## Annex I. Composition of Private Capital Flows to Low-Income Countries

## Source details

**Canonical URL:** [Annex I. Composition of Private Capital Flows to Low-Income Countries](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026179-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026179-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026179-source-pdf.pdf.json)

---

### I. Main findings
- Sample and period:
  - panel regressions for a sample of 69 LICs over 2012-2024.
- Private flows:
  - Post-GFC private capital flows to LICs share a statistically significant and sizable global component captured through US dollar movements against other advanced economy currencies.
  - US dollar appreciation has a statistically and economically significant negative impact on net FDI and on net non-bank (trade) financing in LICs.
  - Bank financing may also be affected by US dollar movements, but this estimate is sensitive to model specifications and insignificant for small developing states (SDS).
  - Global economic cycle has a statistically significant positive effect on FDI inflow to LICs but the size of this effect is relatively small.
- Public flows:
  - Post-GFC external government borrowing by LICs has a statistically significant counter-cyclical component, although its size is relatively small.
  - The counter-cyclical effect is largely coming from multilateral financing, which is inversely related to the global economic and financial cycles.
  - Slower global economic growth and US dollar appreciation against advanced economy currencies are both associated with statistically significant increases in multilateral loans.
  - The multilateral response is stronger in frontier LICs and weaker in fragile and conflict-affected states (FCS).
  - Government borrowing from private creditors increases in periods of low non-fuel commodity prices.
- Comparative scale and interpretation:
  - Estimated coefficients on standardized regressors are comparable in magnitude for public and private flows; smaller GDP-scaled effects for public borrowing reflect smaller scale and volatility of public borrowing relative to GDP.
- Policy implications summarized:
  - strengthen capacity to monitor private capital flows and prevent buildup of imbalances;
  - be ready to address economic and financial shocks supported by counter-cyclical public borrowing while ensuring debt sustainability;
  - policy toolkit guided by the IMF’s Institutional View on the Liberalization and Management of Capital Flows and the IMF’s Integrated Policy Framework (IPF) may include: monetary policy, exchange rate policy, fiscal adjustment, structural policies, FX interventions (underpinned by international reserves), macroprudential and capital flow management measures;
  - FX interventions backed by reserves are important to support fixed exchange rate regimes, address disruptions in shallow FX markets, mitigate FX mismatches, and guard against de-anchored inflation expectations;
  - multilateral and official bilateral lending provides important counter-cyclical support to international reserves, especially if it does not induce additional fiscal spending.

### II. Background and stylized statistics
- Definitions and scope:
  - capital flows refer to the financial account of the balance of payments; they do not reflect grants or remittances.
  - analysis uses net capital flows disaggregated by FDI and other investments (banks, non-banks, governments and central banks, excluding IMF financing).
- Stylized patterns (post-GFC):
  - median net FDI inflow to LICs: increased from early 1990s to 2008, dropped in 2009, and stabilized in the range of 2-3.5 percent of GDP since 2010.
  - net “other investments” through banks: median fluctuated between -0.5 and 0 percent of GDP since the GFC.
  - net “other investments” through non-bank sector: median in range -0.5 to +0.5 percent of GDP, largely foreign trade financing.
  - median net government borrowing (other investments: loans and some other operations): declined before the GFC, increased thereafter and stabilized in the range of 1-2 percent of GDP since 2012.
  - portfolio flows to LICs were negligible in most LICs.
- Structural shifts post-GFC:
  - FDI source composition shifted with a post-GFC rise in share of China, India, and Gulf countries.
  - recipient composition shifted from commodity exporters to exporters of more diverse goods and services.
  - more than half of HIPC countries reached HIPC completion point and received debt relief by 2012, increasing scope for aligning government loans with macroeconomic needs.

### III. Methodology and key model features
- Empirical model:
  - standard push-pull panel regression estimated separately for each type of financial flow:
    - Kflow_{i,t} = β1 × PUSH_t + β2 × PULL_{i,t-1} + γ_i + ε_{i,t}
  - PUSH_t includes: VIX (log-transformed prior to standardization), US dollar nominal effective exchange rate (USD NEER) against advanced economies (AEs), Fed policy rate, non-fuel commodity prices, and global (World) GDP growth.
  - PULL_{i,t-1} includes lagged country-specific variables: real GDP growth, fiscal balance, and current account balance.
- Data treatment and estimation choices:
  - all push and pull variables standardized; dependent variables standardized by country-specific mean and standard deviation (percent of GDP).
  - estimated coefficients on standardized regressors scaled by the median standard deviation in the LIC sample to interpret effects in percent of GDP.
  - main sample: 69 LICs for 2012-2024.
  - disaggregation: net FDI and net other investments by banks, non-banks, and governments and central banks; government flows further disaggregated by source (multilateral, official bilateral, private) using World Bank IDS data.
  - estimation steps: (1) full specification including all explanatory variables (clustered robust standard errors); (2) parsimonious specification retaining USD NEER plus statistically significant push and pull factors from step 1; (3) parsimonious specification used to test heterogeneity (interaction terms for SDS, FCS, frontier, other economies) and robustness checks (including lagged capital flows and COVID-19 controls via yearly dummies for 2020 and 2021 or a single 2020-2021 dummy).
- Identification and endogeneity:
  - push factors treated as exogenous to recipient LIC i.
  - use USD NEER against AEs and lagged pull factors to avoid endogeneity.
  - coefficients on PULL may reflect both lagged policy impact and inertia in capital flows.

