## WP/17/52 — The Drivers of Capital Flows in Emerging Markets Post Global Financial Crisis (Sections 1–3)

## Source details

**Canonical URL:** [WP/17/52 — The Drivers of Capital Flows in Emerging Markets Post Global Financial Crisis (Sections 1–3)](https://www.imf.org/-/media/files/publications/wp/2017/wp1752.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1752.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1752.pdf.json)

---

### Abstract / Key summary
- Sample: 34 emerging markets and developing economies (EMDEs).
- Period: 2009Q3-2015Q4.
- Method: Panel framework with country fixed effects; Driscoll-Kraay standard errors.
- Coverage: Net and gross capital flows across instruments — private, FDI, portfolio, portfolio debt, portfolio equity, and other investment flows.
- Episode analysis: High and low episodes defined as quarters with flows one standard deviation above/below mean (computed over 2009Q3-2015Q4).
- Main high-level conclusion: The capital flow slowdown since the GFC is due to a combination of lower growth prospects of recipient countries and worse global risk sentiment; determinants vary substantially across instruments and between net and gross flows; sensitivities change during high and low flow episodes.

### Motivation and conceptual framing
- Context:
  - Post-GFC: recovery and subsequent volatility; FDI dominates total flows but portfolio and other investment flows increased.
  - Need to analyze both net and gross flows and disaggregate by instrument.
- Push vs. pull factors:
  - Push: global risk aversion (log), commodity prices (growth), global liquidity (growth of G7 M2), U.S. corporate spread, U.S. yield gap (10-year minus 3-year yields).
  - Pull: growth differential vis-à-vis the U.S., rate differential vis-à-vis the U.S., trade openness (share of GDP), reserves (share of GDP), exchange rate regime, institutional quality, income per capita, capital openness, financial development.
- Importance of gross vs. net decomposition: foreign and domestic investor responses can offset or amplify net flows.

### Empirical strategy / Model setup
- Dependent variable: y_i,t = ratio of flows (total or instrument) to country i nominal GDP.
- Regressors: country fixed effects; push and pull factors listed above.
- Estimation:
  - Panel of 34 EMDEs.
  - Quarterly data 2009Q3-2015Q4.
  - Driscoll-Kraay standard errors to control serial correlation and cross-section dependence.
- High/low episodes:
  - High episodes: flows ≥ one standard deviation above mean (mean and standard deviation computed over 2009Q3-2015Q4).
  - Low episodes: flows ≤ one standard deviation below mean.

