## _wp15157

## Source details

**Canonical URL:** [_wp15157](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15157.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15157.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15157.pdf.json)

---

### Objective and contribution
- Main objective: empirically assess the impact of IMF-supported programs on FDI inflows.
- Focus: exclusively on low-income countries (LICs).
- Sample: unbalanced panel data for 73 LICs over the period 1980–2012.
- IMF facilities considered: Extended Credit Facility (ECF) and its two predecessors (PRGF, ESAF), Structural Adjustment Facility (SAF), Exogenous Shocks Facility (ESF), Standby Credit Facility (SCF), and Policy Supported Instrument (PSI) arrangements.
- Identification strategy: address self-selection into IMF programs using:
  - A two-step estimation method based on a treatment effect model.
  - A non-parametric estimation method based on propensity score matching (PSM).

### The empirical model and estimation methods
- Baseline linear specification:
  - y_{i,t} = x′_{i,t-1} β + δ D_{i,t-τ}^{IMF} + η_{i} + ε_{i,t}
  - y: FDI inflows as a percentage of the host country’s GDP.
  - D_{IMF}: dummy = 1 if country i participates in an IMF-supported program in a given year for at least five months, 0 otherwise.
  - All independent variables lagged one year.
- Selection issues and correction:
  - Counterfactual problem and endogeneity/selection of D_{IMF}.
  - Two econometric approaches:
    - Regression-based treatment effect model (Maddala (1983)), estimated by MLE or Heckman two-step.
    - Propensity Score Matching (PSM) with Nearest Neighbor, Radius, and Kernel matching.
- Treatment/selection equations:
  - Outcome equation controls: GDP per capita (log), real GDP growth, broad money (M2) as percent of GDP, (exports+imports)/GDP, inflation rate, FDI stock/GDP, political risk, institutional quality, corruption, democracy.
  - Selection equation determinants: level of development, real GDP growth, foreign reserves, inflation, current account balance, terms of trade, external debt services, years under past IMF programs, democratic institutions, world real GDP growth.

### Data sources and variable definitions
- Sample: 73 LICs, 1980–2012.
- Data sources: UNCTAD World Investment Report (2013); IMF World Economic Outlook (2013); World Bank WDI (2013); ICRG Political Risk Groups (2013); Freedom House (2013); IMF database on arrangements; IMF Finance Department.
- Key variable definitions preserved exactly:
  - FDI Inflows: Foreign Direct Investment inflows as a percentage of GDP.
  - FDI stock: end-period total FDI stock in the host countries as a percentage of GDP.
  - Financial Development: Money and quasi money (M2) as a percentage of GDP.
  - Openness: The sum of exports and imports of goods and services measured as a share of GDP.
  - IMF participation: a dummy variable equals one if a member country is under an IMF-supported programs at least five months of the year and zero otherwise.
  - Democratic institutions: index defined as [14 - (political rights + civil rights) / 12] ranging from 0 (least free) to 1 (most free).
- Appendix Table 1 summary statistics (selected):
  - FDI /GDP: Obs 2024, Mean 3.83, Std. Dev 6.42, Min -14.68, Max 90.46
  - IMF: Obs 2052, Mean 0.37, Std. Dev 0.48, Min 0.00, Max 1.00 (IMF =1: 758; IMF = 0: 1294)
  - Real GDP growth: Obs 2052, Mean 3.96, Std. Dev 6.05, Min -50.25, Max 71.19
  - log (GDP per capita): Obs 2052, Mean 6.49, Std. Dev 0.96, Min 4.28, Max 10.09
  - Inflation rate: Obs 2048, Mean 58.78, Std. Dev 705.55, Min -72.73, Max 23773
  - Openness: Obs 2052, Mean 76.10, Std. Dev 43.67, Min 4.19, Max 403.92
  - FDI stock /GDP: Obs 1979, Mean 27.42, Std. Dev 34.06, Min 0.03, Max 248.56
  - Period of IMF participations: Obs 2052, Mean 4.65, Std. Dev 5.47, Min 0.00, Max 25.00

### Key empirical findings — treatment effect model (baseline and robustness)
- Main finding:
  - Positive relationship between IMF-supported programs and FDI flows to LICs.
  - Estimated average treatment effects: a country with an IMF-supported program was able to attract one to four times more FDI as a percentage of GDP than a country without a program (summary statement from paper).
- Treatment effect model baseline (Table 1; dependent variable: FDI as a percentage of GDP, 1980-2012):
  - IMF participation coefficient: positive and significant at the 1 percent level across specifications.
  - Estimated IMF participation coefficient values across columns: 4.9*, 4.2*, 4.3*, 4.7*, 4.2*, 4.2* (robust z-values: (2.99), (2.25), (4.97), (3.82), (5.12), (5.22)).
  - Log(GDP per capita): coefficients 0.09* in all columns (robust z-values: (11.7), (11.7), (11.2), (9.65), (11.3), (11.2)).
  - Openness coefficients: 0.03*, 0.03*, 0.03*, 0.04*, 0.03*, 0.03* (robust z-values: (5.39), (5.05), (5.88), (5.72), (5.84), (5.87)).
  - Real GDP growth rate coefficients reported as: 0.07***, 0.09**, 0.05**, 0.03, 0.05**, 0.05*** (robust z-values: (1.68), (1.96), (2.00), (0.94), (1.96), (1.87)).
  - Inflation rate coefficients: -0.01*, -0.02*, -0.02*, -0.02**, -0.02*, -0.02* (robust z-values: (-2.86), (-2.71), (-3.07), (-2.19), (-3.07), (-3.09)).
  - FDI stock/GDP coefficients: 0.89**, 0.72***, 1.0*, 0.66**, 0.97*, 0.97* (robust z-values: (2.38), (1.75), (3.99), (1.97), (4.05), (4.07)).
  - Financial development coefficients: 0.0001 across columns (robust z-values: (1.50), (1.18), (1.46), (0.95), (1.42), (1.41)).
  - Democratic Institutions coefficients: -0.89, -0.47, 0.44, 0.74, 0.45, 0.46 (robust z-values: (0.93), (0.46), (0.85), (1.20), (0.86), (0.88)).
  - Number of observations by column: 1763, 1751, 1551, 551, 1092, 1550.
  - Number of countries by column: 73, 73, 71, 71, 71, 71.
  - Wald test statistics (bottom of Table 1): 4/28.6, 19.7, 39.6, 39.6, 41.2, 42.7 with P-values: 0.000, 0.000, 0.000, 0.000, 0.000, 0.000 — indicating statistical significance of selection correction.
- Additional robustness:
  - Controlling for corruption, institutional quality, political risk (Appendix Table 5) — IMF participation coefficient remains significantly positive.
  - Heckman’s two-step estimator (Appendix Table 6) — IMF participation coefficients reported: 3.6*, 3.3*, 3.6*, 3.7*, 4.2*, 3.8* (robust z-values: (5.477), (5.122), (6.114), (6.317), (5.404), (6.487)).
  - Maximum Likelihood estimates (Appendix Table 5) report IMF Participation: 5.2*, 5.2*, 5.1*, 4.5*, 5.1*, 5.1* (robust z-values: (4.870), (4.838), (4.668), (3.130), (4.884), (4.842)).
  - Heterogeneity: Appendix Table 7 (ECF-focused) shows IMF-ECF Participation coefficients: 5.4*, 4.4*, 4.4*, 5.3*, 4.3*, 4.3* (robust stats reported).

