## wp1792 - Section 3 discusses the data used, and section 4 provides descriptive evidence on the

## Source details

**Canonical URL:** [wp1792 - Section 3 discusses the data used, and section 4 provides descriptive evidence on the](https://www.imf.org/-/media/files/publications/wp/2017/wp1792.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1792.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1792.pdf.json)

---

### Literature review — key empirical findings summarized
- GDP growth rate differentials and global risk aversion have typically emerged as the most robust statistically significant determinants of aggregate capital flows to emerging market economies (see e.g. Ahmed and Zlate 2013; Nier, Sedik, and Mondino 2014; Koepke 2015, and IMF 2016a).
- Early transition years: dramatic increase in private capital flows (especially FDI and portfolio equity investment) in the first decade of transition (Lankes and others 1999; Claessens and others 2000).
- Lane and Milesi-Ferretti (2006): large accumulation of net external liabilities with FDI flows prominent in financing external current account imbalances.
- Boom-bust cycle of the New Member States: external/global push factors (high global liquidity, low risk aversion) plus domestic policy failures drove credit booms; responsiveness to global factors varied across recipient countries (Bakker and Gulde 2010; Jevčák and others 2010; Globan 2015).
- Sectoral composition: capital flows into real estate had a greater impact on swings in GDP than other sectors (Mitra 2011).
- Western Balkans literature: increasing (mainly FDI) inflows in boom years and relative stability post-crisis (Murgasova and others 2015); EU an important source of FDI (Ganić 2013); evidence that changes in net capital flows preceded changes in relative unit labor costs (Gabrisch 2015); delayed transition and low FDI limited sectoral diversification in the Western Balkans (Kovtun and others 2014).
- Contribution of this paper: integrates analysis of flows and stock positions, comparative analysis relative to the New Member States, and examines common and country-specific factors in boom years and aftermath — first to apply this framework to the Western Balkans.

### Data — sources, definitions, and variable construction
- Data sources:
  - Quarterly data on capital flows from the Financial Flows Analytics database.
  - Annual data on stocks from the External Wealth of Nations database.
- Sample coverage: the Western Balkans, the New Member States and the EU15 for comparison, over the period 1995–2014.
- Definitions:
  - Capital inflows = net acquisition of domestic assets by nonresidents.
  - Capital outflows = net acquisition of foreign assets by residents, excluding reserve assets.
  - Net capital inflows = capital inflows − capital outflows.
  - Net capital inflows and changes in reserve assets together constitute the financial account balance (IMF BPM definition).
  - Total gross inflows and outflows exclude derivatives flows.
  - Equity flows = FDI + portfolio equity.
  - Debt flows = portfolio debt + other flows.
  - All flows measured as shares of GDP.
- Constructed variables and measurement:
  - Country-specific forecasted growth and interest rate differentials = country rate − simple average of EU14 (or NMS or WB) rates.
  - Real interest rates = policy rates deflated using one-year ahead WEO inflation forecasts.
  - Institutional quality = World Governance Indicators rule of law measure.
  - Capital account openness = Chinn and Ito (2006) index.
  - Large IMF-supported adjustment program = growing IMF borrowing above 100 percent of the respective country quota.
  - Fixed and floating exchange rates = AREAER classification.
  - Regional growth differentials = simple averages of Western Balkans (or NMS) and EU14 growth rates.
  - Global risk aversion = logarithm of the VXO.
  - Change in oil price = year-on-year change in West Texas Intermediate oil price.
- Note: EU14 sometimes excludes Luxembourg as a financial centre.

### Descriptive evidence — evolution, composition, and comparative patterns
- Timing and magnitude of inflows:
  - Capital inflows as a share of GDP increased from about 10 percent in 2003 to around 35 percent at their peak in 2007 in both the Western Balkans and Central, Eastern and Southeastern Europe; the Baltics increased from about 15 percent to 40 percent of GDP.
  - Montenegro peaked at 84 percent of GDP in capital inflows in 2007 (outlier).
- Composition of inflows (1995–2014):
  - FDI accounted for around half of all inflows.
  - “Other investment flows” (mostly bank lending) constituted a further 40 percent.
  - Portfolio inflows were small or even negative: portfolio equity (debt) flows averaged around 0.2 (1) percent of GDP in the Western Balkans.
- Gross outflows:
  - Gross capital outflows played a relatively minor role in the Western Balkans and were volatile with no clear patterns.
  - In Central and Eastern Europe and the Baltics, the boom included increasing outflows, consistent with a positive correlation between inflows and outflows.
- External stock positions and integration:
  - Booming inflows translated into accumulation of large net external liabilities in both the Western Balkans and the New Member States; both net equity and net debt positions worsened by 2007.
  - Financial integration (external assets + liabilities as a share of GDP) increased in the Western Balkans but remains below levels in the New Member States.
  - Today (as of the sample) liabilities as percent of GDP are broadly comparable to the New Member States, while assets remain somewhat lower.
  - Equity shares in liabilities are comparable to the New Member States; equity shares in assets remain lower.
  - Reserve shares are much higher in the Western Balkans than in the New Member States.
- Heterogeneity within the region:
  - Montenegro: much higher liabilities and worse net positions.
  - Kosovo: better net debt positions.
- Capital account openness:
  - Capital account openness increased in both regions in late 1990s/early 2000s.
  - Trend continued in the New Member States until the global financial crisis, reaching levels close to EU14.
  - Western Balkans has, on average, remained much more closed.
  - Implication: similarity in inflows despite lower openness suggests factors like relatively cheap skilled labor and expectations of future EU membership were important for investment decisions.