### Key regression coefficients and statistics (selected, 2009Q3-2015Q4)
- Sample and estimation:
  - Observations: 809 (most regressions); Number of groups: 34; Country Fixed Effects: YES; Period: 2009Q3-2015Q4.
- Table 2 (Net Capital Flows, Share of GDP) — selected coefficients (standard errors in parentheses)
  - Growth differential: 0.25*** (0.08) for Total; 0.38*** (0.11) for Private; 0.13* (0.07) for Other.
  - Interest rate differential: 0.51*** (0.13) for Total; 0.40*** (0.11) for Other; 0.12** (0.06) for Portfolio Debt.
  - Trade openness: -0.07** (0.03) for Total; 0.03*** (0.01) for FDI; -0.05*** (0.02) for Portfolio.
  - Reserves: 0.06*** (0.02) for Total; 0.05*** (0.01) for Portfolio.
  - Institutional Quality: -0.77 (2.60) for Total; 1.70*** (0.42) for Portfolio Equity.
  - Income per capita: 0.00** (0.00) for Total.
  - Capital account openness: -3.37 (2.58) for Total; -1.67** (0.75) for FDI.
  - Financial Development: -18.70 (18.83) for Total; 11.43** (4.65) for FDI; -11.85*** (4.10) for Portfolio Equity; -33.55** (14.99) for Other.
  - Global risk aversion (log): -2.38 (1.65) for Total; -4.81*** (1.65) for Private; -1.70*** (0.52) for Portfolio Debt.
  - Commodity prices (growth): -0.01 (0.03) for Total; 0.04** (0.02) for Private; 0.02** (0.01) for Portfolio Debt; 0.02** (0.01) for Portfolio Equity.
  - U.S. Corporate Spread: 2.17** (0.80) for Total; 3.18*** (1.11) for Private; 0.37* (0.21) for FDI; 0.95** (0.42) for Portfolio.
  - U.S. Yield Gap: 1.01 (1.86) for Total; 0.31** (0.14) for FDI.
- Table 3 (Capital Inflows, Share of GDP) — selected coefficients
  - Growth differential: 0.21* (0.12) for Total; 0.32** (0.12) for Private.
  - Interest rate differential: 0.31* (0.16) for Total.
  - Trade openness: -0.01 (0.05) for Total; 0.06* (0.03) for FDI; -0.03** (0.02) for Portfolio Debt.
  - Reserves: 0.06** (0.02) for Total; 0.07** (0.03) for Private.
  - Institutional Quality: -8.85*** (2.68) for Total; -6.90* (3.37) for Private.
  - Global risk aversion (log): -0.56 (2.72) for Total; 1.96** (0.81) for FDI; -2.41** (1.17) for Portfolio.
  - Commodity prices (growth): -0.04 (0.04) for Total; -0.02** (0.01) for FDI.
- Table 4 (Capital Outflows, Share of GDP) — selected coefficients
  - Growth differential: -0.04 (0.08) for Total.
  - Interest rate differential: -0.21 (0.14) for Total; -0.28* (0.14) for Private; -0.08** (0.03) for Portfolio.
  - Capital account openness: 3.80** (1.78) for Total; 3.89** (1.86) for Private; 1.39*** (0.46) for Portfolio.
  - Global risk aversion (log): 2.17 (1.64) for Total; 1.92** (0.74) for FDI.
  - Commodity prices (growth): -0.04*** (0.01) for Total; -0.02** (0.01) for FDI.
  - U.S. Corporate Spread: -2.17** (1.01) for Total; -2.28* (1.22) for Private; -1.31** (0.58) for FDI.
  - U.S. Yield Gap: 1.10 (0.82) for Total; 1.75* (0.91) for Private.

### Results — Entire sample (baseline) and interpretation
- General:
  - Both push and pull factors are important determinants of capital flows.
  - Considerable heterogeneity across instruments and across net vs. gross flows.
- Exact magnitude examples preserved from the source:
  - A one percentage point increase in real GDP growth differential vis-à-vis the U.S. would increase net total flows by 0.25 percent of GDP.
  - A one percentage point increase in interest rate differential vis-à-vis the U.S. would increase net total flows by 0.51 percent of GDP.
- Instrument-specific findings:
  - Growth and interest rate differentials are not statistically significant for net FDI flows.
  - Growth and interest rate differentials matter for net portfolio and net other investment flows.
  - Gross total capital inflows and gross FDI inflows show results similar to net flows.
  - Growth differentials are statistically significant for gross portfolio inflows.
  - Interest rate differentials matter for gross portfolio outflows (search for yield by residents).
- Interpretation:
  - Growth differentials attract non-resident portfolio investment.
  - Interest rate differentials can drive resident portfolio outflows and influence other-investment flows on both inflow and outflow margins.

### Economic magnitudes and time paths
- Contribution exercise (coefficients multiplied by average values over 2009Q3-2015Q4) indicates:
  - Positive interest rate differential has been an important contributor to flows into EMDEs.
  - Global risk aversion has been a key factor in reducing capital flows to EMDEs.
  - Contribution from commodity prices is low.
- Growth differential contribution over time (net private flows, 2009Q3-2015Q4):
  - Impact fell from a local peak of around 4 percent of GDP towards the beginning of the period to less than 1 percent of GDP in recent times.
  - Median, maximum, and minimum growth differential contributions also declined over time.