### Propensity Score Matching (PSM) and non-parametric results
- PSM implementation:
  1. Estimate propensity scores using pooled panel probit (selection equation).
  2. Match IMF-participating countries to non-participating countries using estimated propensity scores.
  3. Estimate average treatment effect on matched sample; compute standard errors.
- Matching algorithms used: Nearest Neighbor (1, 3, 5), Kernel, Radius matching.
- Non-parametric estimated treatment effects (Table 2):
  - Nearest Neighbor Matching (1): 2.3* (Robust t-stat (3.11))
  - Nearest Neighbor Matching (3): 2.2* (Robust t-stat (3.70))
  - Nearest Neighbor Matching (5): 1.9* (Robust t-stat (2.67))
  - Kernel Matching: 2.7* (Robust t-stat (2.75))
  - Radius Matching: 1.6* (Robust t-stat (3.06))
  - No. of Observations: 1954 (reported for each estimator)
  - No. of Countries: 73
- Interpretation:
  - PSM results suggest a member country under an IMF-supported program attracts two times more FDI as a percentage of GDP than a country not under such a program (significant at the 1 percent level).

### Sensitivity to hidden bias (Rosenbaum bounds)
- Purpose: test sensitivity of matching estimates to an unobserved confounder using parameter Γ.
- Results (Table 3: Rosenbaum's Sensitivity Analysis):
  - The p-critical value from the Wilcoxon signed rank test maintains the 5 percent significance up to Γ = 5, suggesting robustness to hidden bias.
  - Table entries (selected rows):
    - Γ = 1.4: P-Value -0.82  Max 0.65  Min -1.22  CI+ 1.07
    - Γ = 1.8: P-Value -1.40  Max 1.26  Min -1.85  CI+ 1.70
    - Γ = 2.2: P-Value -1.89  Max 1.75  Min -2.43  CI+ 2.20
    - Γ = 2.6: P-Value -2.36  Max 2.15  Min -2.97  CI+ 2.64
    - Γ = 3.0: P-Value -2.80  Max 2.51  Min -3.46  CI+ 3.06
    - Γ = 3.4: P-Value -3.20  Max 2.83  Min -3.91  CI+ 3.45
    - Γ = 3.8: P-Value -3.57  Max 3.15  Min -4.37  CI+ 3.81
    - Γ = 4.2: P-Value -3.91  Max 3.44  Min -4.83  CI+ 4.16
    - Γ = 4.6: P-Value -4.25  Max 3.71  Min -5.26  CI+ 4.50
    - Γ = 5.0: P-Value -4.59  Max 3.98  Min -5.75  CI+ 4.83
  - Note: Robsenbaum Bounds calculated using the command rbounds in Stata; P-value is the upper bond (sig+) of the Wilcoxon's signed rank test.

### Robustness checks and alternative estimations
- Excluding largest FDI recipients:
  - Two thresholds: 10 percent and 5 percent of GDP; Appendix Table 8 columns 1 and 2 show results not sensitive to these exclusions.
- Sample-period sensitivity:
  - Periods considered: 1990-2012, 1995-2012, 2000-2012 (Appendix Table 8 columns 3–5); findings robust across periods.
- Alternative estimators:
  - POLS, Fixed Effects (FE), and system GMM (Appendix Table 9) — IMF-supported programs still exert positive effects on FDI inflows.
- Additional lag tests:
  - Lagging IMF participation variable by two periods (results not reported) leaves empirical results qualitatively similar.

### Theoretical channels and interpretation
- Underlying rationale: IMF-supported programs help restore macroeconomic stability and address structural constraints, rebuilding investor confidence and encouraging long-term FDI.
- Three channels through which IMF programs can facilitate FDI inflows:
  1. Macroeconomic channel — program conditionality restores macroeconomic stability, strengthens fiscal and external positions, boosts growth; successful program reviews send a credibility “signal”.
  2. Structural reforms channel — non-quantitative structural conditions (trade reforms, price liberalization, privatization, banking/financial sector reforms) improve market efficiency and promote private-sector led growth; privatization can create direct FDI opportunities.
  3. Institutional/governance channel — technical assistance and program-supported reforms improve institutional capacity, policy design, governance, transparency, and AML/CFT frameworks.

### Conclusions and suggested avenues for future research
- Main conclusion:
  - Countries participating in IMF-supported programs attracted more FDI inflows than countries not under such programs (based on 1980-2012 unbalanced panel of 73 LICs and two estimation approaches addressing self-selection).
- Recommendations for future research:
  - Distinguish between IMF programs by strength of macroeconomic adjustment rather than using a simple binary program indicator.
  - Differentiate between successfully completed programs vs. unsuccessful ones.

*Source: _wp15157*

### 1. Introduction ________________________________________________________________________________5

### 1. Introduction

### Objective and contribution
- Main objective: empirically assess the impact of IMF-supported programs on FDI inflows.
- Focus: exclusively on low-income countries (LICs).
- Data/sample: unbalanced panel data for 73 LICs over the period 1980–2012.
- IMF facilities considered: all concessional financial and non-financial programs available for LICs (Extended Credit Facility (ECF) and its two predecessors (PRGF, ESAF), Structural Adjustment Facility (SAF), Exogenous Shocks Facility (ESF), Standby Credit Facility (SCF), and Policy Supported Instrument (PSI) arrangements).
- Identification strategy: address self-selection into IMF programs using two econometric approaches:
  - A two-step estimation method based on a treatment effect model.
  - A non-parametric estimation method based on propensity score matching (PSM).

### Key empirical finding
- Positive relationship between IMF-supported programs and FDI flows to LICs.
- Estimated average treatment effects: a country with an IMF-supported program was able to attract one to four times more FDI as a percentage of GDP than a country without a program.
- Robustness: result is very robust under different specifications and sample periods.