### IV. Empirical results (high-level)
- Role of USD NEER:
  - USD NEER appreciation has a statistically significant negative impact on net private capital inflows to LICs, particularly net FDI and net non-bank trade financing.
  - estimates for bank financing are sensitive to specification and insignificant for SDS.
- Global cycle and FDI:
  - global economic cycle has a statistically significant positive albeit relatively small effect on FDI inflows to LICs.
- Public borrowing dynamics:
  - multilateral lending increases when global economic growth slows and the USD appreciates (counter-cyclical).
  - multilateral financing response: stronger in frontier LICs, weaker in FCS.
  - private creditors’ lending responds counter-cyclically to non-fuel commodity price declines.
- Robustness:
  - results robust across full and parsimonious specifications; full annex of results in Annex II and Annex III.

### V. Quantified key estimates and elasticities
- USD NEER (one standard deviation = 8.2 percent USD appreciation) effects (selected):
  - a one standard deviation appreciation in USD NEER leads to a decline in net inflows through non-banks by 0.46 percent of GDP for a median LIC.
  - a one standard deviation appreciation in USD NEER leads to about 0.8 percent of GDP reduction in net private inflows for a median LIC (sum of impacts on FDI and on non-banks).
- Panel coefficients (selected from Table 2 and Annex tables; robust standard errors in parentheses):
  - USD NEER coefficients across specifications: -0.18* (0.093); -0.19*** (0.058); -0.12 (0.086); -0.13* (0.073); -0.24** (0.091); -0.15** (0.063).
  - world GDP growth coefficient for FDI: 0.07* (0.042). One standard deviation for world GDP is equivalent to 2.0 percentage points.
  - log(VIX) reported examples: -0.06 (0.050); 0.003 (0.065); 0.10 (0.062).
  - lagged CA balance: -0.15*** (0.053).
  - lagged GDP growth: 0.09** (0.045) and 0.12*** (0.037) in some specifications.
- Annex Table 1 (lagged capital flows controlled):
  - Lagged capital flows (lag): Column (1): 0.36*** (0.054); Column (2): 0.09** (0.040); Column (3): 0.29*** (0.050).
  - USD NEER examples: Column (1): -0.19*** (0.058); Column (2): -0.10* (0.054); Column (3): -0.13* (0.073); Column (6): -0.21*** (0.077).
  - Fed rate examples: Column (2): 0.14* (0.078); Column (4): 0.21** (0.084).
  - World GDP growth examples: Column (1): 0.12*** (0.027); Column (3): 0.12*** (0.039).

### VI. Public borrowing: coefficients and interpretations
- Public borrowing (selected coefficients from Annex Table 2 and Table 3):
  - USD NEER (All sources / Multilateral / Bilateral / Private): 0.11 (0.084) ; 0.13* (0.066) ; 0.12 (0.084) ; 0.15** (0.056) across relevant columns.
  - Fed rate (All sources): -0.19** (0.086); multilateral: -0.22*** (0.058); private: -0.20*** (0.075).
  - world GDP growth (Multilateral): -0.13** (0.051); -0.12*** (0.044) in some columns.
  - non-fuel commodity price index (Private): -0.24** (0.113); -0.17*** (0.059) in some columns.
  - lagged fiscal balance: examples like -0.17*** (0.051).
- Point estimates to interpret:
  - a one percentage point reduction in world GDP growth → decline in net FDI financing of 0.07 percent of GDP for the median LIC.
  - a one percentage point reduction in global growth → increase in net multilateral inflows by 0.05 percent of GDP for the median LIC.
  - a one standard deviation appreciation in USD NEER → increase in net multilateral inflows by 0.09 percent of GDP for the median LIC (parsimonious specification).
  - a one standard deviation reduction in non-fuel commodity prices (16.9 percent deflation) → increases private government borrowing by 0.15 percent of GDP for the median LIC.

### VII. Balance of payments and macro implications
- A one standard deviation appreciation of USD NEER (8.2 percent USD appreciation) leads to about 0.8 percent of GDP reduction in net private inflows for a median LIC, though total balance-of-payments impact may be lower if imports decline linked to reduced FDI.
- Such capital flow shocks likely:
  - put pressure on LICs’ foreign exchange markets;
  - tighten domestic financing conditions;
  - reduce investments amid lower resource availability.
- Estimated shock magnitude is lower than a typical oil price shock but comparable to the effect of other current account shocks.

