediate 434 (52.6 percent), Pegged 289 (35.0 percent); Total 825 (100)
- Interpretation noted in text:
  - Under de facto categorization, there are only half as many floating regimes compared with de jure, indicative of a "fear of floating" in Asian countries.
  - Pegged regimes slightly prevail in the de facto sample.
- Country-level de facto regime patterns (Table B2 highlights, counts as presented):
  - Countries identified as never having floating regimes: Bangladesh, Brunei Darussalam, China, Fiji, Hong Kong SAR, Lao P.D.R, Maldives, Myanmar, Nepal, Singapore, Vietnam.
  - Countries that never had pegged regimes: Japan, Korea, Taiwan, Province of China.
  - Countries with no intermediate regime at any time: Nepal and Pakistan.
  - Countries that had intermediate regimes most of the time: Sri Lanka, Singapore, Korea, Indonesia, Cambodia.
- Table B2 excerpt (de facto regime distribution by country; sample totals at bottom):
  - Total counts shown: Floating 102, Intermediate 347, Pegged 376, Total 825, and "Floats in Percent of Total" column ending with 12
  - (Table rows list country-specific counts for Floating, Intermediate, Pegged, Floats in Percent of Total for each country as provided.)
- Frequency distribution figures:
  - Figure B1: Frequency Distribution of Exchange Rate Regimes, 1980-2012 (de jure classification, in percent) — Sources: AREAER and IMF staff calculations.
  - Figure B2: Frequency Distribution of Exchange Rate Regimes, 1980-2012 (de facto classification, in percent) — Sources: AREAER and IMF staff calculations.

*Source: _wp15195 - References (PDF), IMF staff estimates and cited IMF publications.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15195.pdf_