### Determinants of capital flows — empirical strategy
- Two complementary estimation strategies:
  1. Aggregate (region-average) regression of average capital flows to the Western Balkans (and for comparison the New Member States) on common factors:
     - Specification:
       Kflows̅t = γ0 + γ1( g̅t_WB − g̅t_EU14 ) + γ2 ir̅t_EU14 + γ3 riskaversiont + γ6 ΔPt_oil + S_t + u_t
       - (g̅t_WB − g̅t_EU14) is the common growth rate differential between the Western Balkans and the EU14.
       - ir̅t_EU14 is a simple average of real policy rates in the EU14.
       - riskaversiont = log(VXO).
       - ΔPt_oil = percent change in WTI oil price (yoy).
       - S_t = seasonal dummy variables.
  2. Cross-country panel regression (country fixed effects) of country inflows on country-specific factors:
     - Specification:
       Kflows_it = θ0 + θ1(g_it − g̅it_EU14) + θ2 institutional quality_it + θ3 capital controls_it + θ4 IMF loan_it + θ5 Δterms of trade_it + T_t + ε_it
       - (g_it − g̅it_EU14) = country growth differential vs EU14 average.
       - T_t = quarter dummy variables.
     - Inflows and outflows examined separately given distinct roles for systemic risk.

### Main empirical findings — aggregate and country regressions
- Aggregate (average) regressions:
  - Average growth rate differentials between the Western Balkans and the EU14, interest rates in the EU14, and global investor risk appetite are estimated to be statistically significant determinants of average capital inflows (Table 1).
  - Results are qualitatively similar for the New Member States, with:
    - Somewhat larger effects of EU14 interest rates.
    - Somewhat smaller effects of the growth differential.
  - These common factors explain more of the variation in the New Member States than in the Western Balkans.
- Robustness and specification notes:
  - Examining growth differentials relative to the New Member States rather than the EU14 and controlling for interest rates in the New Member States yields similar results.
  - EU14 and New Member States growth and interest rates followed similar trends over the period, making separation of their effects difficult.
  - Baseline specification excludes interest rates in the Western Balkans due to inconsistent long time series; adding them yields very similar results (Annex Tables A.1 and A.2), with high regional interest rates often absorbing effects of increasing global risk aversion.
  - Interest rates are included in levels rather than differentials due to collinearity with growth differentials.

### Regression evidence — key coefficient estimates and statistics
- Aggregate capital inflows (Table 1; sample sizes and statistics):
  - Growth differential (WB-EU14): 2.648***  (0.540)
  - Growth differential (NMS-EU14): 1.755***  (0.632)
  - Growth differential (WB-NMS): 1.152**  (0.564) and 0.571  (0.540)
  - Interest rates (EU14): 2.008***  (0.628); 3.684***  (0.663); 1.770*  (0.998)
  - Interest rates (NMS): 3.406***  (0.880); 1.024  (0.860)
  - Global risk aversion (log): -3.783*  (2.146); -5.785***  (2.117); -4.342  (2.818); -5.150  (3.188)
  - Change in the oil price: -0.0142  (0.0312); 0.0348  (0.0270); -0.00646  (0.0288); -0.0194  (0.0341)
  - Sample: WB, NMS, WB, WB
  - Number of obs.: 63, 63, 63, 63
  - Adjusted R-squared: 0.325, 0.572, 0.192, 0.318
  - Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal dummy variables and a constant are included but not reported.
- Aggregate capital outflows (Table 2):
  - Growth differential (WB-EU14): 0.219  (0.291); 0.0295  (0.354)
  - Growth differential (NMS-EU14): 0.508  (0.350)
  - Growth differential (WB-NMS): 0.453*  (0.258); 0.421  (0.278)
  - Interest rates (EU14): -0.3771  (0.455); 1.347***  (0.501); -0.209  (0.490)
  - Interest rates (NMS): 0.0295  (0.366); 0.141  (0.514)
  - Global risk aversion (log): 0.164  (1.344); -5.792***  (1.293); -1.188  (1.494); -0.962  (1.939)
  - Change in the oil price: 0.00103  (0.0164); 0.00196  (0.0183); -0.00644  (0.0152); -0.00409  (0.0171)
  - Sample: WB, NMS, WB, WB
  - Number of obs.: 63, 63, 63, 63
  - Adjusted R-squared: 0.242, 0.393, 0.264, 0.239
  - Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal dummy variables and a constant are included but not reported.
- Country-specific regressions (Tables 3 and 4):
  - Table 3 — Gross capital inflows (WB vs NMS):
    - Growth differential (country-EU14): 0.618*  (0.213); 0.660  (0.530)
    - Institutional quality: 17.64*  (6.147); 12.46  (7.279)
    - Capital account openness: -0.580  (0.924); 1.031  (0.869)
    - IMF program: 7.010  (3.772); 2.996  (3.922)
    - Change in terms of trade: 0.0925  (0.312); -0.00867  (0.374)
    - Sample: WB, NMS
    - Number of obs.: 196, 560
    - Adjusted R-squared: 0.177, 0.299
  - Table 4 — Gross capital outflows (WB vs NMS):
    - Growth differential (country-EU14): 0.542*  (0.192); -0.279  (0.509)
    - Institutional quality: 1.962*  (0.776); 8.637  (5.072)
    - Capital account openness: 0.0413  (1.084); 1.222  (0.695)
    - IMF program: 0.781  (1.340); -0.0673  (1.774)
    - Change in terms of trade: 0.111  (0.129); -0.260  (0.305)
    - Sample: WB, NMS
    - Number of obs.: 220, 560
    - Adjusted R-squared: -0.021, 0.098
  - Note for country tables: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal and quarter dummy variables and a constant are included but not reported.