### High and low capital-flow episodes — sensitivity and selected coefficients
- Definition for Section 1/2 analysis:
  - High episodes: quarters where flows ≥ one standard deviation above mean.
  - Low episodes: quarters where flows ≤ one standard deviation below mean.
- General episode findings:
  - Sensitivity of some instruments to push and pull factors increases during high and low episodes.
  - Variables insignificant in baseline can become important during extreme episodes and vice versa.
- Selected numeric episode results:
  - Growth differential:
    - Baseline (net private): 0.38*** (0.11).
    - High episodes (net private): a one percentage point increase in growth differentials vis-à-vis the U.S. increases net private flows by 0.73 percent of GDP during high episodes.
  - Interest rate differential for net portfolio equity:
    - Baseline: 0.12 percent of GDP.
    - High episodes: 0.68 percent of GDP.
    - Low episodes: 0.21 percent of GDP.
  - U.S. yield gap:
    - Not statistically significant in baseline for many flows but can become important during high episodes; example baseline: 1.01 (1.86) for Total; 0.31** (0.14) for FDI.
  - Trade openness:
    - Coefficients higher during high episodes across most instruments (except FDI); more trade-open economies receive higher net flows during high episodes and tend to have negative net flows during low episodes.
  - Financial development:
    - During low episodes, more financially developed economies tend to have higher net portfolio debt flows but less net other investment flows.
  - Global risk aversion:
    - Statistically significant for many instruments during low and high episodes; net portfolio debt coefficient more negative during low episodes.
    - Global risk aversion for gross FDI inflows increases significantly during high episodes.

### Robustness checks
- Alternative specifications:
  - Growth and interest rate differentials vis-à-vis advanced economies (instead of U.S.) — results robust.
  - U.S. shadow rate instead of U.S. policy rate — results robust.
- Alternative high/low episode definitions:
  - Three consecutive quarters where flows are 0.5 standard deviation higher than mean (vice versa for low).
  - Top 30th and bottom 30th percentile of capital flow distribution.
  - Key messages remain broadly robust across these specifications.

### Policy messages
- The determinants of capital flows can be considerably different across instruments and across the type of flows considered, net or gross.
- Both push and pull factors matter for normal times as well as surges and busts.
- Determinants of flows can be different during normal times versus surges and busts; indicators like the U.S. yield gap — not significant during normal times — can be an important driver during high episodes.

### Caveats and limitations
- Results are influenced by focusing on the post-global financial crisis period (2009Q3-2015Q4); some coefficients might reflect this specific period and may not represent longer-term relationships.
- Suggestion for future research: investigate longer time perspectives and deeper analysis of observed puzzles.

### Appendices (selected)
- Appendix I — List of Countries: Albania, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Ecuador, Egypt, El Salvador, Guatemala, Hungary, India, Indonesia, Jordan, Kazakhstan, Latvia, Lithuania, Macedonia, FYR, Malaysia, Mexico, Paraguay, Peru, Philippines, Poland, Russia, Saudi Arabia, South Africa, Sri Lanka, Thailand, Turkey, Ukraine, Uruguay.
- Appendix II — Data sources: Financial Flow Analytics Database (IMF BOP, IFS, WEO), World Bank WDI, Haver Analytics, CEIC; VIX for global risk aversion; G7 M2 growth for global liquidity; Federal Reserve for U.S. corporate spread; Wu-Xia shadow federal funds rate for U.S. shadow rate.
- Appendix III — High/Low episode figures and tables: Gross Capital Inflows and Gross Capital Outflows regressions split into High Episodes and Low Episodes; statistical significance indicated by filled shapes (at least 10 percent level).

*Source: IMF Working Paper WP/17/52, "The Drivers of Capital Flows in Emerging Markets Post Global Financial Crisis" (Sections 1–3).*

### Section 1

### WP/17/52 — The Drivers of Capital Flows in Emerging Markets Post Global Financial Crisis (Section 1)

### Abstract / Key summary
- Sample: 34 emerging markets and developing economies (EMDEs).
- Period: 2009Q3-2015Q4.
- Method: Panel framework with country fixed effects; Driscoll-Kraay standard errors.
- Coverage: Net and gross capital flows across instruments — private, FDI, portfolio, portfolio debt, portfolio equity, and other investment flows.
- Episode analysis: High and low episodes defined as quarters with flows one standard deviation above/below mean (computed over 2009Q3-2015Q4).
- Main high-level conclusion: The capital flow slowdown since the GFC is due to a combination of lower growth prospects of recipient countries and worse global risk sentiment; determinants vary substantially across instruments and between net and gross flows; sensitivities change during high and low flow episodes.

