## 1. Cross-Country Current Account-Growth Relationship in Developing Countries

## Source details

**Canonical URL:** [1. Cross-Country Current Account-Growth Relationship in Developing Countries](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1190.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1190.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1190.pdf.json)

---

### Introduction and research motivation
- Developing countries experienced large international capital flows in the 1980s and 1990s, but empirical evidence shows little robust long-run growth benefits from global capital inflows.
- Some studies find a negative correlation between growth and capital flows in developing countries; a larger current account surplus (capital outflow) is associated with higher growth.
- This paper advances the thresholds research agenda by:
  - Identifying growth volatility as a robust threshold variable.
  - Distinguishing country regimes (low-volatility vs high-volatility) rather than only differentiating countries.
  - Showing that shifts in a country’s growth volatility are primarily the outcome of external factors.

### Key empirical observations and sample
- Unit of observation: growth over five-year periods.
- Observed growth volatility (standard deviation of growth) ranges between 0.27 percent and 25.65 percentage points.
- Estimated volatility threshold separating regimes: 5.35 percent (Hansen (1996, 2000) methodology).
- Core sample:
  - 61 countries in main analysis.
  - 59 countries were in the low volatility regime for at least one five-year period (1980–2003).
  - 35 countries were in the high volatility regime for at least one five-year period.
- Sample period emphasized: 1980 to 2003 (24-year period).
- Panel construction: five non-overlapping intervals 1980-84, 1985-89, 1990-94, 1995-99, 2000-03; 24 annual observations per country; 1,464 total annual observations.

### Main findings on volatility regimes and capital inflows
- Relationship by regime:
  - Low-volatility regime: positive relationship between capital inflows and growth (private capital particularly productive).
    - Threshold-regression estimate: in low-volatility regime an inflow of capital of one percent of GDP is associated with a 0.094 percent increase in the growth rate.
  - High-volatility regime: the positive relationship disappears and capital inflows can be associated with lower growth.
    - Top quartile (highest 25 percent) volatility: a one-percentage point increase in capital inflows-to-GDP ratio is associated with 0.16 percentage points lower growth (bottom three quartiles: a one-percentage point increase is associated with 0.07 percentage points higher growth; significance at the 7–8 percent confidence level).
- Composition shifts in high-volatility periods:
  - Private capital inflows tend to decline.
  - Official aid share increases; official aid is not associated with higher growth (Deaton (2010) summary).
- Institutions are not an effective standalone threshold for capital inflow productivity:
  - Institutional quality is one among several factors determining volatility.
  - Volatility varies significantly over time while institutions are more time-invariant; institution-based thresholds do not track regime changes well.
- Volatility is driven by interaction of location, institutional features, and global/external factors (for example, US interest rates and commodity prices).

### Empirical strategy and methodological points
- Baseline: standard cross-country growth equation where long-run growth rate of real GDP per capita is a function of controls and average capital inflow measure (current account balance).
- Two departures from Prasad et al. (2006) replication:
  - Focus on 1980–2003 rather than 1970–2000.
  - Openness measured as total trade (exports plus imports normalized by GDP) rather than Sachs and Warner dummy.
- Heteroscedasticity and panel weighting:
  - Annual growth g_{i,t} modeled with country-specific covariates x_i and time-varying innovation standard deviation σ_{i,t} that varies across five periods.
  - Maximum likelihood (weighted) estimation down-weights high-volatility episodes (weights proportional to reciprocal of volatility) and can change inference relative to OLS.
  - Prasad et al. (2006) reported a 0.1 percentage point drop in average annual growth from capital inflows; authors find coefficient 0.06 in some specifications but when allowing time-varying volatility and using maximum likelihood, capital inflows show a positive relationship with higher growth significant at the 5-percent level in medium-term specifications.

### Nonlinearities and volatility threshold tests
- Nonlinear threshold model following Hansen (1996; 2000):
  - Null of no threshold rejected with p-value = 0.02.
  - Estimated volatility threshold Γ = 5.35 percent (standard deviation of growth).
  - Only 25 percent of observations are in the high-volatility regime (σ > 5.35 percent).
- Regime-specific point estimates:
  - Low-volatility regime: Current Account Balance coefficient -0.094 (-(2.86)) reported for σ ≤ 5.35 in Table 5 (note: sign convention depends on current account definition; main text interprets inflows/deficits as promoting growth in low-volatility settings).
  - High-volatility regime: Current Account Balance coefficient 0.177 ((2.07)) reported for σ > 5.35 in Table 5 (table shows regime-specific coefficient heterogeneity across specifications and sign interpretation tied to definition).
- Institutional and income threshold tests:
  - Institutional threshold p-value = 0.54 (cannot reject null of no institutional threshold).
  - Income threshold p-value = 0.2 (cannot reject null of no income threshold).
  - In some splits, poor countries (per-capita income less than $2,350) show negative and statistically significant effects of capital flows on growth, but threshold tests do not strongly reject nulls.

### Panel regressions and endogeneity (identification)
- Concern: causality may run from growth to capital flows; valid instruments for average current account hard to find.
- Identification strategy: panel methods controlling for country fixed effects; estimation via system-GMM (Arellano and Bover (1995); Blundell and Bond (1998)), one-step variant reported.
- Main panel-GMM findings:
  - Low-volatility regime: a one percent increase in the current account deficit relative to GDP is associated with a 0.18 percent increase in average annual growth rates in GMM estimates (compared with 0.09 percent in OLS).
  - High-volatility regime: GMM point estimate on capital flows coefficient is indistinguishable from zero.
  - Interpretation: negative shocks to growth may attract capital inflows (smoothing), complicating causal interpretation; still, evidence supports a growth dividend from private inflows in low-volatility settings.

