## wpiea2024134-print-pdf

## Source details

**Canonical URL:** [wpiea2024134-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024134-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024134-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024134-print-pdf.pdf.json)

---

### Overview
- Sections layout: Section 4 details the data; Section 5 describes the main econometric results; Section 6 presents a sensitivity analysis; Section 7 provides a summary and conclusions.
- Focus: stylized facts on ODA and IMF lending to LICs, correlation between IMF program participation and ODA, and an IV-based econometric identification strategy to estimate the IMF’s (ex-post) catalytic effect on ODA.

### Stylized facts on ODA and IMF lending to LICs
- Role: ODA is a principal source of external financing for LICs with limited access to private capital markets and low domestic savings.
- Aggregate and relative trends:
  - ODA flows to LICs increased from USD 13.4 billion to 52.4 billion from 1990-2019.
  - For the average LIC, ODA fell from 12.7 to 6.3 pp of GDP over 1990-2019.
  - ODA flows have consistently remained above 5 pp of GDP.
- Composition shifts (1990 to 2019):
  - Traditional bilateral donors’ share fell from 63 to 43 percent of ODA.
  - Multilateral donors’ share rose from 32 to 51 percent of ODA.
  - Non-traditional bilateral donors’ share increased since 2009 but remained 4 percent of total ODA flows in 2019.
- External financing composition for LICs in 2019:
  - ODA: 28 percent of external financing flows.
  - Remittances: 48 percent.
  - FDI: 21 percent.
  - Private portfolio flows: 4 percent.

### IMF concessional lending and the catalytic concept
- Purpose: IMF concessional lending to LICs assists countries with protracted balance of payments needs and supports implementation of policy actions to restore external balance and support growth and poverty reduction.
- Program design and disbursements:
  - Program size, length, disbursement schedule, and policy actions are agreed prior to Executive Board approval.
  - Disbursements can be front-, even-, or back-loaded depending on country circumstances and phasing of adjustment measures.
  - Program reviews verify programs are “on-track”; off-track programs lead to stopped disbursements.
- IMF Financing Assurances Policy:
  - Approval and reviews require the financing assurances policy to hold: the program must be (ex-ante) fully financed for the next 12 months with good prospects for the remainder of the program period.
  - Indications of donor financing are required but are not binding.
- Definition used in this study:
  - Catalytic effect is defined as the (ex-post) materialized support from the donor community in response to IMF financing (i.e., actual ODA flows realized after IMF program approval).

### Channels through which IMF programs may catalyze ODA
- Conditionality channel: IMF program conditionality signals authorities’ intentions and strengthens donor perceptions of aid effectiveness.
- Liquidity channel: IMF program liquidity eases adjustment burdens, reinforcing donor willingness to provide financing.
- IMF programs act as a coordination mechanism among donors.

### Correlation evidence between IMF programs and ODA
- Cross-sectional distribution:
  - LICs engaged in an IMF-supported program receive on average 10.2 pp of GDP in annual ODA.
  - LICs not in IMF programs receive on average 5.8 pp of GDP in annual ODA.
  - ODA-to-GDP ratio displays wide variation across LICs with a long right tail.
- Country cases:
  - Guinea-Bissau: five IMF programs between 1990 and 2019; ODA-to-GDP ratio higher in program periods with amounts exceeding 20 pp of GDP in years prior to 2000.
  - Similar dynamics observed in Togo.

### Methodological challenge and identification approach
- Endogeneity/selection bias concern:
  - LICs request IMF programs when external financing needs rise and alternative sources are unavailable.
  - Common drivers (recessions, current account deficits, terms of trade shocks, reform-minded governments) affect both IMF participation and ODA.
- Identification approach:
  - Instrumental Variables (IV) strategy to address selection into programs and program-size bias.

### Instruments and rationale
- IV 1 (addresses bias from selection into programs):
  - Construction: IMFProbability_it × log(IMFLiquidity_t).
    - IMFProbability_it = past number of years under IMF programs (20-year rolling probability).
    - IMFLiquidity_t = (log of) disposable amounts of IMF resources available for lending (GRA / FCC).
  - Rationale: past participation predicts future participation; interaction with IMFLiquidity leverages IMF resource availability variation that is plausibly exogenous to idiosyncratic LIC shocks.
  - Empirical correlations reported:
    - Correlation between uncommitted PRGT assets and the GRA: 0.57 (years with overlapping data 2001-2019).
    - Correlation between non-borrowed PRGT resources and available liquidity in the GRA: 0.7.
- IV 2 (addresses bias related to program size):
  - Construction: AccessLimits_it / GDP_it = IMFQuota_it × Limits_t / GDP_it.
    - IMFQuota_it = country i’s IMF quota in year t in USD.
    - Limits_t = maximum percent of quota borrowable in a single arrangement without triggering exceptional access.
  - Rationale: AccessLimits/GDP reflects potential borrowing capacity in GDP terms; quotas and limits change periodically and are largely orthogonal to short-run ODA drivers.

### Empirical specifications (summary)
- Baseline 2SLS second-stage (equation (3)):
  - log(ODAit/GDPit) = β1 · (Disbursementit/GDPit) + β2 · IMFProbabilityit + δi + γt + εit
  - β1 interpretable as percent change in ODA flows triggered by a one pp change in annual disbursements (total catalytic effect).
- First-stage (equation (4)):
  - Disbursementit/GDPit = α1 · (IMFProbabilityit × IMFLiquidityt) + α2 · IMFProbabilityit + α3 · (AccessLimitsit/GDPit) + δi + γt + uit
  - Expected signs: α2 > 0, α1 < 0, α3 > 0.
- Average marginal effects (AME) computed from equation (5): reported in pp of GDP and computed at sample means and over time.

### Key empirical results (Table 1 highlights)
- Sample: annual data 1990–2019 for 63 countries; observations = 17,461.
- OLS estimates:
  - Pooled OLS (column 1): Disbursement/GDP coefficient = 0.136*** (0.046); Marginal Effects = 0.69 (pp of GDP).
  - FE OLS (column 3): Disbursement/GDP coefficient = 0.070** (0.029); Marginal Effects = 0.35 (pp of GDP).
- 2SLS estimates:
  - Instrumenting with IMFProbability × IMFLiquidity (column 4): Disbursement/GDP coefficient = 0.612*** (0.204); Marginal Effects = 3.0 (pp of GDP); Kleiberg-Paap Wald F-stat = 14.24.
  - Instrumenting with AccessLimit/GDP (column 5): Disbursement/GDP coefficient = 0.528*** (0.161); Marginal Effects = 2.67 (pp of GDP); Kleiberg-Paap Wald F-stat = 13.16.
  - Preferred specification using both instruments (column 6): Disbursement/GDP coefficient = 0.534*** (0.158); Marginal Effects = 2.70 (pp of GDP); First-stage F-stat = 13.16; Hansen J-stat (p-value) = 0.63.
- Interpretation:
  - An increase in disbursements of 1 pp of GDP catalyzes ODA by 2.7 pp of GDP (53 percent increase in ODA-to-GDP ratio).
  - Average yearly disbursements (in percent of GDP) = 0.8.
  - Using preferred 2SLS estimate, a program-requesting country could expect to receive additional ODA of 2.16 pp of GDP through each year under the program (calculated from 0.8 × 2.7 pp).
- Time evolution:
  - AME broadly constant at around 2.7 pp of GDP in LICs during 1990–2019 with a slightly larger impact in 1990 (not statistically different).