### VIII. Policy implications and recommendations (detailed)
- Strengthen capacity to monitor private capital flows and prevent buildup of imbalances.
- Use IMF’s Institutional View on Liberalization and Management of Capital Flows and the IMF’s Integrated Policy Framework (IPF) to guide policy.
- Policy toolkit may include:
  - monetary policy;
  - exchange rate adjustment;
  - fiscal policy adjustment;
  - structural policies;
  - FX interventions supported by international reserves (important for fixed exchange rate regimes and shallow FX markets);
  - macroprudential and capital flow management measures.
- FX interventions supported by reserves can:
  - support fixed exchange rate regimes that prevail in LICs;
  - address disruptions from exchange rate movements in shallow FX markets;
  - mitigate financial stability risks from FX mismatches;
  - guard against de-anchored inflation expectations.
- Counter-cyclical external government borrowing (multilateral or official bilateral) can support FX reserves, especially when it replaces domestic financing without increasing the fiscal deficit.
- IMF financing is highlighted as the only external element of the Global Financial Safety Net widely available to most LICs to address balance of payments needs; it should remain catalytic and complemented by other sources.
- Given projected declines and shifts in official development assistance (toward project- and loan-based financing), LICs may need to place additional emphasis on increasing FX reserves.
- Ensure debt sustainability: increases in borrowing should occur when fiscal space allows and be offset through tighter fiscal policy in "good times".

### IX. Additional empirical notes, robustness, and data composition
- FDI composition:
  - net FDI flows predominantly reflect FDI equity liabilities; FDI equity liability flows significantly exceed FDI debt liability flows in most countries; FDI assets remain close to zero.
  - median correlation between FDI asset and liability flows in 2012–2023 was 0.43 for countries with significant FDI asset accumulation (exceeding 0.7 percent of GDP on average per year).
- Bank flows:
  - median correlation between bank asset and liability flows in 2012–2023 was 0.23.
  - bank flow estimates sensitive to specification; USD NEER effect on bank flows becomes statistically insignificant in Arellano–Bond dynamic specification.
- Robustness to COVID-19:
  - main USD NEER and world GDP growth effects on FDI and non-bank flows remain statistically significant under COVID specifications (one-dummy and two-dummy), though significance for bank flows and world GDP growth can vary.
  - Annex Table 3a and 3b report covid20_21, covid20, covid21 coefficient examples.
- Model selection and stationarity:
  - AIC/BIC: for private flows both favor full specification; for public borrowing AIC favors full, BIC favors parsimonious.
  - Fisher-type panel unit root tests and KPSS tests provide reassurance that main variables are stationary in levels for most variables given wide-panel (N > T) structure.

### X. Annex IV — country coverage (sample 2012–2024)
- Country list (all 2012–2024): Afghanistan; Bangladesh; Benin; Bhutan; Burkina Faso; Burundi; Cabo Verde; Cambodia; Cameroon; Central African Republic; Chad; Comoros; Congo, Democratic Republic of the; Congo, Republic of; Côte d'Ivoire; Djibouti; Dominica; Eritrea; Ethiopia; Gambia, The; Ghana; Grenada; Guinea; Guinea-Bissau; Haiti; Honduras; Kenya; Kiribati; Kyrgyz Republic; Lao P.D.R.; Lesotho; Liberia; Madagascar; Malawi; Maldives; Mali; Marshall Islands; Mauritania; Micronesia; Moldova; Mozambique; Myanmar; Nepal; Nicaragua; Niger; Papua New Guinea; Rwanda; Samoa; Senegal; Sierra Leone; Solomon Islands; Somalia; South Sudan; St. Lucia; St. Vincent and the Grenadines; Sudan; São Tomé and Príncipe; Tajikistan; Tanzania; Timor-Leste; Togo; Tonga; Tuvalu; Uganda; Uzbekistan; Vanuatu; Yemen; Zambia; Zimbabwe.

*Source: IMF Working Paper Riding the Global Financial Cycle: How Capital Flows into LICs (Working Paper No. WP/2026/179).*

### Annex I. Composition of Private Capital Flows to Low-Income Countries ................................................. 

### wpiea2026179-source-pdf - Annex I. Composition of Private Capital Flows to Low-Income Countries .................................................

### Annex I
- Annex I. Composition of Private Capital Flows to Low-Income Countries ................................................. 15

### Annex II
- Annex II. Model Estimates under Different Specifications ............................................................................ 16

### Annex III
- Annex III. Other Robustness Checks and Tests ............................................................................................ 19

### Annex IV
- Annex IV. Country List and Sample Periods .................................................................................................. 21

*Source: wpiea2026179-source-pdf - Annex I. Composition of Private Capital Flows to Low-Income Countries .........................................*

### References .............................................................................................................