### Rationale and theoretical channels
- Underlying assumption: IMF-supported programs help restore macroeconomic stability and address structural constraints to growth, thereby rebuilding confidence and encouraging foreign investors to undertake long-term projects.
- Three channels through which IMF-supported programs can facilitate FDI inflows into LICs:
  1. Macroeconomic channel — a sound macroeconomic framework underpinned by program conditionality that restores macroeconomic stability, strengthens fiscal and external positions, and boosts economic growth; successful program reviews send a credibility “signal” to investors.
  2. Structural reforms channel — non-quantitative structural conditions (trade reforms, price liberalization, privatization, banking/financial sector reforms) that improve market efficiency and promote private-sector led growth; privatization can create direct opportunities for FDI.
  3. Institutional/governance channel — technical assistance and program-supported reforms to institutional capacity, policy design, governance, transparency, and AML/CFT frameworks.

### Situating the paper in the literature
- Most prior empirical studies focus on IMF programs’ effects on macroeconomic performance or on access to international financial markets; relatively little attention to FDI, and especially to LICs.
- Prior findings on IMF programs and FDI are mixed:
  - Bird and Rowlands (2002): no significant support for increased FDI; study did not control for selection bias.
  - Jensen (2004): after controlling for self-selection bias, finds countries with IMF-supported programs attracted 25 percent less FDI than countries without such programs.
  - Biglasier and DeRouen (2010): find countries with IMF-supported programs attracted more U.S. FDI; concessional PRGF lending does not have a positive effect while other facilities such as SAB have positive effects.
- This paper’s contributions relative to the literature:
  - Exclusive focus on LICs (73 countries, 1980–2012).
  - Consideration of all IMF concessional facilities available to PRGT-eligible countries.
  - Use of two complementary methods (treatment effect model and PSM) to mitigate selection bias.

### Paper organization
- Section 2: determinants of FDI location.
- Section 3: role of IMF-supported programs.
- Section 4: empirical literature.
- Section 5: empirical model, estimation methods, and data sources.
- Section 6: empirical results.
- Section 7: conclusions.

*Source: _wp15157 - 1. Introduction*

### 5. E

### 5. E MPIRICAL FRAMEWORK

### The Model
- Baseline linear specification:
  - y_{i,t} = x′_{i,t-1} β + δ D_{i,t-τ}^{IMF} + η_{i} + ε_{i,t} (Equation (1))
  - y: FDI inflows as a percentage of the host country’s GDP.
  - x: vector of exogenous variables (lagged one year).
  - D_{IMF}: dummy = 1 if country i participates in an IMF-supported program in a given year for at least five months, 0 otherwise.
  - η: time invariant country-specific effect; ε: random disturbance.
- Identification challenges:
  - Counterfactual problem: cannot observe y_{i}|D_{i}=1 and y_{i}|D_{i}=0 simultaneously for the same country.
  - Endogeneity/selection bias: D_{IMF} is endogenous (country choice to seek IMF assistance), requiring modeling of selection.
- Methodological approach:
  - Two econometric methods used to correct selection bias and estimate δ consistently:
    - Regression-based treatment effect model (Maddala (1983)).
    - Non-parametric Propensity Score Matching (PSM).

### The Treatment Effect Model
- Selection (first) stage (Equation (2)):
  - D^{*}_{i,t} = z′_{i,t} γ + μ_{i,t}  (latent variable)
  - Observed D_{i,t}^{IMF} generated from D^{*}_{i,t}.
  - z: vector of exogenous variables; μ: error term.
- Estimation logic:
  - Two-step: estimate probability of participating (selection equation), then estimate outcome equation (Equation (1)) using selection correction.
  - Consistent estimation requires error terms of selection and outcome equations (μ and ε) to be correlated; if uncorrelated, OLS is consistent.
  - Model can be estimated by Maximum Likelihood Estimator (MLE) or Heckman’s two-step estimator.
- Determinants included:
  - Outcome (FDI) equation controls:
    1. Market size and potential growth: GDP per capita, real GDP growth.
    2. Financial development: broad money as percent of GDP.
    3. Trade openness: (exports + imports) as percent of GDP.
    4. Macroeconomic stability: inflation rate.
    5. Agglomeration: existing FDI stock as percent of GDP.
    6. Political risk.
    7. Institutional quality, corruption, democracy.
  - Selection equation (determinants of IMF participation):
    1. Macroeconomic factors: level of development, real GDP growth, foreign reserves, inflation, current account balance, terms of trade, external debt services.
    2. Institutional factors: years under past IMF programs, democratic institutions.
    3. Global factors: world real GDP growth.
- Notes:
  - The model is over-identified since several variables in the selection equation do not appear in the outcome equation.
  - All independent variables are lagged by one year.

### Propensity Score Matching (PSM)
- Setup:
  - Potential outcomes: Y_{1i} (if treated), Y_{0i} (if not treated).
  - Treatment indicator D_{i} = 1 for participation, 0 otherwise.
  - Observed triple per country: (D_{i}, Y_{i}, X_{i}).
  - ATT = E[Y_{1i}|D_{i}=1] – E[Y_{0i}|D_{i}=1]; ATE = E[Y_{1i} – Y_{0i}].
  - Propensity score: P(s) = P(D_{i}=1 | X = x).
- Key identifying assumptions:
  1. Conditional independence (unconfoundedness): (Y_{1i}, Y_{0i}) ┴ D | X.
  2. Common support (overlap): 1 > P(D=1|X) > 0 for all X.
- Implementation objective:
  - Match treated and untreated countries on propensity scores to estimate average treatment effects despite selection bias.

### The Data
- Sample: unbalanced panel of 73 LICs over 1980-2012.
- Data sources (as used in the study):
  - FDI inflows, FDI stock, foreign aid, external debt services, total external debt: UNCTAD’s World Investment Report (2013).
  - Real GDP growth, GDP per capita, inflation, terms of trade, foreign reserves (months of imports), current account balances, world real GDP growth: IMF World Economic Outlook (2013).
  - Exports+imports/GDP, broad money (M2): World Bank WDI (2013).
  - Corruption, institutional quality, political risk: International Country Risk Guide, Political Risk Groups (2013).
  - Democracy index: Freedom House (2013).
  - IMF-supported programs: IMF’s database on arrangements.
- IMF participation dummy definition:
  - Equals 1 if member country is under one of SAF, ECF, SCF, ESF, or PSI for at least five months in a year, 0 otherwise.
- Data treatment:
  - All independent variables lagged one year to reduce simultaneity.
  - Appendix Table 1: descriptive statistics; Appendix Tables 2A and 2B: correlation matrices.

