## Effects of IMF-Supported Programs on Gender Inequality (wpiea2024247-print-pdf)

## Source details

**Canonical URL:** [Effects of IMF-Supported Programs on Gender Inequality (wpiea2024247-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024247-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024247-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024247-print-pdf.pdf.json)

---

### Executive summary
- Motivation and research question
  - How would gender disparities have evolved post-crisis without a program with the IMF? Credible control groups are essential to separate crisis effects from program effects.
- Data and scope
  - Sample: over 150 IMF-supported programs from 1994 to 2022.
  - Gender indicators analyzed: labor force participation; primary, secondary, and tertiary education; maternal mortality; adolescent fertility.
  - SCM matches on gender inequality trends and economic conditions such as fiscal spending, external debt, and reserves at crisis onset.
- Methodology
  - Synthetic Control Methods (SCM) used to construct tailored control groups for each program country.
  - Includes bias correction (regression-based SCM bias-corrected estimator) and inference via in-space placebo permutations and p-values based on pre-/post-treatment MSPE ratios.
- Main headline finding
  - No systematic differences in evolution of gender indicators among crisis countries with and without IMF-supported programs: in 87-90 percent of examined cases the program impact is not statistically significantly different from controls.

### Data, sample, and methodology details
- Data horizon: 1990-2023.
- Universe of possible programs: 480.
- Programs analyzed with proper SCM control groups: up to 151 programs (depends on indicator and data availability).
- Seven gender indicators (used across analyses):
  - female labor force participation,
  - ratio of female to male labor force participation,
  - adolescent fertility rate,
  - maternal mortality rate,
  - gender parity index (GPI) for primary enrollment,
  - GPI for secondary enrollment,
  - GPI for tertiary enrollment.
- SCM design features
  - Baseline predictors: lagged gender indicator and reserves in months of imports.
  - Pre- and post-program window: 4 years pre- and 4 years post-program (program year treated as year T0).
  - Exclusion criteria: countries that received another program in the 4 years prior to the current program; donor-pool countries must not receive a program during the entire eight-year window.
  - Robustness predictor set (yields smaller sample): adds government consumption and reserves as percent of external debt.

### Main empirical findings (aggregate and indicator-level)
- Aggregate summary (baseline reserve-coverage predictors; totals reflect Table 2 baseline)
  - Totals: sum 653 observations; 41 positive significant (6%); 44 negative significant (7%); 568 not significant (87%).
- Indicator-level baseline counts (reserve coverage as predictor)
  - Adolescent Fertility Rate: 149 programs; 10 positive significant; 5 negative significant; 134 not significant (90%).
  - Female/Male Labor Force Participation: 142 programs; 8 positive significant; 9 negative significant; 125 not significant (88%).
  - Female Labor Force Participation: 151 programs; 10 positive significant; 13 negative significant; 128 not significant (85%).
  - Maternal Mortality Rate: 91 programs; 5 positive significant; 6 negative significant; 80 not significant (88%).
  - Primary Enrollment GPI: 65 programs; 2 positive significant; 7 negative significant; 56 not significant (86%).
  - Secondary Enrollment GPI: 27 programs; 6 positive significant; 3 negative significant; 18 not significant (67%).
  - Tertiary Enrollment GPI: 28 programs; 0 positive significant; 1 negative significant; 27 not significant (96%).
- Robustness (extended predictor set: reserve coverage, government consumption, reserves as percent of external debt)
  - Aggregate robustness counts: 433 observations; 19 positive significant (4%); 26 negative significant (4%); 388 not significant (90%).
- Pattern of significant findings
  - When statistically significant effects occur (about 10 to 13 percent of cases in baseline), they are mixed in direction — roughly equally split between positive and negative.
  - No program or time period shows significance at the 1 percent level in aggregate; significant findings do not surpass the 5 percent threshold in aggregate analysis.
- Meta-analytic aggregation
  - Aggregated common effect across individually matched SCM estimates is essentially zero and not statistically significant (not significant at the 90 percent level for aggregated effects).