### Cross-country heterogeneity and mechanisms
- Economies with more open capital accounts are more exposed to the common trend in capital inflows: a higher share of the total variance of capital inflows is explained by the common time effect in countries with more open capital accounts (controlling for growth performance).
- Larger, less financially developed, less liquid countries are less exposed to common trends.
- More flexible exchange rates reduce the share of variance in capital inflows explained by common factors by allowing immediate currency depreciations that make domestic assets cheaper and attract capital.
- Countries with higher reserves and lower public debt tend to have a lower percentage of fluctuations in capital inflows attributable to common factors.
- Country-by-country regressions show considerable heterogeneity in the association between growth and capital inflows; the effect is largest in Montenegro, and particularly large in Serbia and Bosnia and Herzegovina.

### Broader patterns and geographic distribution (summary)
- Financial integration in the Western Balkans increased rapidly over the past 25 years, taking off later than in the NMS but reaching comparable capital inflows as a share of GDP.
- FDI and bank lending account for the bulk of inflows and outflows in both regions; outflows remain small in the Western Balkans.
- Geographic proximity matters: Western European countries are dominant external investors in the region.
  - Most portfolio investment in the Western Balkans comes from the EU-15: 60–80 percent of total portfolio investment on average, similar to the ratio for the NMS.
  - Investment from the NMS remained below 5 percent even at its peak, directed mostly to Croatia and Montenegro.
  - Intra-regional investment is negligible in comparison; within the NMS it is mostly below 5 percent, with exceptions (Latvia peak at 10 percent; Slovakia peak at over 20 percent).
- Bilateral patterns and correlation with migration documented for Albania, Bosnia and Herzegovina, Croatia, and Macedonia (migration stocks and portfolio investment shares detailed in the source tables).

### Policy implications and interpretation
- Capital inflows are pro-cyclical and strongly linked to growth differentials; this linkage can amplify external shocks and raise vulnerabilities.
- Exchange rate flexibility can reduce sensitivity of capital inflows to global factors by allowing currency adjustment that attracts capital.
- Higher reserves and lower public debt can make capital inflows more resilient to shifts in global factors.
- Better governance and stronger competition policy are associated with smaller drops in capital flows.
- The composition of inflows matters for resilience: differences in the composition of flows (FDI vs bank flows vs portfolio) can alter sensitivity to global cycles.

### Annex findings — robustness and model variations
- Annex Figure 1: Chinn-Ito openness index and financial integration (Percent of GDP) series for WB, NMS, EU14 (1995–2014).
- Annex Table A.1 — role of global factors in explaining average gross capital inflows (selected estimates):
  - Number of obs.: 63 across reported specifications.
  - Adjusted R-squared reported: 0.425, 0.562, 0.319, 0.424.
  - Selected coefficients:
    - Growth differential (WB-EU14): 2.138***  (0.625)
    - Growth differential (NMS-EU14): 1.765***  (0.650)
    - Interest rates (EU14): 3.697***  (1.071) and variants up to 4.695***  (0.955)/(1.076)
    - Interest rates (WB): -1.764*** and -2.428***  (0.645)/(0.648) where reported
    - Global risk aversion (log): -5.799***  (1.678) in at least one specification
- Annex Table A.2 — role of global factors in explaining average gross capital outflows (selected estimates):
  - Number of obs.: 63 across reported specifications.
  - Adjusted R-squared reported: 0.323, 0.381, 0.346, 0.340.
  - Selected coefficients:
    - Interest rates (EU14): 1.237***  (0.625) in one specification
    - Interest rates (WB): -1.068*** and -1.007**  (0.399)/(0.424) where reported
    - Global risk aversion (log): -5.915***  (1.277) in one specification