### Decomposition: intensive vs extensive margins (Section 5.2, Table 2)
- Specification decomposes effects into:
  - Intensive margin: Disbursement/GDP (instrumented with AccessLimits/GDP).
  - Extensive margin: IMFParticipation dummy (instrumented with IMFProbability × IMFLiquidity).
- Main results:
  - OLS baseline (column 1): Disbursement/GDP = 0.070** (0.029) — Marginal Effect = 0.35 (pp of GDP).
  - Adding IMFParticipation (column 2): Disbursement/GDP = 0.044** (0.021); IMFParticipation = 0.284*** (0.031) — Marginal Effect = 0.22 (pp of GDP).
  - 2SLS isolating disbursement (column 3): Disbursement/GDP = 0.539*** (0.177) — Marginal Effect = 2.72 (pp of GDP).
  - 2SLS instrumenting both margins (column 4): Disbursement/GDP = 0.502*** (0.184); IMFParticipation = 0.199 (0.418) — Marginal Effect = 2.53 (pp of GDP).
- Interpretation: Intensive margin (amounts disbursed) is the primary driver of catalysis; IMFParticipation dummy is less informative once disbursement size is accounted for.

### Decomposition by donor and aid type (Section 5.3)
- Donor-type catalytic effects (IV estimates):
  - Total: Disbursement/GDP = 0.534*** (0.158); Marginal Effect = 2.70 (pp of GDP).
  - Traditional bilateral: 0.331*** (0.109); Marginal Effect = 0.65 (pp of GDP).
  - Multilateral organizations: 0.558*** (0.166); Marginal Effect = 1.40 (pp of GDP).
  - Non-traditional bilateral: 1.271** (0.498); Marginal Effect = 0.05 (pp of GDP).
- Multilateral breakdown (Table 4):
  - World Bank Group: 0.871*** (0.240); Marginal Effect = 0.72 (pp of GDP).
  - EU: 0.339** (0.113); Marginal Effect = 0.17 (pp of GDP).
  - Regional Development Banks: 0.162 (0.140); Marginal Effect = -0.04 (pp of GDP) — not significant.
  - UN agencies: 0.227** (0.113); table formatting ambiguous; text notes a small amount of ODA from the UN around 0.04 pp of GDP.
- Aid-type decomposition (Table 5):
  - Concessional loans: Disbursement/GDP = 0.531*** (0.192); Marginal Effect = 0.66 (pp of GDP).
  - Grants: Disbursement/GDP = 0.372*** (0.116); Marginal Effect = 1.17 (pp of GDP).
- Interpretation:
  - Multilateral organizations and traditional bilaterals are main contributors to catalytic effects.
  - Catalytic effect pertains more to grants (1.17 pp of GDP) than concessional loans (0.66 pp of GDP).

### Within-program heterogeneity and non-linearities (Section 5.4, Table 6)
- Sample: countries actively receiving disbursements (51 countries; 1990–2019).
- Key coefficients:
  - IV (linear) (col 2): Disbursement/GDP = 0.257*** (0.097); Marginal Effect = 1.80 (pp of GDP).
  - IV with quadratic (col 4): Disbursement/GDP = 0.662*** (0.135); (Disbursement/GDP)^2 = -0.026*** (0.007).
- Interpretation:
  - Evidence of an inverted U-shape: larger disbursements catalyze more ODA but with diminishing marginal returns.
  - Conservative threshold where additional disbursements could reduce catalytic effects in absolute terms: around 12 pp of GDP annual disbursements (not reached in practice).

### Data sources and construction (Section 4)
- ODA: OECD Development Assistance Committee dataset (OECD, 2023); dependent variable = gross disbursements of ODA.
  - Exclusions: food and humanitarian aid, technical cooperation, debt forgiveness and debt rescheduling’s forgiven and rescheduled, and all loans and debt relief provided by the IMF.
  - Loan qualification: grant element of at least 25 percent calculated at a discount rate of 10 percent to qualify as ODA.
- IMF disbursements: Finance Department of the IMF (accounting records); recorded in SDR and transformed to USD at recorded SDR/USD exchange rate.
  - Program coverage: Upper Credit Tranche quality programs (Stand-by and Extended Credit Facilities and older versions); excludes emergency financing programs.
- Instruments and constructions:
  - IMFProbabilityit: 20-year rolling probability IMFProbabilityit = (1/20) Στ=t−19^t IMFParticipationiτ.
  - IMFLiquidity: measured using IMF’s forward commitment capacity (FCC).
  - AccessLimits: interaction between IMF quotas and cumulative access limits to the PRGT; sourced from historic IMF records.
- Controls and additional data:
  - GDP growth and current account balance: IMF World Economic Outlook.
  - Capital account openness: Chinn-Ito index (Chinn and Ito, 2023).
  - Political stability proxy: percentage of veto players who drop from a government in a given year (World Bank Database on Political Institutions, World Bank, 2020).
- Sample:
  - Final data sample covers 63 LICs from 1990 to 2019 (list of 63 countries provided in Appendix A).
  - Focus on LICs eligible to receive concessional loans from the PRGT in 2019; excludes countries that graduated from PRGT eligibility prior to 2019.

### Sensitivity analyses (Section 6 and Appendices)
- Instrument exclusion and robustness checks (Table 7):
  - Disbursement/GDP coefficient across augmented specifications remains robust:
    - Column 1: 0.534*** (0.158) — Marginal Effect = 2.70.
    - Column 2 (adds Real GDP growth): 0.537*** (0.175) — Marginal Effect = 2.70.
    - Column 3 (adds Current Account balance): 0.534*** (0.159) — Marginal Effect = 2.99.
    - Column 4 (adds Capital Account Openness): 0.601*** (0.155) — Marginal Effect = 2.67.
    - Column 5 (adds Political Stability): 0.501*** (0.149) — Marginal Effect = 2.51.
    - Column 6 (all controls): 0.569*** (0.157) — Marginal Effect = 2.75.
- Special events and exclusions (Appendix D):
  - Results robust to excluding GFC years (2008–2010), excluding precautionary programs, and controlling for IMF-provided debt relief.
  - Regional heterogeneity:
    - Excluding Sub-Saharan Africa reduces catalytic effect from 2.7 to 1.3 pp of GDP (larger effects in Sub-Saharan Africa).
  - Export-type heterogeneity:
    - Results not driven by any particular export-earnings group.
- GDP denominator robustness (lagged GDP) yields preferred IV Marginal Effects around 2.58–3.71 in alternative specifications.
- Tests for spurious correlations (Appendix B):
  - Splitting LICs by IMFProbability percentile shows no apparent spurious trend between group ODA means and IMF liquidity (FCC).

### Conclusions (Section 7: summary of main findings)
- Quantified catalytic impact:
  - An increase in IMF disbursements of 1 pp of GDP catalyzes additional ODA flows worth 2.7 pp of GDP.
  - Given typical IMF disbursement sizes (average yearly disbursements = 0.8 pp of GDP), on average a LIC engaged in an IMF-supported program can expect to receive an additional 2.2 pp of GDP annually in ODA (calculated as 0.8 × 2.7 = 2.16 pp).
- Donor and modality breakdown:
  - Multilateral organizations (World Bank and EU) and traditional bilateral donors are primary contributors to the catalytic effects.
  - Catalytic impact applies more to grants than concessional loans (grants: 1.17 pp of GDP; concessional loans: 0.66 pp of GDP).
- Margin dominance and non-linearities:
  - The intensive margin (disbursement size) dominates the extensive margin (program participation).
  - Catalytic effects are declining in the size of IMF programs (decreasing marginal effects); conservative threshold where additional disbursements could reduce catalytic effects in absolute terms is around 12 pp of GDP annual disbursements (far from being reached in practice).

