## wpiea2019028 - Section 2–6 (summary of Sections 2–5 and conclusions)

## Source details

**Canonical URL:** [wpiea2019028 - Section 2–6 (summary of Sections 2–5 and conclusions)](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019028.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019028.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019028.pdf.json)

---

### Data and sample
- Main dataset: Dealogic Loan Analytics database — micro data at the level of tranches of all syndicated loans of 127 developing countries from 1993 to 2017.
- Final sample after excluding loans to public authorities: 21,373 syndicated loans to 117 countries.
- Geographic distribution of loan deals in the sample: 51 percent Asia, 27 percent Americas, 15 percent Europe, 7 percent Africa, 0.3 percent Oceania.
- Countries with the majority of syndicated loans across the period: India, Brazil, Russia, Indonesia, Turkey, Mexico.
- Countries with the greatest share of loans supported by MDBs: Afghanistan, Tajikistan, Vanuatu, Belize, Kyrgyzstan, Moldova.
- MDB participation:
  - Deals that involve MDBs are 11.1 percent of the sample.
  - Largest MDB players in the sample: European Investment Bank (EIB), European Bank for Reconstruction and Development (EBRD), International Finance Corporation (IFC).
  - Deals with MDB participation are, on average, significantly larger, with longer maturities, and with fewer non-MDB banks involved.
- Industry grouping: 9 sectors — agriculture, construction and real estate, finance, government, infrastructure, manufacturing, mining and metals, oil and gas, services.
  - Majority of loans belong to finance, infrastructure, and manufacturing sectors; these are also the sectors with majority of loans with MDB participation.
- Scaling and cleaning:
  - Total value of syndicated loan scaled by nominal GDP in USD from the World Economic Outlook (WEO).
  - Winsorize top and bottom 1 percent of total amount scaled by GDP of syndicated loans for data cleaning.
- Panel construction:
  - Balanced panel at the country-sector level: 26,325 observations corresponding to 1,053 country-sectors (117 countries × 9 sectors) observed for 25 years.
- Main outcome variables (baseline focus excludes deals with MDB participation):
  1. Number of syndicated loans in the country-sector (excluding deals with MDB participation).
  2. Total size of syndicated loans (as a share of GDP), excluding amounts lent by MDBs and by their partners in the syndicate.
  3. Number of banks per loan excluding those partnering only in deals with MDB participation.
  4. Average loan maturity.
- Additional outcome variables to explore direct mobilization effects:
  5. Number of banks that participate in syndicated lending per loan (including MDB partners).
  6. Total size of syndicated loans (as a share of GDP), excluding amount lent by MDBs themselves, but including amounts lent by banks co-investing with MDBs’ participation.
- Key summary statistics (country-sector-year averages):
  - On average there are 0.7 syndicated loans in each country-sector-year (Panel A).
  - Average total value of loans in a country-sector-year is 0.11 percent of GDP.
  - Average number of banks (excluding MDBs) involved in syndicated loans is 1.56.
  - Average maturity of loans, without counting loans with MDB participation, is 8.7 months.
- Corporate bonds data for crowding out test:
  - Dealogic bond issuance information for 67 developing countries over 1993-2017.
  - Bond issuances more numerous than syndicated loans in the country-sector pair, but size in percent of GDP is smaller.

### Empirical strategy
- Estimation framework (equation 1): panel at country-sector-year level evaluating whether presence of an MDB attracts private capital.
- Model:
  - y_{cs,t} = sum_{k=0}^2 β_k MDB_{cs,t−k} + δ_{c,t} + ζ_{s,t} + α_{cs} + ε_{cs,t}
- Outcome variables y_{cs,t}: number of syndicated loans; total size of syndicated loans (scaled by GDP); average number of banks involved per syndicated loan; average loan maturity (years).
- Key explanatory variable: MDB_{cs,t} — dummy equal to one if at least one deal supported by MDBs in country-sector pair cs at time t.
  - β_0 measures contemporaneous indirect mobilization effect; up to two lags included. Cumulative effect between t−2 and t computed as (β_0 + β_1 + β_2).
  - Robustness checks include up to two leads of MDB dummy to test anticipation effects.
- Fixed effects:
  - Time-varying country (δ_{c,t}) and sector (ζ_{s,t}) fixed effects; country×sector fixed effects (α_{cs}).
- Standard errors:
  - Clustered at the country level (primary approach). Robustness check: clustering at the country-sector level.