### Breakdown by type of flows
- Flow measures:
  - Private capital: sum of net direct investment, net portfolio flows, and other net private capital flows (World Economic Outlook series BFXP), averaged over five-year periods; may include small element of official flows.
  - Foreign aid: net official development assistance (World Development Indicators, series DT.ODA.ALLD.CD), averaged over five-year periods.
- Results:
  - Private capital inflows:
    - Low-volatility regime: Arellano-Bover estimates show a one percent increase in private capital inflows/GDP associated with growth increases between 0.14 and 0.24 percent annually (Table 9, columns 2 and 4).
    - High-volatility regime: positive correlations exist but are not statistically significant or robust.
  - Foreign aid:
    - Appears to retard growth during low-volatility periods but coefficients are not statistically significant at 5 percent.
    - Negative aid coefficients somewhat larger in high-volatility regime but never significant at 5 percent; coefficients sensitive to exclusion of outliers (Suriname, Republic of Congo).
  - Overall: private inflows stimulate growth in low-volatility settings; official/aid inflows show no robust positive effect on growth in either regime.

### Determinants of volatility and external drivers
- Conceptual view: volatility proxies for combination of factors (institutions, geography, production structure) affecting investment climate and responsiveness to foreign investment.
- Burnside and Tabova (2010) approach:
  - Regress annual GDP growth on six global factors (US growth, three-month T-Bill real rate, excess returns on US stocks, and three commodity price shocks) plus median growth.
  - Volatility of predicted growth rates used as a summary metric of exposure to external volatility.
  - Re-estimation yields coefficient on volatility of predicted growth rates of 1.48 (t-statistic = 7.3) and R-squared = 0.72 (Table 10, column 1), confirming external-factor-generated variation explains much cross-country volatility.
- Malik and Temple (2009) determinants:
  - Output more volatile in smaller countries.
  - Volatility rises non-linearly with distance from the coast.
  - Institutional measures matter unless government type included; democracies show lower volatility and render other institutional variables insignificant when included.
  - Ethnic fractionalization and external wars raise volatility; terms-of-trade volatility sometimes significant.
  - Malik and Temple variables account for 45 percent of cross-country variation in growth volatility.
- Joint findings:
  - Adding Burnside and Tabova measure to Malik and Temple regressions increases explanatory power (Burnside and Tabova variable remains highly significant but point estimate shrinks), implying external drivers capture important components of volatility beyond standard domestic determinants.
  - Conclusion: Burnside and Tabova variable explains 72 percent of cross-country variation alone; R-squared rises to 79 percent when Malik and Temple variables added — inclusion of Malik and Temple variables adds little incremental explanatory power.

### Measurement, within-country variation, and implications
- Volatility measurement:
  - Burnside and Tabova measure: standard deviation of annual real GDP growth rate.
  - This paper follows Ramey and Ramey (1995) and estimates volatility of innovations to growth allowing volatility to vary across countries; estimated from regression in Table 1, column 6.
- Within-country variation:
  - Regressions of actual volatility on predicted volatilities with time dummies show actual volatility variations almost exactly mirror predicted volatility variations—implying volatility of external influences strongly conditions country volatility changes.
  - Terms-of-trade volatility varies over time but largely captured by Burnside and Tabova measure.
  - Institutional improvements associated with reduced volatility (ICRG measure) though results not fully reported.
- Capital flows and volatility:
  - Regressions including average current account, average net private flows, and foreign aid over five-year windows find:
    - Aggregate capital inflows/GDP unrelated to output volatility.
    - Private capital flows negatively associated with volatility; official flows positively associated with volatility—findings have low statistical significance and are inconclusive.
  - Non-linearities and regimes matter:
    - Standard deviation of innovations to growth of 5.35 percent demarcates low and high regimes.
    - In high regime, capital inflows negatively correlated with growth and drive negative pooled correlations when regimes are conflated.
    - Private flows deliver the growth dividend primarily in low-volatility regimes; aid increases in high-volatility regimes and has negative relationship with growth in both regimes.
    - Commodity-dependent enclave economies may attract large private flows yet see limited growth or volatility mitigation benefits.

### Sample facts and long-run interpretation
- Of 61 countries, only 26 were in the low-volatility regime throughout 1980–2003; 35 switched into high-volatility periods at least once.
- For countries that move between regimes, long-run effect of capital inflows on growth is a weighted combination of regime-specific effects; pooled cross-country claims about “long-run” effects are misleading without accounting for regime dynamics.

### Appendix I — Variables and Data Sources (selected definitions)
- Growth: log difference of chained real GDP (rgdpch, Penn World Tables 6.2); dependent variable either annual growth or averages over 1980–2003 or five non-overlapping periods.
- Current Account: percent of GDP (BCA, World Economic Outlook); averaged over sample or subperiods.
- Private Capital (Private Flows): sum of net FDI, net portfolio flows, and other net private capital flows (BFXP, World Economic Outlook); averaged over five subperiods.
- Foreign Aid: net official development assistance (DT.ODA.ALLD.CD, World Development Indicators, 2008); averaged as above.
- Institutional Quality: Hall and Jones (1999) measure constructed as average of five ICRG variables, measured around 1985.
- Openness: total trade flows as percent of GDP (NE.TRD.GNFS.ZS, WDI, 2008).
- Volatility of Terms of Trade: standard deviation 1980–2003 of log difference in net terms of trade (TT.PRI.MRCH.XD.WD, WDI).
- External factors (Burnside and Tabova inputs): US real GDP growth; US real interest rates (3 month T-bill minus US PPI inflation); US excess returns; oil/metals/agriculture price indices converted using US PPI.