### Selected country examples (illustrative SCM results reported in Annex A)
- Dominican Republic (2003 Stand-By Arrangement)
  - Program: $600 million, immediate disbursement $120 million.
  - Crisis context: GDP growth contraction from 5.9 percent in 1999 to -1.3 percent in 2003; 80 percent currency depreciation (DOP 15.83 per USD in 1999 to DOP 29.38 per USD in 2003); central bank lost half of foreign currency reserves.
  - SCM donor pool: 39 countries; synthetic weights for female labor force participation: Mauritius (0.591), Belize (0.298), Saint Vincent and the Grenadines (0.097), Algeria (0.014).
  - Result: bias-corrected gender impact gap statistically significant at the 5 percent level for all post-crisis years in the example; program effect was negative relative to synthetic control because recovery in the Dominican Republic was less pronounced than in the synthetic control.
- Representative Annex A excerpts (selected indicator-country-year impact gaps and significance markers)
  - Adolescent Fertility: Sierra Leone 1994: t+1 = 2.22*; Venezuela 1996: t+1 = 3.87*; Chad 2000: t+1 = -2.56**.
  - Female/Male LFP ratio: Sierra Leone 2001: t+1 = -0.01**; Djibouti 2008: t+1 = 0.02**; Tajikistan 2009: t+1 = 0.02**.
  - Female LFP: Korea 1997: t+2 = -2.41**; Pakistan 2019: t+1 = 1.24*; Moldova 2016: t+1 = -1.41.
  - Maternal mortality: Sierra Leone 2006: t+1 = -83.30**; Guinea 2012: t+1 = -51.38*; Kenya 2011: t+4 = 111.83*.
  - Primary GPI: Georgia 2001: t+1 = 0.06*; Burundi 2004: t+2 = 0.03**.
  - Secondary GPI: Mozambique 2009: t+1 = 0.02**; Cameroon 2017: t+5 = 0.48*.
  - Tertiary GPI: Moldova 2016: t+4 = 0.13*; Rwanda 2016: t+5 = -0.23**.
- Significance notation in Annex A: *, **, and *** denote significance at the 0.10, 0.05, and 0.01 levels, respectively.

### Robustness, limitations, and diagnostics
- SCM requirements and sample selection
  - SCM requires well-matching pre-crisis trajectories; cases with non-unique weighting vectors or inadequate donor pools are excluded.
  - Balanced panel requirement and donor-pool constraints reduce the candidate sample from 480 to at most 151 programs for some indicators.
- Robustness checks
  - Adding predictors (government consumption, reserves as percent of external debt) reduces sample sizes but yields qualitatively similar conclusions.
- Limitations
  - The study examines programs from 1994–2022 and reports gender effects through 1994–2023; programs adopted after 2022 (including those under updated IMF gender strategy) could not be evaluated with a full four-year post-program window.
  - The analysis does not identify the causal effect of specific conditionality items on gender outcomes; it assesses whether program countries experienced different gender trajectories relative to synthetic controls.
- Diagnostic evidence
  - Table 2 and Figures 4.a to 4.g (reported in the paper) highlight match quality between each program country’s pre-crisis gender indicator and its synthetic control.

### Interpretation, policy implications, and conclusion
- Primary interpretation
  - The vast majority of unusual gender impacts observed after economic crises cannot be attributed to IMF-supported programs once credible counterfactuals are applied; many observed changes are driven by the crisis and recovery dynamics rather than program conditionality.
  - Where program effects are statistically significant, effects are mixed (both positive and negative), indicating context- and design-dependent outcomes.
- Contrast with prior literature
  - Prior studies likely conflated crisis effects with program effects due to lack of credible counterfactuals.
- Policy implications
  - Deliberate attention to program policy design can produce improved gender outcomes; the existence of positive cases indicates policy design matters.
  - Recent IMF strategies cited:
    - IMF 2019: strategy on social spending.
    - IMF 2022: strategy on mainstreaming gender into the IMF’s core operations.
  - These strategies are too recent to be fully evaluated in this study given the four-year post-program window requirement.
- Conclusion
  - Using SCM to establish credible counterfactuals, the study finds no systematic, statistically robust effect of IMF-supported programs on key gender inequality indicators since 1994. Most post-crisis gender outcomes are statistically indistinguishable from outcomes observed in matched countries that did not adopt programs.