### Main empirical findings — indirect mobilization (baseline, excludes loans with MDB participation)
- Number of syndicated loans:
  - Having at least one syndicated loan supported by at least one MDB is associated with an increase of 0.82 syndicated loans in the following year (MDB_{cs,t-1} ~ 0.8166* reported; MDB_{cs,t} ~ 0.8514 reported).
  - Second lag indicates increase persists into the second year after MDB participation.
  - Contemporaneous effect similar in magnitude but not always statistically significant.
  - Cumulative effect from t to t+2 is equal to 2.448 (∑k=0^2 MDB_{cs,t-k}) and is statistically significant.
  - Context: average 0.76 syndicated loans in a country-sector-year — catalytic effect is economically large.
- Total amount lent to private borrowers (percent of GDP) — indirect mobilization:
  - Country-sector pairs with at least one MDB participating experience a volume of syndicated loans in the same year which is 0.07% of GDP higher than pairs without MDB presence (MDB_{cs,t} 0.0724***).
  - Cumulative effect between t−2 and t is equal to 0.158% of GDP (∑k=0^2).
  - 0.16% corresponds to about 150 percent of the average size of syndicated loans to private creditors in a country-sector-year (average = 0.11% of GDP).
- Number of banks participating per loan:
  - MDB participation associated with on average 0.55 extra banks participating in syndicated lending to the country-sector pair in the same year (MDB_{cs,t} 0.5475**).
  - Effect is persistent; cumulative effect over three years equals almost 1.975 additional lending banks per loan.
  - Context: average 1.52 banks per syndicated loan per year at the country-sector level.
- Loan maturity:
  - MDB participation increases average weighted loan maturity in a country-sector pair by 0.3581 years in the same year (MDB_{cs,t} 0.3581***).
  - Significant lengthening at t+2 as well; cumulative effect close to 0.790 years of additional loan maturity.
- Table 3 reported coefficients and cumulative effects:
  - MDB_{cs,t}: Number 0.8514; Size (% GDP) 0.0724***; Banks 0.5475**; Maturity 0.3581***
  - MDB_{cs,t-1}: Number 0.8166*; Size 0.0293; Banks 0.7406***; Maturity 0.1413
  - MDB_{cs,t-2}: Number 0.7803*; Size 0.0561***; Banks 0.6865**; Maturity 0.2908***
  - ∑k=0^2 MDB_{cs,t-k}: Number 2.448*; Size (% GDP) 0.158***; Banks 1.975***; Maturity 0.790***

### Robustness checks and threats to identification
- Anticipation effects:
  - Model with leads (k = −2,...,+2) shows no anticipation effects: coefficients on MDB_{cs,t+1} and MDB_{cs,t+2} not significantly different from zero for number of loans, size of loans, average number of banks per loan, average maturity (Table 4).
- Confounding factors — presence of top private banks:
  - Control for presence of top 10 largest private banks by S&P; presence predictive for outcomes except number of loans, but MDB mobilization estimates remain statistically significant and qualitatively similar (Table 5).
- Confounding factors — Chinese lending and foreign aid:
  - Chinese official finance dummy (AidData) shows no evidence that past Chinese lending is associated with an increase in syndicated lending; coefficient on Chinese lending dummy negative and significant in some specifications for maturity (Table 6).
  - OECD DAC sectoral official development assistance dummy shows a significant but small positive effect on number of loans only.
  - Joint control for top 10 banks, Chinese lending, and sectoral foreign aid yields qualitatively similar findings.
- Sensitivity to outliers:
  - Excluding countries with fewer than 100 syndicated loans (1993-2017) → sample of 32 countries: effects on loan size, banks per loan, and loan maturity similar; coefficients on number of loans not significantly different from zero in this subset.
  - Excluding sectors with fewer than 1,000 syndicated loans → left with 5 sectors: results mostly not significantly different.
  - Combined exclusions (~15% of whole sample): results consistent with baseline.
- Sensitivity to model specification:
  - Monte Carlo simulations with 1,000 replications: treatment randomly re-assigned while keeping original distribution. Average cumulative effects from t−2 to t across simulations never statistically significant from zero (Table A7).
- Clustering of standard errors:
  - Baseline clusters at country level; alternative clustering at country-sector level tested. Clustering at country-sector level yields more precisely estimated coefficients and more likely significant compared to baseline.

### Extensions — 5.1 Direct and Indirect Effects (total mobilization)
- Mobilization channels:
  - Signaling and demonstration effects.
  - Risk mitigation and better information (preferred creditor status, monitoring, risk sharing).
- Indirect mobilization: private lenders enter syndicated loans different from those with MDB participation.
- Direct mobilization: private investors partner directly with the MDB in the same syndicated loan.
- Empirical decomposition:
  - Redefine outcomes to measure total mobilization (including loan amounts with MDB participation but excluding MDBs’ own share) and compare to indirect mobilization (excluding loans with MDB participation).
- Key quantitative findings:
  - A country-sector with MDB lending experiences total syndicated loan volume (including MDB-partner amounts) which is 0.31% of GDP higher than in country-sectors without MDB participation.
  - In sectors with MDB participation there are almost 3 more banks, either partnering with the MDB (direct mobilization) or part of other deals (indirect mobilization).
  - Total mobilization effects are larger than indirect effects alone, implying both signaling and risk mitigation channels operate.
- Table 7 reported:
  - Total Size (%GDP) including MDB loan: 0.3107***
  - Total Banks including MDB loan: 2.8218***
  - Indirect Size (%GDP) excluding MDB loan: 0.0785***
  - Indirect Banks excluding MDB loan: 0.6971***

### Extensions — 5.2 Crowding out of corporate bonds
- Rationale: corporate bond issuances as closest substitute to syndicated loans.
- Dependent variables: number of bond issuances and total amount of bond issuances (in percent of GDP) in the country-sector pair at time t. Bond size winsorized at 1%.
- Findings:
  - No significant effect on the number of bonds nor on the size of corporate bond financing at the same time and in the two years after MDBs’ participation (columns 1 and 5, Table 8).
  - Results robust to controls for presence of the top 10 banks and official development assistance in the previous 2 years.
  - When controlling for Chinese lending, there is some marginally significant increase in the number of bonds and size of bond financing at t+1.
  - Cumulative effects are positive for the number of issuances, while not significantly different from zero for the size of inflows.
  - Overall: no evidence of crowding out; suggestive evidence that MDBs’ participation in syndicated lending may crowd in capital inflows in the corporate bond market.

