## 3.1 Instrumenting capital inflows

## Source details

**Canonical URL:** [3.1 Instrumenting capital inflows](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024015-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024015-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024015-print-pdf.pdf.json)

---

### Introduction and motivation
- World gross external liabilities have hovered above 200% of globalgdp for more than a decade, raising vulnerability of recipient countries, especially emerging markets and developing economies (emdes), to capital flow shocks.
- The current wave of monetary tightening may have effects similar to the 2013 Fed tapering episode, which forcefully affected emdes’ capital inflows.
- Conventional view: capital inflows ease financial conditions, boost credit and domestic demand, and support growth; but they also raise exposure to reversals, volatility, boom-bust cycles, and debt sustainability risks.
- Many countries pursue gradual capital account liberalization and use capital controls to limit capital flows.

### Research question and scope
- Primary question: whether capital inflows have a positive, causal impact on the economic performance of recipient economies.
- Focus: emerging markets and developing economies (emdes).
- Time horizons: short and medium term effects (local projections for horizons h ∈ H ≡ {0,...,4}).

### Methodological contribution and identification strategy
- Introduces a Bartik-style shift-share instrumental variable (iv) exploiting heterogeneous bilateral investment positions.
- Reduced-form panel regression (equation (1)):
  - ∆U_i,t = β F→i,t Y_i,t−1 + γ_i + γ_t + ε_i,t
  - Main coefficient of interest: β for each flow type—equity, bond, fdi, and other investment.
  - Estimation by two-stage least square (tsls) with country-clustered standard errors.
- Shift-share instrument construction (equation (3)):
  - Z_i,t ≡ Σ_{j≠i} (P_{j→i,t−1} / P→i,t−1) · g(P_{j→W\i,t}) · p→i,t−1
  - Instrument uses investors’ shares in country i’s position and growth of j’s position in rest of world, excluding contemporaneous P_{j→i,t}, to mitigate endogeneity.
- Dynamic analysis via local projections with instruments (Jordà, 2005):
  - U_{i,t+h} − U_{i,t−1} = β^{(h)} ( ∆P→i,t / P→i,t−1 p→i,t−1 ) + lags + γ^{(h)}_i + γ^{(h)}_t + ε^{(h)}_{i,t}
  - Chosen lag structure: 2 lags (robustness with 1 lag noted).

### Data sources and instrument variants
- Outcome variables U from IMF and World Bank databases; inward positions p→i,t from IMF IIP statistics; inflows F→i,t from IMF BOP data or restated flows by Coppola et al. (2021) (rbep).
- Bilateral positions and IVs by instrument type:
  - Portfolio investments (equity and bonds): IMF CPIS end-year positions (2001–2020), extended using IMF-derived liabilities for non-respondents.
  - Direct investments: IMF CDIS (2009–2020), using derived liabilities from outward positions.
  - Other investments (mostly cross-border loans): two IV versions:
    - BIS Locational Banking Statistics (lbs): bilateral outstanding claims/liabilities since 1977; reporting countries grew from 28 in 2000 to 48 in 2020; estimated to cover more than 90 percent of cross-border banking activity during 2000–2020. lbs-based IV uses BIS-provided break- and exchange-rate-adjusted changes approximating F^{(OI)}_{j→i,t}.
    - BIS Consolidated Banking Statistics (cbs): nationality-based; reporting countries grew from 23 in 2000 to 31 in 2020.
- Restated external portfolios (rbep) by Coppola et al. (2021):
  - rbep provides restated bilateral portfolio investment positions for 2007–2017, correcting tax-haven distortions via the “Fund Holdings” methodology.
  - rbep provides positions but not flows; approximation used: F→i,t ≈ P→i,t − P→i,t−1, yielding model ∆U_i,t = β g(P→i,t) p→i,t−1 + γ_i + γ_t + ε_i,t.
  - Valuation effects η→i,t are approximated and distributed across investors proportional to opening bilateral positions to construct rbep-based instruments.

### Samples and key summary statistics
- Two samples:
  - Reported-flows-based regressions: 2009–2020 (cpis, lbs, cbs).
  - rbep-based regressions: 2007–2017.
- Sample-level means (in percent of GDP):
  - Mean fdi inflows: 3.5 percent of GDP.
  - Mean other investments: 2.6 percent of GDP.
  - Mean bond portfolio investments: 0.9 percent of GDP.
  - Mean equity portfolio investments: 0.09 percent of GDP.
- rbep-based sample summary (selected):
  - Bond inflow (rbep): N = 550, Mean = 0.59, SD = 2.06, p5 = -1.64, p25 = -0.10, p50 = 0.20, p75 = 1.22, p95 = 3.70.
  - Equity inflow (rbep): N = 550, Mean = 0.22, SD = 3.80, p5 = -1.48, p25 = -0.12, p50 = 0.01, p75 = 0.32, p95 = 1.97.
- BOP-based sample summary (selected):
  - Bond inflow (bop): N = 616, Mean = 0.90, SD = 1.87, p5 = -1.25, p25 = -0.06, p50 = 0.38, p75 = 1.82, p95 = 4.36.
  - Other investment inflow (bop): N = 616, Mean = 2.59, SD = 4.78, p5 = -1.87, p25 = 0.24, p50 = 1.71, p75 = 3.59, p95 = 10.01.
  - Real gdp growth (percent): N = 616, Mean = 3.26, SD = 3.96, p5 = -4.96, p25 = 1.73, p50 = 3.87, p75 = 5.64, p95 = 8.52.

### Main empirical findings
- Instrument strength and identification:
  - lbs-based IV for other investment flows is strong throughout the local projection horizon—first-stage robust F-statistics in Table B.4 are all above 10 and above the Montiel Olea and Pflueger (2013) critical value 15 for 20% worst case bias.
  - cbs-based IV strong for h = 0,1,2 but not in h = 3.
  - IVs for portfolio investments and fdi are weak, with F-statistics below 4 in many cases (Tables B.1–B.3).
- Causal effects (reported-flows; lbs-based IV for other investment):
  - One percentage point of GDP of exogenous other investment inflows generates 0.35 and 0.49 percentage point cumulative realgdpgrowth on impact (h = 0) and after one year (h = 1), respectively; coefficient becomes insignificant in h = 2 and h = 3.
  - A standard deviation of other investment inflow in the sample is 4.5 percent of GDP.
  - Other macro effects:
    - Private consumption and investment: significantly and positively impacted in the first three years (investment coefficient at h = 0 significant only at the 10 percent level), then effects dissipate.
    - Real imports increase and real exports decrease in the first year.
    - Other investment inflows also cause inflation throughout the horizon.
  - Credit and credit quality:
    - Credit-to-GDP rises by around 1.0 percentage point of GDP for the first three years and grows further in the fourth year.
    - Nonperforming loan to total loan ratio declines in h = 0,1, then picks up in h ≥ 2 (latter positive effect lacks statistical significance).
- Causal effects (rbep-restated flows):
  - The IV for bond portfolio investment becomes substantially stronger with rbep; robust F-statistics around or above 10 (Table B.6).
  - Estimated causal effect on real GDP gradually increases, reaching 1.05 in h = 2 (significantly different from 0), then dropping to 0.58 in h = 3.
  - Real private consumption estimates positive throughout the horizon; neither real investment nor real private consumption receive a significant effect from a bond portfolio inflow in Figure A.3.
  - Equity portfolio investment causal effect remains unidentified even with rbep—the IV is too weak, potentially due to larger valuation effects in equity.
- Exchange rate and policy transmission:
  - In floating exchange rate regimes (sample covering h = 0..2), exogenous bond inflows cause a lasting appreciation of the exchange rate (0.8 log point in h = 0) while expansionary effects on real GDP remain significant, likely via lower lending interest rates.
- Comparison with prior studies:
  - Contrasts with Blanchard et al. (2017) who find bond inflows contractionary; differences may reflect sample size and country composition, rbep inclusion of longer-term bonds, weaker monetary transmission in some emdes, and policy responses that mute contractionary exchange rate effects.