*IMF Working Paper (source: sections III–V of the supplied content).*

### Section 3 discusses the data used, and section 4 provides descriptive evidence on the

### wp1792 - Section 3 discusses the data used, and section 4 provides descriptive evidence on the

### Literature review — key empirical findings summarized
- GDP growth rate differentials and global risk aversion have typically emerged as the most robust statistically significant determinants of aggregate capital flows to emerging market economies (see e.g. Ahmed and Zlate 2013; Nier, Sedik, and Mondino 2014; Koepke 2015, and IMF 2016a).
- Early transition years: dramatic increase in private capital flows (especially FDI and portfolio equity investment) in the first decade of transition (Lankes and others 1999; Claessens and others 2000).
- Lane and Milesi-Ferretti (2006): large accumulation of net external liabilities with FDI flows prominent in financing external current account imbalances.
- Boom-bust cycle of the New Member States: external/global push factors (high global liquidity, low risk aversion) plus domestic policy failures drove credit booms; responsiveness to global factors varied across recipient countries (Bakker and Gulde 2010; Jevčák and others 2010; Globan 2015).
- Sectoral composition: capital flows into real estate had a greater impact on swings in GDP than other sectors (Mitra 2011).
- Western Balkans literature: increasing (mainly FDI) inflows in boom years and relative stability post-crisis (Murgasova and others 2015); EU an important source of FDI (Ganić 2013); evidence that changes in net capital flows preceded changes in relative unit labor costs (Gabrisch 2015); delayed transition and low FDI limited sectoral diversification in the Western Balkans (Kovtun and others 2014).
- Contribution of this paper: integrates analysis of flows and stock positions, comparative analysis relative to the New Member States, and examines common and country-specific factors in boom years and aftermath — first to apply this framework to the Western Balkans.

### Data — sources, definitions, and variable construction
- Data sources:
  - Quarterly data on capital flows from the Financial Flows Analytics database.
  - Annual data on stocks from the External Wealth of Nations database.
- Sample coverage: the Western Balkans, the New Member States and the EU15 for comparison, over the period 1995–2014.
- Definitions:
  - Capital inflows = net acquisition of domestic assets by nonresidents.
  - Capital outflows = net acquisition of foreign assets by residents, excluding reserve assets.
  - Net capital inflows = capital inflows − capital outflows.
  - Net capital inflows and changes in reserve assets together constitute the financial account balance (IMF BPM definition).
  - Total gross inflows and outflows exclude derivatives flows.
  - Equity flows = FDI + portfolio equity.
  - Debt flows = portfolio debt + other flows.
  - All flows measured as shares of GDP.
- Constructed variables and measurement:
  - Country-specific forecasted growth and interest rate differentials = country rate − simple average of EU14 (or NMS or WB) rates.
  - Real interest rates = policy rates deflated using one-year ahead WEO inflation forecasts.
  - Institutional quality = World Governance Indicators rule of law measure.
  - Capital account openness = Chinn and Ito (2006) index.
  - Large IMF-supported adjustment program = growing IMF borrowing above 100 percent of the respective country quota.
  - Fixed and floating exchange rates = AREAER classification.
  - Regional growth differentials = simple averages of Western Balkans (or NMS) and EU14 growth rates.
  - Global risk aversion = logarithm of the VXO.
  - Change in oil price = year-on-year change in West Texas Intermediate oil price.
- Note: EU14 sometimes excludes Luxembourg as a financial centre.