### Extensions — 5.3 Sector heterogeneity
- Spillovers across sectors:
  - Testing whether MDB lending in one sector affects private flows to other sectors within the same country-year.
  - Controlling for MDB presence in another sector leaves mobilization effects in the focal sector unaltered.
  - MDB participation in one sector does not significantly prevent private resources from flowing into other sectors.
  - Evidence of positive spillovers on the number of creditors participating in other sectors.
- Infrastructure sector focus:
  - Infrastructure needs: UN estimates total investment in economic infrastructure under $1 trillion per year, but will need to reach between $1.6-2.5 trillion a year over the period 2015-30.
  - Subsample (infrastructure only):
    - No significant effects on number of loans nor on size (possible loss of power).
    - Effect on number of creditors similar to baseline.
    - MDB participation in infrastructure increases average maturity of syndicated loans by 0.81 years (subsample results show larger maturity effects; Table 10 reports cumulative maturity 2.605*** for infrastructure subsample).

### Extensions — 5.4 Country heterogeneity
- Role of country risk (investment grade):
  - Investment grade defined via S&P (BBB- or higher); if missing, Moody’s (Baa3 or higher).
  - Cumulative effects from t to t+2 are significantly larger than zero for all outcomes only when countries are rated as investment grade.
  - For riskier countries, cumulative effects significant only for number of creditors and loan maturity.
  - Interpretation: easier for MDBs to mobilize syndicated loans and larger credit flows to less risky (investment grade) countries; in riskier countries MDBs attract lenders and improve loan terms via risk mitigation.
  - Table 11 examples:
    - With Investment Grade Rating ∑k=0^2 MDB: Number 3.657**; Size 0.323*; Maturity 0.592**
    - Without Investment Grade Rating ∑k=0^2 MDB: Banks 1.610**; Maturity 0.843***
- Role of financial development:
  - Countries split by average domestic credit to private sector (% of GDP) above/below sample median.
  - Results show significantly larger mobilization effects in countries more financially developed: cumulative size 0.224***; banks 2.241***; maturity 0.672***.
  - In less financially developed countries, most MDB mobilization effects are nullified; average maturity is the only exception (cumulative maturity 0.993***).
  - Interpretation: financial development and market depth are necessary conditions for access to loan syndications; MDB mobilization is stronger where financial markets are deeper.
- Low-income and low-growth countries:
  - Low income countries:
    - No mobilization effects at t nor at t+1; at t+2 significant positive increase in number of loans, number of creditors, and maturity, but cumulative effects not significant.
    - Interpretation: greater difficulty to mobilize private capital in low income countries.
  - Low-growth countries:
    - No significant changes in number of loans nor lending banks per loan.
    - MDB participation significantly associated with higher credit.
    - Loan maturity increases over three-year window by around 4 months per year (average maturity of syndicated loans without MDB support is almost 5.5 years).

### Conclusions (section 6)
- MDB entry in a country-sector via participation in loan syndication increases the number of syndicated loans and associated capital inflows, and improves access to credit: average number of lending banks per loan and average loan maturity increase.
- Effects are economically sizable and persist up to at least three years.
- Evidence supports both indirect (signaling/demonstration) and direct (risk mitigation/information provision) mobilization channels.
- Mobilization effects are heterogeneous:
  - Concentrated in countries more economically and financially developed, and with better credit ratings.
  - Less effective in low-income, less financially developed, and risky countries—where MDBs face significant challenges to attract private resources despite larger financing needs.
- Policy implication: MDBs can play a fundamental role in attracting private sector investment, but effectiveness varies with country risk, financial development, and income level, indicating potential limits in mobilizing private finance where needs are greatest.