### Annex B — Data sources (key database codenames used)
- Gender indicators (WDI DATABASE — World Bank World Development Indicator)
  - Adolescent Fertility Rate: WB: SP.ADO.TFRT
  - Female/Male Labor Force Participation: WB: SL.TLF.CACT.FM.ZS
  - Female Labor Force Participation: WB: SL.TLF.CACT.ZS
  - Maternal Mortality Rate: WB: SH.STA.MMRT
  - Primary Enrollment GPI: WB: SE.ENR.PRIM.FM.ZS
  - Secondary Enrollment GPI: WB: SE.ENR.SECO.FM.ZS
  - Tertiary Enrollment GPI: WB: SE.ENR.TERT.FM.ZS
- Control variables (WDI DATABASE)
  - Total reserves in months of imports: WB: FI.RES.TOTL.MO
  - Total reserves (% of total external debt): WB: FI.RES.TOTL.DT.ZS
  - Government final consumption expenditure (% of GDP): WB: NE.CON.GOVT.ZS
- IMF program sources referenced
  - IMF commitment database — IMF-supported Programs (IMF LENDING DATABASE).
- Additional cited data references include World Bank World Development Indicators Data Files and IMF Monitoring of Fund Arrangements Database (MONA).

*Executive Summary, Introduction, Conclusion, Annex A, and Annex B excerpts from “Effects of IMF-Supported Programs on Gender Inequality,” IMF Working Paper (wpiea2024247-print-pdf).*

### Executive Summary ......................................................................................................

### Executive Summary

### Motivation and research question
- Countries facing economic crises often implement extensive fiscal and monetary consolidations to stabilize their economies; such crises may disproportionately impact populations in vulnerable situations which often include women.
- The paper asks: how would gender disparities have evolved post-crisis without a program with the IMF? To answer this, credible control groups are essential to avoid conflating the impact of crises with the impact of IMF-supported programs.

### Data and scope
- Sample: over 150 IMF-supported programs from 1994 to 2022.
- Gender inequality indicators analyzed (chosen for best data availability across countries and time):
  - labor force participation;
  - primary, secondary, and tertiary education;
  - maternal mortality;
  - adolescent fertility.
- At the onset of the crisis, SCM matches on gender inequality trends and economic conditions such as fiscal spending, external debt, and reserves.

### Methodology
- Method used: Synthetic Control Methods (SCM), producing custom-tailored control groups that closely match each program country’s pre-program trends.
- SCM is applied to address the counterfactual: the evolution of gender disparities had the IMF-supported program not been adopted.
- The approach includes bias correction and inference procedures (not detailed in the Executive Summary).

### Main findings
- No systematic differences in the evolution of gender indicators among crisis countries with and without IMF-supported programs.
- Specific quantitative finding:
  - In 87-90 percent of the examined cases, the impact of IMF-supported programs on gender inequality indicators is not statistically significantly different from the control group that did not adopt an IMF-supported program.
- In the minority of cases with statistically significant effects:
  - Effects are observed in both directions — some IMF-supported programs are associated with positive changes in gender inequality indicators, others with negative changes.
- The analysis indicates that the effects of economic crises alone were the primary drivers of observed changes in gender inequality in many instances.

### Interpretation and implications
- Previous literature associating IMF conditionality with gender inequality likely suffered from contaminated estimates that conflated crisis effects with program effects; credible control groups are necessary to separate these.
- Caution against generalizing from individual country experiences or small-sample studies, because statistically significant outcomes vary substantially across countries in both size and sign.
- The finding that gender inequality has declined in some cases under IMF-supported programs suggests that policy design matters and that attention to design can lead to better gender outcomes.

_Executive Summary, “Effects of IMF-Supported Programs on Gender Inequality,” IMF Working Paper (Executive Summary pages provided)._

### Introduction

### Introduction

### Background and motivation
- Economic crises disproportionately impact populations in vulnerable situations, which often include women. Recovery trajectories involving fiscal and monetary consolidations aim to stabilize macroeconomic environments and improve development prospects.
- Consolidations can, depending on design, disproportionately (positively or negatively) impact women and girls.
- Credible counterfactuals are crucial to determine whether observed changes in gender inequality are due to crises, economic adjustment, or conditionality in IMF-supported programs.
- Synthetic Control Methods (SCM) are applied to construct tailored control groups to isolate program effects from crisis effects.