*Source: wpiea2024134-print-pdf*

### Section 4 details the data employed for the econometric analysis.  Section 5 describes the

### wpiea2024134-print-pdf - Section 4 details the data employed for the econometric analysis.  Section 5 describes the

### Overview
- Sections layout: Section 4 details the data; Section 5 describes the main econometric results; Section 6 presents a sensitivity analysis; Section 7 provides a summary and conclusions.
- Focus: stylized facts on ODA and IMF lending to LICs, correlation between IMF program participation and ODA, and an IV-based econometric identification strategy to estimate the IMF’s (ex-post) catalytic effect on ODA.

### Stylized facts on ODA and IMF lending to LICs
- ODA is a principal source of external financing for LICs with limited access to private capital markets and low domestic savings.
- Aggregate and relative trends:
  - ODA flows to LICs increased from USD 13.4 billion to 52.4 billion from 1990-2019.
  - For the average LIC, ODA fell from 12.7 to 6.3 pp of GDP over 1990-2019.
  - ODA flows have consistently remained above 5 pp of GDP.
- Composition shifts (1990 to 2019):
  - Traditional bilateral donors’ share fell from 63 to 43 percent of ODA.
  - Multilateral donors’ share rose from 32 to 51 percent of ODA.
  - Non-traditional bilateral donors’ share increased since 2009 but remained 4 percent of total ODA flows in 2019.
- External financing composition for LICs in 2019:
  - ODA: 28 percent of external financing flows.
  - Remittances: 48 percent.
  - FDI: 21 percent.
  - Private portfolio flows: 4 percent.

### IMF concessional lending and the catalytic concept
- Purpose: IMF concessional lending to LICs assists countries with protracted balance of payments needs and supports implementation of policy actions to restore external balance and support growth and poverty reduction.
- Program design and disbursements:
  - Program size, length, disbursement schedule, and policy actions are agreed prior to Executive Board approval.
  - Disbursements can be front-, even-, or back-loaded depending on country circumstances and phasing of adjustment measures.
  - Program reviews verify programs are “on-track”; off-track programs lead to stopped disbursements.
- IMF Financing Assurances Policy:
  - Approval and reviews require the financing assurances policy to hold: the program must be (ex-ante) fully financed for the next 12 months with good prospects for the remainder of the program period.
  - Indications of donor financing are required but are not binding; some pledges may not materialize and some donors may provide financing without prior announcement.
- Definition used in this study:
  - Catalytic effect is defined as the (ex-post) materialized support from the donor community in response to IMF financing (i.e., actual ODA flows realized after IMF program approval).

### Channels through which IMF programs may catalyze ODA
- Two main channels:
  - Conditionality channel: IMF program conditionality signals authorities’ intentions and strengthens the recipient government’s ability to withstand shocks, increasing donors’ perception of aid effectiveness.
  - Liquidity channel: IMF program liquidity eases adjustment burdens, reinforcing donor willingness to provide financing.
- IMF programs act as a coordination mechanism among donors.

### Correlation evidence between IMF programs and ODA
- Cross-sectional distribution:
  - LICs engaged in an IMF-supported program receive on average 10.2 pp of GDP in annual ODA.
  - LICs not in IMF programs receive on average 5.8 pp of GDP in annual ODA.
  - ODA-to-GDP ratio displays wide variation across LICs with a long right tail.
- Country-level time series examples:
  - Guinea-Bissau participated in five IMF programs between 1990 and 2019; ODA-to-GDP ratio is on average higher in program periods than off-program years, with amounts exceeding 20 pp of GDP in years prior to 2000.
  - Similar dynamics observed in Togo.

### Methodological challenge
- Simple OLS of ODA flows on IMF financing would likely be biased:
  - LICs request IMF programs when external financing needs rise and alternative sources are unavailable.
  - The same drivers (recessions, widening current account deficits, adverse terms of trade shocks, new reform-minded governments) can increase both IMF program participation and ODA, creating endogeneity/selection bias.
- Identification approach: Instrumental Variables (IV) strategy to address selection into programs and bias from program size.

### Identification strategy and instruments
- Two IVs built to address two biases: selection into programs and program size bias. Approach follows Lang (2021), Gehring and Lang (2020), and Krahnke (2023).

- IV 1 (addresses bias from selection into programs):
  - Construction:
    - Interaction: IMFProbability_it × log(IMFLiquidity_t). (Equation labeled (1))
    - IMFProbability_it = past number of years under IMF programs (country-specific probability of past participation).
    - IMFLiquidity_t = (log of) disposable amounts of IMF resources available for lending (from the IMF’s General Resources Accounts, GRA).
  - Rationale:
    - Past participation predicts future participation; interaction with IMFLiquidity adds predictive power because the IMF’s willingness to lend varies with available resources.
    - Using GRA liquidity (rather than PRGT liquidity) reinforces exogeneity because GRA is largely quota-financed and driven by factors like quota reviews and predetermined large repayments.
  - Exogeneity and exclusion restriction:
    - Variations in GRA availability are largely driven by factors independent of idiosyncratic LIC shocks (e.g., quota reviews, large repayments), making IMFProbability × IMFLiquidity plausibly exogenous conditional on IMFProbability.
  - Additional notes:
    - The paper controls for other drivers of ODA in baseline results and shows in Appendix B that the IV method does not suffer from spurious trends.
  - Empirical correlations reported:
    - Correlation between uncommitted PRGT assets and the GRA: 0.57 (years with overlapping data 2001-2019).
    - Correlation between non-borrowed PRGT resources and available liquidity in the GRA: 0.7.

- IV 2 (addresses bias related to program size):
  - Construction:
    - AccessLimits_it / GDP_it = IMFQuota_it × Limits_t / GDP_it. (Equation labeled (2))
    - IMFQuota_it = size of country i’s IMF quota in year t expressed in USD.
    - Limits_t = the maximum amount of resources (in percent of quota) that a country can borrow in a single IMF arrangement without triggering exceptional access policies.
  - Interpretation:
    - AccessLimits/GDP reflects the real amount (in GDP terms) of lending a country can potentially borrow in a single arrangement.
    - AccessLimits is an institutional nominal cap (in SDR) and is a strong predictor of program size and disbursements.
  - Exogeneity and exclusion restriction:
    - Quotas and Limits are reviewed periodically (at least every five years) and follow formulas tied to macro characteristics like economy size and trade openness, making revisions largely orthogonal to short-run macro drivers of ODA.
    - Limits apply equally to all PRGT-eligible LICs and can be adjusted faster than quotas in response to demand (e.g., during COVID-19).

### Empirical strategy summary
- Use IV1 to predict program participation (selection bias).
- Use IV2 to predict program size and disbursements (size bias).
- Control for other drivers of ODA in baseline specifications and conduct sensitivity analyses (Section 6) to assess robustness.