### Introduction / Motivation
- Context:
  - Post-GFC era: capital flows experienced recovery and subsequent volatility; FDI dominates total flows but portfolio and other investment flows increased.
  - Need to analyze both net and gross flows and disaggregate by instrument.
- Extensions relative to prior work:
  - Later sample start (2009Q3) to capture post-crisis dynamics.
  - Broader instrument coverage.
  - Inclusion of both net and gross inflows/outflows.
  - Explicit analysis of high and low capital flow episodes (± one standard deviation from mean).

### Literature review — conceptual framing
- Theoretical basis: portfolio theory — expected returns, risk, and risk preferences.
- Push vs. pull factors:
  - Push: external conditions attracting investors (global risk aversion, global liquidity, commodity prices, U.S. corporate spreads, U.S. yield gap).
  - Pull: domestic characteristics affecting risks/returns (growth differential vis-à-vis U.S., rate differential vis-à-vis U.S., trade openness, reserves, exchange rate regime, institutional quality, income per capita, capital openness, financial development).
- Empirical strands:
  - Extreme episodes literature (sudden stops and surges) — push factors influence occurrence and riskiness; pull factors influence direction and magnitude.
  - Full-sample literature — identifies longer-term determinants and changing relationships over time.
- Importance of gross vs. net decomposition: foreign and domestic investors can respond differently, potentially offsetting or amplifying net flows.

### Empirical strategy / Model setup
- Baseline empirical model:
  - Dependent variable y_i,t = ratio of flows (total or instrument) to country i nominal GDP.
  - Regressors: country fixed effects, external push factors, domestic pull factors.
- Push factors included: global risk aversion (log), commodity prices (growth), global liquidity (growth), U.S. corporate spread, U.S. yield gap (gap between U.S. long- and short-term government bond yields; 10-year and 3-year yields used).
- Pull factors included: growth differential vis-à-vis U.S., rate differential vis-à-vis U.S., trade openness (share of GDP), reserves (share of GDP), exchange rate regime, institutional quality, income per capita, capital openness, financial development.
- Estimation details:
  - Panel of 34 EMDEs (see Appendix I in source).
  - Quarterly data 2009Q3-2015Q4.
  - Driscoll-Kraay standard errors to control serial correlation and cross-section dependence.
- High/low episodes identification:
  - High episodes: quarters where flows ≥ one standard deviation above mean (mean and standard deviation computed over 2009Q3-2015Q4).
  - Low episodes: quarters where flows ≤ one standard deviation below mean.

### Results — Entire sample (baseline)
- General:
  - Both push and pull factors are important determinants of capital flows.
  - Considerable heterogeneity across instruments and across net vs. gross flows.
- Key quantitative findings preserved exactly from the source:
  - A one percentage point increase in real GDP growth differential vis-à-vis the U.S. would increase net total flows by 0.25 percent of GDP.
  - A one percentage point increase in interest rate differential vis-à-vis the U.S. would increase net total flows by 0.51 percent of GDP.
- Instrument-specific highlights:
  - Growth and interest rate differentials are not statistically significant for net FDI flows.
  - Growth and interest rate differentials matter for net portfolio and net other investment flows.
  - Gross total capital inflows and gross FDI inflows show results similar to net flows.
  - Growth differentials are statistically significant for gross portfolio inflows (implying growth differential matters for non-residents’ portfolio investment).
  - Interest rate differentials matter for gross portfolio outflows (implying search for yield influences residents’ decision to invest abroad).
- Interpretation:
  - Growth differentials tend to attract non-resident portfolio investment.
  - Interest rate differentials can drive resident portfolio outflows (search for yield) and influence other-investment flows on both inflow and outflow margins.