### Scope of assessment
- Data horizon: 1990-2023.
- Universe of possible programs: 480.
- Programs analyzed with proper SCM control groups: up to 151 programs (depending on indicator and data availability).
- Gender indicators analyzed (seven): 
  - female labor force participation,
  - ratio of female to male labor force participation,
  - adolescent fertility rate,
  - maternal mortality rate,
  - gender parity index (GPI) for primary enrollment,
  - GPI for secondary enrollment,
  - GPI for tertiary enrollment.

### Key methodological points
- SCM constructs a custom synthetic control (weighted donor-pool average) for each program country to match nonlinear pre-crisis gender and selected macroeconomic trajectories (baseline predictors: lagged gender indicator and reserves in months of imports).
- Pre- and post-program window: 4 years pre- and 4 years post-program (program year treated as year T0, with T0 = 5 in nine-year observation example).
- Exclusion criteria: countries that received another program in the 4 years prior to the current program; donor-pool countries must not receive a program during the entire eight-year window (four years before and four years after).
- Bias correction: regression-based adjustment (SCM bias-corrected estimator) applied to address predictor discrepancies; inferential procedure uses in-space placebo permutations and p-values based on pre-/post-treatment MSPE ratios.
- Robustness checks: extended predictor set including government consumption and reserves as percent of external debt (yields smaller sample but qualitatively similar results).

### Data sources and program selection
- Primary data: World Bank World Development Indicators (WDI 2024), covering 1990–2023.
- Program list: IMF commitment database with cross-checks to MONA database; included all IMF-supported programs with ex-post conditionality (excluding RCF, RFI, FCL).
- Balanced panel requirement and donor-pool constraints reduce sample from 480 to at most 151 programs for some indicators.

### Example: Dominican Republic (2003 program)
- 2003 Stand-By Arrangement: $600 million, immediate disbursement $120 million.
- Crisis context cited: GDP growth contraction from 5.9 percent in 1999 to -1.3 percent in 2003; 80 percent currency depreciation (DOP 15.83 per USD in 1999 to DOP 29.38 per USD in 2003); central bank lost half of foreign currency reserves.
- SCM donor pool: 39 countries; synthetic weights for Dominican Republic female labor force participation: Mauritius (0.591), Belize (0.298), Saint Vincent and the Grenadines (0.097), Algeria (0.014).
- Result for Dominican Republic female labor force participation: bias-corrected gender impact gap statistically significant at the 5 percent level for all post-crisis years in the example; program effect was negative relative to synthetic control because recovery in the Dominican Republic was less pronounced than in the synthetic control.

### Main empirical findings (aggregated)
- Across programs and gender indicators, about 90 percent of IMF-supported programs do not exhibit statistically significant program-driven changes in gender inequality (i.e., gender impact gaps are statistically insignificant and thus indistinguishable from control-group experience).
- When statistically significant effects are observed (about 10 to 13 percent of cases), they are roughly equally split between positive and negative impacts.
- No program or time period shows significance at the 1 percent level; significant findings do not surpass the 5 percent threshold in aggregate analysis.
- Summary numeric results (Table 2 baseline with reserve coverage as predictor):
  - Adolescent Fertility Rate: 149 programs; 10 positive significant; 5 negative significant; 134 not significant (90%).
  - Female/Male Labor Force Participation: 142 programs; 8 positive significant; 9 negative significant; 125 not significant (88%).
  - Female Labor Force Participation: 151 programs; 10 positive significant; 13 negative significant; 128 not significant (85%).
  - Maternal Mortality Rate: 91 programs; 5 positive significant; 6 negative significant; 80 not significant (88%).
  - Primary Enrollment GPI: 65 programs; 2 positive significant; 7 negative significant; 56 not significant (86%).
  - Secondary Enrollment GPI: 27 programs; 6 positive significant; 3 negative significant; 18 not significant (67%).
  - Tertiary Enrollment GPI: 28 programs; 0 positive significant; 1 negative significant; 27 not significant (96%).
  - Totals (baseline reserve-coverage predictors): sum 653 observations; 41 positive significant (6%); 44 negative significant (7%); 568 not significant (87%).
- Robustness case (reserve coverage, government consumption, reserves as percent of external debt) yields similar patterns with slightly different counts (aggregate: 433 observations; 19 positive significant (4%); 26 negative significant (4%); 388 not significant (90%)).