*Source: wpiea2024134-print-pdf*

### 3.2    Econometric Specifications

### 3.2    Econometric Specifications

### 3.2.1    Total catalytic effects
- Baseline 2SLS second-stage specification (equation (3)):
  - log(ODAit/GDPit) = β1 · (Disbursementit/GDPit) + β2 · IMFProbabilityit + δi + γt + εit
  - ODA/GDP: total flows of ODA (in percent of nominal GDP) that country i received in year t.
  - Disbursement/GDP: total IMF disbursements measured in percent of nominal GDP borrower i received in year t.
  - IMFProbabilityit: country i’s rolling probability of requesting an IMF program based on past IMF programs.
  - δi and γt: country and year fixed effects.
- Interpretation:
  - β1 measures the percentage change in ODA flows triggered by a one pp change in annual disbursements; it captures the total catalytic effect (intensive + extensive margins).
- First-stage specification (equation (4)):
  - Disbursementit/GDPit = α1 · (IMFProbabilityit × IMFLiquidityt) + α2 · IMFProbabilityit + α3 · (AccessLimitsit/GDPit) + δi + γt + uit
  - IMFProbability included in both stages; the sole source of exogenous variation for program participation is the interaction IMFProbability × IMFLiquidity.
  - Expected signs:
    - α2 > 0 (past participation raises current probability of requesting a new program).
    - α1 < 0 (when IMF resources are low (high), more (less) frequent users request programs).
    - α3 > 0 (higher AccessLimits/GDP increases capacity to borrow, leading to larger disbursements).
- Average marginal effects (AME) in pp of GDP computed from equation (5):
  - AME = ∂E[ODA/GDP]/∂(Disbursement/GDP) = ∂ exp(β1 · (Disbursement/GDP) + β2 · IMFProbability + bδi + bγt) / ∂(Disbursement/GDP)
  - AME computed at sample means and over time to express catalytic effects in pp of GDP.

Key empirical results (Table 1, columns referenced as in source):
- OLS estimates:
  - Column (1) pooled OLS: Disbursement/GDP coefficient = 0.136*** (0.046)
    - Marginal Effects (in pp of GDP) = 0.69
  - Column (3) with country and year fixed effects: Disbursement/GDP coefficient = 0.070** (0.029)
    - Marginal Effects (in pp of GDP) = 0.35
- 2SLS (instrumental variables) estimates:
  - Column (4) instrumenting with IMFProbability × IMFLiquidity:
    - Disbursement/GDP coefficient = 0.612*** (0.204)
    - Marginal Effects (in pp of GDP) = 3.0
    - Kleiberg-Paap Wald F-stat = 14.24
  - Column (5) instrumenting with AccessLimit/GDP:
    - Disbursement/GDP coefficient = 0.528*** (0.161)
    - Marginal Effects (in pp of GDP) = 2.67
    - Kleiberg-Paap Wald F-stat = 13.16
  - Column (6) using both instruments (preferred specification):
    - Disbursement/GDP coefficient = 0.534*** (0.158)
    - Marginal Effects (in pp of GDP) = 2.70
    - First-stage F-stat = 13.16 (reported under columns)
    - Hansen J-stat (p-value) = 0.63 (over-identification cannot be rejected)
    - Interpretation: an increase in disbursements of 1 pp of GDP catalyzes ODA by 2.7 pp of GDP (53 percent increase in ODA-to-GDP ratio).
- Additional first-stage findings:
  - IMFProbability coefficient on Disbursement/GDP: positive in columns 4 and 6 (IMFProbability = 0.145*** in one specification).
  - IMFProbability × IMFLiquidity interaction: negative and highly significant (consistent with intuition about liquidity-driven selection).
  - AccessLimit/GDP: positive and significant (0.095*** in one specification).
- Sample and estimation notes:
  - Panel: annual data 1990–2019 for 63 countries; observations = 17,461.
  - Dependent variable log-transformed due to right-skewness and to obtain normally distributed residuals.
  - The 2SLS indicates a downward bias in OLS estimates.
  - Average yearly disbursements (in percent of GDP) = 0.8.
  - Using the preferred 2SLS estimate, a program-requesting country could expect to receive additional ODA of 2.16 pp of GDP through each year under the program (calculated from 0.8 × 2.7 pp).

- Time evolution of catalytic effects (Figure 5):
  - Using column 6 coefficients, AME computed over time shows catalytic effects of IMF programs on ODA broadly constant at around 2.7 pp of GDP in LICs during 1990–2019.
  - Slightly larger catalytic impact in 1990, but not statistically different from other years.

### 3.2.2    Catalytic effects decomposed by margin
- Second-stage specification for decomposition (equation (6)):
  - log(ODAit/GDPit) = β1 · (Disbursementit/GDPit) + β2 · IMFParticipationit + β3 · IMFProbabilityit + δi + γt + εit
  - IMFParticipationit: dummy = 1 when country is in an active IMF program.
  - Interpretation:
    - β1 quantifies catalytic impact of disbursements (intensive margin) controlling for program participation.
    - β2 measures catalytic effect of participation in programs (extensive margin) conditional on disbursement amounts.
- Instrumenting strategy for decomposition:
  - IMFParticipation instrumented with IMFProbability × IMFLiquidity.
  - Disbursement/GDP instrumented with AccessLimits/GDP.
  - One instrument employed at a time to identify each margin’s catalytic effect.

### 3.2.3    Decomposition of ODA and non-linear effects
- Decompositions performed by donor groups and aid type:
  - Donor groups distinguished:
    - i) traditional bilateral donors,
    - ii) non-traditional bilateral donors,
    - iii) multilateral organizations.
  - Traditional donors refer to countries with OECD membership.
  - Multilateral category further decomposed into: European Union (EU), Regional Development Banks, the United Nations, the World Bank, and other multilateral organizations.
  - Aid types decomposed between concessional loans and grants.
- Non-linearity (returns to scale) analysis (equation (7)):
  - Conditional on active IMF programs (IMFProgram = 1), second-stage specification:
    - log(ODAit/GDPit | IMFProgram = 1) = β1 · (Disbursementit/GDPit) + β2 · (Disbursementit/GDPit)^2 + δi + γt + εit
  - Objective: determine whether size of disbursement has increasing, constant, or decreasing returns to scale by including a quadratic term and conditioning on strictly positive disbursement years.
  - Instrument: AccessLimits/GDP (single IV) used because the IMFProbability × IMFLiquidity interaction is not usable under this conditional sample.
  - Estimation approach for quadratic with one IV follows Wooldridge (2010): linear projection of Disbursements/GDP on AccessLimits/GDP, square the projection, and use it as an instrument for (Disbursement/GDP)^2.