### Heterogeneity: role of economic fundamentals
- Motivation: absorptive capacity (economic, political, social quality) conditions how inflows translate into growth.
- Interaction regressions use w equal to initial public debt or the risk premium (spread between lending rate and government rate); binary large-w indicator equals one when w is above top 30th percentile.
- Key heterogeneous findings:
  - Countries with higher financing costs benefit more from loan and bond inflows (Table B.8).
    - Interaction coefficients OI Inflow/gdp×large spread show consistently positive and often significant estimates (examples: 0.34∗∗∗, 0.56∗∗∗, 0.73∗∗∗ across horizons).
  - Loan inflows are less favorable for countries with high public debt (Table B.9): interaction OI inflow/gdp×large public debt often negative and significant (examples: -0.66∗∗∗, -1.06∗∗∗, -1.10∗∗∗ across horizons).
  - Interaction with financial development (Sahay et al., 2015 index) is tentative: expansionary impact of other investment inflows is dampened when financial markets are more developed; results weaker and less significant than main findings.

### Robustness and limitations
- Results robust to trimming choices, country selection (e.g., dropping small countries), and adding other capital inflows as lagged controls.
- Limitations:
  - Weak IVs for portfolio equity and fdi prevent causal identification for those flow types in many specifications.
  - Valuation effects complicate reconstruction of flows from positions, notably for equity.
  - rbep provides restated positions but not flows; approximation F→i,t ≈ P→i,t − P→i,t−1 and allocation of valuation effects are used to construct instruments.

### Policy implications and takeaways
- Not all capital flows impact emdes alike; composition matters in addition to aggregate size.
- Other investment (cross-border loans) and bond portfolio investment show positive causal effects on real GDP and, in many specifications, on real investment and consumption.
- Equity portfolio investment causal effects remain unidentified due to weak instruments and valuation issues.
- Country fundamentals matter for absorption:
  - High financing costs: external loans and bonds can be complementary and more beneficial.
  - High public debt: loan inflows are less favorable, suggesting limited absorptive capacity.
  - Financial development: more developed financial systems may better convert portfolio inflows into growth, while loan inflows may be more expansionary where domestic markets are shallow.
- Policy relevance: global monetary tightening in advanced economies may negatively affect emdes through reduced bond and loan inflows; emdes should consider composition of inflows (including debt instruments) alongside efforts to attract FDI.

_Italic: Fueling or following growth? Causal effects of capital inflows on recipient economies — Working Paper No. WP/2024/015 (Chapter 3.1 "Instrumenting capital inflows"; Appendices A–D)_

### 3.1    Instrumenting capital inflows   .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  

### 3.1    Instrumenting capital inflows

### Introduction and motivation
- With the world’s gross external liabilities hovering above 200% of globalgdp for more than a decade, countries, especially emerging markets and developing economies (emdes), are thought to be vulnerable to capital flow shocks.
- The current wave of monetary tightening may have similar consequences to the 2013 Fed tapering episode, which had a forceful impact on emdes’ capital inflows.
- Conventional wisdom: capital inflows ease financial conditions, boost credit and domestic demand, and support growth; but they also increase exposure to reversals, volatility, boom-bust cycles, and debt sustainability risks.
- Authorities in a number of countries have adopted gradual capital account liberalization and employ capital controls to limit capital flows.

### Research question and scope
- Primary question: whether capital inflows have a positive, causal impact on the economic performance of recipient economies.
- Focus: emerging markets and developing economies (emdes).
- Time horizons of interest: short and medium term effects.

### Contributions of the paper
- Methodological contribution: proposes a new empirical methodology to ascertain causality using an instrumental variable (iv) that is a shift-share instrument `a la Bartik (1991) exploiting heterogeneous bilateral investment positions.
- Scope contribution: examines effects for each main type of capital flow—portfolio investment (debt and equity), fdi, and “other investments” (a bop category largely comprising cross-border loans).
- Data contribution: first paper to use capital flow data that filter out the distorting effect of tax heavens, leveraging Coppola et al. (2021) to restate flows by ultimate investors.

### Identification challenges addressed
- Reverse causality: booming economies attract capital; the iv strategy is designed to alleviate reverse causality concerns.
- Confounding global factors: potential confounders associated with global financial conditions are addressed by the shift-share iv and inclusion of time fixed effects in regressions to filter out business and financial cycles.

### Key empirical findings (summary)
- Bond inflows and “other investment inflows” have a significantly positive effect on output and investment, with effects taking a few years to fully materialize.
- Result contrasts with Blanchard et al. (2017), who find bond inflows contractionary; potential reasons for difference include sample size and country composition (Blanchard et al.: 19 emerging markets; this paper: more than 80 countries, including less developed economies).
- Some evidence that absorption capacity depends on fundamentals: public debt, interest rates, and financial development.
- Results are inconclusive with respect to fdi.
- Using restated flows that filter out tax haven distortions is expected to strengthen the shift-share iv and correct the equity/debt breakdown (example statistic: 15 percent of equity flows from the United States into the United Kingdom in 2017 ultimately arrived into third countries, per Coppola et al., 2021).

### Relation to existing literature and methodological choices
- The paper situates itself within literature that documents short-term currency appreciation vs. cheaper financing channels (Mundell-Fleming) and longer-term growth effects via capital accumulation, while acknowledging financial frictions and potential misallocation.
- Empirical literature shows mixed evidence: e.g., Reinhart and Reinhart (2008) find temporary acceleration during capital flow bonanzas; Blanchard et al. (2017) find bond flows have a small negative effect on output; Davis (2015) finds foreign debt increases output gap and other macro variables.
- Comparison to Sanders et al. (2020): both find equity and debt inflows expansionary through different channels; this paper extends their work by covering additional flow types and using restated flows to capture ultimate recipients.
- Comparison to Aldasoro et al. (2023): addresses concerns that shift-share ivs based on lbs are correlated with the global financial cycle by including time fixed effects and using all publicly available lbs data; finds shift-share ivs can remain strong after controlling for fixed effects.
- Methodological distinction from Blanchard et al. (2017): limits the number of ivs to one via heterogeneous investor shares across recipients, reducing weak-iv concerns associated with having as many ivs as countries.