### Key empirical coefficients and table highlights (selected)
- Table 1 — Current Account Balance coefficients across specifications: 0.098; 0.060; 0.012; 0.012; -0.013; -0.027. Associated t-statistics where shown: (2.13), (1.45), (0.13), (0.15), -(0.14), -(0.27).
- Table 2 — Weighted regressions (maximum likelihood) Current Account Balance coefficients: -0.082; -0.068; -0.070; -0.091. t-statistics: -(1.66), -(1.99), -(2.01), -(3.38).
- Table 3 — Volatility summary: Overall Mean 4.43; Standard Deviation 3.26 (Overall); Minimum 0.27 (Overall); Maximum 25.65 (Overall).
- Table 4 — Interaction and regime splits: Top-quartile volatility link shows negative growth effect for capital inflows; regime threshold examples: Low Volatility (< 2.54%) coefficient 0.600 (1.95); Medium Low (< 3.86%) coefficient 0.896 (2.22); Medium High (< 5.35%) coefficient 1.064 (2.42).
- Table 5 — Hansen threshold estimates: Volatility threshold 5.35 (95% CI (3.21, 6.96)); Institutional quality threshold 0.61 (95% CI (0.24, 0.79)); Income threshold 7.76 (95% CI (6.13, 10.34)). Regime-specific current account coefficients vary across splits (examples reported in main text and tables).
- Table 8 — Panel (Arellano-Bover) low-regime GMM: current account -0.112 (-(2.99)) in one low-regime specification; low-volatility GMM finds a one percent increase in current account deficit associated with 0.18 percent higher growth in reported GMM estimates (text summary).
- Table 9 — Private flows in low-volatility regime: coefficients in Arellano-Bover range 0.14 to 0.24 (significant in Table 9 columns cited); foreign aid coefficients generally negative and not robustly significant.
- Table 10 — Determinants of volatility: Constant/Std Dev of Predicted Growth examples 1.469 (7.92); 1.128 (6.79); 1.024 (10.46); R-squared values show high explanatory power for Burnside and Tabova predictor (e.g., R-squared = 0.72 in re-estimation).

*Source: _wp1190 — 1. Cross-Country Current Account-Growth Relationship in Developing Countries (excerpt).*

### 1. Cross-Country Current Account-Growth Relationship in Developing Countries ...............28

### 1. Cross-Country Current Account-Growth Relationship in Developing Countries

### Introduction and research motivation
- Developing countries stood to benefit from large international capital flows of the 1980s and 1990s, but empirical evidence shows little robust long-run growth benefits from global capital inflows.
- Some studies find a negative correlation between growth and capital flows in developing countries; a larger current account surplus (capital outflow) is associated with higher growth.
- The paper seeks to advance the thresholds research agenda by:
  - Identifying growth volatility as a robust threshold variable.
  - Distinguishing country regimes (low-volatility vs high-volatility) rather than only differentiating countries.
  - Showing that shifts in a country’s growth volatility are primarily the outcome of external factors.

### Key empirical observations and statistics
- Unit of observation: growth over five-year periods.
- Observed growth volatility (standard deviation of growth) ranges between 0.27 percent and 25.65 percentage points.
- A GDP growth volatility of 5.35 percent distinguishes low and high volatility regimes (based on Hansen (1996 and 2000) methodology).
- Core sample: 61 countries.
  - 59 countries were in the low volatility regime for at least one five-year period (1980–2003).
  - 35 countries were in the high volatility regime for at least one five-year period.
- Sample period emphasized in this paper: 1980 to 2003 (24-year period), chosen because sizeable global capital flows are a post-1980s characteristic.
- The paper contrasts its 1980–2003 focus with Prasad et. al. (2006), who covered 1970–2000.

### Main findings on volatility regimes and capital inflows
- The positive relationship between capital inflows and growth (predicted by standard theory) is clear in the low volatility regime and disappears in the high-volatility regime.
- Countries move in and out of volatility regimes; conflating regimes can be misleading because determinants of growth change across regimes.
- High volatility periods are associated with:
  - Smaller private capital inflows (greater uncertainty deters investors).
  - Deferred investment in equipment and technology and larger planning errors.
  - Depressed complementary investment in human capital.
- In high volatility regimes, aggregate capital inflows/GDP ratios have not necessarily been lower; composition shifts toward official aid rather than private flows.
  - Official aid, as summarized by Deaton (2010), is not associated with higher growth.
- Institutions alone do not serve as an effective threshold for capital inflow productivity:
  - Institutional quality is only one among several factors determining a country’s overall volatility.
  - Volatility changes significantly over time, while institutions tend to be more time-invariant; therefore institutions do not track regime changes well.
- Volatility is driven by the interaction of location, institutional features, and global/external factors (for example, US interest rates and commodity prices).

### Empirical strategy and methodological points
- Baseline growth specification: standard cross-country growth equation where long-run growth rate of real GDP per capita is a function of controls and average capital inflow measure (current account surplus in Prasad et. al. (2006) framework).
- Two departures from Prasad et. al. (2006) in replication:
  - Focused on 1980–2003 rather than 1970–2000.
  - Increased country coverage by measuring openness as volume of trade (exports plus imports normalized by GDP) rather than the Sachs and Warner dummy.
- Statistical note: pooled cross-country regression misspecifies the capital inflows–growth relationship when regimes are conflated.
  - Maximum likelihood estimates place lower weight on high growth-volatility observations and produce a strongly positive role for capital inflows compared with OLS.
  - The difference between OLS and maximum likelihood signals misspecification rather than a correction for bias.