### 4    Data (data sources and construction relevant to econometric specifications)
- ODA data:
  - Source: OECD Development Assistance Committee dataset (OECD, 2023).
  - Dependent variable: gross disbursements of ODA.
  - Exclusions from ODA: food and humanitarian aid, technical cooperation, debt forgiveness and debt rescheduling’s forgiven and rescheduled, and all loans and debt relief provided by the IMF.
  - Loan qualification: grant element of at least 25 percent calculated at a discount rate of 10 percent to qualify as ODA.
- IMF disbursements:
  - Source: Finance Department of the IMF (accounting records of actual financial transactions).
  - Yearly measurement: sum of all disbursements to each country in a calendar year; augmentations recorded in disbursement variable.
  - Currency: recorded in SDR and transformed to USD at recorded SDR/USD exchange rate.
  - Program coverage: only Upper Credit Tranche quality programs (Stand-by and Extended Credit Facilities and older versions); excludes emergency financing programs.
  - Less than one percent of sample are disbursements to blended borrowers.
- Instruments and constructions:
  - IMFProbabilityit:
    - 20-year rolling probability: IMFProbabilityit = (1/20) Στ=t−19^t IMFParticipationiτ
    - IMFParticipationit = 1 if country i is in an IMF program in year t, 0 otherwise.
  - IMFLiquidity:
    - Measured using IMF’s forward commitment capacity (FCC) as in Krahnke (2023); FCC measures annual available resources from the GRA adjusted for already committed resources and prudential balances.
  - AccessLimits:
    - Interaction between IMF quotas and cumulative access limits to the PRGT; sourced from historic IMF records on country quotas and cumulative access limits to PRGT resources.
    - Cumulative Access Limits are caps on borrowing amounts for each country, as a percent of quota, before triggering Exceptional Access policies.
- Additional controls and sensitivity data:
  - GDP growth and current account balance: IMF World Economic Outlook indicators.
  - Degree of capital account openness: Chinn-Ito index (Chinn and Ito, 2023).
  - Political stability proxy: percentage of veto players who drop from a government in a given year (World Bank Database on Political Institutions, World Bank, 2020). The control variable is a share with 0 representing no exits and 1 representing exit and replacement of all veto players.
- Sample:
  - Final data sample covers 63 LICs from 1990 to 2019.
  - Focus on LICs eligible to receive concessional loans from the PRGT in 2019; excludes countries that graduated from PRGT eligibility prior to 2019.

*Source: wpiea2024134-print-pdf - 3.2    Econometric Specifications*

### 5.2    Decomposing the Catalytic Effects by the Intensive and Extensive

### 5.2    Decomposing the Catalytic Effects by the Intensive and Extensive Margins

### Key empirical approach
- Two IVs are used to separately identify: (i) the extensive margin (IMFParticipation) and (ii) the intensive margin (Disbursement/GDP).
- IMFParticipation is instrumented with IMFProbability × IMFLiquidity (eq. 1).
- Disbursement/GDP is instrumented with AccessLimits/GDP (eq. 2).
- Panel: annual data between 1990 and 2019 for 63 countries. Country and year fixed effects included. Robust standard errors. Kleiberg-Paap Wald F-stat reported.

### Main estimation results (Table 2 highlights)
- OLS baseline (column 1): Disbursement/GDP coefficient = 0.070** (0.029).
- Adding IMFParticipation (column 2): Disbursement/GDP = 0.044** (0.021); IMFParticipation = 0.284*** (0.031).
- 2SLS isolating disbursement (column 3): Disbursement/GDP = 0.539*** (0.177) when instrumenting Disbursement/GDP with AccessLimits/GDP (IMFParticipation not instrumented).
- 2SLS instrumenting both margins (column 4): Disbursement/GDP = 0.502*** (0.184); IMFParticipation = 0.199 (0.418) (not statistically significant at 10 percent).
- Marginal effects (in pp of GDP) reported: column 1 = 0.35; column 2 = 0.22; column 3 = 2.72; column 4 = 2.53.
- Interpretation: amounts disbursed (intensive margin) play a more important role in attracting ODA than the binary program participation dummy (extensive margin).

---

### 5.3    Decomposing the Catalytic Effects by Donor and Aid Type

### Catalytic effects by donor type (Table 3)
- IV estimates regressing log(ODA/GDP) on Disbursement/GDP with donor-type breakdown; panel 1990–2019, 63 countries.
- Disbursement/GDP coefficients (with standard errors):
  - Total (column 1): 0.534*** (0.158); Marginal Effect = 2.70 (pp of GDP).
  - Traditional bilateral (column 2): 0.331*** (0.109); Marginal Effect = 0.65 (pp of GDP).
  - Multilateral organizations (column 3): 0.558*** (0.166); Marginal Effect = 1.40 (pp of GDP).
  - Non-traditional bilateral (column 4): 1.271** (0.498); Marginal Effect = 0.05 (pp of GDP).
- Interpretation:
  - A 1 pp increase in disbursements (in percent of GDP) leads to an average increase of ODA of 0.65 pp of GDP from traditional bilaterals.
  - Largest percent impact on multilateral organizations: a 1 pp increase in disbursements raises ODA from multilaterals by 56 percent; in level terms, 1.4 pp of GDP.
  - Very large percent change for non-traditional bilaterals (127 percent) but small AME (0.05 pp of GDP) due to their small baseline ODA.

### Catalytic effects by multilateral organization (Table 4)
- Disbursement/GDP coefficients (with standard errors) and marginal effects:
  - Total multilateral (col 1): 0.558*** (0.166); Marginal Effect = 1.40 (pp of GDP).
  - EU (col 2): 0.339** (0.113); Marginal Effect = 0.17 (pp of GDP).
  - Regional Development Banks (col 3): 0.162 (0.140); Marginal Effect = -0.04 (pp of GDP) — not significant.
  - UN agencies (col 4): 0.227** (0.113); Marginal Effect = 0.72 (pp of GDP)? (table reports 0.720.04 formatting; text states IMF programs catalyze a small amount of ODA from the UN at around 0.04 pp of GDP).
  - World Bank Group (col 5): 0.871*** (0.240); Marginal Effect = 0.72 (pp of GDP).
  - Others (col 6): 0.231* (0.139); Marginal Effect = 0.04 (pp of GDP).
- Interpretation:
  - Positive and significant catalytic effect on EU and World Bank Group contributions.
  - No significant catalytic effect detected for Regional Development Banks (RDBs); possible explanation: RDBs provide more project financing than budget support.

### Catalytic effects by aid type (Table 5)
- Disbursement/GDP coefficients:
  - Total (col 1): 0.534*** (0.158); Marginal Effect = 2.70 (pp of GDP).
  - Concessional loans (col 2): 0.531*** (0.192); Marginal Effect = 0.66 (pp of GDP).
  - Grants (col 3): 0.372*** (0.116); Marginal Effect = 1.17 (pp of GDP).
- Interpretation:
  - A 1 pp increase in disbursements increases concessional loans by 53.1 percent and grants by 37.2 percent.
  - In level terms, a 1 pp increase in disbursements catalyzes on average 0.66 pp of GDP in concessional loans and 1.17 pp of GDP in grants.
  - LICs receive more grants than concessional loans from the catalytic effect.

---

### 5.4    Decomposing the Catalytic Effects within IMF Programs

### Role of disbursement size (Table 6)
- Sample restricted to countries actively receiving disbursements (51 countries; 1990–2019).
- OLS and IV estimates of log(ODA/GDP) on Disbursement/GDP and (Disbursement/GDP)^2 with country and year FE.
- Key coefficients and marginal effects:
  - OLS (col 1): Disbursement/GDP = 0.0272** (0.013); Marginal Effect = 0.19 (pp of GDP).
  - IV (col 2): Disbursement/GDP = 0.257*** (0.097); Marginal Effect = 1.80 (pp of GDP).
  - OLS with quadratic (col 3): Disbursement/GDP = 0.101*** (0.032); (Disbursement/GDP)^2 = -0.003*** (0.001).
  - IV with quadratic (col 4): Disbursement/GDP = 0.662*** (0.135); (Disbursement/GDP)^2 = -0.026*** (0.007).
- Interpretation:
  - OLS suffers downward bias; 2SLS shows a stronger effect: a 1 pp increase in disbursements increases ODA flows by 26 percent, amounting to 1.8 pp of GDP.
  - Evidence of a quadratic (inverted U-shape) relationship: larger disbursements catalyze more ODA, but marginal effects diminish.
  - Conservative estimate of threshold where additional disbursements could reduce catalytic effects in absolute terms: around annual disbursements worth 12 pp of GDP (not reached in practice).