### Interpretation and implications
- The vast majority of unusual gender impacts observed after economic crises cannot be attributed to IMF-supported programs once credible counterfactuals are applied; many observed changes are driven by the crisis and its recovery economics rather than program conditionality.
- Where program effects are statistically significant, there is no consistent direction—effects can be beneficial or adverse depending on country-specific circumstances.
- Aggregated (meta-analytic) common effect across individually matched SCM estimates is essentially zero and not statistically significant (not significant at the 90 percent level for aggregated effects).
- The analysis does not identify how specific conditionality items affect gender outcomes; it only assesses whether program countries experienced different gender trajectories compared to their synthetic controls.

### Robustness and limitations
- SCM requires well-matching pre-crisis trajectories; program cases with non-unique weighting vectors or inadequate donor pools are excluded.
- Robustness checks adding predictors (government consumption, external debt measures) reduce sample sizes due to additional data requirements but yield qualitatively similar conclusions.
- The study examines programs from 1994–2022 and reports gender effects through 1994–2023; programs adopted after 2022 (including those under updated IMF gender strategy) could not be evaluated with a full four-year post-program window in this analysis.

### Conclusion
- Using SCM to establish credible counterfactuals, the study finds no systematic, statistically robust effect of IMF-supported programs on key gender inequality indicators since 1994. Most post-crisis gender outcomes are statistically indistinguishable from outcomes observed in matched countries that did not adopt programs, implying crisis recovery dynamics drive the bulk of observed gender changes.

*Source: IMF Working Paper — Introduction (wpiea2024247-print-pdf).*

### Conclusion

### Conclusion

### Methodology and identification
- Objective: isolate the effects of IMF-supported programs on gender inequality by disentangling:
  - the economic crisis and the policy response to it, and
  - the impact of IMF-supported programs.
- Key methodological requirement: construct a tailored control group for each program country to examine the counterfactual—how the country’s gender indicator would have evolved with economic crisis but without a program.
- Approach used: Synthetic Control Methods to construct credible, individually tailored control groups for each IMF-supported program.
- Empirical scope: over 150 programs from 1994 to 2022.
- Diagnostic evidence of match quality: Table 2 and summary Figures 4.a to 4.g highlight the quality of the match between each program country’s pre-crisis gender indicator and economic trends with its control group.

### Main findings
- Indicators examined: seven different gender inequality indicators.
- Aggregate result: about 90 percent of program countries do not exhibit gender effects that are significantly different from those observed by the control group.
- Statistically significant effects:
  - A minority of programs show statistically significant effects on gender inequality.
  - Among those, effects are mixed: there are both negative and positive effects.
- Primary driver: overwhelming evidence that post-crisis economic contractions, rather than IMF-supported programs, are the primary drivers of observed gender disparities.
- Interpretation of association: the simple presence of an IMF-supported program may be associated with (negative) changes in gender inequality, but these changes are not produced by the program because the control group produced similar effects.

### Interpretation and contrast with prior literature
- Critique of previous literature: prior studies often lacked credible counterfactuals or specified them imprecisely, likely conflating effects of economic crises with effects of IMF-supported programs.
- Implication: much of the previous literature’s discussion of negative impacts of programs on gender may reflect conflated estimates that do not properly separate crisis effects from program effects.
- Positive cases: the decline in gender inequality observed in some program cases indicates that attention to policy design can produce better gender outcomes.

### Policy implications and limits
- Recent IMF strategies:
  - IMF 2019: strategy on social spending.
  - IMF 2022: strategy on mainstreaming gender into the IMF’s core operations.
- These strategies are designed to mitigate potential adverse effects on vulnerable populations.
- Limitations of this study:
  - The IMF policies cited (IMF 2019 and IMF 2022) are too recent to be examined for effects in this study due to data constraints.
  - The mainstreaming of gender into IMF operations has so far taken a cautious approach to integrating gender into programs, limiting the ability to assess their full impact to date.
- Policy lesson: deliberate attention to the design of policies within IMF-supported programs can lead to improved gender outcomes.