### Results — High and low episodes (overview from Section 1)
- Sensitivity changes:
  - Certain types of flows show increased sensitivity to push and pull factors during high and low episodes.
  - Some variables insignificant in baseline can become important during extreme episodes and vice versa.
- Specific result noted in Section 1:
  - The U.S. yield gap (not statistically significant during the baseline case) can become an important driver during high episodes.
- Implication: drivers of capital flows can differ between normal times and surge/bust episodes; indicators that matter in extremes may be muted in baseline analysis.

### Policy messages (as stated in the source)
- Message 1: The determinants of capital flows can be considerably different across instruments and across the type of flows considered, net or gross.
- Message 2: Both push and pull factors matter for normal times as well as surges and busts.
- Message 3: The determinants of flows can be different during normal times versus surges and busts; indicators like the U.S. yield gap — not significant during normal times — can be an important driver during high episodes.

### Caveats and limitations (from source)
- Time-frame limitation:
  - Results are influenced by focusing on the post-global financial crisis period (2009Q3-2015Q4).
  - Some coefficients might reflect this specific period and may not be representative of longer-term relationships.
- Suggestion for future research:
  - Investigate longer time perspectives and deeper analysis of observed puzzles.

_Italic source: IMF Working Paper WP/17/52, "The Drivers of Capital Flows in Emerging Markets Post Global Financial Crisis" (Section 1)._

### Section 2

### wp1752 - Section 2

### Summary of main findings: pull, structural, and push factors
- Pull factors
  - Interest rate differential is statistically significant for other investment flows, both gross inflows and gross outflows.
  - The positive interest rate differential, particularly on the back of ultra-low interest rates in the U.S., has helped EMDEs to attract flows. This effect appears in net as well as gross inflows and outflows, particularly for other investment flows.
- Structural factors
  - Trade openness: a one percentage point increase in trade openness increases net FDI flows by 0.03 percent of GDP.
    - The response of net FDI flows to trade openness is mainly due to the increase in gross FDI inflows (statistically significant) rather than gross FDI outflows (statistically insignificant).
  - Financial development matters for net FDI and net portfolio debt (see table coefficients).
  - Capital account openness results are caveated: statistically significant for some instruments for gross capital outflows, not significant for gross inflows, and significant for net FDI and net portfolio equity flows but with a negative sign. Results are influenced by the period considered and the short-sample nature of slow-moving variables.
  - Many structural factors lack statistical significance for high and low episodes, indicating episodes are not necessarily driven by structural factors.
- Push factors
  - Global risk aversion is statistically significant for net private flows and net portfolio debt flows. While not statistically significant for net flows, both gross inflows and gross outflows of FDI are sensitive to global risk aversion—more global risk aversion leads to higher inflows and higher outflows.
  - Global commodity prices are statistically significant for several instruments (see tables).
  - U.S. corporate spread and U.S. yield gap affect various components; their influence is more pronounced during high and low episodes (see section on episodes).