---

### 6    Sensitivity Analysis

### Tests of instrument exclusion restrictions and robustness
- Baseline specification (3) augmented with typical determinants of ODA to test whether IVs correlate with other ODA drivers.
- Controls added across specifications in Table 7: Real GDP growth; Current Account balance; Capital Account Openness (Chinn and Ito index); Political Stability; and all controls together.
- Disbursement/GDP coefficient across columns remains:
  - Column 1: 0.534*** (0.158).
  - Column 2 (adds Real GDP growth): 0.537*** (0.175).
  - Column 3 (adds Current Account balance): 0.534*** (0.159).
  - Column 4 (adds Capital Account Openness): 0.601*** (0.155).
  - Column 5 (adds Political Stability): 0.501*** (0.149).
  - Column 6 (all controls): 0.569*** (0.157).
- Marginal Effects (in pp of GDP) range reported: 2.70, 2.70, 2.99, 2.67, 2.51, 2.75 across columns.
- Interpretation:
  - Inclusion of multiple standard determinants of ODA does not materially change the instrumented coefficient on Disbursement/GDP, supporting the validity of the IV strategy.
- Additional sensitivity analyses conducted (not tabulated here) include excluding countries by region and export structure, and using lagged GDP; results are presented in Appendix D (referenced).

---

### 7    Conclusions

### Summary of main findings
- The study quantifies a significant catalytic impact of IMF lending on ODA in LICs.
- Main quantified effects:
  - An increase in IMF disbursements of 1 pp of GDP catalyzes additional ODA flows worth 2.7 pp of GDP.
  - Given typical IMF disbursement sizes, on average a LIC engaged in an IMF-supported program can expect to receive an additional 2.2 pp of GDP annually in ODA.
- Donor and modality breakdown:
  - Multilateral organizations (World Bank and EU) and traditional bilateral donors are primary contributors to the catalytic effects.
  - The catalytic effect pertains more to grants than to concessional loans.
- Intensive margin dominates:
  - The amount of lending (intensive margin) is more statistically relevant for attracting ODA than mere program participation (extensive margin).
- Nonlinearities:
  - Catalytic effects are declining in the size of IMF programs (decreasing marginal effects).
  - A conservative estimate of the threshold where additional disbursements could reduce catalytic effects in absolute terms is around 12 pp of GDP annual disbursements (far from being reached in practice).

*Source: 5.2 Decomposing the Catalytic Effects by the Intensive and Extensive Margins (excerpt from wpiea2024134-print-pdf)*

### References

### References

### Key cited works
- Al-Sadiq, Ali J. (2015). “The Impact of IMF-Supported Programs on FDI in Low-income Countries”. In: IMF Working Papers 2015/157, pp. 1–38.
- Bird, Graham, Mumtaz Hussain, and Joseph Joyce (2004). “Many Happy Returns? Recidivism and the IMF”. In: Journal of International Money and Finance 23(2), pp. 231–51.
- Bird, Graham and Dane Rowlands (2000). “The Catalysing Role of Policy-Based Lending by the IMF and World Bank: Fact or Fiction?” In: Journal of International Development 12, pp. 951–73.
- Bird, Graham and Dane Rowlands (2007). “The IMF and the Mobilisation of Foreign Aid”. In: Journal of Development Studies 43(5), pp. 856–70.
- Chahine, Salim, Ugo Panizza, and Guilherme Suedekum (2024). “IMF Programs and Borrowing Costs: Does Size Matter?” In: CEPR Discussion Paper 19015, pp. 1–27.
- Chinn, Menzie David and Hiro Ito (2023). “Measuring Financial Integration: More Data, More Countries, More Expectations”. In: NBER Working Paper 31505, pp. 1–25.
- Christian, Paul and Christopher B. Barrett (2017). “Revisiting the Effect of Food Aid on Conflict: A Methodological Caution”. In: World Bank Policy Research Working Paper 8171, pp. 1–81.
- Cohen-Setton, Jérémie and Francesco Toni (2024). “The Catalytic Effect of IMF Emergency Financing during COVID-19: Evidence from Official and Private Capital Flows”. In: IMF Working Papers, forthcoming.
- Conway, Patrick (2007). “The Revolving Door: Duration and Recidivism in IMF Programs”. In: The Review of Economics and Statistics 89(2), pp. 205–20.
- Dabla-Norris, Era, Camelia Minoiu, and Luis-Felipe Zanna (2010). “Business Cycle Fluctuations, Large Shocks, and Development Aid: New Evidence”. In: IMF Working Papers 2010/240, pp. 1–39.
- Díaz-Cassou, Javier, Alicia García-Herrero, and Luis Molina (2006). “What Kind of Flows Does the IMF Catalyze and When?” In: Bank of Spain Working Paper 0617, pp. 1–72.
- Erce, Aitor and Daniel Riera-Crichton (2015). “Catalytic IMF? A Gross Flows Approach”. In: Federal Reserve Bank of Dallas Globalization and Monetary Policy Institute Working Paper 254, pp. 1–19.
- Gehring, Kai and Valentin Lang (2020). “Stigma or Cushion? IMF Programs and Sovereign Creditworthiness”. In: Journal of Development Economics 146(102507), pp. 1–14.
- Gündüz, Yasemin Bal and Masyita Crystallin (2018). “Do IMF-Supported Programs Catalyze Donor Assistance to Low-Income Countries?” In: The Review of International Organizations 13(3), pp. 359–93.
- International Monetary Fund (2023a). “2023 Handbook of IMF Facilities for Low-Income Countries”. In: IMF Policy Papers 2023/020, pp. 1–190.
- International Monetary Fund (2023b). “Sub-Saharan Africa: The Big Funding Squeeze”. In: Regional Economic Outlook Apr. 2023, pp. 1–20.
- Jensen, Nathan M. (2004). “Crisis, Conditions, and Capital: The Effect of International Monetary Fund Agreements on Foreign Direct Investment Inflows”. In: The Journal of Conflict Resolution 48(2), pp. 194–210.
- Krahnke, Tobias (2023). “Doing More with Less: The Catalytic Function of IMF Lending and the Role of Program Size”. In: Journal of International Money and Finance 135(102856), pp. 1–31.
- Lang, Valentin (2021). “The economics of the democratic deficit: The effect of IMF programs on inequality”. In: The Review of International Organizations 16(3), pp. 599–623.
- Maurini, Claudia and Alessandro Schiavone (2021). “The Catalytic Role of IMF Programs”. In: Bank of Italy Working Papers 1331, pp. 1–34.
- Mody, Ashoka and Diego Saravia (2006). “Catalysing Private Capital Flows: Do IMF Programmes Work as Commitment Devices?” In: Economic Journal 116(513), pp. 843–67.
- OECD (2023). “Official Development Assistance”. In: OECD International Development Statistics, database.
- Rabehajaina, Njato, Jean-Pierre Gueyie, and Komlan Sedzro (2023). “Determinants of Bilateral Official Development Assistance”. In: Applied Economics 55(54), pp. 6345–59.
- Schiavone, Alessandro and Claudia Maurini (2023). “The Catalytic Role of IMF Programs to Low Income Countries”. In: Bank of Italy Occasional Papers 782, pp. 1–29.
- Stock, James H. and Motohiro Yogo (2005). “Testing for Weak Instruments in Linear IV Regression”. In: Identification and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, pp. 80–108.
- Stubbs, Thomas H., Alexander E. Kentikelenis, and Lawrence P. King (2016). “Catalyzing Aid? The IMF and Donor Behavior in Aid Allocation”. In: World Development 78, pp. 511–28.
- Sturm, Jan-Egbert, Helge Berger, and Jackob De Haan (2005). “Which Variables Explain Decisions on IMF Credit? An Extreme Bounds Analysis”. In: Economics Politics 17(2), pp. 177–213.
- Woo, Byungwon (2013). “Conditional on Conditionality: IMF Program Design and Foreign Direct Investment”. In: Taylor Francis Journals 39(3), pp. 292–315.
- Wooldridge, Jeffrey M. (2010). Econometric Analysis of Cross Section and Panel Data (2nd ed.) MIT Press.
- World Bank (2020). “Database of Political Institutions”. In: database.
- World Bank (2024). “World Development Indicators”. In: database.