### Implications highlighted in the chapter
- Heterogeneous effects across flow types imply policy relevance for the composition of inflows, not just their aggregate size.
- Findings suggest that country fundamentals (public debt, interest rates, financial development) matter for absorption capacity and therefore for how capital inflows translate into output and investment.

*Source: wpiea2024015-print-pdf (Chapter 3.1 "Instrumenting capital inflows")*

### 3.  Empirical approach

### 3.  Empirical approach

### 3.1.  Instrumenting capital inflows
- Reduced-form panel regression (equation (1)):
  - ∆U_i,t = β F→i,t Y_i,t−1 + γ_i + γ_t + ε_i,t
  - ∆ is the first difference operator; U_i,t denotes the economic variable of interest (e.g., log GDP, investment, export, import, or the exchange rate).
  - F→i,t ≡ Σ_{j∈W\{i}} F_{j→i,t} is the total investment flow into country i, scaled by GDP, Y.
  - γ_i and γ_t are country and time fixed effects.
  - Main coefficients of interest: β for each type of flow—equity, bond, fdi, and other investment.
  - Estimation method: two-stage least square (tsls) with country-clustered standard errors.

- Endogeneity concern and solution:
  - Reverse causality and pull factors imply E[F→i,t ε_i,t] ≠ 0, so OLS is biased.
  - Use a shift-share instrument (Bartik-style) constructed from bilateral investment positions to capture push factors and be exogenous to recipient-country pull factors.
  - Relation between positions and flows:
    - P→i,t ≡ Σ_{j∈W\{i}} P_{j→i,t}; ∆P→i,t = F→i,t + η→i,t where η→i,t captures valuation effects (exchange rate, market price, measurement errors).
    - p→i,t ≡ P→i,t / Y_i,t (inward investment position in percent of GDP).
  - Equivalent expression (equation (2)):
    - ∆U_i,t = β [ ∆P→i,t / P→i,t−1 − η→i,t / P→i,t−1 ] p→i,t−1 + γ_i + γ_t + ε_i,t

- Shift-share instrument construction (equation (3)):
  - Z_i,t ≡ Σ_{j≠i} (P_{j→i,t−1} / P→i,t−1) · g(P_{j→W\i,t}) · p→i,t−1
  - P_{j→W\i,t} ≡ Σ_{ℓ∉{i,j}} P_{j→ℓ,t}
  - Instrument uses shares of each investor j in country i’s position and growth of j’s position in the rest of the world (excluding i).
  - Instrument excludes contemporaneous P_{j→i,t} to mitigate endogeneity from pull factors.

- Relevance and exclusion:
  - Relevance argued from persistence and gravity determinants of international investment allocations (distance, language, culture).
  - Exclusion restriction: investment flows toward third-party countries P_{j→ℓ,t} should not be correlated with ε_i,t conditional on controls.
  - Country and year fixed effects used to capture global cycles and common push factors; sample largely consists of relatively closed economies, reducing risk of residual confounding.

### 3.2.  Dynamic panel regressions
- Objective: analyze medium-term dynamics of capital flow impacts using local projections with instruments (Jordà, 2005).
- Estimation equation for horizons h ∈ H ≡ {0,...,4} (equation (4)):
  - U_{i,t+h} − U_{i,t−1} = β^{(h)} ( ∆P→i,t / P→i,t−1 p→i,t−1 ) + lags + γ^{(h)}_i + γ^{(h)}_t + ε^{(h)}_{i,t}
  - Left-hand side: accumulated change in U between t−1 and t+h.
  - β^{(h)} interpretable as impact h years after a capital inflow.
- Controls and lags:
  - Include lags of both ∆U and the main explanatory variable as common in local projections.
  - Chosen lag structure: 2 lags (best tradeoff between parsimony and fit); robustness check with 1 lag yields similar results.
  - For parsimony, lagged capital flows are not instrumented; lags help neutralize endogeneity risk.
- Instrument exogeneity conditions for included lags:
  - E[Z_{i,t} · g(P→i,t−n) p→i,t−n−1] = 0 and E[Z_{i,t} · ∆U_{i,t−n}] = 0 for all included lags n.

### 3.3.  Data
- Outcome variables U: from IMF and World Bank databases.
- Inward investment positions p→i,t: from IMF International Investment Position (iip) Statistics.
- Inflows F→i,t: from IMF balance of payments (bop) data or restated flows by Coppola et al. (2021).
  - Inflows defined as net sales of domestic financial instruments to foreign residents; positive when liabilities increase, negative when liabilities decrease.

- Bilateral investment positions P_{j→i,t} sources and IV construction:
  - Portfolio investments (equity and bonds):
    - IMF Coordinated Portfolio Investment Survey (cpis): end-year positions from 2001 onward; 67 economies responded for end-2001, 86 for end-2020.
    - Use liability data derived by the IMF based on reported assets to extend coverage to non-participating economies.
  - Direct investments:
    - IMF Coordinated Direct Investment Survey (cdis): surveys outward and inward positions for 61 countries in 2009 up to 85 economies in 2020.
    - Use fdi liability derived from outward positions to maximize coverage and reliability.
  - Other investments (mostly cross-border loans):
    - Two IV versions:
      - BIS Locational Banking Statistics (lbs): bilateral outstanding claims and liabilities since 1977; reporting countries grew from 28 in 2000 to 48 in 2020; estimated to cover more than 90 percent of cross-border banking activity during 2000–2020.
        - BIS provides break- and exchange-rate-adjusted changes approximating F^{(OI)}_{j→i,t}, enabling an IV free of valuation effects (equation (5)):
          - Z^{(OI)}_{i,t} ≡ ( Σ_{j≠i} P^{(OI)}_{j→i,t−1} / P^{(OI)}→i,t−1 · F^{(OI)}_{j→W\i,t} / P^{(OI)}_{j→W\i,t−1} ) p^{(OI)}→i,t−1
      - BIS Consolidated Banking Statistics (cbs): nationality-based measure; reporting countries grew from 23 in 2000 to 31 in 2020; may yield a stronger IV if push factors originate at bank headquarters.

- Restated external portfolios (rbep) by Coppola et al. (2021):
  - rbep provides restated bilateral portfolio investment positions from 2007 to 2017, correcting flaws in official cpis data (notably tax-haven distortions) via the “Fund Holdings” methodology.
  - Reclassifies some iip liabilities that behave like portfolio investments, improving accuracy for large emerging markets.
  - rbep provides restated positions but not restated flows; approximation used: F→i,t ≈ P→i,t − P→i,t−1, implying empirical model becomes (equation (6)):
    - ∆U_i,t = β g(P→i,t) p→i,t−1 + γ_i + γ_t + ε_i,t
  - Valuation effects: η→i,t = ∆P→i,t|_{iip} − F→i,t|_{bop}; these aggregate valuation effects are distributed across investor countries proportional to opening bilateral positions to approximate bilateral flows when injecting valuation effects into rbep-based instruments.