### Results robustness and endogeneity
- Distinction between low and high volatility regimes remains robust across:
  - Cross-section analyses.
  - GMM models that address endogeneity concerns.
- In low volatility observations:
  - Countries receive capital flows primarily as private capital.
  - Higher private inflows are associated with higher growth.
- In high volatility observations:
  - Capital arrives mainly in the form of official aid, which is not associated with higher growth.

### Determinants of growth volatility and external drivers
- Drawing on Malik and Temple (2009) and Burnside and Tabova (2010):
  - Structural characteristics (e.g., location and production diversity) are important for cross-country differences in volatility.
  - Conditional on structural characteristics, growth volatility is a function of external drivers that are largely exogenous (e.g., US interest rates, commodity prices).
- Because volatility can change for exogenous reasons, the effectiveness of capital inflows can change over time even if domestic institutions remain constant.

### Conclusion (as organized in the paper)
- The pooled cross-country finding of no relationship between capital inflows and growth is driven principally by observations in high volatility regimes.
- When countries are in the lower volatility zone, foreign capital—especially private capital—is productive and associated with higher growth.
- Policy and empirical implications emphasize accounting for volatility regimes and external drivers when assessing the growth effects of capital flows.

*Source: _wp1190 - 1. Cross-Country Current Account-Growth Relationship in Developing Countries (excerpt) — 1980–2003 analysis and findings.*

### 0.1 percentage point drop in the average annual growth rate. From this they conclude that capital

### _wp1190 - 0.1 percentage point drop in the average annual growth rate. From this they conclude that capital

### A. Dealing with Heteroscedasticity Reveals Misspecification
- Findings and methodological points
  - Cross-country growth regressions commonly regress average growth over the entire period on initial values or averages of covariates; when data are heteroscedastic, averaging is not necessary and is inefficient.
  - OLS is standard but GLS or maximum likelihood that exploit panel (time-series) variation can improve efficiency by modeling time-varying volatility.
  - Adaptation of Ramey and Ramey (1995) framework: annual growth g_{i,t} for i=1,...,61 and t=1980,...,2003 is modeled with country-specific covariates x_i and time-varying innovation standard deviation σ_{i,t} that varies across five non-overlapping intervals: 1980-84, 1985-89, 1990-94, 1995-99, and 2000-03.
  - Sample construction: 24 observations per country and 1,464 observations in total.
- Key empirical contrasts and coefficients
  - Prasad et al. (2006) find capital inflows hurt growth (0.1 percentage point drop in average annual growth); authors find a coefficient of 0.06 (smaller t-statistic), supporting that result in some specifications.
  - Allowing volatility to vary over time (five-year intervals) and using maximum likelihood changes inference: in medium-term (five-year varying covariates) specifications, capital inflows show a positive relationship with higher growth, significant at the 5-percent level.
  - When volatility is allowed to vary only across countries (not over time), the current account coefficient changes sign but appears to be statistical noise.
- Implications
  - Maximum likelihood weights observations by the reciprocal of volatility; high-volatility episodes are down-weighted, potentially isolating relationships that hold in low-volatility country-periods.
  - Results suggest misspecification in models that ignore time-varying volatility: the relationship between current account and growth is conditional on volatility and may only hold during low-volatility episodes.

### B. Alternate Specification
- Model re-specification
  - Regress average growth in country i in period k (periods: 1980–84, 1985–89, 1990–94, 1995–99, 2000–03) on controls, average current account k_{i,k}, volatility σ_{i,k}, and an interaction between k_{i,k} and σ_{i,k}.
  - Volatility measure obtained from a previously estimated model (column 4 in Table 2).
- Key interaction findings
  - Simple linear interaction between current account and volatility is not statistically meaningful.
  - Nonlinear specification using volatility dummies: below 25th percentile, below 50th percentile, below 75th percentile.
  - Top quartile (highest 25 percent volatility): a one-percentage point increase in the capital inflows-GDP ratio is associated with 0.16 percentage points lower growth.
  - Bottom three quartiles: a one-percentage point increase in capital inflows-to-GDP ratio is associated with 0.07 percentage points higher growth.
  - These differentiated coefficients are significant at the 7–8 percent confidence level (not the conventional 5 percent).
- Conclusions
  - Time-varying volatility conditions the capital flows–growth relationship.
  - A threshold effect is suggested: volatility must be in the top quartile to materially weaken or reverse the growth benefits of capital inflows.

### III. VOLATILITY THRESHOLDS
- Nonlinear threshold model specification
  - Growth g_{i,t} modeled allowing the effect of current account k_{i,t} on growth to change when volatility σ_{i,t} exceeds an estimated threshold Γ, following Hansen (1996; 2000).
- Statistical test and threshold estimate
  - Null of no threshold is rejected with p-value = 0.02.
  - Estimated threshold: standard deviation of growth crosses 5.35 percent of real GDP per capita.
  - Only 25 percent of observations belong to the high-volatility regime (i.e., above 5.35 percent).
- Regime-specific effects
  - Low-volatility regime: an inflow of capital of one percent of GDP is associated with a 0.094 percent increase in the growth rate.
  - High-volatility regime: capital inflows accompany a sharp reduction in growth (negative coefficient).
  - Institutional threshold split: coefficient on capital flows not significant at 5 percent in either low- or high-institutions regime; p-value for institutional threshold = 0.54 (cannot reject null).
  - Income threshold split: in poor countries (per-capita income less than $2,350) capital flows have negative and statistically significant effects on growth; p-value for income threshold = 0.2 (cannot reject null).
- Distributional observations
  - Panel of 298 observations (61 countries * 5 periods - 7 missing).
  - Many countries switch regimes over time; 26 countries were in the low-volatility regime throughout, 35 experienced high volatility for at least one period.
  - Countries with more periods of high volatility tend to be poorer with weaker institutions; private capital flows to GDP decrease as number of high-volatility periods increases (except a small group with commodity dependence).