### Key regression coefficients and statistics (selected, 2009Q3-2015Q4)
- Sample and estimation
  - Observations: 809 (most regressions); Number of groups: 34; Country Fixed Effects: YES; Period: 2009Q3-2015Q4.
- Table 2 (Net Capital Flows, Share of GDP) — selected coefficients (standard errors in parentheses)
  - Growth differential: 0.25*** (0.08) for Total; 0.38*** (0.11) for Private; 0.13* (0.07) for Other.
  - Interest rate differential: 0.51*** (0.13) for Total; 0.40*** (0.11) for Other; 0.12** (0.06) for Portfolio Debt.
  - Trade openness: -0.07** (0.03) for Total; 0.03*** (0.01) for FDI; -0.05*** (0.02) for Portfolio.
  - Reserves: 0.06*** (0.02) for Total; 0.05*** (0.01) for Portfolio.
  - Institutional Quality: -0.77 (2.60) for Total; 1.70*** (0.42) for Portfolio Equity.
  - Income per capita: 0.00** (0.00) for Total.
  - Capital account openness: -3.37 (2.58) for Total; -1.67** (0.75) for FDI.
  - Financial Development: -18.70 (18.83) for Total; 11.43** (4.65) for FDI; -11.85*** (4.10) for Portfolio Equity; -33.55** (14.99) for Other.
  - Global risk aversion (log): -2.38 (1.65) for Total; -4.81*** (1.65) for Private; -1.70*** (0.52) for Portfolio Debt.
  - Commodity prices (growth): -0.01 (0.03) for Total; 0.04** (0.02) for Private; 0.02** (0.01) for Portfolio Debt; 0.02** (0.01) for Portfolio Equity.
  - U.S. Corporate Spread: 2.17** (0.80) for Total; 3.18*** (1.11) for Private; 0.37* (0.21) for FDI; 0.95** (0.42) for Portfolio.
  - U.S. Yield Gap: 1.01 (1.86) for Total; 0.31** (0.14) for FDI.
- Table 3 (Capital Inflows, Share of GDP) — selected coefficients
  - Growth differential: 0.21* (0.12) for Total; 0.32** (0.12) for Private.
  - Interest rate differential: 0.31* (0.16) for Total.
  - Trade openness: -0.01 (0.05) for Total; 0.06* (0.03) for FDI; -0.03** (0.02) for Portfolio Debt.
  - Reserves: 0.06** (0.02) for Total; 0.07** (0.03) for Private.
  - Institutional Quality: -8.85*** (2.68) for Total; -6.90* (3.37) for Private.
  - Global risk aversion (log): -0.56 (2.72) for Total; 1.96** (0.81) for FDI; -2.41** (1.17) for Portfolio.
  - Commodity prices (growth): -0.04 (0.04) for Total; -0.02** (0.01) for FDI.
- Table 4 (Capital Outflows, Share of GDP) — selected coefficients
  - Growth differential: -0.04 (0.08) for Total.
  - Interest rate differential: -0.21 (0.14) for Total; -0.28* (0.14) for Private; -0.08** (0.03) for Portfolio.
  - Capital account openness: 3.80** (1.78) for Total; 3.89** (1.86) for Private; 1.39*** (0.46) for Portfolio.
  - Global risk aversion (log): 2.17 (1.64) for Total; 1.92** (0.74) for FDI.
  - Commodity prices (growth): -0.04*** (0.01) for Total; -0.02** (0.01) for FDI.
  - U.S. Corporate Spread: -2.17** (1.01) for Total; -2.28* (1.22) for Private; -1.31** (0.58) for FDI.
  - U.S. Yield Gap: 1.10 (0.82) for Total; 1.75* (0.91) for Private.

### Economic magnitude and time paths
- Contribution exercise (coefficients multiplied by average values over 2009Q3-2015Q4) indicates:
  - Positive interest rate differential has been an important contributor to flows into EMDEs.
  - Global risk aversion has been a key factor in reducing capital flows to EMDEs.
  - Contribution from commodity prices is low (in line with IMF (2016a) finding cited in the source).
- Growth differential contribution over time (net private flows, 2009Q3-2015Q4)
  - The impact of growth differential fell from a local peak of around 4 percent of GDP towards the beginning of the period to less than 1 percent of GDP in recent times.
  - Median, maximum, and minimum growth differential contributions also declined over time.
  - These results align with the view that much of the decline in inflows can be explained by narrowing growth differentials between EMDEs and advanced economies.