- Practical construction for rbep-instrument:
  - Assume ∀j,i,t,k: η_{j→i,t} ≈ η→i,t · P_{j→i,t} / P→i,t−1, allowing F^{(k)}_{j→i,t} ≈ ∆P_{j→i,t} − η→i,t · P_{j→i,t} / P→i,t−1.
  - Instruments then constructed similarly to other-investment case using adjusted bilateral flows.

### 3.4.  Sample and summary statistics
- Two samples to maximize consistency and size:
  - Reported-flows-based regressions: 2009–2020; encompasses cpis, lbs, and cbs data.
  - rbep-based regressions: 2007–2017.
- Key summary statistics (sample-level observations):
  - fdi and “other investment” dominate capital flows into the EMDEs:
    - Mean fdi inflows: 3.5 percent of GDP.
    - Mean other investments: 2.6 percent of GDP.
    - Mean bond portfolio investments: 0.9 percent of GDP.
    - Mean equity portfolio investments: 0.09 percent of GDP.
  - Standard deviations are substantially larger for fdi and other investments than for portfolio investments.
  - Median sizes are similar in the rbep-based sample.
  - In the rbep-based sample:
    - Standard deviations are slightly smaller for bond inflows and substantially greater for equity inflows, possibly reflecting valuation changes being more prevalent in equity than bond positions.

*Source: wpiea2024015-print-pdf - 3.  Empirical approach*

### 4.  Regression results

### 4.  Regression results

### 4.1.  Regressions with officially reported flows
- Correlations between realgdpgrowth and capital inflows vary across investment instruments.
- OLS regression patterns (first four columns of Tables B.1–B.5):
  - Direct investment inflows: strongly and persistently correlated with growth, with significantly positive coefficients for all horizons, even after controlling for year and country fixed effects, lagged values of the dependent variable and lagged inflows (Table B.3).
  - Bond portfolio inflows and other investment inflows: coefficients insignificant, positive in the short run (h= 0,1) and negative thereafter (Tables B.4 and B.1).
  - Equity portfolio investment: significantly positive coefficient in h= 0 and remains positive (but not statistically significant) for the rest of the horizon (Table B.2).
- Instrument validity and strength:
  - The lbs-based iv for other investment flows appears strong throughout the local projection horizon—first stage robust F-statistics in Table B.4 are all above the rule-of-thumb threshold 10 and also above the critical value for 20% worst case bias in Montiel Olea and Pflueger (2013), which is 15.
  - The cbs-based iv is strong for h= 0,1,2, but not in h= 3.
  - The difference in strength between the lbs and the cbs is likely partly due to lbs data being free of methodological breaks and exchange rate fluctuations.
  - IVs for portfolio investments and fdi are weak, with F-statistics below 4 (Tables B.1–B.3).
    - Possible reasons: valuation effects changing instrument Z_i,t; gravity factors weaker drivers of portfolio investments versus loans; offshore intermediaries/tax havens bundling flows and obscuring investor locational information.
- TSLS estimates and causal findings for other investment inflows (lbs-based instruments; last four columns of Table B.5):
  - One percentage point of gdp of exogenous other investment inflows generates 0.35 and 0.49 percentage point in cumulative realgdpgrowth on impact (h= 0) and after one year (h= 1), respectively.
  - Coefficient becomes insignificant in h= 2 and h= 3.
  - Economic magnitude: a standard deviation of other investment inflow in the sample is 4.5 percent of gdp.
- Other macroeconomic effects (Figure A.2):
  - Private consumption and investment: significantly and positively impacted in the first three years (investment coefficient at h= 0 significant only at the 10 percent level), then effects dissipate.
  - Real imports increase and real exports decrease in the first year—consistent with a jump in domestic demand.
  - Other investment inflows also cause inflation throughout the horizon.
- Credit and credit quality dynamics:
  - Credit-to-gdp is estimated to rise by around 1.0 percentage point of gdp for the first three years and grow further in the fourth year.
  - Estimated impact in h= 0 matches the size of the exogenous inflow, suggesting exogenous foreign loans do not crowd out domestic credit when they flow in.
  - In following years both gdp and credit-to-gdp increase, implying the increase of foreign loans causes additional credit supply.
  - Nonperforming loan to total loan ratio declines in h= 0,1, then picks up in h≥2 (latter positive effect lacks statistical significance), which could suggest exogenous addition of loans driven by push factors is not used effectively and eventually undermines credit quality—ultimately offsetting initial gdpgains and explaining dissipating impacts on realgdp, real investment, and real consumption.
- Comparison to literature:
  - Results resemble findings that growth in low- and middle-income countries rises during capital bonanzas then returns to pre-bonanza levels.
  - Aldasoro et al. (2023) identify positive causal effect of bank lending flows; Blanchard et al. (2017) and Davis (2015) find positive impact on macro variables of portions of capital flows that include “other investments” (using broader aggregates).
- Robustness note:
  - Results are robust to variations in trimming choices, country selection (e.g., dropping small countries), and the addition of other types of capital inflows as lagged control variables.

### 4.2.  Regressions with restated flows
- With therbep data:
  - The iv for bond portfolio investment becomes substantially stronger; robust F-statistics are all around or above 10 (Table B.6).
  - Estimated causal effect on real GDP gradually increases and reaches 1.05 in h= 2, which is significantly different from 0, before dropping to 0.58 in h= 3.
  - For other macro variables (Figure A.3):
    - Neither real investment nor real private consumption receive a significant effect from a bond portfolio inflow.
    - Estimates of real private consumption are positive throughout the local projection horizon.
- Equity portfolio investment:
  - Causal effect cannot be identified even with the restated rbep data—the iv proves too weak, possibly because valuation effects (larger for equity than bonds) invalidate approximations to rebuild capital flows.
- Comparison with prior studies:
  - Similar to Sanders et al. (2020), bond investment has a positive and long-lasting effect on realgdpgrowth.
  - Contrasts with Blanchard et al. (2017), who find bond inflows have small, negative impacts on output due to contractionary exchange rate appreciation effects not being compensated by easier financing conditions.
- Exchange rate and country-subsample findings:
  - Filtering to countries where the exchange rate is allowed to respond (floating exchange rate regime; sample covering horizons h= 0..2) shows a lasting appreciation of the exchange rate as a consequence of exogenous bond inflows (0.8 log point in h= 0); despite this, expansionary effects remain significant, transiting through both real investment and real private consumption, which seem stimulated by lower lending interest rates.
  - Possible explanations for contrast with Blanchard et al. (2017):
    - Sample here is larger and covers less developed countries.
    - rbeptypically includes longer-term bonds; bond investments are not necessarily determined by the monetary policy rate.
    - In countries with weaker monetary policy transmission and shallower financial markets, bond inflows may trigger expansionary improvements in financing conditions.
    - Policy responses (monetary policy, foreign exchange interventions, capital flow management measures) may mute contractionary effects of inflows on the exchange rate.