### Empirical Results — Overview
- Selection equation (pooled panel probit) results (Appendix Table 3):
  - LICs more likely to seek IMF assistance if they have:
    - Low levels of economic development,
    - Large current account deficits,
    - Large fiscal deficits,
    - High public debt-to-GDP ratios,
    - High external debt services-to-exports ratios,
    - High inflation rates,
    - Low foreign reserves.
  - Past participation in IMF-supported programs increases likelihood of seeking another program.
  - Variables with unexpected signs: real GDP growth, changes in terms of trade, world real GDP growth, and foreign aid inflows.
  - Robustness: results similar under random effects probit and correlated random effects probit (Appendix Table 4).
- Possible explanations for unexpected signs:
  1. Engagement motivations: some countries seek IMF policy/conditionality rather than financing.
  2. Group heterogeneity: LICs are heterogeneous in vulnerability to shocks.
  3. Nonlinearity: relationships may be nonlinear and misrepresented by linear model.

### Empirical Results — Treatment Effect Model (Table 1 and robustness)
- Baseline findings (Table 1; dependent variable: FDI as a percentage of GDP, 1980-2012):
  - IMF participation coefficient:
    - Positive and robustly significant at the 1 percent level across specifications.
    - Estimated average treatment effect: a member country under an IMF-supported program attracts four times more FDI as a percentage of GDP than a country not under such a program.
  - Control variables:
    - Log(GDP per capita): positive and highly significant at the 1 percent level.
    - Real GDP growth rate: positive and statistically significant at the 10 percent level.
    - Openness: positive and significant at the 1 percent level.
    - FDI stock/GDP (agglomeration): positive and highly significant.
    - Financial development, inflation rates, and democratic institutions: exceptions where coefficients did not align with expectations in some specifications.
  - Model diagnostics:
    - Wald tests reported at the bottom of Table 1 indicate statistical significance of the correlation between selection and outcome errors; treatment effect model is appropriate.
- Robustness checks:
  - Controlling for corruption, institutional quality, and political risk (Appendix Table 5): IMF participation coefficient remains significantly positive.
  - Re-estimation by Heckman’s two-step estimator (Appendix Table 6): results do not change materially.
- Table 1 reported statistics (selected):
  - IMF participation coefficient estimates across columns: 4.9*, 4.2*, 4.3*, 4.7*, 4.2*, 4.2* (robust z-values shown in parentheses: (2.99), (2.25), (4.97), (3.82), (5.12), (5.22)).
  - Log(GDP per capita) coefficients: 0.09*, 0.09*, 0.09*, 0.09*, 0.09*, 0.09* (robust z-values: (11.7), (11.7), (11.2), (9.65), (11.3), (11.2)).
  - Openness coefficients: 0.03*, 0.03*, 0.03*, 0.04*, 0.03*, 0.03* (robust z-values: (5.39), (5.05), (5.88), (5.72), (5.84), (5.87)).
  - Real GDP growth rate coefficients (row labeled “Real GDP growth rate”): 0.07***, 0.09**, 0.05**, 0.03, 0.05**, 0.05*** (robust z-values: (1.68), (1.96), (2.00), (0.94), (1.96), (1.87)).
  - Inflation rate coefficients: -0.01*, -0.02*, -0.02*, -0.02**, -0.02*, -0.02* (robust z-values: (-2.86), (-2.71), (-3.07), (-2.19), (-3.07), (-3.09)).
  - FDI stock/GDP coefficients: 0.89**, 0.72***, 1.0*, 0.66**, 0.97*, 0.97* (robust z-values: (2.38), (1.75), (3.99), (1.97), (4.05), (4.07)).
  - Financial development coefficients: 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001 (robust z-values: (1.50), (1.18), (1.46), (0.95), (1.42), (1.41)).
  - Democratic Institutions coefficients: -0.89, -0.47, 0.44, 0.74, 0.45, 0.46 (robust z-values: (0.93), (0.46), (0.85), (1.20), (0.86), (0.88)).
  - Number of observations by column: 1763, 1751, 1551, 551, 1092, 1550 (as shown in Table 1).
  - Number of countries: 73, 73, 71, 71, 71, 71.
  - Wald test statistics: 4/28.6, 19.7, 39.6, 39.6, 41.2, 42.7 with P-values: 0.000, 0.000, 0.000, 0.000, 0.000, 0.000.
- Heterogeneity of programs:
  - Authors note potential differences across IMF facilities (short-term vs. longer-term arrangements).
  - Constructed alternative dummy including only ECF and its predecessor ESAF/PRGF to test differentiated effects (analysis continued beyond the provided content).

* _wp15157 - 5. E_

### Appendix Table 7, as expected, the impact of the IMF-ECF programs have a stronger

### _wp15157 - Appendix Table 7, as expected, the impact of the IMF-ECF programs have a stronger

### Robustness checks on baseline regressions
- Excluding countries with the largest FDI inflows:
  - Two exclusion thresholds considered: 10 percent and 5 percent of GDP.
  - Results (Appendix Table 8, columns 1 and 2) indicate results are not sensitive to exclusion of largest recipients.
- Sample-period sensitivity:
  - Post-1990s regressions considered three periods: 1990-2012, 1995-2012, and 2000-2012.
  - Results (Appendix Table 8, columns 3, 4, and 5) indicate findings are not sensitive to different sample periods.
- Alternative estimators (abstracting from self-selection):
  - Pooled OLS (POLS), Fixed Effects (FE), and system GMM estimated (Appendix Table 9).
  - IMF-supported programs are still found to exert positive effects in encouraging FDI inflows into participant countries.
- Additional lag test:
  - Lagging IMF participation variable by two periods (results not reported) leaves empirical results qualitatively similar.

### Propensity Score Matching (PSM) procedure and main non-parametric results
- PSM estimation steps:
  1. Estimate propensity scores using a pooled panel probit regression (selection equation) with observed pre-program covariates.
  2. Match IMF-participating countries to non-participating countries using estimated propensity scores.
  3. Estimate average treatment effect with matched sample and compute standard errors.
- Matching algorithms used:
  - Nearest Neighbor, Radius, Kernel matching.
  - Matches of observations within the common support sample were used in all estimates.
- Determinants of program participation specified as in Equation (2) and estimated by pooled panel probit regressions.
  - In unreported regressions, pooled panel logit model used; results do not change.
- Table 2: IMF-Supported Programs and FDI Inflows — Non-Parametric Estimations (summary):
  - Estimated treatment effects (IMF Participation) across estimators:
    - Nearest Neighbor Matching (1): 2.3* (Robust t-stat (3.11))
    - Nearest Neighbor Matching (3): 2.2* (Robust t-stat (3.70))
    - Nearest Neighbor Matching (5): 1.9* (Robust t-stat (2.67))
    - Kernel Matching: 2.7* (Robust t-stat (2.75))
    - Radius Matching: 1.6* (Robust t-stat (3.06))
  - Notes:
    - No. of Observations: 1954 (reported for each estimator)
    - No. of Countries: 73
    - Robust t-stat in parentheses.
    - *, ** and *** indicate statistical significance at 1%, 5% and 10% levels, respectively.
- Interpretation of PSM results:
  - The PSM results suggest that a member country under an IMF-supported program attracts two times more FDI as a percentage of GDP than a country not under such a program.
  - This result is significant at the one percent level.