### High and low capital-flow episodes (selected results)
- Definition and charts: high/low episode regressions show coefficients for selected variables when dependent variable is net flows (Appendix III contains gross inflow/outflow analyses).
- Growth differential
  - Sensitivity increases during high episodes.
  - A one percentage point increase in growth differentials vis-à-vis the U.S. increases net private flows by 0.73 percent of GDP during high episodes, compared to 0.38 percent of GDP in the baseline.
  - The effect during high episodes mainly reflects gross capital inflows rather than outflows.
  - Coefficients are not significant for high/low episodes for net portfolio or net other investment flows; coefficient is significant for low episodes for net FDI.
- Interest rate differential
  - Sensitivity increases during both high and low episodes for net portfolio equity:
    - Baseline: 0.12 percent of GDP.
    - High episodes: 0.68 percent of GDP.
    - Low episodes: 0.21 percent of GDP.
  - This higher sensitivity is mainly due to gross portfolio equity inflows during high episodes.
  - Sensitivity of net other investment flows towards interest rate differentials decreases during high and low episodes.
- Structural factors in episodes
  - Trade openness coefficients are higher during high episodes across most instruments (except FDI): more trade-open economies receive higher net flows during high episodes and tend to have negative net flows during low episodes (significant for total, private, and FDI flows).
  - Financial development: during low episodes, more financially developed economies tend to have higher net portfolio debt flows but less net other investment flows.
  - Reserves: significant for net total flows in baseline, not significant during high episodes.
- Push factors in episodes
  - Global risk aversion coefficients are statistically significant for many instruments during low and high episodes, but magnitude is often lower or similar to baseline; exception: net portfolio debt where coefficient is more negative during low episodes.
  - Global risk aversion for gross FDI inflows increases significantly during high episodes (investors’ preference for FDI when risk aversion increases).
  - U.S. yield gap: coefficients are more negative for net flows during high episodes; not statistically significant in baseline except for net FDI. Impact predominantly due to U.S. yield gap effects on gross capital inflows (gross outflows also contribute).
  - U.S. corporate spread and U.S. yield gap drive increased sensitivity of some instruments during high/low episodes more than global risk aversion in some cases.

### Robustness checks
- Alternative specifications tested and found to be robust:
  - Growth and interest rate differentials vis-à-vis advanced economies (instead of the U.S.).
  - U.S. shadow rate instead of the U.S. policy rate (shadow rate not bounded below by zero).
- Alternative definitions of high/low episodes:
  - High episode defined as three consecutive quarters where flows are 0.5 standard deviation higher than mean (vice versa for low).
  - High/low episodes defined as top 30th and bottom 30th percentile of capital flow distribution.
  - Key messages remain broadly robust across these specifications.

### Conclusion (selected)
- Both push and pull factors are important drivers for capital flows.
- The analysis covers net and gross terms across instruments: private, FDI, portfolio, portfolio debt, portfolio equity, and other investment flows for the post-crisis period 2009Q3-2015Q4.
- Findings highlight heterogeneity across instruments, time-varying sensitivities during high/low episodes, and the limited role of some structural factors in explaining episode dynamics.

*Source: wp1752 - Section 2 (2009Q3-2015Q4).*

### Section 3

### wp1752 - Section 3

### Main findings and interpretation
- There is considerable variation across the type of instruments and the type of flows (net or gross).
- The sensitivity of some instruments towards push and pull factors increases during periods of high and low capital flows.
- Some variables may not necessarily be significant during normal times, but can be important drivers during high and low episodes, and vice versa.
- Overall, the results imply that the capital flows slowdown witnessed in recent years is due to a combination of low growth prospects of recipient countries and worse global risk sentiment.