### 4.3.  Role of economic fundamentals
- Motivating idea:
  - Impact of capital inflows may depend on absorptive capacity—quality of economic, political, and social environment; heterogeneity may explain lack of consensual empirical results.
- Regression specification (equation (7)):
  - U_i,t+h − U_i,t−1 = β^(h)_g(P→i,t) p→i,t−1 + β^(h)_+ w g(P→i,t) p→i,t−1 + lags + γ^(h)_i + γ^(h)_t + ε^(h)_i,t
  - w is either the initial level of public debt or the risk premium (the latter is the spread between the lending rate to the private sector and the government).
- Estimation approach:
  - Regressions run on entire sample and with binary w variables equal to one when w is above the top 30th percentile (sub-sample keeping only the top and low 30 percentiles for w).
  - For ease of interpretation, only latter reported; Figure A.5 plots evolution of β_+ when h varies.
- Findings:
  - Even when differentiating across country fundamentals, instruments remain weak and cannot identify an impact for fdi or portfolio equity inflows.
  - Countries with higher financing costs seem to benefit more than others from loan and bond inflows (Table B.8).
    - External finance appears to more usefully complement domestic markets when domestic financing conditions suffer; otherwise external finance may simply fund less productive projects.
  - Loan inflows are less favorable for countries with high public debt (Table B.9)—such countries have less absorptive capacity and may use external investors to compensate for less public investment rather than complement it.
  - Interaction with financial development:
    - Bottom panel of Figure A.5: the coefficient of interaction between other investment inflows and financial development (summary measure by Sahay et al., 2015) shows expansionary impact is dampened when financial markets are more developed.
    - Tentative results: F-statistics are smaller and estimates less significant than main findings, but indicate that not all capital inflows are more expansionary when financial development is lower; portfolio investments may yield more growth dividends in countries better equipped with financial markets and institutions.

### 5.  Conclusion (key takeaways and policy implications)
- Main empirical findings:
  - Other investment and bond portfolio investment have positive causal effects on macroeconomic variables such as realgdp and real investment.
  - Equity portfolio investment causal effects remain unidentified due to weak instruments/valuation effects.
- Policy implications for emdes:
  - Not all flows impact emdes alike.
  - While emdes often try to attract fdi for stability, they should also consider attracting debt instruments (bonds and loans), which seem associated with more significant growth dividends.
  - Other flows may be more pro-cyclical—coming into countries already growing rather than fueling growth.
  - The economically significant positive effects of an exogenous capital inflow imply that global monetary tightening in advanced economies may negatively affect emdes, especially for bonds and loan inflows.