### Descriptive evidence — evolution, composition, and comparative patterns
- Timing and magnitude of inflows:
  - Capital inflows as a share of GDP increased from about 10 percent in 2003 to around 35 percent at their peak in 2007 in both the Western Balkans and Central, Eastern and Southeastern Europe; the Baltics increased from about 15 percent to 40 percent of GDP.
  - Montenegro peaked at 84 percent of GDP in capital inflows in 2007 (outlier).
- Composition of inflows (1995–2014):
  - FDI accounted for around half of all inflows.
  - “Other investment flows” (mostly bank lending) constituted a further 40 percent.
  - Portfolio inflows were small or even negative: portfolio equity (debt) flows averaged around 0.2 (1) percent of GDP in the Western Balkans.
- Gross outflows:
  - Gross capital outflows played a relatively minor role in the Western Balkans and were volatile with no clear patterns.
  - In Central and Eastern Europe and the Baltics, the boom included increasing outflows, consistent with a positive correlation between inflows and outflows.
- External stock positions and integration:
  - Booming inflows translated into accumulation of large net external liabilities in both the Western Balkans and the New Member States; both net equity and net debt positions worsened by 2007.
  - Financial integration (external assets + liabilities as a share of GDP) increased in the Western Balkans but remains below levels in the New Member States.
  - Today (as of the sample) liabilities as percent of GDP are broadly comparable to the New Member States, while assets remain somewhat lower.
  - Equity shares in liabilities are comparable to the New Member States; equity shares in assets remain lower.
  - Reserve shares are much higher in the Western Balkans than in the New Member States.
- Heterogeneity within the region:
  - Montenegro: much higher liabilities and worse net positions.
  - Kosovo: better net debt positions.
- Capital account openness:
  - Capital account openness increased in both regions in late 1990s/early 2000s.
  - Trend continued in the New Member States until the global financial crisis, reaching levels close to EU14.
  - Western Balkans has, on average, remained much more closed.
  - Implication: similarity in inflows despite lower openness suggests factors like relatively cheap skilled labor and expectations of future EU membership were important for investment decisions.

### Determinants of capital flows — empirical strategy and main results
- Two complementary estimation strategies:
  1. Aggregate (region-average) regression of average capital flows to the Western Balkans (and for comparison the New Member States) on common factors:
     - Specification (as in source):
       Kflows̅t = γ0 + γ1( g̅t_WB − g̅t_EU14 ) + γ2 ir̅t_EU14 + γ3 riskaversiont + γ6 ΔPt_oil + S_t + u_t
       - (g̅t_WB − g̅t_EU14) is the common growth rate differential between the Western Balkans and the EU14.
       - ir̅t_EU14 is a simple average of real policy rates in the EU14.
       - riskaversiont = log(VXO).
       - ΔPt_oil = percent change in WTI oil price (yoy).
       - S_t = seasonal dummy variables.
  2. Cross-country panel regression (country fixed effects) of country inflows on country-specific factors:
     - Specification (as in source):
       Kflows_it = θ0 + θ1(g_it − g̅it_EU14) + θ2 institutional quality_it + θ3 capital controls_it + θ4 IMF loan_it + θ5 Δterms of trade_it + T_t + ε_it
       - (g_it − g̅it_EU14) = country growth differential vs EU14 average.
       - T_t = quarter dummy variables.
     - Inflows and outflows examined separately given distinct roles for systemic risk.
- Main empirical findings:
  - Average growth rate differentials between the Western Balkans and the EU14, interest rates in the EU14, and global investor risk appetite are estimated to be statistically significant determinants of average capital inflows (Table 1 in source).
  - Results are qualitatively similar for the New Member States, with:
    - Somewhat larger effects of EU14 interest rates.
    - Somewhat smaller effects of the growth differential (though separation empirically difficult).
  - These common factors explain more of the variation in the New Member States than in the Western Balkans.
  - Robustness notes:
    - Examining growth differentials relative to the New Member States rather than the EU14 and controlling for interest rates in the New Member States yields similar results.
    - EU14 and New Member States growth and interest rates followed similar trends over the period, making separation of their effects difficult.
    - Baseline specification excludes interest rates in the Western Balkans due to inconsistent long time series; adding them yields very similar results (Annex Tables A.1 and A.2), with high regional interest rates often absorbing effects of increasing global risk aversion.
    - Interest rates are included in levels rather than differentials due to collinearity with growth differentials.

### Summary of substantive findings
- Western Balkans caught up rapidly with New Member States in capital inflows during 2003–07 despite later start and generally lower capital account openness.
- Capital inflows to the Western Balkans were primarily FDI and bank lending; FDI ≈ 50 percent of inflows, other investment ≈ 40 percent (1995–2014).
- Portfolio inflows were small: portfolio equity ≈ 0.2 percent of GDP, portfolio debt ≈ 1 percent of GDP (Western Balkans averages).
- Peak aggregate capital inflows ≈ 35 percent of GDP in 2007 for Western Balkans; Montenegro peak = 84 percent of GDP in 2007.
- Net external liabilities increased for the region; equity shares in liabilities rose and are high relative to development level, while equity shares in assets remain lower than in New Member States.
- Key drivers of average inflows: regional growth differential vs EU14, EU14 interest rates, and global investor risk appetite (log VXO).