---

### Appendix A: Details on Sample Coverage

### A.1 Key variables and sources of data (Table A1)
- ODA: Coverage 1990-2019 — Source: OECD DAC database
- Disbursements: Coverage 1990-2019 — Source: IMF Financial Data Query Tool
- PRGT Participation: Coverage 1990-2019 — Source: IMF Financial Data Query Tool
- IMF Probability: Coverage 1971-2019 — Source: IMF Annual Financial Reports
- IMF Liquidity (FCC): Coverage 1990-2019 — Source: IMF
- Quotas: Coverage 1990-2019 — Source: IMF Financial Data Query Tool
- Cumulative Access Limits to the PRGT: Coverage 1990-2019 — Source: IMF
- Nominal GDP: Coverage 1990-2019 — Source: WB World Development Indicators

### A.2 List of 63 LICs in the baseline sample
- Afghanistan, Bangladesh, Benin, Burkina Faso, Burundi, Cambodia, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Democratic Republic of Congo, Republic of Congo, Cote d’Ivoire, Djibouti, Dominica, Ethiopia, The Gambia, Ghana, Grenada, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, Kenya, Kyrgyz Republic, Lao People’s Democratic Republic, Lesotho, Liberia, Madagascar, Malawi, Maldives, Mali, Mauritania, Republic of Moldova, Mozambique, Myanmar, Nepal, Nicaragua, Niger, Papua New Guinea, Rwanda, Samoa, Sao Tome & Principe, Senegal, Sierra Leone, Solomon Islands, Somalia, South Sudan, St. Lucia, St. Vincent and the Grenadines, Sudan, Republic of Tajikistan, Tanzania, Togo, Tonga, Uganda, Republic of Uzbekistan, Vanuatu, Republic of Yemen, Zambia, Zimbabwe.

### A.3 Composition of donors
- Traditional bilateral donors: Australia, Austria, Belgium, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Lithuania, Luxembourg, Netherlands, New Zealand, Norway, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, United Kingdom and the United States.
- Non-Traditional bilateral donors: Azerbaijan, Bulgaria, Chinese Taipei, Croatia, Cyprus, Estonia, Israel, Kazakhstan, Kuwait, Latvia, Liechtenstein, Lithuania, Malta, Monaco, Qatar, Romania, Russian Federation, Saudi Arabia, Thailand, Timor-Leste, Türkiye and United Arab Emirates.
- Multilateral Organizations:
  - EU Institutions
  - Regional Development Banks: African Development Bank (AfDB), Asian Development Bank (AsDB), Inter-American Development Bank (IDB), Asian Infrastructure Investment Bank (AIIB), Arab Bank for Economic Development in Africa (BADEA), Central American Bank for Economic Integration (CABEI), Caribbean Development Bank (CarDB), Council of Europe Development Bank (CEB), European Bank for Reconstruction and Development (EBRD), Islamic Development Bank (IsDB), North American Development Bank (NADB).
  - United Nations: Central Emergency Response Fund (CERF), Food and Agriculture Organisation (FAO), IFAD, International Atomic Energy Agency (IAEA), International Labour Organisation (ILO), Joint Sustainable Development GOals Fund (Joint SDG Fund), UN Capital Development Fund (UNCDF), UN Institute for Disarmament Research (UNIDIR), UN Peacebuilding Fund (UNPBF), UN Women, UN-AIDS, UNDP, UNECE, UNEP, UNFPA, UNHCR, UNICEF, United Nations Conference on Trade and Development (UNCTAD), United Nations Industrial Development Organization (UNIDO), UNRWA, UNTA, WFP, WHO-Strategic Preparedness and Response Plan (SPRP), World Health Organisation (WHO), World Tourism Organisation (UNWTO), WTO - International Trade Centre (ITC).
  - World Bank Group: International Bank for Reconstruction and Development (IBRD), International Development Association (IDA) and the International Finance Corporation (IFC).
  - Other Multilateral: Adaptation Fund, Arab Fund (AFESD), Asian Forest Cooperation Organisation (AFoCO), Center of Excellence in Finance (CEF), CGIAR, Climate Investment Funds (CIF), Global Alliance for Vaccines and Immunization (GAVI), Global Environment Facility (GEF), Global Fund, Global Green Growth Institute (GGGI), Green Climate Fund (GCF), International Centre for Genetic Engineering and Biotechnology (ICGEB), International Commission on Missing Persons (ICMP), Montreal Protocol, Nordic Development Fund (NDF), OPEC Fund for International Development (OPEC Fund), OSCE, World Organisation for Animal Health (WOAH).

---

### Appendix B: Checking for Spurious Correlations in the Identification Strategy

- Concern: First IV may suffer from spurious trends if countries with higher IMFProbability receive proportionally more ODA as IMF liquidity increases over time (Christian and Barrett, 2017).
- Test: Split LICs into two groups by IMFProbability percentile (below 50 percentile = low; above 50 percentile = high).
- Finding:
  - Figure A1 (described) shows no apparent overlap in long-run trends between either group’s ODA means and the IMF liquidity time series (logarithm of FCC).
  - Using other percentiles yields similar results.
- Conclusion: Evidence suggests absence of spurious trends in the IV approach.

Note: Figure A1 description — blue line: mean of ODA for low probability group; red line: mean of ODA for high probability group; green line: logarithm of the FCC variable. Source: Authors’ calculations based on OECD ODA database (OECD, 2023).

---

### Appendix C: Empirical Results (Additional Tables)

### C.1 First Stage Regressions of Table 2 (Table A2)
- Panel: annual data between 1990 and 2019 for 63 countries; Observations 1746.
- Columns (1) and (2) report first-stage OLS results with dependent variables Disbursement/GDP and IMFParticipation respectively.
- Coefficients (standard errors in parentheses):
  - IMFProbability: 0.009** (0.004) in (1); 0.012*** (0.002) in (2).
  - IMFProbability × IMFLiquidity: -0.004*** (0.001) in (1); -0.002*** (0.000) in (2).
  - AccessLimit/GDP: 0.094*** (0.026) in (1); 0.012*** (0.004) in (2).
- R2: 0.18 in (1); 0.41 in (2).
- Country FE: Yes; Year FE: Yes.
- Significance: *p <0.10, **p <0.05, ***p <0.01.