*Source: 4.  Regression results (wpiea2024015-print-pdf)*

### References

### References

### Bibliographic references
- Addison, T., G. Mavrotas, and M. McGillivray (2005). Development assistance and development finance: evidence and global policy agendas. Journal of International Development: The Journal of the Development Studies Association 17(6), 819–836.
- Aizenman, J., Y. Jinjarak, and D. Park (2011, October). Capital flows and economic growth in the era of financial integration and crisis, 1990-2010. Working Paper 17502, nber.
- Aldasoro, I., P. Beltrán, F. Grinberg, and T. Mancini-Griffoli (2023). The macro-financial effects of international bank lending on emerging markets. Journal of International Economics 142, 103733.
- Baharumshah, A. Z., L. Slesman, and E. S. Devadason (2017). Types of foreign capital inflows and economic growth: New evidence on role of financial markets. Journal of International Development 29(6), 768–789.
- Bartik, T. J. (1991). Who benefits from state and local economic development policies? Technical report, WE Upjohn Institute for Employment Research.
- Becker, T., O. Jeanne, P. Mauro, J. D. Ostry, and R. Ranci`ere (2007). Country insurance: The role of domestic policies. In C. W. Crowe, S. Johnson, J. D. Ostry, and J. Zettelmeyer (Eds.), Macrofinancial Linkages, Chapter 15, pp. 429–472. Washington, dc: imf.
- Bertaut, C. C., B. Bressler, and S. E. Curcuru (2019, 9). Globalization and the geography of capital flows. Fed notes, Fed Board of Governors, Washington, dc.
- Blanchard, O., J. D. Ostry, A. R. Ghosh, and M. Chamon (2015, October). Are capital inflows expansionary or contractionary? theory, policy implications, and some evidence. Working Paper 21619, nber.
- Blanchard, O., J. D. Ostry, A. R. Ghosh, and M. Chamon (2016). Capital flows: expansionary or contractionary? American Economic Review 106(5), 565–69.
- Blanchard, O., J. D. Ostry, A. R. Ghosh, and M. Chamon (2017). Are capital inflows expansionary or contractionary? theory, policy implications, and some evidence. imf Economic Review 65, 563–585.
- Blomström, M., A. Kokko, and J.-L. Mucchielli (2003). The economics of foreign direct investment incentives. In Foreign direct investment in the real and financial sector of industrial countries, pp. 37–60. Springer.
- Borensztein, E., J. De Gregorio, and J.-W. Lee (1998). How does foreign direct investment affect economic growth? Journal of International Economics 45(1), 115—-135.
- Caballero, J. (2012, May). Do Surges in International Capital Inflows Influence the Likelihood of Banking Crises?: Cross-Country Evidence on Bonanzas in Capital Inflows and Bonanza-Boom-Bust Cycles. Working Paper 3971, Inter-American Development Bank, Washington, dc.
- Calderón, C. and M. Kubota (2012). Gross inflows gone wild : Gross capital inflows, credit booms and crises. Policy Research Working Paper 6270, World Bank Group, Washington, dc.
- Cardarelli, R., S. Elekdag, and M. A. Kose (2010). Capital inflows: Macroeconomic implications and policy responses. Economic Systems 34(4), 333–356.
- Carkovic, M. and R. Levine (2005). Does foreign direct investment accelerate economic growth. Does foreign direct investment promote development 195, 220.
- Choe, J. I. (2003). Do foreign direct investment and gross domestic investment promote economic growth? Review of development economics 7(1), 44–57.
- Copelovitch, M. and D. A. Singer (2017). Tipping the (im)balance: Capital inflows, financial market structure, and banking crises. Economics & Politics 29(3), 179–208.
- Coppola, A., M. Maggiori, B. Neiman, and J. Schreger (2021). Redrawing the map of global capital flows: The role of cross-border financing and tax havens. The Quarterly Journal of Economics 136(3), 1499–1556.
- Correia, S. (2018, September). ivreghdfe: Stata module for extended instrumental variable regressions with multiple levels of fixed effects. Technical report, Boston College Department of Economics.
- Davis, J. S. (2015). The macroeconomic effects of debt- and equity-based capital inflows. Journal of Macroeconomics 46, 81–95.
- Fernández, A., M. W. Klein, A. Rebucci, M. Schindler, and M. Uribe (2016). Capital control measures: A new dataset. IMF Economic Review 64, 548–574.
- Fernandez-Arias, E. (1996). The new wave of private capital inflows: push or pull? Journal of development economics 48(2), 389–418.
- Gabaix, X. and R. S. J. Koijen (2020, December). Granular instrumental variables. Working Paper 28204, nber.
- Gopinath, G., Ş. Kalemli-Özcan, L. Karabarbounis, and C. Villegas-Sanchez (2017). Capital allocation and productivity in south europe. The Quarterly Journal of Economics 132(4), 1915–1967.
- Hines, James R., J. (2010, December). Treasure islands. Journal of Economic Perspectives 24(4), 103–26.
- imf (2013). Unconventional monetary policies—recent experience and prospects. Technical report, imf, Washington, dc.
- imf (2018). Review of the institutional view on the liberalization and management of capital flows. Policy Paper 2022/008, imf, Washington, dc.
- imf (2021). Exchange arrangements and exchange restrictions. Annual report, imf, Washington, dc.
- Jordà, Ò. (2005). Estimation and inference of impulse responses by local projections. American economic review 95(1), 161–182.
- Jordà, Ò., M. Schularick, and A. M. Taylor (2020). The effects of quasi-random monetary experiments. Journal of Monetary Economics 112, 22–40.
- Jordà, O. and A. M. Taylor (2016). The time for austerity: Estimating the average treatment effect of fiscal policy. The Economic journal 126(590), 219–255.
- Korinek, A. (2012). Capital flows, crises, and externalities. In F. Allen, M. Aoki, J.-P. Fitoussi, N. Kiyotaki, R. Gordon, and J. E. Stiglitz (Eds.), The Global Macro Economy and Finance, pp. 98–117. London: Palgrave Macmillan.
- Lane, P. R. and P. McQuade (2014). Domestic credit growth and international capital flows. The Scandinavian Journal of Economics 116(1), 218–252.
- Li, X. and X. Liu (2005). Foreign direct investment and economic growth: an increasingly endogenous relationship. World development 33(3), 393–407.
- Lim, J. J., S. Mohapatra, and M. Stocker (2014). Tinker, taper, qe, bye? the effect of quantitative easing on financial flows to developing countries. Policy Research Working Paper 6820, World Bank, Washington, dc.
- Magud, N. E., C. M. Reinhart, and E. R. Vesperoni (2014). Capital inflows, exchange rate flexibility and credit booms. Review of Development Economics 18(3), 415–430.
- Massa, I. (2011). Impact of multilateral development finance institutions on economic growth. Technical report, ODI, London.
- Mendoza, E. G. and M. E. Terrones (2008, May). An anatomy of credit booms: Evidence from macro aggregates and micro data. Working Paper 14049, nber.
- Mendoza, E. G. and M. E. Terrones (2012, September). An anatomy of credit booms and their demise. Working Paper 18379, nber.
- Mercado, R. (2018). Bilateral capital flows: gravity, push, and pull. Trinity Economics Papers 0818, Trinity College Dublin.
- Montiel Olea, J. L. and C. Pflueger (2013). A robust test for weak instruments. Journal of Business & Economic Statistics 31(3), 358–369.
- Obstfeld, M., J. D. Ostry, and M. S. Qureshi (2019). A tie that binds: Revisiting the trilemma in emerging market economies. Review of Economics and Statistics 101(2), 279–293.
- Ostry, J. D., A. R. Ghosh, M. Chamon, and M. S. Qureshi (2011). Capital controls: when and why? IMF Economic Review 59(3), 562–580.
- Portes, R. and H. Rey (2005). The determinants of cross-border equity flows. Journal of international Economics 65(2), 269–296.
- Rajan, R. G. and A. Subramanian (2008). Aid and growth: What does the cross-country evidence really show? The Review of economics and Statistics 90(4), 643–665.
- Reinhart, C. M., V. Reinhart, and C. Trebesch (2016). Global cycles: Capital flows, commodities, and sovereign defaults, 1815–2015. American Economic Review 106(5), 574–580.
- Reinhart, C. M. and V. R. Reinhart (2008). Capital flow bonanzas: An encompassing view of the past and present. nber International Seminar on Macroeconomics 5(1), 9–62.
- Reis, R. (2013). The portuguese slump and crash and the euro crisis. Technical report, National Bureau of Economic Research.
- Rey, H. (2015). Dilemma not trilemma: the global financial cycle and monetary policy independence. Technical report, National Bureau of Economic Research.
- Sahay, R., M. Cihak, P. N’Diaye, A. Barajas, D. Ayala Pena, R. Bi, Y. Gao, A. Kyobe, L. Nguyen, and C. Saborowski (2015). Rethinking financial deepening: Stability and growth in emerging markets. imf staff discussion note 15/08, imf, Washington, dc.
- Sanders, N. J., A. I. Barreca, and M. J. Neidell (2020). Estimating causal effects of particulate matter regulation on mortality. Epidemiology 31(2), 160.
- Stock, J. and M. Yogo (2005). Testing for weak instruments in linear iv regression. In D. W. Andrews (Ed.), Identification and Inference for Econometric Models, pp. 80–108. New York: Cambridge University Press.
- Te Velde, D. W. (2011). The role of development finance institutions in tackling global challenges. Technical report, ODI, London.