### Riding the Global Financial Cycle: How Capital Flows into LICs

### I. Main findings
- Sample and period: panel regressions for a sample of 69 LICs over 2012-2024.
- Private flows:
  - Post-GFC private capital flows to LICs share a statistically significant and sizable global component captured through US dollar movements against other advanced economy currencies.
  - US dollar appreciation has a statistically and economically significant negative impact on net FDI and on net non-bank (trade) financing in LICs.
  - Bank financing may also be affected by US dollar movements, but this estimate is sensitive to model specifications and insignificant for small developing states (SDS).
  - Global economic cycle has a statistically significant positive effect on FDI inflow to LICs but the size of this effect is relatively small.
- Public flows:
  - Post-GFC external government borrowing by LICs has a statistically significant counter-cyclical component, although its size is relatively small.
  - The counter-cyclical effect is largely coming from multilateral financing, which is inversely related to the global economic and financial cycles.
  - Slower global economic growth and US dollar appreciation against advanced economy currencies are both associated with statistically significant increases in multilateral loans.
  - The multilateral response is stronger in frontier LICs and weaker in fragile and conflict-affected states (FCS).
  - Government borrowing from private creditors increases in periods of low non-fuel commodity prices (reflecting higher financing needs during negative terms-of-trade shocks).
- Comparative scale and interpretation:
  - Although the effect of global push factors on public borrowing appears modest in percent of GDP, estimated coefficients on the standardized regressors are comparable in magnitude to those for private capital flows; the smaller apparent GDP-scaled effect reflects smaller scale and volatility of public borrowing relative to GDP.
- Policy implications summarized:
  - LICs need to strengthen capacity to monitor private capital flows and prevent buildup of imbalances.
  - Be ready to address economic and financial shocks supported by counter-cyclical public borrowing while ensuring debt sustainability.
  - Policy toolkit guided by the IMF’s Institutional View on the Liberalization and Management of Capital Flows and the IMF’s Integrated Policy Framework (IPF) may include: monetary policy, exchange rate policy, fiscal adjustment, structural policies, FX interventions (underpinned by international reserves), macroprudential and capital flow management measures.
  - FX interventions backed by reserves are important to support fixed exchange rate regimes that prevail in LICs, address disruptions from exchange rate movements in shallow FX markets, mitigate financial stability risks from FX mismatches, and guard against de-anchored inflation expectations.
  - Multilateral and official bilateral lending provides important counter-cyclical support to international reserves, especially if it does not induce additional fiscal spending.

### II. Background and stylized statistics
- Definitions and scope:
  - Capital flows refer to the financial account of the balance of payments; they do not reflect grants or remittances.
  - The analysis uses net capital flows disaggregated by FDI and other investments (banks, non-banks, governments and central banks, excluding IMF financing).
- Stylized patterns (post-GFC):
  - Median net FDI inflow to LICs increased from early 1990s to 2008, dropped in 2009, and stabilized in the range of 2-3.5 percent of GDP since 2010.
  - Net “other investments” flows through banks: median fluctuated between -0.5 and 0 percent of GDP in most years since the GFC.
  - Net “other investments” through the non-bank sector: median in the range from -0.5 to +0.5 percent of GDP and largely reflects foreign trade financing.
  - Median net government borrowing (other investments: loans and some other operations) declined before the GFC but increased thereafter and stabilized in the range of 1-2 percent of GDP since 2012.
  - Portfolio flows to LICs were negligible in most LICs.
- Structural shifts post-GFC:
  - Composition of FDI sources shifted with a post-GFC rise in the share of China, India, and Gulf countries.
  - Recipient composition shifted from commodity exporters to exporters of more diverse goods and services.
  - More than half of heavily indebted poor countries (HIPC) reached the HIPC completion point and received debt relief by 2012, increasing scope for aligning government loans with macroeconomic needs.

### III. Methodology and key model features
- Empirical model:
  - Standard push-pull panel regression estimated separately for each type of financial flow:
    - Kflow_{i,t} = β1 × PUSH_t + β2 × PULL_{i,t-1} + γ_i + ε_{i,t}
  - PUSH_t vector includes: VIX (log-transformed prior to standardization), US dollar nominal effective exchange rate (USD NEER) against advanced economies (AEs), Fed policy rate, non-fuel commodity prices, and global (World) GDP growth.
  - PULL_{i,t-1} includes lagged country-specific variables: real GDP growth, fiscal balance, and current account balance.
- Data treatment and estimation choices:
  - All push and pull variables standardized; dependent variables standardized by country-specific mean and standard deviation (percent of GDP) to limit undue influence from volatile observations.
  - Estimated coefficients on standardized regressors are scaled by the median standard deviation in the LIC sample to interpret effects in percent of GDP.
  - Main sample: 69 LICs for 2012-2024 (post-GFC period).
  - Disaggregation: net FDI and net other investments by banks, non-banks, and governments and central banks; government flows further disaggregated by source (multilateral, official bilateral, private) using World Bank IDS data.
  - Estimation steps:
    1. Full specification including all explanatory variables (clustered robust standard errors).
    2. Parsimonious specification retaining USD NEER plus statistically significant push and pull factors from step 1.
    3. Parsimonious specification used to test heterogeneity (interaction terms for SDS, FCS, frontier, other economies) and robustness checks (including lagged capital flows and COVID-19 controls via yearly dummies for 2020 and 2021 or a single 2020-2021 dummy).
- Identification and endogeneity considerations:
  - Push factors treated as exogenous to recipient LIC i.
  - To avoid endogeneity, use USD NEER against AEs (not against all trading partners or bilateral rates) and lagged pull factors.
  - Coefficients on PULL may reflect both lagged policy impact and inertia in capital flows.

### IV. Empirical results (high-level)
- Role of USD NEER:
  - USD NEER appreciation (capturing reduction in risk-taking and tightening global financial conditions) has a statistically significant negative impact on net private capital inflows to LICs, particularly for net FDI and net non-bank trade financing.
  - Estimates for bank financing are sensitive to specification and insignificant for SDS.
- Global cycle and FDI:
  - Global economic cycle has a statistically significant positive albeit relatively small effect on FDI inflows to LICs.
- Public borrowing dynamics:
  - Multilateral lending increases when global economic growth slows and the USD appreciates (counter-cyclical).
  - Multilateral financing response varies by country type: stronger in frontier LICs, weaker in FCS.
  - Private creditors’ lending responds counter-cyclically to non-fuel commodity price declines (terms-of-trade shocks).
- Robustness:
  - Results are robust across full and parsimonious specifications, with full annex of results in Annex II (detailed coefficients and sensitivity tests).