### Sensitivity analysis for hidden bias (Rosenbaum bounds)
- Motivation:
  - Matching estimators can be biased if unobserved heterogeneity (hidden bias) remains between treatment and control groups.
  - Rosenbaum bounds method used to test sensitivity to an unobserved confounder.
- Sensitivity parameter Γ:
  - If Γ = 1, assignment of participation is effectively random conditional on observables (no hidden bias).
  - Larger Γ implies greater potential for hidden bias to overturn conclusions.
- Results (Table 3: Robsenbaum's Sensitivity Analysis):
  - The p-critical value from the Wilcoxon signed rank test maintains the 5 percent significance up to Γ = 5, suggesting robustness to hidden bias.
  - Table 3 entries (Γ, P-Value, Max, Min, CI+, CI- presented as rows):
    - Γ = 1.4: P-Value -0.82  Max 0.65  Min -1.22  CI+ 1.07
    - Γ = 1.8: P-Value -1.40  Max 1.26  Min -1.85  CI+ 1.70
    - Γ = 2.2: P-Value -1.89  Max 1.75  Min -2.43  CI+ 2.20
    - Γ = 2.6: P-Value -2.36  Max 2.15  Min -2.97  CI+ 2.64
    - Γ = 3.0: P-Value -2.80  Max 2.51  Min -3.46  CI+ 3.06
    - Γ = 3.4: P-Value -3.20  Max 2.83  Min -3.91  CI+ 3.45
    - Γ = 3.8: P-Value -3.57  Max 3.15  Min -4.37  CI+ 3.81
    - Γ = 4.2: P-Value -3.91  Max 3.44  Min -4.83  CI+ 4.16
    - Γ = 4.6: P-Value -4.25  Max 3.71  Min -5.26  CI+ 4.50
    - Γ = 5.0: P-Value -4.59  Max 3.98  Min -5.75  CI+ 4.83
  - Notes:
    - Robsenbaum Bounds calculated using the command rbounds in Stata.
    - P-value is the upper bond (sig+) of the Wilcoxon's signed rank test.
    - Columns labelled: Point estimate, 95% Confidence Interval, Hodges Lehmann.

### Conclusions (study-level findings and suggestions for future research)
- Main empirical finding:
  - Countries participating in IMF-supported programs attracted more FDI inflows than countries not under such programs (based on 1980-2012 unbalanced panel of 73 LICs and two estimation approaches addressing self-selection).
- Theoretical channels identified through which IMF-supported programs facilitate FDI to LICs:
  - Program conditionality aimed to restore macroeconomic stability and create conditions for sustainable and inclusive economic growth.
  - Market efficiency improvements brought by programs’ structural reforms.
  - Technical assistance aimed to improve a country’s institutional capacity.
- Estimation approaches used:
  - Two-step regression-based method estimating outcome and selection equations simultaneously.
  - Non-parametric propensity score matching estimating average treatment effect.
- Scope for future research recommended:
  - Distinguish between IMF programs by strength of macroeconomic adjustment rather than using a simple binary program indicator.
  - Differentiate between successfully completed programs vs. unsuccessful ones.

### Appendix Table 1 — Summary statistics (Sample: 73 Low-Income Host Countries 1980-2012)
- Variables (Obs, Mean, Std. Dev, Min, Max):
  - FDI /GDP: Obs 2024, Mean 3.83, Std. Dev 6.42, Min -14.68, Max 90.46
  - IMF: Obs 2052, Mean 0.37, Std. Dev 0.48, Min 0.00, Max 1.00
    - IMF =1: 758
    - IMF = 0: 1294
  - Real GDP growth: Obs 2052, Mean 3.96, Std. Dev 6.05, Min -50.25, Max 71.19
  - log (GDP per capita): Obs 2052, Mean 6.49, Std. Dev 0.96, Min 4.28, Max 10.09
  - Inflation rate: Obs 2048, Mean 58.78, Std. Dev 705.55, Min -72.73, Max 23773
  - Financial development: Obs 1893, Mean 33.90, Std. Dev 22.42, Min 4.53, Max 151.55
  - Openness: Obs 2052, Mean 76.10, Std. Dev 43.67, Min 4.19, Max 403.92
  - FDI stock /GDP: Obs 1979, Mean 27.42, Std. Dev 34.06, Min 0.03, Max 248.56
  - Reserves: Obs 2052, Mean 4.26, Std. Dev 3.55, Min 0.00, Max 23.05
  - Log (terms of trade): Obs 2052, Mean 4.70, Std. Dev 0.40, Min 3.30, Max 7.15
  - Current Account Balance / GDP: Obs 2052, Mean -7.74, Std. Dev 11.08, Min -154.03, Max 35.48
  - External Debt /GDP: Obs 1957, Mean 3.09, Std. Dev 3.25, Min 0.00, Max 30.18
  - Debt Services/ total Exports: Obs 1726, Mean 15.12, Std. Dev 13.81, Min 0.25, Max 156.86
  - Total period of IMF participations: Obs 2052, Mean 4.65, Std. Dev 5.47, Min 0.00, Max 25.00
  - Fiscal balance /GDP: Obs 1417, Mean -2.68, Std. Dev 6.79, Min -46.21, Max 25.4
  - Quota /GDP: Obs 2052, Mean 194.43, Std. Dev 322.70, Min 0.02, Max 770.5
  - log (foreign Aid): Obs 1962, Mean 5.30, Std. Dev 1.41, Min -3.22, Max 9.34
  - World Real GDP Growth: Obs 2052, Mean 2.78, Std. Dev 1.37, Min -2.11, Max 4.59
  - Democratic institutions: Obs 2052, Mean 0.43, Std. Dev 0.28, Min 0.00, Max 1.00
  - Law and Order: Obs 1271, Mean 2.71, Std. Dev 1.07, Min 0.00, Max 5.00
  - Corruption: Obs 1271, Mean 2.29, Std. Dev 0.98, Min 0.00, Max 5.00
  - Political risk: Obs 1263, Mean 52.87, Std. Dev 11.57, Min 9.00, Max 76.00
- Source: Author's calculation

*Source: _wp15157 - Appendix Table 7, as expected, the impact of the IMF-ECF programs have a stronger.*