### IV. PANEL REGRESSIONS AND ENDOGENEITY
- Identification strategy
  - Concern: causality may run from growth to capital flows; difficulty finding valid instruments for average current account.
  - Attempted instruments such as distance/distantness found weak for average current account; instead use panel identification controlling for country effects.
  - Estimation via system-GMM (Arellano and Bover (1995); Blundell and Bond (1998)), one-step variant reported.
- Main panel-GMM findings
  - Using volatility threshold from Table 5 to split samples into high- and low-volatility regimes.
  - Low-volatility regime: a one percent increase in the current account deficit relative to GDP is associated with a 0.18 percent increase in average annual growth rates in GMM estimates (compared with 0.09 percent in OLS).
  - High-volatility regime: GMM point estimate on capital flows coefficient is indistinguishable from zero (capital does not foster growth, nor does it significantly hurt growth in GMM).
  - Interpretation: negative shocks to growth may attract capital inflows to smooth shocks; e.g., Bluedorn (2005) finds a shock valued at one year’s GDP pushes current account into deficit by 5 percent.

### IV.A Breakdown by Type of Flows
- Measures and limitations
  - Distinguish private capital inflows (average net inflow of private capital) and foreign aid (average inflow of aid) over five-year periods.
  - World Economic Outlook private capital flow measures may include a small element of official flows; aid measures include grants and other official flows, so sum of private + official does not equal current account deficit.
- Results for private and official flows
  - Private capital inflows:
    - Low-volatility regime: in Arellano-Bover estimates a one percent increase in private capital inflows relative to GDP is associated with a growth increase between 0.14 and 0.24 percent annually (Table 9, columns 2 and 4).
    - High-volatility regime: private capital and growth are positively correlated but relationship is not statistically significant or robust (Table 9, column 7).
  - Foreign aid:
    - Appears to retard growth during low-volatility periods, but not statistically significant.
    - Negative aid coefficients are somewhat larger in high-volatility regime but never significant at 5 percent.
    - Coefficient sensitivity: exclusion of Suriname and Republic of Congo reduces magnitude.
  - Overall: private inflows stimulate growth in low-volatility settings; official/aid inflows show no robust positive effect on growth in either regime.

### V. DETERMINANTS OF VOLATILITY
- Conceptual framing
  - Volatility may proxy for a combination of factors (institutions, geography, production structure) that affect investment climate and responsiveness to foreign investment.
- Empirical proxies and antecedents
  - Two papers inform analysis: Malik and Temple (2009) (cross-country determinants of volatility emphasizing geography and institutions) and Burnside and Tabova (2010) (time-varying global shocks interacting with domestic structure).
  - Burnside and Tabova proxy: regress annual GDP growth on six global factors (US growth, three-month T-Bill real rate, excess returns on US stocks, and three commodity price shocks) plus median growth in sample; volatility of predicted growth rates serves as a summary metric of exposure to external volatility.
- Validation and results
  - Re-estimation yields coefficient on volatility of predicted growth rates of 1.48 (t-statistic = 7.3) and R-squared = 0.72 (Table 10, column 1), confirming Burnside and Tabova’s finding that external-factor-generated variation explains much of cross-country volatility.
  - Malik and Temple determinants:
    - Output more volatile in smaller countries.
    - Volatility rises (non-linearly) with distance from the coast.
    - Measures of institutional quality (Hall and Jones index, ICRG, constraints on executive) matter unless government type is included; including government type finds lower volatility in democracies and renders other institutional variables insignificant.
    - Ethnic fractionalization and external wars raise volatility; terms-of-trade volatility sometimes significant.
    - Malik and Temple variables together account for 45 percent of cross-country variation in growth volatility.
  - When adding Burnside and Tabova measure to Malik and Temple regressions:
    - Burnside and Tabova measure remains highly significant but point estimate shrinks, indicating overlap and that external-exposure proxy captures important components of volatility beyond standard domestic determinants.

*Source: _wp1190 (excerpt provided).*

### conclusion is that the inclusion of the Malik and Temple variables adds little explanatory power,

### _wp1190 - conclusion is that the inclusion of the Malik and Temple variables adds little explanatory power,

### Explanatory power of variables
- The Burnside and Tabova variable by itself explains 72 percent of the cross-country variation in growth volatilities, with the R-squared rising to 79 percent following the inclusion of the Malik and Temple variables.
- Conclusion: the inclusion of the Malik and Temple variables adds little explanatory power.

### Measurement of volatility and within-country variation
- Burnside and Tabova (2010) estimate the actual volatility of growth as the standard deviation of the annual real GDP growth rate.
- This paper follows Ramey and Ramey (1995) and estimates the volatility of innovations to growth, allowing the volatility to vary across countries; the estimate of growth volatility is obtained from the regression equation reported in Table 1, column 6.
- Within-country estimates (column 4) of actual volatility regressed on predicted volatilities show that, after adding time dummies, variations in actual volatility almost exactly mirror variations in predicted volatility—implying that the volatility of external influences strongly conditions changes in a country’s volatility.
- Terms of trade volatility does vary over time but is largely captured by variation in the Burnside and Tabova measure. Institutional variations are captured in the ICRG measure and do show that improved institutions are associated with reduced volatility (results not reported).