*IMF WORKING PAPERS Effects of IMF-Supported Programs on Gender Inequality  
INTERNATIONAL MONETARY FUND*

### Annex A. Country-by-Country Results: Effects of

### Annex A. Country-by-Country Results: Effects of IMF-Supported Programs on Gender Inequality

### A.1 Adolescent Fertility Rate (Table A.1.1 and Robustness Table A.1.2)
- Definition: A positive gender impact gap implies a higher adolescent fertility rate (births per 1,000 women ages 15-19) in program countries compared to the control group.
- Predictors (A.1.1): Adolescent Fertility (lagged), Reserve Coverage (lagged).
- Predictors (A.1.2 robustness): Adolescent Fertility (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (t and subsequent t+1…t+4 shown in table A.1.1):
  - Bolivia 1994: t = 2.16; t+1 = 0.08; t+2 = 1.52; t+3 = 6.68; t+4 = 9.72; t+5 shown = 10.69
  - Chad 1994: t = 9.56; t+1 = 1.46; t+2 = 11.78; t+3 = 17.71; t+4 = 22.37; t+5 = 29.10
  - Sierra Leone 1994: t = 0.19; t+1 = 2.22*; t+2 = 0.51; t+3 = 5.83; t+4 = 10.46; t+5 = 12.08
  - Venezuela 1996: t = -0.78; t+1 = 3.87*; t+2 = 6.06*; t+3 = 8.68*; t+4 = 14.86*; t+5 = 13.76*
  - Guinea 1997: t = -0.45; t+1 = -2.89*; t+2 = -8.52**; t+3 = -8.27**; t+4 = -12.64**; t+5 = -12.63*
  - Chad 2000 (A.1.1): t = -0.16; t+1 = -2.56**; t+2 = 2.23**; t+3 = 1.59*; t+4 = 3.80*; t+5 = 9.64**
- Selected robustness results (Table A.1.2):
  - Chad 2000 (robust): t = 3.51; t+1 = -4.60; t+2 = -4.28; t+3 = -2.26; t+4 = 5.99; t+5 = 26.48
  - Rwanda 1998 (robust): t = 0.20; t+1 = -2.32*; t+2 = -6.28*; t+3 = -9.65*; t+4 = -11.80*; t+5 = -15.57*
  - Bangladesh 2012 (robust): t = 3.33; t+1 = 3.88; t+2 = 9.25; t+3 = 6.03; t+4 = 4.01; t+5 = 2.72
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

### A.2 Ratio of Female to Male Labor Force Participation Rate (Table A.2.1 and Robustness Table A.2.2)
- Definition: A positive gender impact gap implies a higher labor force participation of women relative to men in IMF-supported programs compared to the control group.
- Predictors (A.2.1): Ratio of Female to Male Labor Force (lagged), Reserve Coverage (lagged).
- Predictors (A.2.2 robustness): Ratio of Female to Male Labor Force (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (Table A.2.1):
  - Sierra Leone 2001: t = 0.00; t+1 = -0.01**; t+2 = -0.00**; t+3 = -0.00**; t+4 = -0.01**; t+5 = -0.01**
  - Djibouti 2008: t = 0.00; t+1 = 0.02**; t+2 = 0.02**; t+3 = 0.02**; t+4 = 0.04**; t+5 = 0.04**
  - Tajikistan 2009: t = 0.00; t+1 = 0.02**; t+2 = 0.03**; t+3 = 0.05**; t+4 = 0.05**
  - Gambia 2012: t = 0.00; t+1 = 0.01; t+2 = 0.04; t+3 = 0.06; t+4 = 0.08; t+5 = 0.10
  - Moldova 2021: t = 0.00; t+1 = -0.02**; t+2 = 0.01; t+3 = 0.02*
- Selected robustness results (Table A.2.2):
  - Colombia 1999 (robust): t = 0.00; t+1 = -0.03**; t+2 = -0.03**; t+3 = -0.03**; t+4 = -0.06**; t+5 = -0.06**
  - Turkey 1999 (robust): t = 0.00; t+1 = 0.08; t+2 = 0.09; t+3 = 0.12; t+4 = 0.21; t+5 = 0.26
  - Gambia 2012 (robust): t = 0.00; t+1 = 0.00; t+2 = 0.03; t+3 = 0.06; t+4 = 0.09; t+5 = 0.12
  - Mongolia 2017 (robust): t = 0.00; t+1 = 0.01*; t+2 = -0.05**; t+3 = -0.03**; t+4 = -0.03**; t+5 = -0.04**
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