### Appendix Table 2A: Correlation Matrix

### _wp15157 - Appendix Table 2A: Correlation Matrix

### Correlation matrices: key pairwise correlations (Appendix Table 2A)
- Matrix variables: IMF, Real GDP growth, Log (GDP per capita), CAB /GDP, External debt/GDP, Real world GDP growth, Debt Services /exports, Inflation rate, Log (terms of trades), Foreign Reserves, Period of IMF programs, log (foreign Aid), Democratic Institutions.
- Selected correlations (row variable listed first, column variable second where applicable):
  - IMF with:
    - IMF: 1.00
    - Real GDP growth: 0.11
    - Log (GDP per capita): -0.19
    - CAB /GDP: -0.03
    - External debt/GDP: -0.08
    - Real world GDP growth: 0.02
    - Debt Services /exports: 0.04
    - Inflation rate: -0.06
    - Log (terms of trades): -0.01
    - Foreign Reserves: 0.10
    - Period of IMF programs: 0.54
    - log (foreign Aid): 0.35
    - Democratic Institutions: 0.07
  - Real GDP growth with:
    - Log (GDP per capita): 0.09
    - CAB /GDP: 0.07
    - External debt/GDP: -0.04
    - Real world GDP growth: 0.09
    - Debt Services /exports: -0.14
    - Inflation rate: -0.09
    - Log (terms of trades): -0.02
    - Foreign Reserves: 0.13
    - Period of IMF programs: 0.13
    - log (foreign Aid): 0.12
    - Democratic Institutions: 0.03
  - Log (GDP per capita) with:
    - CAB /GDP: -0.05
    - External debt/GDP: 0.08
    - Real world GDP growth: -0.07
    - Debt Services /exports: -0.34
    - Inflation rate: -0.05
    - Log (terms of trades): -0.11
    - Foreign Reserves: 0.08
    - Period of IMF programs: -0.08
    - log (foreign Aid): -0.35
    - Democratic Institutions: 0.50
  - Period of IMF programs correlations (selected):
    - Foreign Reserves: 0.22
    - log (foreign Aid): 0.42
    - Democratic Institutions: 0.17

- Source line in table: "Source: Author's calculation"

### Correlation matrices: key pairwise correlations (Appendix Table 2B)
- Matrix variables: Real GDP growth, Log (GDP per capita), Openness, Inflation rate, Democratic Institutions, FDI Stock/GDP, Financial Development, Law & Order, Corruption, Political risk, IMF participation.
- Selected correlations (row variable listed first):
  - FDI/GDP with:
    - Real GDP growth: 0.59
    - Log (GDP per capita): 0.11
    - Openness: 0.42
    - Inflation rate: -0.02
    - Democratic Institutions: 0.06
    - Financial Development: 0.04
    - Law & Order: 0.02
    - Corruption: 0.02
    - Political risk: 0.07
    - IMF participation: -0.04
  - Real GDP growth with:
    - Log (GDP per capita): 0.09
    - Openness: 0.10
    - Inflation rate: -0.10
    - Democratic Institutions: 0.06
    - FDI Stock/GDP: 0.11
    - Financial Development: -0.06
    - Law & Order: 0.20
    - Corruption: -0.02
    - Political risk: 0.22
    - IMF participation: 0.13
  - Political risk correlations (selected):
    - Law & Order: 0.62
    - Corruption: 0.31
    - IMF participation: 0.21

- Source line in table: "Source: Author's calculation"

### Variables explicitly listed in these appendices
- IMF participation
- Real GDP growth
- Log (real GDP per capita)
- Current Account Balance / GDP
- External Debt / GDP
- External Debt Services / Exports
- Real world GDP growth
- Inflation rate
- Changes in Terms of Trades (Log (terms of trades))
- Reserves in months of imports (Foreign Reserves)
- Periods under IMF programs
- Foreign Aid (log (foreign Aid))
- Democratic Institutions
- Openness
- Financial Development
- FDI Stock/GDP
- Law & Order
- Corruption
- Political risk

---

### Determinants of Participations in IMF-Supported Programs (Appendix Table 3 — pooled Probit)
- Dependent variable: IMF participations: 1980-2012 (dummy = 1 if under IMF-programs at least five months, zero otherwise).
- Estimation: pooled Probit regressions; all independent variables lagged by one year; robust z-values in parentheses.
- Selected coefficient estimates across specifications (columns (1)–(6)):
  - Reserves in months of imports: -0.001 (-0.266), -0.001 (-0.388), -0.005 (-1.464), -0.01** (-2.040), -0.01** (-2.448), -0.003 (-1.540)
  - Current Account Balance / GDP: 0.02* (3.190), 0.02* (3.469), 0.03* (4.611), 0.01 (1.369), 0.03* (3.620), 0.03* (4.483)
  - Real GDP growth: -0.50* (-11.53), -0.50* (-11.19), -0.45* (-9.372), -0.53* (-9.029), -0.40* (-7.940), -0.45* (-9.270)
  - log (real GDP per capita): -0.009 (-0.984), -0.007 (-0.722), -0.008 (-0.796), -0.02*** (-1.733), -0.007 (-0.724), -0.012 (-0.72)
  - Fiscal balance / GDP: -0.002* (-2.832), -0.002* (-2.768), -0.002* (-2.829), -0.002** (-2.161), -0.002* (-2.834), -0.002** (-2.188)
  - World real GDP growth: 0.045 (0.558), 0.059 (0.709), 0.048 (0.561), 0.167 (1.483), 0.047 (0.543), 0.102 (1.124)
  - Foreign aid /GDP (log): 0.11* (15.42), 0.12* (15.58), 0.11* (15.26), 0.11* (13.51), 0.11* (15.25), 0.13* (15.72)
  - Inflation rate: 0.71* (5.421), 0.67* (4.929), 0.61* (4.370), 0.66* (3.876), 0.60* (4.317), 0.37** (2.533)
  - Periods under IMF programs: 0.03* (3.175) in one specification
  - Democratic Institutions: 0.02** (2.511), 0.016 (1.419), 0.02** (2.495), 0.004 (0.508) in various specs
  - External Debt / GDP and External Debt Services / Exports appear in alternative columns with coefficients reported.
- No. of Observations by column: 1976, 1953, 1976, 1884, 1275, 1867, 1884 (as presented in sequence)
- No. of Countries: 73 (all columns)
- Pseudo R2: 0.22, 0.22, 0.22, 0.26, 0.22, 0.24
- Notes:
  - All independent variables lagged by one year.
  - Models estimated by pooled Probit regressions; robust z-values in parentheses.
  - *, ** and *** indicate statistical significance at 1%, 5% and 10% levels, respectively.