### Determinants of volatility (cross-section and time variation)
- Externally-induced volatility differentiates country volatility, but the level of volatility is also a function of domestic structure as implied by the Burnside and Tabova method.
- The Malik and Temple approach helps identify some structural factors; geographical remoteness is particularly important.
- The considerably greater explanatory power of the Burnside and Tabova variable suggests additional structural features are at play beyond those captured by Malik and Temple.
- Changes in volatility over time within a country, conditioned by unchanging structural features, are largely dependent on changes in global drivers.
- Weak institutions raise volatility but are only part of the explanation; institutions are not a good proxy for volatility on their own.
- Volatility is an inclusive measure, encompassing several country features that deter international capital and render them less effective.

### Capital flows and volatility
- Columns 5 to 8 include the average size of current account balances, average net private flows and foreign aid, over five-year windows as regressors.
- Results: an inflow of capital (i.e., a current account deficit) is unrelated to output volatility.
- Private capital flows are negatively associated with volatility and official flows are positively associated; given the low statistical significance, these findings are not conclusive.
- Aggregate capital flows (or private and official flows separately) are not correlated with country growth volatility.

### Capital inflows and growth: non-linearities and regimes
- Non-linearities are important: countries are differentiated by their level of growth volatility; some countries stay in a volatility regime while others move between regimes, causing shifts in the productivity of capital inflows.
- A standard deviation of innovations to growth of 5.35 percent of real GDP per-capita demarcates low and high volatility regimes.
- In the high regime, capital inflows are negatively correlated with growth, which drives the negative correlation between capital inflows and aggregate growth when all countries are pooled.
- Failure to distinguish between volatility regimes can lead to the incorrect conclusion that capital inflows are damaging in all situations.
- Even when growth and capital flows are positively correlated, causality could run from growth to capital flows; recognizing limitations of GMM estimates in dealing with endogeneity, the results suggest capital inflows do spur growth.
- The action comes mainly from private capital flows:
  - In low volatility regimes, countries attract larger private capital flows and gain a growth dividend from these flows.
  - In high volatility settings, private capital flows diminish and have an insignificant relationship with growth.
- Aid-flows increase in high volatility regimes; whether in a low or high volatility period, aid has a negative relationship with growth.
- Some heavily commodity-reliant economies attract large private capital flows relative to the economy, but being of an enclave nature these flows have limited impact on growth or in mitigating volatility.

### Implications for long-run effects and sample facts
- Statements about “long-run” effects of foreign capital do not apply when countries move in and out of regimes.
- Of the 61 countries in the sample, only 26 were in the low volatility regime throughout the period analyzed; for the other 35 countries, the effective long-run relationship between capital inflows and growth will be a combination of effects in low and high volatility regimes.

*IMF Working Paper content unit _wp1190 — conclusion excerpt provided in source content.*

### Appendix I. Variables and Data Sources

### _wp1190 - Appendix I. Variables and Data Sources

### Variables and Data Sources (Definitions and Measurement)
- Growth
  - Calculated as the log difference of chained real GDP (series rgdpch, Penn World Tables 6.2). Dependent variable in regressions is either the annual growth rate or the average over the entire sample period 1980 to 2003, or over non-overlapping periods: 1980-84, 1985-89, 1990-94, 1995-99, and 2000-03.
- Current Account
  - Expressed as a percent of GDP (series BCA, World Economic Outlook). Appears as a country average over the entire sample period or over 1980-84, 1985-89, 1990-94, 1995-99, and 2000-03.
- Ethnic Fractionalization
  - One minus the Herfindahl index of ethnolinguistic group share (Alesina et al., 2003).
- Financial Development
  - Private credit as a percent of GDP intermediated through the financial sector (World Bank, Financial Structure Database).
- Financial Integration
  - Sum of stocks of (the absolute value of) external assets plus external liabilities as a percentage of GDP (Lane and Milesi-Ferretti, 2007).
- Fiscal Balance
  - General government balance expressed as a percent of GDP (series GGB, World Economic Outlook). Due to data limitations appears only as an average for the entire sample period in cross-country regressions.
- Foreign Aid
  - Net official development assistance (series DT.ODA.ALLD.CD, World Development Indicators, 2008). Averaged over full sample or the five non-overlapping periods above.
- Government Type
  - Calculated as the difference between democracy and autocracy scores (Polity IV project).
- Institutional Quality
  - Hall and Jones (1999) measure of government anti-diversion policies; constructed as an average of five ICRG variables: “Corruption in Government,” the “Rule of Law,” “Expropriation Risk,” “Repudiation of Contracts by Government,” and “Quality of the Bureaucracy.” Measured for each country around 1985.
- Initial Income
  - Log of real per capita GDP adjusted for PPP (series rgdpl, Penn World Tables 6.2). Either 1980 value or initial value for each period.
- Life Expectancy
  - Measured in years (series SP.DYN.LE00.IN, World Development Indicators, 2008). Either 1980 value or initial period value.
- Land Area from Coast
  - Proportion of total area within 100 km of the ocean or ocean-navigable river (Gallup, Sachs and Mellinger, 1999).
- Landlocked
  - Dummy for landlocked country (Gallup, Sachs and Mellinger, 1999).
- Openness
  - Total trade flows as a percent of GDP (series NE.TRD.GNFS.ZS, World Development Indicators, 2008). Averaged over full sample or five subperiods.
- Population
  - Logarithm of initial population in 1980 (series SP.POP, World Development Indicators, 2008).
- Population Growth
  - Population growth (series SP.POP.GROW, World Development Indicators, 2008). Averaged over the five subperiods.
- Private Capital (Private Flows)
  - Sum of net direct investment, net portfolio flows, and other net private capital flows (series BFXP, World Economic Outlook). “Other private capital flows, net” computed by subtracting liabilities to official creditors from total other investment; may include some official investment. Averaged over the five subperiods.
- Primary Commodities Price Indexes
  - Oil, metals, and agricultural products price indices, converted to relative prices using the US PPI (Burnside and Tabova, 2010).
- US Real GDP Growth
  - Log difference of US real GDP per capita in constant 2000 dollars (Burnside and Tabova, 2010).
- US Real Interest Rates
  - Difference between the 3 month T-bill rate and the rate of inflation of the US producer price index (Burnside and Tabova, 2010).
- US Excess Returns
  - Excess return on the US stock market (Burnside and Tabova, 2010; original source: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html).
- Volatility of Terms of Trade
  - Standard deviation from 1980 to 2003 of the log difference in net terms of trade (series TT.PRI.MRCH.XD.WD, World Development Indicators; some gaps filled with IFS).
- War
  - Dummy for whether a country had an external war between 1960 and 1985 (Gallup, Sachs and Mellinger, 1999).