*IMF Working Paper (source: sections III–V of the supplied content).*

### 1. Average real GDP growth rates

### 1. Average real GDP growth rates

### Key empirical findings on capital inflows
- Panels 1 and 3 (Figure 8) show a tight empirical link between actual and predicted capital inflows for the Western Balkans and the New Member States (NMS).
- Panels 2 and 4 suggest the decline in inflows is strongly associated with the shrinking real GDP growth differential relative to the EU14; diminished growth prospects counterbalance the effect of decreasing risk aversion, which would predict an increase in capital inflows.
- Predictions somewhat underestimate the boom in both the Western Balkans and the NMS, pointing to some overoptimism beyond what was warranted by growth and interest rate differentials.
- On average, predictions match the slowdown in capital inflows since 2007 in the Western Balkans quite well, but tend to overpredict capital inflows to the NMS.
- Continued FDI inflows to the Western Balkans could include finalizations of projects launched before the global financial crisis and new greenfield investments attracted by lower factor costs and, in some cases, lower exchange rates.

### Regression evidence — capital inflows (Table 1)
- Growth differential (WB-EU14): 2.648***  (0.540)
- Growth differential (NMS-EU14): 1.755***  (0.632)
- Growth differential (WB-NMS): 1.152**  (0.564) and 0.571  (0.540)
- Interest rates (EU14): 2.008***  (0.628); 3.684***  (0.663); 1.770*  (0.998)
- Interest rates (NMS): 3.406***  (0.880); 1.024  (0.860)
- Global risk aversion (log): -3.783*  (2.146); -5.785***  (2.117); -4.342  (2.818); -5.150  (3.188)
- Change in the oil price: -0.0142  (0.0312); 0.0348  (0.0270); -0.00646  (0.0288); -0.0194  (0.0341)
- Sample: WB, NMS, WB, WB
- Number of obs.: 63, 63, 63, 63
- Adjusted R-squared: 0.325, 0.572, 0.192, 0.318
- Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal dummy variables and a constant are included but not reported.

### Regression evidence — capital outflows (Table 2)
- Growth differential (WB-EU14): 0.219  (0.291); 0.0295  (0.354)
- Growth differential (NMS-EU14): 0.508  (0.350)
- Growth differential (WB-NMS): 0.453*  (0.258); 0.421  (0.278)
- Interest rates (EU14): -0.3771  (0.455); 1.347***  (0.501); -0.209  (0.490)
- Interest rates (NMS): 0.0295  (0.366); 0.141  (0.514)
- Global risk aversion (log): 0.164  (1.344); -5.792***  (1.293); -1.188  (1.494); -0.962  (1.939)
- Change in the oil price: 0.00103  (0.0164); 0.00196  (0.0183); -0.00644  (0.0152); -0.00409  (0.0171)
- Sample: WB, NMS, WB, WB
- Number of obs.: 63, 63, 63, 63
- Adjusted R-squared: 0.242, 0.393, 0.264, 0.239
- Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal dummy variables and a constant are included but not reported.

### Country-specific determinants (Tables 3 and 4)
- Table 3 — Gross capital inflows (WB vs NMS)
  - Growth differential (country-EU14): 0.618*  (0.213); 0.660  (0.530)
  - Institutional quality: 17.64*  (6.147); 12.46  (7.279)
  - Capital account openness: -0.580  (0.924); 1.031  (0.869)
  - IMF program: 7.010  (3.772); 2.996  (3.922)
  - Change in terms of trade: 0.0925  (0.312); -0.00867  (0.374)
  - Sample: WB, NMS
  - Number of obs.: 196, 560
  - Adjusted R-squared: 0.177, 0.299
  - Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal and quarter dummy variables and a constant are included but not reported.

- Table 4 — Gross capital outflows (WB vs NMS)
  - Growth differential (country-EU14): 0.542*  (0.192); -0.279  (0.509)
  - Institutional quality: 1.962*  (0.776); 8.637  (5.072)
  - Capital account openness: 0.0413  (1.084); 1.222  (0.695)
  - IMF program: 0.781  (1.340); -0.0673  (1.774)
  - Change in terms of trade: 0.111  (0.129); -0.260  (0.305)
  - Sample: WB, NMS
  - Number of obs.: 220, 560
  - Adjusted R-squared: -0.021, 0.098
  - Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal and quarter dummy variables and a constant are included but not reported.

### Cross-country heterogeneity and mechanism (Figures 10 and 11)
- Economies with more open capital accounts are more exposed to the common trend in capital inflows: a higher share of the total variance of capital inflows is explained by the common time effect in countries with more open capital accounts (controlling for growth performance).
- Larger, less financially developed, less liquid countries are less exposed to common trends.
- More flexible exchange rates reduce the share of variance in capital inflows explained by common factors by allowing immediate currency depreciations that make domestic assets cheaper and attract capital.
- Countries with higher reserves and lower public debt tend to have a lower percentage of fluctuations in capital inflows attributable to common factors.
- Country-by-country regressions (Figure 11) show considerable heterogeneity in the association between growth and capital inflows; the effect is largest in Montenegro, and particularly large in Serbia and Bosnia and Herzegovina.