### Reduced-form evidence of catalytic effects and program size (Table A3)
- Dependent variable: log (ODA/GDP).
- Sample: observations with positive disbursements; panel 1990–2019 for 51 countries; Observations 702.
- Coefficients:
  - AccessLimit/GDP: 0.0425*** (0.00956) in (1); 0.0913*** (0.0102) in (2).
  - (AccessLimit/GDP)2: -0.000949*** (0.000177) in (2).
- Country FE: Yes; Year FE: Yes.

---

### Appendix D: Sensitivity Analysis (Additional Results)

### D.1 Special events (Table A4)
- IV estimates regressing log (ODA/GDP) on Disbursement/GDP; panel 1990–2019 for 63 countries.
- Instruments: IMFProbability × IMFLiquidity and AccessLimit/GDP.
- Columns:
  - Baseline (1): Disbursement/GDP 0.534*** (0.158); Marginal Effects 2.70 (pp of GDP); F-stat 8.23; Countries 63; Observations 1746.
  - Excluding GFC years (2008–2010) (2): Disbursement/GDP 0.510*** (0.149); Marginal Effects 2.59; F-stat 8.25; Observations 1631.
  - Excluding precautionary programs (3): Disbursement/GDP 0.533*** (0.159); Marginal Effects 2.70; F-stat 8.02; Observations 1738.
  - Controlling for IMF debt relief (4): Disbursement/GDP 0.523*** (0.155); IMF Debt Relief/GDP 0.039*** (0.011); Marginal Effects 2.64; F-stat 8.15; Observations 1746.
- Note: Baseline results robust to excluding GFC years, excluding precautionary programs, and controlling for IMF-provided debt relief.

### D.2 Regional and export-type heterogeneity (Tables A5, A6)
- Table A5: Excluding regions one at a time (panel 1990–2019 for 63 countries).
  - Baseline (None excluded): Disbursement/GDP 0.534*** (0.158); Marginal Effects 2.70; F-stat 8.23; Observations 1746.
  - Excluding Europe and Central Asia: 0.544*** (0.166); Marginal Effects 2.83; F-stat 8.77; Observations 1654.
  - Excluding Latin America and Caribbean: 1.205* (0.658); Marginal Effects 6.46; F-stat 3.90; Observations 1510.
  - Excluding Middle East North Africa: 0.530*** (0.158); Marginal Effects 2.68; F-stat 8.05; Observations 1746.
  - Excluding South Asia: 0.554*** (0.165); Marginal Effects 2.86; F-stat 7.63; Observations 1687.
  - Excluding Sub-Saharan Africa: 0.338*** (0.095); Marginal Effects 1.32; F-stat 11.66; Observations 1638.
- Interpretation: Removing Sub-Saharan Africa reduces the magnitude of the catalytic effect from 2.7 to 1.3 pp of GDP, implying larger catalytic effects in sub-Saharan African countries.

- Table A6: Excluding countries by main source of export earnings (panel 1990–2019 for 63 countries).
  - Baseline (None excluded): Disbursement/GDP 0.534*** (0.158); Marginal Effects 2.70; F-stat 8.23; Observations 1746.
  - Excluding Fuel: 1.012* (0.546); Marginal Effects 5.30; F-stat 5.34; Observations 1628.
  - Excluding Manufactures: 0.523*** (0.155); Marginal Effects 2.75; F-stat 8.14; Observations 1601.
  - Excluding Primary Products: 0.442*** (0.125); Marginal Effects 1.99; F-stat 18.65; Observations 1701.
  - Excluding Services: 0.498*** (0.156); Marginal Effects 2.40; F-stat 6.13; Observations 1220.
  - Excluding Diversified: 0.459*** (0.137); Marginal Effects 2.45; F-stat 6.86; Observations 1357.
- Interpretation: Results are not driven by any particular export-earnings group.

### D.3 GDP denominator (Table A7)
- Re-estimation using a one-year lag of GDP in variables measured in GDP terms; panel 1990–2019 for 63 countries; Observations 1682.
- Second stage — Dependent variable: log (ODA/GDP(t-1)):
  - OLS columns (1)-(3): Disbursement/GDP(t-1) coefficients 0.122*** (0.040), 0.074*** (0.026), 0.061** (0.026); Marginal Effects 0.64, 0.39, 0.32 (pp of GDP(t-1)).
  - IV columns (4)-(6): Disbursement/GDP(t-1) coefficients 0.701*** (0.263), 0.471*** (0.170), 0.487*** (0.171); Marginal Effects 3.71, 2.49, 2.58.
- First stage (selected):
  - IMFProbability: 0.011** (0.006) in Column 1; -0.005*** (0.002) in Column 2; 0.007 (0.005) in Column 3.
  - IMFProbability × IMFLiquidity: -0.003*** (0.001) in Column 2; -0.002*** (0.001) in Column 3.
  - AccessLimit/GDP(t-1): 0.086*** (0.027) in Column 2; 0.084*** (0.027) in Column 3.
- F-stat: 10.11, 10.14, 6.03 across first-stage specifications.
- Note: In the preferred specification (Column 6), an increase of one pp in disbursements (percent of GDP in time (t-1)) is associated with an additional ODA flows worth 2.58 pp of GDP in time (t-1), compared to 2.7 in Column 6 of Table 1.

### D.4 Intensive margin and government debt (Tables A8, A9)
- Approach: Use reduced-form AccessLimit/GDP as proxy for disbursements to assess interaction with government Debt/GDP.
- Table A8 (reduced-form OLS, 51 countries, Observations 702/681):
  - AccessLimit/GDP coefficients:
    - 0.0425*** (0.00956) in (1)
    - 0.0913*** (0.0102) in (2)
    - 0.0854*** (0.0105) in (3)
    - 0.110*** (0.0163) in (4)
  - (AccessLimit/GDP)2:
    - -0.000949*** (0.000177) in (2)
    - -0.000921*** (0.000179) in (3)
    - -0.00201*** (0.000546) in (4)
  - Debt/GDP: 0.000991* (0.000512) in (3); 0.000222 (0.000671) in (4).
  - Debt/GDP × (AccessLimit/GDP)2: 0.00000259** (0.00000115) in (4).
- Interpretation: Program size has a positive and significant relationship with outcomes; the square of program size has a negative and significant effect (concavity). The interaction term indicates the decreasing catalytic effects become “less concave” when debt levels are high.

- Table A9 (by high and low government debt levels; 51 countries; Observations 702, 346, 351):
  - Baseline OLS (1): Disbursement/GDP 0.101*** (0.0318); (Disbursement/GDP)2 -0.003*** (0.001).
  - Baseline IV (2): Disbursement/GDP 0.662*** (0.135); (Disbursement/GDP)2 -0.026*** (0.007).
  - Low Debt/GDP IV (3): Disbursement/GDP 1.153** (0.524); (Disbursement/GDP)2 -0.033** (0.015).
  - High Debt/GDP IV (4): Disbursement/GDP 0.758** (0.343); (Disbursement/GDP)2 -0.062** (0.0256).
- Interpretation: For both low- and high-debt groups, program size coefficients are positive and significant; squared program size coefficients are negative and significant.

---

*The Catalytic Impact of IMF lending on Official Development Assistance — Working Paper No. WP/2024/134*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024134-print-pdf.pdf_