### Alternative estimations: Random Effects and Correlated Random Effects Probit (Appendix Table 4)
- Dependent variable: IMF participations: 1980-2012.
- Random Effects Probit and Correlated Random Effects Probit specifications (columns (1)–(3) for each estimator).
- Selected coefficient estimates:
  - Reserves in months of imports: -0.001 (-0.49), -0.001 (-0.45), -0.01 (-1.75) and similar in correlated RE columns (-0.001 (-0.44), -0.001 (-0.43), -0.01 (-1.92))
  - Current Account Balance / GDP: 0.02* (3.57), 0.02* (3.65), 0.03* (3.61) and 0.02* (3.59), 0.02* (3.70), 0.02* (3.63)
  - Real GDP growth: -0.65* (-8.80), -0.63* (-8.30), -0.65* (-6.57) and -0.54* (-9.57), -0.53* (-9.24), -0.46* (-6.68)
  - log (real GDP per capita): 0.03*** (1.79), 0.03*** (1.73), 0.04* (2.61) and 0.01, 0.01, 0.03** (1.30, 1.23, 2.08)
  - Fiscal balance / GDP: -0.003* (-2.64), -0.003* (-2.6), -0.002*** (-1.76) and -0.002* (-2.57), -0.002* (-2.52), -0.002** (-2.25)
  - Foreign Aid / GDP: 0.07* (7.83), 0.07* (7.71), 0.02** (2.16) and 0.08* (10.7), 0.08* (11.01), 0.04* (4.83)
  - Inflation rate: 1.2* (5.68), 1.2* (5.40), 1.1* (4.60) and 0.90* (5.60), 0.90* (5.31), 0.97* (5.07)
  - Foreign Aid / GDP reported as 0.7* and 0.4* in some columns (with parentheses (8.42), (9.00))
- No. of Observations: 1976, 1953, 1867 (for both RE and correlated RE blocks)
- No. of Countries: 73
- Rho hat (p-value): 0.00, 0.00, 0.00 (reported)
- Wald test (p-value): 0.00, 0.00, 0.00 (reported)
- Notes:
  - All independent variables lagged by one year; constant term included; *, ** and *** indicate statistical significance at 1%, 5% and 10% levels.

---

### IMF-Supported Programs and FDI inflows: Maximum Likelihood and Heckman estimations (Appendix Tables 5–6)
- Dependent variable: FDI as a percentage of GDP: 1980-2012.
- Appendix Table 5 — Maximum Likelihood Estimations (selection models; models differ in selection equation specifications):
  - Selected coefficients across columns (1)–(6):
    - IMF Participation: 5.2* (4.870), 5.2* (4.838), 5.1* (4.668), 4.5* (3.130), 5.1* (4.884), 5.1* (4.842)
    - Real GDP growth rate: 0.07*** (1.793), 0.07*** (1.7924), 0.07 (1.5854), 0.04 (0.4204), 0.07 (1.4593), 0.07 (1.422)
    - Log(GDP per capita): 1.1* (3.252), 1.1* (3.258), 1.0* (3.035), 0.32 (0.772), 1.0* (3.105), 1.0* (3.091)
    - Financial development: -0.02* (3.286), -0.02* (3.253), -0.03* (3.126), -0.03* (3.057), -0.03* (3.262), -0.03* (3.267)
    - Openness: 0.02* (3.819), 0.03* (3.791), 0.03* (3.930), 0.03* (4.486), 0.03* (4.053), 0.03* (4.080)
    - Law and Order: 0.0001*** (1.934), 0.0001*** (1.925), 0.0002** (2.097), 0.0002 (0.363), 0.0002** (2.145), 0.0002** (2.118)
    - FDI stock / GDP: 0.09* (8.060), 0.09* (8.061), 0.09* (7.092), 0.10* (6.213), 0.09* (7.139), 0.09* (7.139)
    - Corruption (example coefficients): -0.50 (0.774), -0.50 (0.770), -0.11 (0.175), 0.31 (0.363), -0.14 (0.211), -0.12 (0.189)
    - Inflation rate: -0.009 (0.499) etc.
  - No. of Observations: 1003, 1003, 946, 682, 945, 945 (per columns)
  - No. of Countries: 43
  - Wald test statistics and p-values reported (e.g., Wald test 4/41.8; p-value 0.00)
  - Note: All independent variables lagged by one year; robust z-values in parentheses; *, ** and *** indicate statistical significance at 1%, 5% and 10% levels.
- Appendix Table 6 — Heckman's Two Step Estimations:
  - Dependent variable: FDI as a percentage of GDP: 1980-2012.
  - Selected coefficients across columns (1)–(6):
    - IMF Participation: 3.6* (5.477), 3.3* (5.122), 3.6* (6.114), 3.7* (6.317), 4.2* (5.404), 3.8* (6.487)
    - Real GDP growth rate: 0.08* (3.756), 0.09* (4.541), 0.06* (2.588), 0.06* (2.558), 0.04 (1.437), 0.06** (2.445)
    - Log (GDP per capita): 0.60* (2.657), 0.50* (2.140), 0.83* (3.726), 0.84* (3.780), 0.52*** (1.797), 0.86* (3.874)
    - Financial development: -0.02* (-2.639) ... -0.02* (-2.727)
    - Openness: 0.03* (10.08), 0.03* (9.554), 0.03* (8.657), 0.03* (8.710), 0.03* (8.189), 0.03* (8.739)
    - Inverse Mills ratio: -1.8* (4.280), -1.6* (3.891), -1.9* (5.165), -2.0* (5.400), -2.6* (5.323), -2.1* (5.602)
  - No. of Observations: 1763, 1751, 1555, 1550, 1092, 1550 (per columns)
  - No. of Countries: 73, 73, 71, 71, 71, 71 (as listed)
  - Notes: All independent variables lagged by one year; robust z-values in parentheses; *, ** and *** indicate statistical significance at 1%, 5% and 10% levels; Inverse Mills ratio reported with strong negative and significant coefficients.

---

### ECF-Supported Programs and FDI inflows: Maximum Likelihood Estimations (Appendix Table 7)
- Dependent variable: FDI as a percentage of GDP: 1980-2012.
- Selected coefficients across columns (1)–(6):
  - IMF-ECF Participation: 5.4* (3.472), 4.4* (1.901), 4.4* (5.055), 5.3* (4.481), 4.3* (5.147), 4.3* (5.313)
  - Real GDP growth rate: 0.06, 0.08, 0.03, 0.02, 0.03, 0.03 (parentheses indicate t/z stats e.g., (1.180), (1.482), (0.929), (0.567), (1.048), (1.037))
  - Log GDP per capita: 0.91* (2.687), 0.69 (1.460), 0.94* (3.943), 0.74** (2.264), 0.92* (3.958), 0.93* (4.024)
  - Financial development: -0.01* (-2.969), -0.02* (-2.786), -0.02* (-2.992), -0.02** (-2.164), -0.02* (-2.992), -0.02* (-3.007)
  - Openness: 0.03* (5.475), 0.03* (5.124), 0.03* (5.973), 0.04* (6.001), 0.03* (5.953), 0.03* (5.976)
  - FDI Stock /GDP: 0.09* (11.783), 0.09* (11.642), 0.09* (11.291), 0.09* (9.580), 0.09* (11.347), 0.09* (11.328)
  - No. of Observations and Countries reported similarly to Heckman tables.
  - Wald test and p-values reported (e.g., Wald test 4/35.7; p-value 0.00).