### Broader patterns and geographic distribution (Box 1 summary)
- Financial integration in the Western Balkans increased rapidly over the past 25 years, taking off later than in the NMS but reaching comparable capital inflows as a share of GDP.
- FDI and bank lending account for the bulk of inflows and outflows in both regions; outflows remain small in the Western Balkans.
- Geographic proximity matters: Western European countries are dominant external investors in the region.
  - Most portfolio investment in the Western Balkans comes from the EU-15: 60–80 percent of total portfolio investment on average, similar to the ratio for the NMS.
  - Investment from the NMS remained below 5 percent even at its peak, directed mostly to Croatia and Montenegro.
  - Intra-regional investment is negligible in comparison; within the NMS it is mostly below 5 percent, with exceptions (Latvia peak at 10 percent; Slovakia peak at over 20 percent).
- Within EU15, Austria, Germany and Greece play prominent roles; France, the UK and Italy play lesser roles (mostly in Albania).

### Policy implications and interpretation
- Capital inflows are pro-cyclical and strongly linked to growth differentials; this linkage can amplify external shocks and raise vulnerabilities.
- Exchange rate flexibility can reduce sensitivity of capital inflows to global factors by allowing currency adjustment that attracts capital.
- Higher reserves and lower public debt can make capital inflows more resilient to shifts in global factors.
- Better governance and stronger competition policy are associated with smaller drops in capital flows.
- The composition of inflows matters for resilience: differences in the composition of flows (FDI vs bank flows vs portfolio) can alter sensitivity to global cycles.

*Source: wp1792 - 1. Average real GDP growth rates (PDF chapter/section).*

### 2. Share of portfolio inflows from selected countries

### 2. Share of portfolio inflows from selected countries

### Source composition of portfolio inflows (Figure 1.1 and Panel averages)
- Data source: Coordinated Portfolio Investment Survey.
- Note: Data not available for the Netherlands on the sending side, and Kosovo on the receiving side. Panel 2 refers to averages over 2001-2012.
- Regions and categories shown in figure:
  - WB inflows from the EU15
  - NMS inflows from the EU15
  - WB inflows from the NMS
  - NMS inflows from the NMS
- Time coverage in figure: 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012.
- Percent of total inflows (scale shown 0–90 in the figure).

### Bilateral patterns and correlation with migration
- Observed bilateral investment patterns:
  - Austria and Germany are especially significant investors in the New Member States (NMS).
  - Denmark and Finland and Sweden are prominent investors in the Baltics.
  - Greece is important for Bulgaria and Romania.
- Correlation with migration:
  - Bilateral investment flows correlate strongly with migrant destinations, with exceptions:
    - Luxembourg (as a financial centre) is an exception.
    - Ireland and the UK are exceptions because their labor markets have remained relatively closed to the Western Balkans.
  - Possible drivers noted: geographic (linguistic) factors, historical ties, or investment by migrants and their children in their home countries.

### Table 1.1: Selected bilateral migration stocks and portfolio inflows (2000 migration stock; 2001-2012 average portfolio investment as share of total inward portfolio investment)
- Sources: Coordinated Portfolio Investment Survey; World Bank, Global Bilateral Migration database; World Bank, World Development Indicators.
- Notes: Bilateral migration data not available for Kosovo, Montenegro and Serbia; bilateral portfolio investment data not available for Kosovo on the receiving side and the Netherlands on the sending side. Red denotes over 10 percent, yellow 5-10 percent, green 1-5 percent.

- Albania
  - Migration stock as a share of population, 2000 (sending countries listed in columns): Austria 0.1, Belgium 0.1, Denmark 0.0, Finland 0.0, France 1.0, Germany 7.2, Greece 13.6, Ireland 0.0, Italy 8.9, Luxembourg 0.0, Portugal 0.0, Sweden 0.0, UK 0.0, NMS 0.0, EU15 31.0
  - Portfolio investment as a share of total inward portfolio investment, 2001-2012 average (sending countries): Austria 0.9, Belgium 5.5, Denmark 0.4, Finland 0.0, France 7.6, Germany 2.3, Greece 24.0, Ireland 7.5, Italy 22.6, Luxembourg 10.4, Netherlands 0.0, Portugal 0.0, Spain 0.0, Sweden 3.9, UK 85.0

- Bosnia and Herzegovina
  - Migration stock as a share of population, 2000: Austria 4.1, Belgium 0.1, Denmark 0.8, Finland 0.0, France 1.0, Germany 5.0, Greece 0.0, Ireland 0.0, Italy 0.7, Luxembourg 0.0, Portugal 0.0, Sweden 0.0, UK 0.0, NMS 1.4, EU15 13.2
  - Portfolio investment as a share of total inward portfolio investment, 2001-2012 average: Austria 5.6, Belgium 0.0, Denmark 7.9, Finland 0.0, France 0.5, Germany 2.9, Greece 0.0, Ireland 1.5, Italy 4.8, Luxembourg 30.5, Netherlands 0.0, Portugal 0.0, Spain 1.1, Sweden 5.8, UK 60.7