*Source — wpiea2019028, Sections 2–6 as provided in the source PDF content.*

### Section 5 introduces a number of extensions and Section 6 concludes.

### wpiea2019028 - Section 5 introduces a number of extensions and Section 6 concludes.

### Data
- Main dataset: Dealogic Loan Analytics database — micro data at the level of tranches of all syndicated loans of 127 developing countries from 1993 to 2017.
- Final sample after excluding loans to public authorities: 21,373 syndicated loans to 117 countries.
- Geographic distribution of loan deals in the sample: 51 percent Asia, 27 percent Americas, 15 percent Europe, 7 percent Africa, 0.3 percent Oceania.
- Countries with the majority of syndicated loans across the period: India, Brazil, Russia, Indonesia, Turkey, Mexico (largest bubbles in Figure 1 in source).
- Countries with the greatest share of loans supported by MDBs: Afghanistan, Tajikistan, Vanuatu, Belize, Kyrgyzstan, Moldova (darker bubbles in Figure 1 in source).
- MDB participation:
  - Deals that involve MDBs are 11.1 percent of the sample.
  - Largest MDB players in the sample: European Investment Bank (EIB), European Bank for Reconstruction and Development (EBRD), International Finance Corporation (IFC).
  - Deals with MDB participation are, on average, significantly larger, with longer maturities, and with fewer non-MDB banks involved (Table 1 in source).
- Industry grouping: 9 sectors — agriculture, construction and real estate, finance, government, infrastructure, manufacturing, mining and metals, oil and gas, services.
  - Majority of loans belong to finance, infrastructure, and manufacturing sectors; these are also the sectors with majority of loans with MDB participation.
- Scaling and cleaning:
  - Total value of syndicated loan scaled by nominal GDP in USD from the World Economic Outlook (WEO).
  - Winsorize top and bottom 1 percent of total amount scaled by GDP of syndicated loans for data cleaning.
- Panel construction:
  - Balanced panel at the country-sector level: 26,325 observations corresponding to 1,053 country-sectors (117 countries × 9 sectors) observed for 25 years.
- Main outcome variables (baseline focus excludes deals with MDB participation):
  1. Number of syndicated loans in the country-sector (excluding deals with MDB participation).
  2. Total size of syndicated loans (as a share of GDP), excluding amounts lent by MDBs and by their partners in the syndicate.
  3. Number of banks per loan excluding those partnering only in deals with MDB participation.
  4. Average loan maturity.
- Additional outcome variables to explore direct mobilization effects (described later, see Section 5.1 in source):
  5. Number of banks that participate in syndicated lending per loan (including MDB partners).
  6. Total size of syndicated loans (as a share of GDP), excluding amount lent by MDBs themselves, but including amounts lent by banks co-investing with MDBs’ participation.
- Summary statistics (Table 2 in source):
  - On average there are 0.7 syndicated loans in each country-sector-year (Panel A).
  - Average total value of loans in a country-sector-year is 0.11 percent of GDP.
  - Average number of banks (excluding MDBs) involved in syndicated loans is 1.56.
  - Average maturity of loans, without counting loans with MDB participation, is 8.7 months.
- Corporate bonds data for crowding out test:
  - Dealogic bond issuance information for 67 developing countries over 1993-2017.
  - Panel B summary: bond issuances more numerous than syndicated loans in the country-sector pair, but size in percent of GDP is smaller.

### Empirical strategy
- Estimation framework (equation 1): panel at country-sector-year level evaluating whether presence of an MDB attracts private capital.
- Model:
  y_{cs,t} = sum_{k=0}^2 β_k MDB_{cs,t−k} + δ_{c,t} + ζ_{s,t} + α_{cs} + ε_{cs,t}
- Outcome variables y_{cs,t}:
  1) number of syndicated loans;
  2) total size of syndicated loans, scaled by GDP;
  3) average number of banks involved per syndicated loan;
  4) average loan maturity of syndicated loans, in years.
- Baseline measures focus on indirect mobilization — excluding loans with MDB participation. Direct mobilization measures described in Section 5.1 (see Data section above).
- Key explanatory variable: MDB_{cs,t} — dummy equal to one if at least one deal supported by MDBs in country-sector pair cs at time t.
  - β_0 measures contemporaneous indirect mobilization effect.
  - Up to two lags included; cumulative effect between t−2 and t computed as (β_0 + β_1 + β_2).
  - Robustness checks include up to two leads of MDB dummy to test anticipation effects.
- Fixed effects:
  - Time-varying country (δ_{c,t}) and sector (ζ_{s,t}) fixed effects.
  - Country×sector fixed effects (α_{cs}).
- Standard errors:
  - Clustered at the country level (primary approach) following Abadie et al. (2017).
  - Robustness check: clustering at the country-sector level.

### Results — Main findings
- Estimation uses country×sector, country×year and sector×year fixed effects.
- Number of syndicated loans (Column 1, Table 3):
  - Having at least one syndicated loan supported by at least one MDB is associated with an increase of 0.82 syndicated loans in the following year.
  - Second lag indicates increase persists into the second year after MDB participation.
  - Contemporaneous effect similar in magnitude but not statistically significant.
  - Cumulative effect from t to t+2 is equal to 2.45 and is statistically significant.
  - Context: average 0.76 syndicated loans in a country-sector-year — catalytic effect is economically large.
- Total amount lent to private borrowers (percent of GDP) — indirect mobilization (Column 2, Table 3):
  - Country-sector pairs with at least one MDB participating experience a volume of syndicated loans in the same year which is 0.07% of GDP higher than pairs without MDB presence.
  - No additional effect in the following year, smaller lagged effect after two years.
  - Cumulative effect between t−2 and t is equal to 0.16% of GDP.
  - 0.16% corresponds to about 150 percent of the average size of syndicated loans to private creditors in a country-sector-year (average = 0.11% of GDP, Table 2).
  - Scaling by the size of the average syndicated loan (0.11% of GDP, Table 1) still shows a large cumulative effect.
- Number of banks participating per loan (Column 3, Table 3):
  - MDB participation associated with on average 0.55 extra banks participating in syndicated lending to the country-sector pair in the same year.
  - Effect is persistent over time with similar magnitudes each year.
  - Cumulative effect over three years equals almost 2 additional lending banks per loan.
  - Context: average 1.52 banks per syndicated loan per year at the country-sector level.
- Loan maturity (Column 4, Table 3):
  - MDB participation increases average weighted loan maturity in a country-sector pair by 0.36 years in the same year.
  - Significant lengthening of average maturity at t+2 as well.
  - Cumulative effect close to 0.8 years of additional loan maturity.