### Appendices — contents and notes
- Appendix A. Charts
  - Figure A.1. Gross capital flows in the world (in usd trillion). Source: imf iipdata.
  - Figure A.2. Responses to other investment inflow. Notes: accumulated change after capital flows in (at year 0); shaded areas represent the 90 percent confidence bands of the tsls estimate; country and year fixed-effects controlled; standard errors clustered by countries.
  - Figure A.3. Responses to bond portfolio investment inflows. Same notes as Figure A.2.
  - Figure A.4. Responses to bond portfolio investment inflows under floating regimes. Same notes as Figure A.2.
  - Figure A.5. Heterogeneity of impact of other investment inflows (lbs-based iv): (a) Marginal effect on countries with high corporate spread; (b) Marginal effect on countries with high public debt; (c) Interaction with financial development. Same notes as Figure A.2.
- Appendix B. Local projection tables (tables B.1–B.9)
  - Table B.1. Effect of bond portfolio inflow on real GDP over time. Key reported coefficients and statistics include:
    - Debt inflow/gdp: 0.047 0.049 -0.12 -0.23 -11.1 -13.3 -18.7 -15.8 (columns (1)–(8))
    - L.Debt inflow/gdp: -0.072 -0.21∗∗ -0.31∗∗ -0.35∗∗ 1.25 1.31 1.74 1.55
    - L2.Debt inflow/gdp: -0.17∗∗∗ -0.33∗∗∗ -0.47∗∗∗ -0.50∗∗ 0.29 0.023 -0.090 -0.47
    - L.gdpgrowth: 0.25∗∗∗ 0.32∗∗∗ 0.31∗∗∗ 0.31∗ 0.29 0.20 0.0099 0.032
    - Method: OLS (cols (1)–(4)), IV (cols (5)–(8)). Horizon h = 0,1,2,3. N = 612 550 496 449 612 550 496 449. 1st stage robust F stat = 0.1 0.2 0.1 0.3.
  - Table B.2. Effect of equity portfolio inflow on real GDP over time. Key entries:
    - Equity inflow/gdp: 0.25∗ 0.34 0.18 0.17 1.09 1.20 0.45 0.27
    - L.Equity inflow/gdp: 0.32∗ 0.55 0.84 1.05 0.24 0.48 0.82 1.04
    - L2.Equity inflow/gdp: -0.27 -0.31 -0.41 -0.44 -0.24 -0.25 -0.40 -0.44
    - L.gdpgrowth: 0.24∗∗∗ 0.29∗∗∗ 0.27∗∗ 0.25 0.28∗∗∗ 0.27∗∗ 0.25
    - 1st stage robust F stat = 2.3 4.8 5.4 4.0.
  - Table B.3. Effect of fdi inflow on real GDP over time. Key entries:
    - FDI inflow/gdp: 0.038 0.053 0.040 0.083 0.14 -0.14 -0.33 0.064
    - L.FDI inflow/gdp: -0.0041 -0.0088 -0.0036 0.053 0.0018 0.073 0.240 0.26∗∗
    - L2.FDI inflow/gdp: 0.019 0.0051 -0.051 -0.29∗∗∗ 0.10 -0.018 -0.070 -0.21∗∗
    - 1st stage robust F stat = 1.2 1.2 0.3 0.7.
  - Tables B.4–B.7. Effects of other investment, portfolio bond, and portfolio equity inflows on real GDP over time (various IV implementations). Selected 1st stage robust F stats reported across tables: 53.95 2.44 0.54 4.0; 26.11 7.41 7.00 0.4; 15.54 7.45 5.09 9.8; 6.7 1.5 1.2 1.1.
  - Table B.8. Other investment inflow effect vs. domestic spread. Key coefficients:
    - OI Inflow/gdp: 0.036 -0.0023 -0.28 -0.69∗∗∗ -1.08∗∗∗ 0.26 -0.93 -1.35 -3.36∗∗∗ -5.20∗∗∗ 0.35 ∗ 0.40 0.21 -0.37 -1.31∗∗∗ (columns (1)–(15))
    - OI Inflow/gdp×large spread: 0.34∗∗∗ 0.56∗∗∗ 0.73∗∗∗ 0.87∗∗∗ 0.97∗∗∗ 1.59∗∗∗ 1.58∗∗ 1.66∗∗∗ 1.08 1.43 0.50∗∗ 0.84∗∗∗ 0.94∗∗∗ 1.21∗∗∗ 1.55∗∗∗
    - 1st stage rob. F stat range reported includes 22.9 20.1 16.2 21.3 19.4 10.1 7.6 10.9 11.8 16.9 15.2 16.1 22.6 17.4.
    - Note: spread variable is the difference between private and sovereign interest rates. Regressions rely on the lbs-based ivs.
  - Table B.9. Other investment inflow effect vs. public debt. Key coefficients:
    - OI inflow/gdp: 0.24∗∗∗ 0.32 0.15 0.18 -0.25 1.27∗ 0.36 -0.34 -1.00 -1.29 0.39∗ 0.60∗ 0.39 0.36 0.19
    - OI inflow/gdp×large public debt: -0.16 -0.25 -0.66∗∗∗ -1.06∗∗∗ -1.10∗∗∗ -0.95∗∗∗ -1.27∗ -1.61∗∗ -1.82∗∗ -2.97∗∗∗ -0.30∗∗∗ -0.47∗∗∗ -0.69∗∗∗ -1.21∗∗∗ -1.77∗∗∗
    - 1st stage rob. F stat range includes 5.2 5.9 6.4 3.1 4.6 4.0 3.4 3.0 2.5 5.2 5.3 4.1 5.4 4.1.
    - Note: regressions rely on the lbs-based ivs.
- Appendix C. Notes on data
  - Dependent and macroeconomic variables:
    - Real gdp and cpi are from the World Economic Outlook database.
    - Lending risk premium, real investment, real consumption, real export, real import, the nominal effective exchange rate, and the npl ratio are from the imf’s international finance statistics (ifs).
    - Financial development index from Sahay et al. (2015).
    - Other financial variables from the World Bank’s world development indicators.
  - Investment flows and positions:
    - From the cbs data, international claims of domestic banks include cross-border claims in any currency plus local claims of foreign affiliates denominated in non-local currencies.
    - rbep restated positions use the “Issuance” methodology for P→i,t; “Fund Holdings” methodology by Coppola et al. (2021) used to derive g(Pj→W\i t) for an iv.
    - Exclusions applied for Australian and Norwegian data: drop recipients where either country represents more than 20 percent of total investment positions at any time; additionally drop observations in 2014 (Norway) and 2017 (Australia) for countries where investment stock share of Norway and Australia, respectively, is greater than 10 percent at any point in time.
    - Data points with negative bilateral investment positions are dropped.
  - Sample selection:
    - Two samples constructed:
      - Reported flows sample: covers 2009–2020 and encompasses the cpis, lbs, and cbs data.
      - rbep-based regressions sample: 2007–2017.
    - Countries included in tax haven list of Coppola et al. (2021) are dropped.
  - Table C.1. Country coverage:
    - bop sample (66 countries): Albania, Argentina, Armenia, Bangladesh, Belarus, Benin, Bolivia, Bosnia and Herzegovina, Botswana, Brazil, Bulgaria, Burkina Faso, Cameroon, Chile, Colombia, Croatia, Cˆote d’Ivoire, Ecuador, Egypt, Georgia, Guinea, Hungary, India, Indonesia, Iraq, Jamaica, Kazakhstan, Kenya, Kuwait, Kyrgyz Republic, Malawi, Malaysia, Mali, Mexico, Moldova, Mongolia, Montenegro, Morocco, Mozambique, Namibia, Nicaragua, Niger, Nigeria, North Macedonia, Pakistan, Peru, Philippines, Poland, Russia, Rwanda, Senegal, Serbia, South Africa, Sri Lanka, Tajikistan, Tanzania, Thailand, Trinidad and Tobago, Tunisia, T ̈urkiye, Uganda, Ukraine, Uruguay, Uzbekistan, Venezuela, and Zambia.