---

### Robustness and alternative sampling results (Appendix Tables 8–10)
- Appendix Table 8 — Maximum Likelihood Estimations: Alternative Sampling
  - Samples: FDI /GDP < 10, FDI /GDP < 5, 1990-2012, 1995-2012, 2000-2012.
  - Selected coefficients:
    - IMF Participation: 2.2* (10.82), 1.3* (8.65), 5.5** (2.28), 6.1** (2.51), 4.8* (3.83)
    - Log (GDP per capita): 0.59* (5.12), 0.19** (2.05), 0.84 (1.53), 0.73 (1.29), 0.65*** (1.88)
    - Openness and Inflation rate coefficients reported (e.g., -0.01** (-2.21) in one column).
  - No. of Observations: 1592, 1310, 1383, 1131, 893 (per columns)
  - No. of Countries: 73, 72, 73, 73, 73
  - Wald tests and p-values reported.
- Appendix Table 9 — Alternative Estimations (POLS, FE, GMM)
  - Estimators: POLS, FE, GMM with different column labels (POLSPOLSFEFEGMMGMM).
  - Selected coefficients:
    - IMF Participation: 0.13 (0.716), 0.31** (2.151)
    - Real GDP growth rate: 1.8* (5.262), 1.5* (3.433), 1.5** (2.190), 1.3*** (1.691), 0.95** (2.138), 0.60 (1.528)
    - Log (GDP per capita): 0.12** (2.396), 0.14*** (1.736), 0.10** (2.317), 0.12** (2.026), 0.12** (2.372), 0.25 (1.630)
    - FDI stock / GDP: 0.04* (4.591), 0.04* (3.143), 0.03* (3.351), 0.03** (2.361), 0.04 (1.554), 0.003 (0.152)
    - Adjusted R-Squared reported: 0.34, 0.34, 0.12, 0.10, ----, ----
  - No. of Observations: 1864, 1035, 1864, 1035, 1859, 1035 (per columns)
  - No. of Countries: 73, 43, 73, 43, 73, 43
- Appendix Table 10 — Non-Parametric Estimations (Propensity Score Matching)
  - Estimators and results:
    - Nearest Neighbor Matching (1): 0.59 (1.39)
    - Nearest Neighbor Matching (3): 0.28 (0.75)
    - Nearest Neighbor Matching (5): 0.47 (1.44)
    - Kernel Matching: 0.28 (1.00)
    - Radius Matching: 0.38 (1.59)
  - No. of Observations: 1954 (for each estimator)
  - No. of Countries: 73
  - Notes: Robust t-stat in parentheses; *, ** and *** indicate statistical significance at 1%, 5% and 10% levels.

---

### List of PRGT-eligibility Countries (Appendix Table 11 — selected entries)
- Table format: Country; Date of First Eligibility; Date of Graduation (where applicable).
- Selected entries:
  - Afghanistan: Date of First Eligibility March 26, 1986
  - Albania: Date of First Eligibility April 7, 1992
  - Armenia 1/: Date of First Eligibility December 15, 1993; Date of Graduation July 8, 2013 (note in source)
  - China: Date of First Eligibility March 26, 1986; Date of Graduation December 20, 2000
  - Macedonia, FYR: Date of First Eligibility February 23, 1994; Date of Graduation June 11, 2003
  - Pakistan: Date of First Eligibility March 26, 1986; Date of Graduation April 10, 2010
  - South Sudan: Date of Graduation August 9, 2012
  - Multiple countries listed with Date of First Eligibility March 26, 1986
- Source: Finance Department, the IMF
- Notes in table:
  - 1/ On April 8, 2013 the Executive Board decided to remove Armenia and Georgia from the list of PRGT eligible countries effective July 8, 2013 or at the time of when their arrangements under PRGT expire, whichever is later. Armenia's ECF expired on June 27, 2013 and Georgia's SCF expired on April 4, 2014 (see SM/13/75, 3/18/2013).
  - 2/ On September 24, 2001, Zimbabwe was removed from the PRGT-eligible list due to its overdue obligations to the PRG Trust (Decision No. 12582-(01/99)).

---

### Definitions of Variables and Data Sources (as listed in appendices)
- Data sources: UNCTAD’s World Investment Report, 2013; The World Bank’s World Development Indicators, 2013; The Political Risk Groups, ICRG; IMF’s World Economic Outlook database, 2013; IMF Finance Department database; Freedom House’s database, 2013.
- Key variable definitions (exact phrasing preserved):
  - FDI Inflows: Foreign Direct Investment inflows as a percentage of GDP.
  - FDI stock: end-period total FDI stock in the host countries as a percentage of GDP.
  - Foreign Aid: total Official financial flows. It consist of the sum of official development assistance net (ODA) and other official flows net (OOF).
  - External Debt/GDP: External long-term debt as a percentage of GDP.
  - Debt Service to Exports: External long-term debt service as a percentage of total exports of goods and services.
  - Financial Development: Money and quasi money (M2) as a percentage of GDP.
  - Openness: The sum of exports and imports of goods and services measured as a share of GDP.
  - Corruption: the International Country Risk Guide (ICRG) corruption index: Countries are scored from (0 = high) to (6 = low).
  - Law & Order: is an index, ranging countries from (0 = very low) to (6 = very high), measuring the strength of Law and Order.
  - Political Risk: is an index, ranging countries from (0 = very low) to (100 = very high), measuring the country’s political stability
  - Real GDP Growth: Annual real GDP growth rate.
  - CAB/GDP: Current Account Balance as a percentage of GDP.
  - World real GDP growth: Annual real GDP growth rate.
  - Inflation rate: consumer price index (annual %).
  - Foreign Reserves: total international reserves in months of total imports
  - Level of Development: Annual GDP per capita.
  - Fiscal balance/GDP: Overall fiscal balance as a percentage of GDP
  - Terms of Trade: Terms of goods and services Trade index.
  - IMF participation: a dummy variable equals one if a member country is under an IMF-supported programs at least five months of the year and zero otherwise.
  - Period of IMF-programs: total number of years a member country has been under IMF- supported programs.
  - Democratic institutions: our own compilation based on data for political rights and civil liberties. Countries are ranked from 1 (most free) to 7 (least free) in both indices. Our index is defined as [14 - (political rights + civil rights) / 12] and so it ranges from 0 (least free) to 1 (most free).

*Source: Author's calculation; Finance Department, the IMF; definitions and sources as listed in the appendices.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15157.pdf_