- Croatia
  - Migration stock as a share of population, 2000: Austria 1.9, Belgium 0.0, Denmark 0.0, Finland 0.0, France 0.2, Germany 5.0, Greece 0.0, Ireland 0.0, Italy 0.6, Luxembourg 0.0, Portugal 0.0, Sweden 0.0, UK 0.0, NMS 0.1, EU15 7.9
  - Portfolio investment as a share of total inward portfolio investment, 2001-2012 average: Austria 15.3, Belgium 0.4, Denmark 1.0, Finland 0.1, France 5.0, Germany 18.4, Greece 0.5, Ireland 2.7, Italy 3.3, Luxembourg 13.3, Netherlands 0.2, Portugal 0.1, Spain 0.7, Sweden 9.9, UK 70.8

- Macedonia
  - Migration stock as a share of population, 2000: Austria 1.0, Belgium 0.1, Denmark 0.1, Finland 0.0, France 0.2, Germany 2.7, Greece 0.0, Ireland 0.0, Italy 2.2, Luxembourg 0.0, Portugal 0.0, Sweden 0.0, UK 0.0, NMS 0.1, EU15 6.5
  - Portfolio investment as a share of total inward portfolio investment, 2001-2012 average: Austria 10.8, Belgium 0.3, Denmark 4.7, Finland 0.0, France 0.4, Germany 14.3, Greece 3.9, Ireland 5.9, Italy 2.2, Luxembourg 18.0, Netherlands 0.0, Portugal 0.0, Spain 0.0, Sweden 4.0, UK 64.6

### Annex: Financial integration, capital account openness, and role of global factors
- Annex Figure 1 (two panels):
  - Panel 1: Chinn-Ito openness index (series span 1995–2013; axis shows -1.5 to 3.0).
  - Panel 2: Financial integration (Percent of GDP) for WB, NMS, EU14 (time series 1995–2014 with vertical scale 0–900 for one chart and -1.5 to 3.0 for another).
  - Sources: External Wealth of Nations database; Chinn and Ito (2006).

- Annex Table A.1. The role of global factors in explaining average gross capital inflows (sample sizes and statistics)
  - Sample: WB and NMS (columns differ by regression specification).
  - Number of obs.: 63 (for each reported specification).
  - Adjusted R-squared reported: 0.425, 0.562, 0.319, 0.424 (across columns).
  - Selected coefficient estimates (standard errors in parentheses; significance denoted as in source):
    - Growth differential (WB-EU14): 2.138*** (0.625)
    - Growth differential (NMS-EU14): 1.765*** (0.650)
    - Growth differential (WB-NMS): 0.885 (0.713) and 0.437 (0.550) in two specifications
    - Interest rates (EU14): 3.577*** (omitted), 3.697*** (1.071), 4.695*** (0.955)/(1.076) as reported across columns
    - Interest rates (NMS): -0.0661 (0.957) where reported
    - Interest rates (WB): -1.764*** (omitted) and -2.428*** (0.645)/(0.648) where reported
    - Global risk aversion (log): -1.083 (2.028); -5.799*** (1.678); -1.363 (2.827); -2.243 (2.622) across specifications
    - Change in the oil price: -0.0109 (0.0303); 0.0341 (0.0282); -0.0168 (0.0296); -0.0164 (0.0266)
  - Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal dummy variables and a constant are included but not reported.

- Annex Table A.2. The role of global factors in explaining average gross capital outflows
  - Sample and size: Number of obs. 63 for reported specifications.
  - Adjusted R-squared reported: 0.323, 0.381, 0.346, 0.340 (across columns).
  - Selected coefficient estimates (standard errors in parentheses; significance denoted as in source):
    - Growth differential (WB-EU14): -0.0956 (0.368); -0.265 (0.317) across two columns
    - Growth differential (NMS-EU14): 0.453 (0.296) where reported
    - Growth differential (WB-NMS): 0.371 (0.266); 0.432* (0.257)
    - Interest rates (EU14): 0.574 (omitted); 1.237*** (0.625); 0.736 (0.416)/(0.478) across columns
    - Interest rates (NMS): 0.242 (0.554) where reported
    - Interest rates (WB): -1.068*** (omitted); -1.007** (0.399)/(0.424) where reported
    - Global risk aversion (log): 1.754 (1.497); -5.915*** (1.277); 0.491 (1.800); 0.609 (1.570) across specifications
    - Change in the oil price: 0.00291 (0.0195); 0.00201 (0.0173); -0.00257 (0.0158); -0.00261 (0.0153)
  - Note: * denotes significant at 10 percent, ** at 5 percent, * at 1 percent. Seasonal dummy variables and a constant are included but not reported.

*Source: IMF Working Paper wp1792 (figures, tables, notes, and annexes as provided in the source content).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1792.pdf_