### Results — Robustness
- Threats tested: anticipation effects and confounding factors; role of extreme values; model specification; alternative clustering of standard errors.
- Anticipation effects:
  - Model with leads: y_{cs,t} = sum_{k=-2}^{+2} β_k MDB_{cs,t−k} + δ_{c,t} + ζ_{s,t} + α_{cs} + ε_{cs,t}.
  - Inclusion of leads shows no anticipation effects (Table 4): coefficients on MDB_{cs,t+1} and MDB_{cs,t+2} not significantly different from zero for number of loans, size of loans, average number of banks per loan, average maturity.
- Confounding factors — presence of top private banks:
  - Control for presence of top 10 largest private banks by S&P (dummy = 1 if at least one of top 10 banks was present in country-sector in previous 2 years).
  - Presence of top banks predictive for outcomes except number of loans, but MDB mobilization estimates remain statistically significant and qualitatively similar (Table 5).
  - Note: top 10 list and ranking details provided in source footnote.
- Confounding factors — Chinese lending and foreign aid:
  - Chinese official finance data from AidData (2000-2014) used to create dummy = 1 if country-sector received Chinese-financed project in previous 2 years.
  - Results (Table 6, columns 1-4): no evidence that past Chinese lending is associated with an increase in syndicated lending, number of loans, number of lenders, and average maturity (coefficient on Chinese lending dummy negative and significant in column 4).
  - OECD (2018) DAC sectoral official development assistance (1993-2015) used to create dummy = 1 if country-sector received aid in previous 2 years.
  - Augmenting baseline with aid dummy: MDB mobilization estimates remain statistically significant and qualitatively similar; aid flows show a significant but small positive effect on number of loans only (Table 6, columns 5-8).
  - Joint control for top 10 banks, Chinese lending, and sectoral foreign aid yields qualitatively similar findings (results available upon request in source).
- Sensitivity to outliers:
  - Exclude countries with fewer than 100 syndicated loans over 1993-2017 → sample of 32 countries. Results mostly robust: effects on loan size, banks per loan, and loan maturity similar; coefficients on number of loans not significantly different from zero in this subset (Table A4).
  - Exclude sectors with fewer than 1,000 syndicated loans in 1993-2017 → left with 5 sectors (oil and gas, construction and real estate, manufacturing, infrastructure, finance). Results mostly not significantly different from Table 3 (Table A5).
  - Combine country and sector exclusions (sample ~15% of whole sample): results consistent with baseline (Table A6).
- Sensitivity to model specification:
  - Monte Carlo simulations with 1,000 replications: treatment randomly re-assigned while keeping original distribution.
  - Average cumulative effects from t−2 to t across simulations never statistically significant from zero (Table A7) — model by itself does not generate mobilization effects.
- Clustering of standard errors:
  - Baseline clusters at country level; alternative clustering at country-sector level tested.
  - Clustering at country-sector level yields more precisely estimated coefficients and more likely significant compared to baseline (Table A8).
  - Choice of clustering depends on assumed level of MDB support assignment; baseline uses most conservative approach (country-level clustering).

### Extensions (introductory note)
- Section 5 of the source introduces a number of extensions; Section 6 concludes. (The supplied content ends at the beginning of Section 5; extensions are referenced but not detailed in the supplied text.)

*Italic: Source — wpiea2019028, Section 2–5 as provided in the source PDF content.*

### 5.1    Direct and Indirect Effects

### 5.1    Direct and Indirect Effects

### Mobilization channels and interpretation
- Two main channels through which MDB lending can mobilize private creditors:
  - signaling and demonstration effects (MDB investment on commercial terms signals good investment opportunities and can encourage governments to adopt reforms);
  - risk mitigation and better information (MDBs’ preferred creditor status, close engagement, monitoring, and risk sharing reduce perceived risks and fill information gaps).
- Indirect mobilization: private lenders enter syndicated loans different from those with MDB participation (attracted by signaling/demonstration).
- Direct mobilization: private investors partner directly with the MDB in the same syndicated loan, benefiting from risk sharing and preferred creditor status.

### Empirical strategy to disentangle effects
- Total mobilization effects are observed (include both direct and indirect).
- To disentangle, outcome variables are redefined to measure total mobilization (including volume of loan in which MDBs participate, excluding MDBs’ own share, and number of private creditors partnering in the loan itself).
- Comparison between total mobilization (including loans with MDB participation) and indirect mobilization (excluding loans with MDB participation) isolates the contribution of direct mobilization.

### Key empirical findings (timet)
- A country-sector with MDB lending experiences a volume of syndicated loans— including both MDB-partner amounts (direct mobilization) and other loans (indirect mobilization)—which is 0.31% of GDP higher than in country-sectors without MDB participation.
- In sectors with MDB participation there are almost 3 more banks, either partnering with the MDB (direct mobilization) or part of other deals (indirect mobilization).
- Total mobilization effects are larger than indirect effects alone, suggesting both signaling and risk mitigation channels are at play.

---

### 5.2    Is There a Crowding Out of Corporate Bonds?

### Rationale and method
- Corporate bond issuances are assessed as the closest substitute to syndicated loans (similar size and maturity).
- Dependent variables: number of bond issuances and total amount of bond issuances (in percent of GDP) in the country-sector paircs at time t.
- Bond size variable is winsorized at 1%.

### Findings
- No significant effect on the number of bonds nor on the size of corporate bond financing at the same time and in the two years after MDBs’ participation (columns 1 and 5, Table 8).
- Results robust to controls for presence of the top 10 banks and official development assistance in the previous 2 years.
- When controlling for Chinese lending (columns 3 and 7), there is some marginally significant increase in the number of bonds and size of bond financing at t+1.
- Cumulative effects are positive for the number of issuances, while not significantly different from zero for the size of inflows.
- Overall: no evidence of crowding out; suggestive evidence that MDBs’ participation in syndicated lending may crowd in capital inflows in the corporate bond market.