    - rbep sample (82 countries): Albania, Algeria, Angola, Argentina, Armenia, Belarus, Benin, Bolivia, Bosnia and Herzegovina, Botswana, Brazil, Brunei Darussalam, Bulgaria, Burundi, Cambodia, Chile, China, Colombia, Dominican Republic, Ecuador, El Salvador, Eswatini, Ethiopia, Georgia, Guatemala, Guyana, Hungary, India, Indonesia, Iraq, Islamic Republic of Iran, Jamaica, Kazakhstan, Kuwait, Kyrgyz Republic, Lao P.D.R., Libya, Malawi, Malaysia, Mali, Mauritania, Mexico, Moldova, Mongolia, Montenegro, Morocco, Mozambique, Namibia, Nepal, Nicaragua, Nigeria, North Macedonia, Oman, Pakistan, Papua New Guinea, Paraguay, Peru, Poland, Qatar, Republic of Congo, Russia, Saudi Arabia, Senegal, Serbia, South Africa, Sri Lanka, Sudan, Tajikistan, Thailand, The Gambia, Togo, Trinidad and Tobago, Tunisia, T ̈urkiye, Uganda, Ukraine, United Arab Emirates, Uruguay, Uzbekistan, Venezuela, Vietnam, Yemen, Zambia, and Zimbabwe.
- Appendix D. Summary statistics
  - Table D.1. Summary statistics for bop-based sample (in percent of gdp):
    - Bond inflow (bop): N = 616, Mean = 0.90, SD = 1.87, p5 = -1.25, p25 = -0.06, p50 = 0.38, p75 = 1.82, p95 = 4.36.
    - Equity inflow (bop): N = 616, Mean = 0.15, SD = 0.73, p5 = -0.56, p25 = -0.04, p50 = 0.02, p75 = 0.24, p95 = 1.25.
    - fdi inflow (bop): N = 489, Mean = 3.56, SD = 7.78, p5 = 0.08, p25 = 1.27, p50 = 2.40, p75 = 3.96, p95 = 9.84.
    - Other investment inflow (bop): N = 616, Mean = 2.59, SD = 4.78, p5 = -1.87, p25 = 0.24, p50 = 1.71, p75 = 3.59, p95 = 10.01.
  - Table D.2. Summary statistics for macro variables in bop-based sample:
    - Real gdp growth (percent): N = 616, Mean = 3.26, SD = 3.96, p5 = -4.96, p25 = 1.73, p50 = 3.87, p75 = 5.64, p95 = 8.52.
    - Real investment growth (percent): N = 406, Mean = 4.05, SD = 12.54, p5 = -15.64, p25 = -2.23, p50 = 4.53, p75 = 10.67, p95 = 24.18.
    - Real consumption growth (percent): N = 446, Mean = 3.09, SD = 7.16, p5 = -4.97, p25 = 1.21, p50 = 3.66, p75 = 5.69, p95 = 10.07.
    - Real export growth (percent): N = 470, Mean = 3.36, SD = 11.79, p5 = -14.95, p25 = -0.64, p50 = 3.93, p75 = 8.19, p95 = 16.62.
    - Real import growth (percent): N = 470, Mean = 4.40, SD = 12.46, p5 = -14.38, p25 = -0.69, p50 = 4.85, p75 = 10.06, p95 = 21.77.
    - Change in credit/gdp (percentage point): N = 501, Mean = 0.99, SD = 4.71, p5 = -5.61, p25 = -0.80, p50 = 0.77, p75 = 2.68, p95 = 7.74.
    - Exchange rate change (log point): N = 248, Mean = -2.47, SD = 8.72, p5 = -19.52, p25 = -4.82, p50 = -0.83, p75 = 2.23, p95 = 7.05.
    - CPI change (log point): N = 613, Mean = 5.57, SD = 6.21, p5 = -0.02, p25 = 2.32, p50 = 4.21, p75 = 7.25, p95 = 13.49.
    - NPL share change (percentage point): N = 377, Mean = 0.03, SD = 2.38, p5 = -2.72, p25 = -0.80, p50 = -0.00, p75 = 0.68, p95 = 3.06.
    - Risk premium change (percentage point): N = 233, Mean = -0.10, SD = 2.43, p5 = -3.44, p25 = -0.73, p50 = -0.10, p75 = 0.61, p95 = 3.16.
  - Table D.3. Summary statistics for rbep-based sample and joint sample (in percent of gdp):
    - Bond inflow (rbep): N = 550, Mean = 0.59, SD = 2.06, p5 = -1.64, p25 = -0.10, p50 = 0.20, p75 = 1.22, p95 = 3.70.
    - Equity inflow (rbep): N = 550, Mean = 0.22, SD = 3.80, p5 = -1.48, p25 = -0.12, p50 = 0.01, p75 = 0.32, p95 = 1.97.
    - Bond (rbep; joint sample): N = 438, Mean = 0.68, SD = 2.15, p5 = -1.68, p25 = -0.12, p50 = 0.36, p75 = 1.35, p95 = 3.73.
    - Equity (rbep; joint sample): N = 438, Mean = 0.28, SD = 4.17, p5 = -1.49, p25 = -0.13, p50 = 0.00, p75 = 0.33, p95 = 2.26.
    - Bond (bop; joint sample): N = 438, Mean = 1.24, SD = 2.39, p5 = -1.02, p25 = 0.00, p50 = 0.36, p75 = 2.11, p95 = 5.35.
    - Equity (bop; joint sample): N = 438, Mean = 0.18, SD = 0.87, p5 = -0.26, p25 = -0.01, p50 = 0.01, p75 = 0.19, p95 = 1.20.
    - Note: last four rows show rbep- and bop-based summary statistics on the sample restricted to observations available for both.
  - Table D.4. Summary statistics for macro variables in rbep-based sample:
    - Real gdp growth (percent): N = 550, Mean = 3.88, SD = 7.50, p5 = -1.92, p25 = 1.78, p50 = 3.80, p75 = 5.81, p95 = 8.98.
    - Real investment growth (percent): N = 321, Mean = 4.85, SD = 11.25, p5 = -11.49, p25 = -1.70, p50 = 4.95, p75 = 10.18, p95 = 24.85.
    - Real consumption growth (percent): N = 336, Mean = 3.45, SD = 4.37, p5 = -3.19, p25 = 1.76, p50 = 3.66, p75 = 5.49, p95 = 10.44.
    - Real export growth (percent): N = 354, Mean = 4.74, SD = 7.97, p5 = -6.34, p25 = 0.46, p50 = 4.19, p75 = 8.41, p95 = 18.03.
    - Real import growth (percent): N = 354, Mean = 4.90, SD = 9.92, p5 = -10.23, p25 = -0.12, p50 = 4.70, p75 = 9.95, p95 = 20.29.
    - Change in credit/gdp (percentage point): N = 518, Mean = 0.76, SD = 8.35, p5 = -6.11, p25 = -1.01, p50 = 0.85, p75 = 2.84, p95 = 8.91.
    - Exchange rate change (log point): N = 270, Mean = -2.33, SD = 9.03, p5 = -19.52, p25 = -4.72, p50 = -0.95, p75 = 2.33, p95 = 9.10.
    - CPI change (log point): N = 548, Mean = 6.12, SD = 11.00, p5 = -0.26, p25 = 2.01, p50 = 4.00, p75 = 7.12, p95 = 16.64.
    - NPL share change (percentage point): N = 387, Mean = 0.19, SD = 2.30, p5 = -2.74, p25 = -0.61, p50 = -0.01, p75 = 0.61, p95 = 3.64.
    - Risk premium change (percentage point): N = 220, Mean = -0.08, SD = 2.56, p5 = -3.34, p25 = -0.74, p50 = -0.12, p75 = 0.34, p95 = 3.65.

_Italic: Fueling or following growth? Causal effects of capital inflows on recipient economies — Working Paper No. WP/2024/015 (References, Appendices A–D)_

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024015-print-pdf.pdf_