*Italic: Source: IMF Working Paper content (References and sections provided).*

### 8.2 percent USD appreciation, leads to a net FDI

### 8.2 percent USD appreciation, leads to a net FDI

### Key empirical findings on global drivers of capital flows to LICs
- US dollar appreciation (USD NEER) against other advanced-economy (AE) currencies had a statistically and economically significant negative effect on net private inflows to LICs in 2012–2024, primarily through FDI and non-bank financing.
- A one standard deviation appreciation in USD NEER against other AE currencies, equivalent to 8.2 percent USD appreciation, leads to:
  - a decline in net inflows through non-banks by 0.46 percent of GDP for a median LIC.
  - about 0.8 percent of GDP reduction in net private inflows for a median LIC (sum of impacts on FDI and on non-banks).
- Net FDI inflows to LICs are procyclical with respect to the global economic cycle:
  - a one percentage point reduction in world GDP growth is associated with a decline in net FDI financing of 0.07 percent of GDP for the median LIC.
- The VIX is used as year-average VIX close reading to capture global risk sentiment; log(VIX) estimates are reported but in many specifications are not statistically significant for FDI.
- Robustness: Estimates for USD NEER and world GDP growth effects on FDI and non-bank flows withstand robustness checks discussed in Annex II and Annex III, while estimates for bank flows are sensitive to specification.

### Panel estimates and reported coefficients (selected)
- From Table 2 (net private capital flows to LICs; dependent variable: net capital inflow as a percentage of GDP):
  - USD NEER coefficients (various columns): -0.18* (0.093); -0.19*** (0.058); -0.12 (0.086); -0.13* (0.073); -0.24** (0.091); -0.15** (0.063).
  - world GDP growth coefficient for FDI: 0.07* (0.042). A one standard deviation for world GDP is equivalent to 2.0 percentage points (used to interpret effects).
  - log(VIX) reported: -0.06 (0.050); 0.003 (0.065); 0.10 (0.062) across specifications.
  - lagged CA balance: -0.15*** (0.053) and similar across columns.
  - lagged GDP growth: 0.09** (0.045) and 0.12*** (0.037) in some specifications.
  - Note: Robust standard errors clustered at country level in parentheses. Statistical significance: *** p<0.01, ** p<0.05, * p<0.1. All specifications include country fixed effects; year fixed effects are not included.

### Non-bank (trade) financing
- Non-bank net inflows (mostly trade financing) show a statistically and economically significant negative impact from US dollar appreciation in 2012–2024.
  - A one standard deviation appreciation in USD NEER (8.2 percent USD appreciation) leads to a decline in net non-bank inflows by 0.46 percent of GDP for a median LIC.
- Non-bank asset and liability flows (mostly trade financing) are similar in size and uncorrelated (Annex I).

### Bank financing and heterogeneity
- Estimates for bank financing are sensitive to model specification and do not provide reliable, robust estimates of capital flow drivers.
  - USD NEER is statistically significant in the parsimonious specification but not in the full specification.
  - Fed rate is statistically significant in both specifications but with an unexpected sign; Fed rate and USD NEER become insignificant under robustness checks (Annex II).
- Institutional heterogeneity:
  - For non-SDS, the coefficient on USD NEER is statistically significant; for SDS it is not significant.
  - For non-FCS, coefficient on Fed rate is not statistically significant; for FCS the coefficient is significant and positive.
  - Bank flows show assets dominating liabilities in median cases; median correlation between bank asset and liability flows in 2012–2023 was 0.23 (Annex I).

### Public borrowing by LICs: procyclical and counter-cyclical elements
- Net financing by governments and central banks in the form of other investments has a statistically significant negative relationship with the Fed rate: higher Fed rate → procyclical decline in financing (Table 3, "All sources").
- Public financing exhibits a statistically significant positive relationship with USD NEER in the parsimonious specification: US dollar appreciation → counter-cyclical increase in public financing.
- Multilateral financing:
  - Exhibited a statistically significant counter-cyclical relationship with global economic growth and with USD NEER in 2012–2024 (Table 3, "Multilateral").
  - A one percentage point reduction in global growth leads to an increase in net multilateral inflows by 0.05 percent of GDP for a median LIC (effect weaker for FCS and stronger for frontier LICs).
  - US dollar NEER appreciation by one standard deviation leads to an increase in net multilateral inflows by 0.09 percent of GDP for a median LIC.
  - Note: the USD NEER effect is statistically significant for the parsimonious specification but insignificant for the full specification (footnote 17).
- Bilateral financing:
  - Net inflows from official bilateral donors tend to be negatively associated with Fed rate increases and with VIX (the latter sensitive to specification).
  - Diverging donor trends: net financing from China increased until 2014 and declined thereafter, reaching negative levels in 2023–2024; Paris Club bilateral financing increased between 2015 and 2021 then moderated (text).
- Private government borrowing:
  - Counter-cyclical with respect to non-fuel commodity prices: a one standard deviation reduction in non-fuel commodity prices (equivalent to 16.9 percent non-fuel commodity price deflation) increases borrowing by 0.15 percent of GDP for a median LIC.
- From Table 3 (selected coefficients, net public borrowing as percent of GDP):
  - USD NEER (All sources / Multilateral / Bilateral / Private): 0.11 (0.084) ; 0.13* (0.066) ; 0.12 (0.084) ; 0.15** (0.056) across columns.
  - Fed rate (All sources): -0.19** (0.086); multilateral: -0.22*** (0.058); private: -0.20*** (0.075).
  - world GDP growth (Multilateral): -0.13** (0.051); -0.12*** (0.044) in some columns.
  - non-fuel commodity price index (Private): -0.24** (0.113); -0.17*** (0.059) in some columns.
  - lagged fiscal balance: coefficients like -0.17*** (0.051) across specifications indicate countries with weaker fiscal positions tend to attract more public borrowing.