### Sample: Countries and Coverage
- 61 Country Sample across regions including East Asia and the Pacific, East and Central Europe, Latin America and the Caribbean, Middle East and North Africa, South Asia, Sub-Saharan Africa, Western Europe.
  - Examples listed: China, Hong Kong, Indonesia, Korea, Malaysia, Philippines, Thailand; Hungary, Poland, Romania, Turkey; Argentina, Brazil, Chile, Colombia, Mexico, Venezuela; Algeria, Egypt, Iran, Israel, Morocco, Saudi Arabia; India, Pakistan, Sri Lanka; Botswana, Cameroon, Cote d`Ivoire, Ethiopia, Ghana, Kenya, Nigeria, South Africa, Zimbabwe; Cyprus. (Full list provided in content.)

### Key Empirical Findings (Selected Coefficients and Statistics)
- Table 1 — Cross-Country Current Account-Growth Relationship
  - Current Account Balance coefficients across specifications: 0.098; 0.060; 0.012; 0.012; -0.013; -0.027. Corresponding t-statistics (where shown): (2.13), (1.45), (0.13), (0.15), -(0.14), -(0.27).
  - Initial Income coefficients: -1.257; -1.584; -1.782; -1.782; -2.248; -2.168. t-statistics: -(6.19), -(4.22), -(3.72), -(4.77), -(4.79), -(4.07).
  - Institutional Quality coefficients: 4.252; 5.084; 6.745; 6.745; 7.644; 7.666. t-statistics: (2.77), (2.99), (3.58), (4.07), (4.68), (4.36).
  - Number of Countries across columns: 60; 87; 61; 61; 61; 61.
  - Number of Observations in panel specifications: 1464; 1464; 1464.
  - Estimation methods include OLS and Maximum Likelihood; clustered standard errors or robust standard errors used as noted.

- Table 2 — Weighted Regressions (Panel, Maximum Likelihood)
  - Current Account Balance coefficients: -0.082; -0.068; -0.070; -0.091. t-statistics: -(1.66), -(1.99), -(2.01), -(3.38).
  - Initial Income coefficients: -1.763; -2.186; -1.986; -2.247. t-statistics: -(3.22), -(3.45), -(2.70), -(2.88).
  - Institutional Quality coefficients: 7.882; 9.613; 9.741; 10.510. t-statistics: (4.83), (5.08), (4.88), (5.48).
  - Number of Countries: 61; Number of Observations: 1464; 1429; 1429; 1429 (varies by column).

- Table 3 — Volatility Summary
  - Overall Mean: 4.43
  - Standard Deviation: 3.26 (Overall); 2.38 (Between); 2.23 (Within)
  - Minimum: 0.27 (Overall); 1.24 (Between); -4.23 (Within)
  - Maximum: 25.65 (Overall); 13.99 (Between); 17.17 (Within)

- Table 4 — Current Account, Growth and Volatility: Interactions
  - Baseline Current Account Balance coefficients across columns: -0.042; 0.058; 0.089; 0.159. t-statistics: -(0.82); (0.86); (1.07); (1.75).
  - Volatility * Current Account Balance: 0.010 (t-statistic (1.18)).
  - Regime interaction impacts (Impact of current account in low / mid-low / mid-high regime): -0.053 -(1.66); -0.062 -(1.83); -0.075 -(1.89).
  - Volatility regime thresholds reported: Low Volatility (< 2.54%) coefficient 0.600 (1.95); Medium Low (<3.86%) coefficient 0.896 (2.22); Medium High (<5.35%) coefficient 1.064 (2.42).
  - Number of Countries: 61; Number of Observations: 298.

- Table 5 — Threshold Effects (Hansen threshold tests)
  - Volatility threshold: 5.35 (95% confidence interval (3.21, 6.96)); p-values reported for threshold significance; reported regime splits and observations.
  - Institutional quality threshold: 0.61 (95% confidence interval (0.24, 0.79)).
  - Income threshold: 7.76 (95% confidence interval (6.13, 10.34)).
  - Examples of current account coefficients by regime (columns shown):
    - Low Regime (σ ≤ 5.35): Current Account Balance -0.094 (-(2.86))
    - High Regime (σ > 5.35): Current Account Balance 0.177 ((2.07))
    - Low Institutions (≤ 0.614): Current Account Balance -0.029 (-(0.44))
    - High Institutions (> 0.614): Current Account Balance 0.115 ((1.80))
    - Low Income (≤ 7.76): Current Account Balance 0.189 ((2.13))
    - High Income (> 7.76): Current Account Balance -0.034 (-(0.75))
  - Joint R-square examples: (0.02); (0.54); (0.20); 0.26; 0.23; 0.24 (by column).