### References (selection as presented in the source)
- Ahmed, Shagil, and Andrei Zlate, 2014, “Capital flows to emerging market economies: A brave new world?” Journal of International Money and Finance, Vol. 48, pp. 221–248.
- Ahmed, Swarnali, 2015, “If the Fed Acts, How Do You React? The Liftoff Effect on Capital Flows,” IMF Working Paper No. 15/256 (Washington: International Monetary Fund).
- Broner, Fernando, Tatiana Didier, Aitor Erce, and Sergio Schmukler, 2013, “Gross capital flows: Dynamics and crises,” Journal of Monetary Economics, Vol. 60(1), pp. 113–133.
- Calvo, Guillermo, 1998, “Capital flows and capital-market crises: the simple economics of sudden stops,” Journal of Applied Economics, Vol. 1(1), pp. 35–54.
- Forbes, Kristin J., and Frank E. Warnock, 2012, “Capital flow waves: surges, stops, flight, and retrenchment,” Journal of International Economics, Vol. 88(2), pp. 235–251.
- Ghosh, Atish R., Mahvash S. Qureshi, Jun II Kim, and Juan Zalduendo, 2014, “Surges,” Journal of International Economics, Vol. 92(2), pp. 266–285.
- IMF, 2011a, “International Capital Flows: Reliable or Fickle?” World Economic Outlook, International Monetary Fund, Chapter 4, April 2011.
- IMF, 2011b, “Recent Experiences in Managing Capital Inflows – Cross-Cutting Themes and Possible Policy Framework,” IMF Board Paper, International Monetary Fund.
- IMF, 2016a, “Understanding the Slowdown in Capital Flows in Emerging Markets,” World Economic Outlook, International Monetary Fund, Chapter 2, April 2016.
- IMF, 2016b, “Capital Flows – Review of Experience with the Institutional View,” IMF Board Paper, International Monetary Fund.
- Koepke, Robin, 2015, “What Drives Capital Flows to Emerging Markets? A Survey of the Empirical Literature,” Institute of International Finance.
- Nier, Erlend, Tahsin S. Sedik, and Tomas Mondino, 2014, “Gross Private Capital Flows to Emerging Markets: Can the Global Financial Cycle Be Tamed?” IMF Working Paper No. 14/196 (Washington: International Monetary Fund).
- Pagliari, Maria S., and Swarnali A. Hannan, 2017 (forthcoming), “The Volatility of Capital Flows in Emerging Markets: Measures and Determinants,” IMF Working Paper (Washington: International Monetary Fund).

### Appendix I – List of Countries
- Albania, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Ecuador, Egypt, El Salvador, Guatemala, Hungary, India, Indonesia, Jordan, Kazakhstan, Latvia, Lithuania, Macedonia, FYR, Malaysia, Mexico, Paraguay, Peru, Philippines, Poland, Russia, Saudi Arabia, South Africa, Sri Lanka, Thailand, Turkey, Ukraine, Uruguay.

### Appendix II – Data Sources (Variables and Sources as listed)
- Capital flow variables: Financial Flow Analytics Database compiled from the IMF’s Balance of Payments Statistics, International Financial Statistics, and World Economic Outlook databases, World Bank’s World Development Indicators database, Haver Analytics, CEIC Asia database, and CEIC China database.
- Real GDP growth, interest rate, trade openness, reserves, income per capita, commodity prices, U.S. Yield Gap: IMF WEO database, IFS, national sources.
- Exchange rate regime: IMF AREAER and Coarse Classification.
- Institutional quality: Rule of law measure from World Bank’s Worldwide Governance Indicators.
- Capital account openness: Chinn and Ito (2006), updated version of the database.
- Financial development: Svirydzenka, K. (2016), “Introducing a New Broad-based Index of Financial Development”, IMF Working Paper No. 16/5.
- Global risk aversion: CBOE Market Volatility Index (VIX), downloaded from Haver Analytics.
- Global liquidity: Growth of G7 M2, IMF and national sources.
- U.S. Corporate Spread: Federal Reserve (FRED).
- U.S. Shadow Rate: Wu-Xia Shadow Federal Funds Rate from Federal Reserve Bank of Atlanta, downloaded from Haver Analytics.

### Appendix III – Results of High/Low Episodes (figures and tables notes)
- Figures AIII.1 and AIII.2 present coefficients of regressions where the dependent variables are Gross Capital Inflows and Gross Capital Outflows, respectively, for the period 2009Q3-2015Q4, split into High Episodes and Low Episodes.
- Footnote for figures and tables: "1/ The y-axis represents coefficients of selected indicators where the underlying regressions explain the relevant flows as a function of all the variables in Table 1. Shapes that are filled are statistically significant at least at the 10 percent level. Unfilled shapes are not statistically significant."
- Tables present regression results for Net Capital Flows (Share of GDP), Capital Inflows (Share of GDP), and Capital Outflows (Share of GDP), across categories: Total, Private, FDI, Portfolio, Portfolio Debt, Portfolio Equity, Other; reported statistics include coefficients, standard errors (in parentheses), significance markers (*** p<0.01, ** p<0.05, * p<0.1), Observations, Number of groups, and Country Fixed Effects indicators.

*Source: wp1752 - Section 3 (wp1752 - Section 3).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1752.pdf_