### Balance of payments and macroeconomic implications
- A one standard deviation appreciation of US dollar NEER (8.2 percent USD appreciation) would lead to about 0.8 percent of GDP reduction in net private inflows for a median LIC, though the total impact on the balance of payments may be lower if imports decline linked to reduced FDI.
- Such capital flow shocks would likely:
  - put pressure on LICs’ foreign exchange markets;
  - tighten domestic financing conditions;
  - reduce investments amid lower resource availability.
- The estimated shock magnitude is lower than the effect of a typical oil price shock but comparable to the effect of other current account shocks (reference to Figure 16 in IMF, 2024b).

### Policy implications and recommendations
- LICs need to strengthen capacity to monitor private capital flows, prevent buildup of imbalances, and be ready to respond to shocks.
- Macroeconomic policy recommendations should be guided by the IMF’s Institutional View on the Liberalization and Management of Capital Flows and the IMF’s Integrated Policy Framework (IPF).
- Policy toolkit may include:
  - monetary policy;
  - exchange rate adjustment;
  - fiscal policy adjustment;
  - structural policies;
  - FX interventions supported by international reserves (important for LICs with fixed exchange rate regimes or shallow FX markets);
  - macroprudential and capital flow management measures.
- FX interventions underpinned by international reserves are likely an important element to support fixed exchange rate regimes, address destabilizing premia in shallow FX markets, mitigate FX mismatches, and anchor inflation expectations.
- Counter-cyclical external government borrowing (especially multilateral or official bilateral) can support FX reserves and address external shocks, particularly when it replaces domestic financing without increasing the fiscal deficit.
- IMF financing is highlighted as the only external element of the Global Financial Safety Net widely available to most LICs to address balance of payments needs, but it should remain catalytic and complemented by other sources.
- Given projected declines and shifts in official development assistance (toward project- and loan-based financing), LICs may need to place additional emphasis on increasing FX reserves.
- Ensure debt sustainability: increases in borrowing should occur when fiscal space allows and be offset through tighter fiscal policy in "good times".

### Additional empirical notes and data composition (Annex I & II)
- In LICs, net FDI flows predominantly reflect FDI equity liabilities; FDI equity liability flows significantly exceed FDI debt liability flows in most countries; FDI assets remain close to zero (Annex I).
- Median correlation between FDI asset and liability flows in 2012–2023 was 0.43 for countries with significant FDI asset accumulation (exceeding 0.7 percent of GDP on average per year) (Annex I, footnote 20).
- Median correlation between bank asset and liability flows in 2012–2023 was 0.23 (Annex I, footnote 21).
- The procyclicality of private financial flows is largely robust to an Arellano–Bond dynamic panel specification including a lagged dependent variable; however, USD NEER effect on bank flows becomes statistically insignificant in that dynamic specification (Annex II).

*Source: IMF Working Paper — Riding the Global Financial Cycle: How Capital Flows into LICs (excerpt provided).*

### Annex Table 1. Estimates for net private flows in LICs, controlling for lagged capital flows

### Annex Table 1. Estimates for net private flows in LICs, controlling for lagged capital flows

### Net private flows — main coefficient estimates
- Dependent variable: net capital inflow as a percentage of GDP. All specifications include country fixed effects; year fixed effects are not included. Robust standard errors clustered at country level in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
- Lagged capital flows (lag):
  - Column (1): 0.36*** (0.054)
  - Column (2): 0.09** (0.040)
  - Column (3): 0.29*** (0.050)
- USD NEER:
  - Column (1): -0.19*** (0.058)
  - Column (2): -0.10* (0.054)
  - Column (3): -0.13* (0.073)
  - Column (4): -0.10 (0.095)
  - Column (5): -0.15** (0.063)
  - Column (6): -0.21*** (0.077)
- Fed rate:
  - Column (2): 0.14* (0.078)
  - Column (4): 0.21** (0.084)
- World GDP growth:
  - Column (1): 0.12*** (0.027)
  - Column (3): 0.12*** (0.039)

Notes on specification:
- Results remain robust when controlling for the lagged variable of capital flows using the Arellano–Bond dynamic panel-data estimation.
- Alternative specification including lagged capital flows retains impact of USD NEER, world GDP growth, Fed policy rate, and non-fuel commodity price index.
- The estimated coefficient on VIX for bilateral flows becomes statistically insignificant under the dynamic specification.
- Most country-specific factors become statistically insignificant in columns (2), (4), (6), and (8) and are suppressed in Annex Table 2.