---

### 5.3    Sector Heterogeneity

### Spillover effects across sectors
- Analysis tests whether MDB lending in one sector affects private flows to other sectors within the same country-year.
- Controlling for MDB presence in another sector leaves mobilization effects in the focal sector unaltered.
- MDB participation in one sector does not significantly prevent private resources from flowing into other sectors.
- Evidence of positive spillovers on the number of creditors participating in other sectors, suggesting MDB participation may create new opportunities across sectors that are not reflected in number of loans, size, or maturity.

### Infrastructure sector focus
- Infrastructure investment needs: UN estimates total investment in economic infrastructure under $1 trillion per year, but will need to reach between $1.6-2.5 trillion a year over the period 2015-30.
- Infrastructure characterized by high perceived risks and long-term financing needs, making MDB mobilization particularly relevant.
- Subsample analysis (infrastructure only, Table 10):
  - No significant effects on number of loans nor on size (possible loss of power).
  - Effect on number of creditors similar to baseline results.
  - MDB participation in infrastructure increases average maturity of syndicated loans by 0.81 years, with effects substantial even in subsequent years.

---

### 5.4    Country Heterogeneity

### The Role of Country Risk (credit ratings)
- Investment grade defined via S&P (BBB- or higher); if missing, Moody’s (Baa3 or higher). Countries never rated by S&P or Moody’s considered non-investment grade.
- Cumulative effects from timet to t+2 are significantly larger than zero for all outcomes only when countries are rated as investment grade (columns 5-8, Table 11).
- For riskier countries, cumulative effects are significant only for the number of creditors and loan maturity (columns 1-4, Table 11).
- Interpretation:
  - Easier for MDBs to mobilize syndicated loans and larger credit flows to less risky (investment grade) countries.
  - MDB participation helps attract other lenders and improve loan terms via risk mitigation, particularly in riskier countries.
  - Cumulative mobilization effects on maturity are larger in countries without investment grade ratings; given these countries have significantly smaller average maturity, the effect on loan terms is larger in risky countries.

### The Role of Financial Development
- Countries split by average domestic credit to the private sector (% of GDP) above/below the sample median.
- Results (Table 12):
  - Significantly larger mobilization effects in countries more financially developed: significant mobilization on size, banks, and maturity (columns 1-4).
  - In less financially developed countries, most MDB mobilization effects are nullified; average maturity is the only exception (columns 5-8).
- Interpretation: financial development and market depth are necessary conditions for access to loan syndications; MDB mobilization is stronger where financial markets are deeper.

### Low-Income and Low-Growth countries
- Low income countries (World Bank definition) subsample (Table 13, columns 1-4):
  - No mobilization effects at t nor at t+1.
  - At t+2, significant positive increase in number of loans, number of creditors, and maturity of loans.
  - Magnitude not enough to make cumulative effects significant.
  - Interpretation: greater difficulty to mobilize private capital in low income countries; MDBs mobilize largest amounts in high income countries per joint MDB report.
- Low-growth countries (below sample median real GDP growth, IMF WEO data) subsample (Table 13, columns 5-8):
  - No significant changes in number of loans nor lending banks per loan.
  - MDB participation significantly associated with higher credit.
  - Loan maturity consistently increases over the three-year window by around 4 months per year (sizable: average maturity of syndicated loans without MDB support is almost 5.5 years).
- Overall: mobilization effects less likely in poor countries (higher country risk); growth performance is less of a constraint and MDBs can catalyze financing with a long-term horizon.

---

### Conclusions (section 6 summary)
- MDB entry in a country-sector via participation in loan syndication increases number of syndicated loans and associated capital inflows, and improves access to credit: average number of lending banks per loan and average loan maturity increase.
- These effects are economically sizable and persist up to at least three years.
- Evidence supports both indirect (signaling/demonstration) and direct (risk mitigation/information provision) mobilization channels.
- Mobilization effects are heterogeneous:
  - Concentrated in countries more economically and financially developed, and with better credit ratings.
  - Less effective in low-income, less financially developed, and risky countries—where MDBs face significant challenges to attract private resources despite larger financing needs.
- Policy implication: MDBs can play a fundamental role in attracting private sector investment, but effectiveness varies with country risk, financial development, and income level, indicating potential limits in mobilizing private finance where needs are greatest.

*Source: wpiea2019028 - 5.1    Direct and Indirect Effects*

### References

### References

### Bibliographic sources and themes
- Core topics covered by the cited literature:
  - Multilateral development banks (MDBs), MDB participation in infrastructure, and MDB risk mitigation.
  - Syndicated loans, syndicated loan pricing, market integration, secondary market pricing, and leveraged loan syndication.
  - Catalytic finance, IMF catalytic effects, and the IMF’s role vis-à-vis private capital markets.
  - Mobilization of private finance, blended finance, and crowding-in of private sector finance.
  - Official financing from China and OECD ODA, and their effects on private flows.
  - Measurement and determinants of private capital flows and corporate borrowing, including debt maturity and spreads on emerging market debt.
- Representative authored works and institutions appearing in the reference list:
  - Abadie, A., Athey, S., Imbens, G. W., and Wooldridge, J. (2017).
  - Altunbaş, Y., Kara, A., and Marqués-Ibáñez, D. (2010).
  - Arezki, R., Bolton, P., Peters, S., Samama, F., and Stiglitz, J. (2017).
  - Blended Finance Taskforce (2018).
  - CEPR, IMF Working Papers, World Bank Policy Research Working Papers, AidData working papers, OECD indicators, and numerous MDB reports and technical notes.
- Data and indicator sources explicitly cited:
  - Dealogic Loan Analytics (for syndicated loan trend series in Figure 2).
  - AidData (Dreher et al., 2017) and OECD (2018) (noted in Table 6 source line).
  - World Bank’s World Development Indicators (used for domestic credit to private sector measure referenced in Table A2 notes).
  - IMF’s World Economic Outlook (used for GDP growth in Table 13 notes).
  - World Bank country income classification for fiscal year 2018 (used in Table A3 notes).