- Table 6 — Descriptive Statistics by Regime (selected summary statistics)
  - Growth (Full Sample): 1.75; Current Account: -2.22; Private Flows: 1.92; Official Flows: 2.21; Income (ln per capita): 8.64; Life Expectancy: 66.54; Trade Volume: 72.32; Institutions: 0.57; Fiscal Balance: -3.91; Volatility: 2.68.
  - High Volatility Regime examples: Growth 0.53; Current Account -2.01; Private Flows -0.04; Official Flows 2.19; Income 8.61; Life Expectancy 69.07; Trade Volume 87.95; Institutions 0.61; Fiscal Balance -3.12; Volatility 7.26.

- Table 7 — Current Account-Growth Relationship by Incidence of High Volatility
  - Current Account Balance coefficients across multiple sample splits (examples):
    - (7-1): -0.099 (-(2.01))
    - (7-2): -0.111 (-(3.45))
    - (7-4): 0.001 ((0.02))
    - (7-7): -0.012 (-(0.22))
    - (7-10): 0.030 ((0.67))
    - (7-12): -0.092 (-(3.54))
  - Initial Income coefficients commonly negative and significant (examples: -2.289; -2.578; -2.707; -1.801).
  - Institutional Quality coefficients large and positive across specifications (examples: 9.716; 11.771; 11.591; 9.470).
  - Number of Observations vary across panels (examples: 604; 959; 1117; 1357) with corresponding numbers of countries reported per specification.

- Table 8 — Panel Regressions (Fixed Effects and Arellano-Bover)
  - Low Regime (8-1): Current Account -0.112 (-(2.99)); Initial Income -1.986 (-(2.74)); Openness 0.017 (0.02? t-statistic (2.76) shown).
  - High Regime (8-3): Current Account 0.251 ((2.19)); Initial Income -6.880 (-(1.76)); Initial Life Expectancy 0.342 (1.47); Openness 0.041 (0.80).
  - Number of Countries and Observations examples: Low Regime: 59 countries, 225 observations; High Regime: 35 countries, 73 observations.
  - Diagnostic tests reported: first and second order serial correlation p-values, Sargan test p-values; Number of Instruments example: 4127.

- Table 9 — Breakdown by Type of Flow (Fixed Effects Arellano-Bover)
  - Private Flows coefficients (selected): 0.098 ((2.54)); 0.242 ((2.76)); 0.064 ((1.63)); 0.144 ((2.29)); 0.193 ((0.98)); 0.166 ((1.55)); 0.003 ((0.02)); -0.007 (-(0.05)).
  - Foreign Aid coefficients (selected): -0.099 (-(1.09)); -0.113 (-(0.74)); -0.086 (-(0.96)); -0.072 (-(0.67)); -0.563 (-(1.68)); -0.210 (-(0.64)); -0.482 (-(1.15)); 0.134 ((0.48)).
  - Other controls and diagnostics: R-squared (within) examples (0.20); (0.24); (0.35); (0.42). Hansen test p-values and serial correlation p-values provided per column; Number of Instruments examples: 50; 77; 29.

- Table 10 — Determinants of Volatility (selected coefficients)
  - Constant/Standard Deviation of Predicted Growth in several columns: 1.469 (7.92); 1.128 (6.79); 1.024 (10.46); 1.009 (9.39); 1.036 (8.73).
  - Initial Population coefficient examples: -0.416 (-(2.15)); -0.103 (-(1.03)).
  - Percent of Land Area 100 km from Coast: -0.117 (-(2.66)); -0.044 (-(1.87)). Quadratic term Percent of Land Area 100 km from Coast^2: 0.001 (2.69); 0.000 (1.70).
  - War coefficient examples: -7.006 (-(3.35)); -2.738 (-(2.10)).
  - Ethnic Fractionalization*War: 0.126 (3.14); 0.047 (1.64).
  - Government type coefficients: -0.103 (-(2.19)); -0.054 (-(1.78)).
  - Number of observations and countries vary by column (examples: observations 615; 353; 298; 298; 266; countries 61; 35; 35; 61; 61; 54).
  - R-squared examples: within 0.25; between 0.74; overall 0.76; other columns show R-squared 0.45; 0.81; 0.50.

### Empirical Patterns and Interpretations Reported in Tables
- Institutional quality is consistently a strong, positive predictor of growth across multiple specifications (coefficients generally large and positive; t-statistics often highly significant).
- Initial income displays consistent negative coefficients (convergence effect) across specifications (e.g., -1.257 through -2.707 in various tables).
- The relationship between the current account balance and growth varies by specification and volatility regime:
  - Cross-sectional regressions sometimes show positive point estimates for current account (e.g., 0.098) but panel and weighted regressions often yield negative coefficients (e.g., -0.082; -0.068; -0.070; -0.091) with varying significance.
  - Threshold and interaction analyses indicate the sign and magnitude of the current account-growth relationship depend on volatility, institutional quality, and income thresholds (examples detailed under Table 5 and Table 4).
- Volatility itself is influenced by geographic and political factors (e.g., percent of land area near coast, war dummy, interactions of ethnic fractionalization with war, government type).

*Italicized source attribution: Appendix I. Variables and Data Sources, _wp1190 - Appendix I. Variables and Data Sources.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1190.pdf_