### Robustness to COVID-19 (private flows)
- Controlling for COVID-19 as one dummy (covid20_21) or two dummies (covid20 and covid21) — main patterns largely robust.
- Annex Table 3a (one COVID dummy):
  - covid20_21:
    - Column (1): -0.24* (0.14)
    - Column (3): -0.22 (0.14)
    - Column (5): -0.01 (0.12)
  - USD NEER:
    - Column (1): -0.19*** (0.058)
    - Column (2): -0.20*** (0.061)
    - Column (3): -0.13* (0.073)
    - Column (4): -0.09 (0.078)
    - Column (5): -0.15** (0.063)
    - Column (6): -0.15** (0.064)
  - Fed rate:
    - Column (2): 0.14* (0.078)
    - Column (4): 0.08 (0.088)
  - World GDP growth:
    - Column (1): 0.12*** (0.027)
    - Column (3): 0.079** (0.033)
- Annex Table 3b (two COVID dummies):
  - covid20:
    - Column (1): -0.31 (0.91)
    - Column (2): -0.40** (0.17)
    - Column (3): -0.02 (0.13)
  - covid21:
    - Column (1): -0.20 (0.47)
    - Column (2): -0.06 (0.17)
    - Column (3): -0.00 (0.15)
  - USD NEER and Fed rate coefficients similar to Annex Table 3a:
    - USD NEER: -0.19*** (0.058); -0.20*** (0.062); -0.13* (0.073); -0.061 (0.079); -0.15** (0.063); -0.15** (0.064)
    - Fed rate: Column (2): 0.14* (0.078); Column (4): 0.064 (0.089)
  - World GDP growth:
    - Column (1): 0.12*** (0.027)
    - Column (3): 0.056 (0.29)

Implication:
- The effect of USD NEER on FDIs and on other sectors remains statistically significant under all COVID specifications but becomes insignificant for banks in some specifications.
- The effect of world GDP growth on FDIs remains statistically significant under the one-dummy specification but becomes statistically insignificant under the two-dummy specification due to increased standard errors.

---

### Annex Table 2. Estimates for net public borrowing in LICs, controlling for lagged capital flows

### Net public borrowing — main coefficient estimates
- Dependent variable: net capital inflow as a percentage of GDP. All specifications include country fixed effects; year fixed effects are not included. Robust standard errors clustered at country level in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
- Lag:
  - Column (1): 0.29*** (0.062)
  - Column (2): 0.32*** (0.053)
  - Column (3): 0.37*** (0.050)
  - Column (4): .15** (0.069)
- log(VIX):
  - Column (3): -0.09** (0.040)
  - Column (4): -0.079 (0.055)
- USD NEER:
  - Column (1): 0.13* (0.066)
  - Column (2): 0.20** (0.079)
  - Column (3): 0.15** (0.056)
  - Column (4): 0.24*** (0.061)
  - Column (5): 0.04 (0.083)
  - Column (6): 0.050 (0.10)
  - Column (7): -0.04 (0.078)
  - Column (8): -0.19* (0.097)
- World GDP growth:
  - Column (2): -0.12*** (0.044)
  - Column (3): -0.10** (0.046)
- Fed rate:
  - Column (1): -0.22*** (0.058)
  - Column (2): -0.12** (0.060)
  - Column (3): -0.22*** (0.058)
  - Column (4): -0.14** (0.055)
- Non-fuel commodity prices:
  - Column (5)/(6)/(7)/(8) where applicable: -0.17*** (0.059); -0.14** (0.069)

Notes:
- The estimated coefficient on VIX for bilateral flows becomes statistically insignificant in alternative model specifications.
- Most country-specific factors become statistically insignificant in columns (2), (4), (6), and (8) and are suppressed in Annex Table 2.

### Robustness to COVID-19 (public borrowing)
- Annex Table 4a (one COVID dummy, covid20_21):
  - covid20_21:
    - Column (1): 0.099 (0.13)
    - Column (2): 0.11 (0.15)
    - Column (3): -0.39** (0.16)
    - Column (4): 0.10 (0.15)
  - log(VIX):
    - Column (3): -0.09** (0.040)
    - Column (4): 0.00 (0.040)
  - USD NEER:
    - Column (1): 0.13* (0.066)
    - Column (3): 0.15** (0.056)
    - Column (4): 0.15** (0.059)
    - Other columns: see main table figures above.
  - World GDP growth:
    - Column (2): -0.12*** (0.044)
    - Column (3): -0.10** (0.049)
  - Fed rate:
    - Column (1): -0.22*** (0.058)
    - Column (2): -0.19** (0.078)
    - Column (3): -0.22*** (0.058)
    - Column (4): -0.31*** (0.079)
  - Non-fuel commodity prices:
    - Column (7)/(8): -0.17*** (0.059); -0.18*** (0.059)
- Annex Table 4b (two COVID dummies, covid20 and covid21):
  - covid20:
    - Column (1): 0.28 (0.17)
    - Column (2): 0.29 (0.82)
    - Column (3): -0.24 (0.19)
    - Column (4): -0.078 (0.17)
  - covid21:
    - Column (1): -0.078 (0.17)
    - Column (2): 0.31 (0.46)
    - Column (3): -0.45** (0.18)
    - Column (4): 0.29 (0.19)
  - log(VIX):
    - Column (3): -0.09** (0.040)
    - Column (4): -0.023 (0.047)
  - USD NEER, world GDP growth, Fed rate, non-fuel commodity price coefficients largely similar to Annex Table 4a with noted changes in statistical significance when two dummies are included.