### Figures (visual summaries)
- Figure 1: Number of Syndicated Loans and Share with MDB Support by Country, 1993-2017
  - Map visualization: bubble area = number of syndicated loans (1993-2017) by country; bubble color (blue intensity) = share of syndicated loans supported by at least one MDB in the period.
- Figure 2: Number of Syndicated Loans
  - Time series chart of total number of syndicated loans. Source: Dealogic Loan Analytics.

### Key summary statistics (Tables)
- Table 1: Summary Statistics of Syndicated Loans (individual loan level)
  - Panel A. All Syndicated Loans:
    - Size (% GDP): Mean 0.108 SD 0.319 Min 0.001 Max 4.295 N 21,164
    - Number of participating banks: Mean 6.153 SD 7.498 Min 1 Max 80 N 21,373
    - Maturity (years): Mean 5.458 SD 4.476 Min 0.083 Max 35.083 N 18,603
  - Panel B. Syndicated Loans with MDB Participation:
    - Size (% GDP): Mean 0.169 SD 0.403 Min 0.001 Max 3.938 N 1,946
    - Number of participating banks: Mean 4.487 SD 5.976 Min 1 Max 59 N 1,967
    - Maturity (years): Mean 7.489 SD 4.499 Min 0.230 Max 1,111 (note: N 1,111)

- Table 2: Summary Statistics in the Aggregate Data (country-sector-year)
  - Panel A. Syndicated Loans (country-sector-year):
    - Number of Loans: Mean 0.737 SD 4.518 Min 0 Max 188 N 26,325
    - Size (% GDP): Mean 0.109 SD 0.693 Min 0 Max 41.514 N 26,325
    - Size w/o Loan w/MDB (% GDP): Mean 0.101 SD 0.672 Min 0 Max 41.514 N 26,325
    - Banks: Mean 1.474 SD 5.185 Min 0 Max 80 N 26,325
    - Banks w/o MDB partners: Mean 1.562 SD 5.309 Min 0 Max 80 N 26,325
    - Maturity (years): Mean 0.724 SD 2.174 Min 0 Max 35 N 26,096
  - Panel B. Corporate Bonds:
    - Number of Bonds: Mean 1.245 SD 8.801 Min 0 Max 336 N 15,075
    - Bond Size (% GDP): Mean 0.037 SD 0.180 Min 0 Max 6.187 N 15,075

### Mobilization and empirical effects (selected econometric findings)
- Table 3: MDB Mobilization Effects (baseline estimates of equation 1)
  - Coefficients (MDBcs,t lags) and cumulative effects (∑k=0^2 MDBcs,t-k):
    - MDBcs,t: Number of loans 0.8514; Size (% GDP) 0.0724***; Banks 0.5475**; Maturity 0.3581***
    - MDBcs,t-1: Number 0.8166*; Size 0.0293; Banks 0.7406***; Maturity 0.1413
    - MDBcs,t-2: Number 0.7803*; Size 0.0561***; Banks 0.6865**; Maturity 0.2908***
    - ∑k=0^2 MDBcs,t-k: Number 2.448*; Size (% GDP) 0.158***; Banks 1.975***; Maturity 0.790***
  - Wald test p-values for cumulative effects:
    - Number: 0.0841
    - Size: 10.00324 (note: appears as 10.003240 in source) — preserve exact representation as in source
    - Banks: 0.001852.40e-06 (source text shows "40.001852.40e-06"; preserve as given? The table shows "0.001852.40e-06" — to adhere to "use ONLY supplied content" and preserve exactly, present p-value line exactly as in table: "Wald test:p value 0.08410.003240.001852.40e-06")
  - Estimation controls: Sector-country FE, Sector-year FE, Country-year FE; Standard errors clustered at country level; significance: * p<0.1, ** p<0.05, *** p<0.01.
  - Sample averages reported:
    - Average MDB 0.0483; Average Number of Loans 0.760; Average Size (%GDP) 0.105; Average Banks 1.526; Average Maturity 0.755

- Table 4: Controlling for Anticipation Effects (equation 2)
  - MDBcs,t and leads/lags included (MDBcs,t+1 and MDBcs,t+2):
    - MDBcs,t: Number 0.8176**; Size 0.0703***; Banks 0.5065**; Maturity 0.3366***
    - MDBcs,t-1: Number 0.9013***; Size 0.0245; Banks 0.6947***; Maturity 0.0781
    - MDBcs,t-2: Number 0.8960***; Size 0.0563***; Banks 0.6318**; Maturity 0.2038**
    - MDBcs,t+1: Number 0.4407; Size 0.0410; Banks 0.3179; Maturity 0.1260
    - MDBcs,t+2: Number 0.3091; Size 0.0602; Banks 0.3627; Maturity 0.1634
  - Sample averages: Average MDB 0.049; Average Number of Loans 0.730; Average Size (% GDP) 0.105; Average Banks 1.545; Average Maturity 0.738

- Table 6: Mobilization Effects Controlling for Chinese Lending and Aid Flows
  - Columns 1–4 control for Chinese lending; Columns 5–8 control for ODA:
    - Example coefficients (columns 1-4):
      - MDBcs,t (col 1): 0.5779***; MDBcs,t-1: 0.8569***; MDBcs,t-2: 0.8006**; Chinese lending: 0.0039; ODA (in other specification): 0.0997**
      - ∑k=0^2 MDBcst-k reported as 2.235*** (col 1) and 2.848** (col 5) among others.
    - Wald test p-values listed across columns; observations and R-squared vary by column (e.g., Observations 11,934 and 21,168; R-squared 0.753 and 0.672 in different specifications).
  - Sources: AidData (Dreher et al., 2017) and OECD (2018).
  - Average MDB reported per specification: e.g., 0.0631 (columns 1-4), 0.0425 (columns 5-8).

- Table 7: Total Mobilization Effects (direct + indirect and indirect only)
  - MDBcs,t coefficients:
    - Total Size (%GDP) including MDB loan: 0.3107***
    - Total Banks including MDB loan: 2.8218***
    - Indirect Size (%GDP) excluding MDB loan: 0.0785***
    - Indirect Banks excluding MDB loan: 0.6971***
  - Observations 26,325; R-squared 0.372 and 0.653 in different columns.

- Table 8: Effects on Bonds (corporate bonds)
  - Results generally imprecise; selected entries:
    - ∑k=0^2 MDB Support cs,t-k: Number of Bonds columns show values such as 0.570, 0.463, 2.397* (varying by column); Bond Size (% GDP) cumulative values such as 0.0421, 0.0388, 0.0858.
    - Observations vary by specification: 13,869; 6,318; 11,718.
  - Inclusion of Top 10 Banks, Chinese lending, and OECD ODA as controls in some specifications.

### Subsample and robustness findings (selected)
- Inter-sectoral spillovers (Table 9):
  - MDBcs,t cumulative ∑k=0^2 effect: Number 2.337*; Size (% GDP) 0.168***; Banks 2.296***; Maturity 0.791***.
  - MDB in other sector c̄s,t coefficients reported (e.g., Number -0.2700; Size 0.0256; Banks 0.7779**).
- Infrastructure sector (Table 10):
  - MDBcs,t coefficients notably larger in count and maturity dimensions:
    - MDBcs,t (Number) 1.8310; MDBcs,t-1 1.9705; MDBcs,t-2 2.3603
    - ∑k=0^2 MDBcs,t-k: Number 6.162; Size 0.0665; Banks 2.198*; Maturity 2.605***
    - Observations 2,691; Average Size (%GDP) 0.363; Average Maturity 2.150
- By investment grade rating (Table 11):
  - For countries With Investment Grade Rating (columns 5-8):
    - ∑k=0^2 MDB Support cs,t-k: Number 3.657**; Size 0.323*; Banks 1.418; Maturity 0.592**
  - For W/o Investment Grade Rating (columns 1-4):
    - ∑k=0^2 MDB Support cs,t-k: Number 0.379; Size 0.0511; Banks 1.610**; Maturity 0.843***
- By financial development (Table 12):
  - More financially developed countries (columns 1-4):
    - ∑k=0^2 MDB Support cs,t-k: Number 3.480; Size 0.224***; Banks 2.241***; Maturity 0.672***
  - Less financially developed countries (columns 5-8):
    - ∑k=0^2 MDB Support cs,t-k: Number 0.815; Size 0.0687; Banks 1.486; Maturity 0.993***
- Low income and low growth subsamples (Table 13):
  - Low Income Countries (columns 1-4): ∑k=0^2 MDB Support cs,t-k: Number 0.144; Size -0.0484; Banks 0.404; Maturity 0.775
  - Low Growth Country-years (columns 5-8): ∑k=0^2 MDB Support cs,t-k: Number -0.121; Size 0.127***; Banks 0.675; Maturity 0.930***

### Online appendix and robustness details (selected numeric items)
- Table A1: Amounts lent 1993-2017 by Multilateral Organizations (USD million)
  - European Investment Bank - EIB 45,528
  - European Bank for Reconstruction and Development - EBRD 45,064
  - International Finance Corporation - IFC 41,062
  - International Bank for Reconstruction & Development - World Bank 35,228
  - Inter-American Development Bank - IADB 24,611
  - African Development Bank 13,070
  - Asian Development Bank 8,293
  - Islamic Development Bank 7,274
  - Corporacion Andina de Fomento - CAF Development Bank of Latin America 4,115
  - (additional MDBs and USD million amounts listed in table)
- Table A2: Countries by Financial Development
  - Lists of High Financial Development Countries and Low Financial Development Countries as provided verbatim.
- Table A3: Countries by Income Levels
  - Low Income Countries list (e.g., Afghanistan, Benin, Burkina Faso, ... Zimbabwe)
  - Middle and High Income Countries list (extensive list as provided)
- Robustness tables (A4–A9) and Monte Carlo results (A7) report alternative samples, clustering choices, and simulation outcomes:
  - Example (Table A7): ∑k=0^2 βk averages from Monte Carlo Simulations: -0.001, -0.001, -0.001, 0.001; No of Replications 1,000.

*This overlay summarizes the References section and associated Figures and Tables from the supplied content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019028.pdf_