Implication:
- Main results for public financial flows remain robust to models controlling for COVID-19.
- Including two yearly dummies for 2020 and 2021 can render world GDP growth coefficients statistically insignificant due to higher standard errors, though point estimates may change (e.g., nearly double in some specifications).
- Non-fuel commodity prices significantly affect public financing from private sources (negative coefficients).

---

### Annex III. Other robustness checks and tests

- Model selection (AIC/BIC):
  - For private capital flows, both AIC and BIC favor the full specification.
  - For public borrowing (all sources and multilaterals), AIC favors the full specification while BIC favors the parsimonious alternative.
  - Given the structural nature of the analysis, the authors present the parsimonious specification along with the full specification. Main results remain qualitatively unchanged across both specifications.

- Panel unit root tests (Fisher-type ADF with 1 lag; *p<0.10, **p<0.05, ***p<0.01):
  - K flow variables (Inv. Chi-sq, Inv. Normal (Z), Mod. Inv. Chi-sq, N):
    - bfd: 303.90*** [0.000], -5.37*** [0.000], 11.21 [0.000], 64
    - bfobanks: 421.76*** [0.000], -9.77*** [0.000], 18.91 [0.000], 64
    - bfofirms: 362.67*** [0.000], -7.78*** [0.000], 15.93 [0.000], 64
    - bfoggcbximf: 315.96*** [0.000], -3.91*** [0.000], 11.97 [0.000], 65
    - bfogg_mlat_ximf: 180.88*** [0.001], -2.81*** [0.002], 3.46 [0.000], 63
    - bfogg_prvt: 204.76*** [0.000], -3.63*** [0.000], 8.80 [0.000], 50
  - Push variables:
    - lVIX: 72.04 [1.000], 1.96 [0.975], -3.97 [1.000], 69
    - NEER_US_AE: 112.49 [0.945], -1.20 [0.115], -1.54 [0.938], 69
    - NFCPI: 40.79 [1.000], 5.45 [1.000], -5.85 [1.000], 69
    - GGDP_W: 475.82*** [0.000], -15.41*** [0.000], 20.33 [0.000], 69
    - IR_FED: 162.98* [0.072], -4.19*** [0.000], 1.50 [0.066], 69
  - Pull variables:
    - CA: 376.15*** [0.000], -6.81*** [0.000], 14.56 [0.000], 68
    - fiscal: 205.05*** [0.000], -3.14*** [0.001], 4.98 [0.000], 63
    - ggdp: 364.94*** [0.000], -9.11*** [0.000], 13.88 [0.000], 68

- Stationarity conclusions:
  - Fisher-type panel unit root tests confirm variables are stationary in levels for most main variables.
  - ADF fails to reject unit root for global log VIX, USD NEER, and non-fuel CPI; attributed to low power with short T = 13.
  - KPSS tests fail to reject stationarity for these global variables through lag orders 0–4.
  - Given many variables are ratios to GDP and the wide-panel structure (N > T), results provide reassurance against spurious regression.

---

### Annex IV. Country list and sample periods
- The list of countries and sample periods reflect data availability, which varies across variables. The reported sample period of 2012–2024 corresponds to the variable with the longest available coverage.
- Country sample period pairs (all 2012–2024):
  - Afghanistan; Bangladesh; Benin; Bhutan; Burkina Faso; Burundi; Cabo Verde; Cambodia; Cameroon; Central African Republic; Chad; Comoros; Congo, Democratic Republic of the; Congo, Republic of; Côte d'Ivoire; Djibouti; Dominica; Eritrea; Ethiopia; Gambia, The; Ghana; Grenada; Guinea; Guinea-Bissau; Haiti; Honduras; Kenya; Kiribati; Kyrgyz Republic; Lao P.D.R.; Lesotho; Liberia; Madagascar; Malawi; Maldives; Mali; Marshall Islands; Mauritania; Micronesia; Moldova; Mozambique; Myanmar; Nepal; Nicaragua; Niger; Papua New Guinea; Rwanda; Samoa; Senegal; Sierra Leone; Solomon Islands; Somalia; South Sudan; St. Lucia; St. Vincent and the Grenadines; Sudan; São Tomé and Príncipe; Tajikistan; Tanzania; Timor-Leste; Togo; Tonga; Tuvalu; Uganda; Uzbekistan; Vanuatu; Yemen; Zambia; Zimbabwe.

*Source: IMF Working Paper Riding the Global Financial Cycle: How Capital Flows into LICs (Working Paper No. WP/2026/179).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026179-source-pdf.pdf_