### A.3 Female Labor Force Participation Rate (Table A.3.1 and Robustness Table A.3.2)
- Definition: A positive gender impact gap implies a higher female labor force participation rate in IMF-supported programs compared to the control group. (Female labor force participation in percent, lagged used as predictor.)
- Predictors (A.3.1): Female Labor Force Participation (in percent, lagged), Reserve Coverage (lagged).
- Predictors (A.3.2 robustness): Female Labor Force Participation (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (Table A.3.1):
  - Turkey 1994: t = -3.59; t+1 = 11.14; t+2 = 10.01; t+3 = 7.64; t+4 = 5.04; t+5 = 5.21
  - Korea 1997: t = 0.07; t+1 = 0.38; t+2 = -2.41**; t+3 = -2.69**; t+4 = -1.40*; t+5 = -0.97*
  - Lithuania 2000: t = -0.11; t+1 = -2.84**; t+2 = -4.41*; t+3 = -5.46*; t+4 = -2.37*; t+5 = -5.09*
  - Sri Lanka 2001: t = -9.10; t+1 = 2.20; t+2 = 3.60; t+3 = 5.68; t+4 = 5.19; t+5 = 7.54
  - Albania 2014: t = -5.58; t+1 = 2.28; t+2 = 7.00; t+3 = 10.34; t+4 = 10.53; t+5 = 12.76
  - Moldova 2016: t = 2.60; t+1 = -1.41; t+2 = -3.56; t+3 = -5.17; t+4 = -2.41; t+5 = -0.48
  - Pakistan 2019: t = 0.17; t+1 = 1.24*; t+2 = 2.63*; t+3 = 2.94*; t+4 = 2.36*; t+5 = 2.28*
- Selected robustness highlights (Table A.3.2):
  - Tanzania 1996 (robust): t = -0.04; t+1 = 0.54*; t+2 = 0.94*; t+3 = 1.25*; t+4 = 2.32*; t+5 = 3.36*
  - Moldova 2006 (robust): t = -1.01; t+1 = 2.61; t+2 = 6.20; t+3 = 6.01; t+4 = 5.73; t+5 = 7.54
  - Haiti 2015 (robust): t = -0.11; t+1 = 1.13*; t+2 = 1.76**; t+3 = 2.87**; t+4 = 2.98**; t+5 = 2.47**
  - Egypt 2016 (robust): t = -0.01; t+1 = -0.31**; t+2 = -1.03**; t+3 = -5.89**; t+4 = -9.15**; t+5 = -11.77**
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

### A.4 Maternal Mortality Rate (Table A.4.1 and Robustness Table A.4.2)
- Definition: A positive gender impact gap implies a higher rate of pregnancy-related deaths (per 100,000 live births) in IMF-supported programs compared to the control group.
- Predictors (A.4.1): Maternal Mortality (lagged), Reserve Coverage (lagged).
- Predictors (A.4.2 robustness): Maternal Mortality (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (Table A.4.1):
  - Burundi 2004: t = 82.17; t+1 = 22.13; t+2 = -6.53; t+3 = 8.61; t+4 = -19.44; t+5 = -70.37
  - Zambia 2004: t = 35.59; t+1 = 29.41; t+2 = 47.87; t+3 = 64.46; t+4 = 77.58; t+5 = 108.21
  - Sierra Leone 2006: t = -3.44; t+1 = -83.30**; t+2 = -164.01**; t+3 = -328.25**; t+4 = -346.15**; t+5 = -419.61**
  - Mozambique 2009: t = -0.92; t+1 = 2.63; t+2 = -14.49; t+3 = -16.84; t+4 = -25.23; t+5 = -42.12*
  - Lesotho 2010: t = 69.67; t+1 = 112.85; t+2 = 170.26; t+3 = 137.18; t+4 = 101.77; t+5 = 181.48
  - Kenya 2011: t = 8.80; t+1 = 35.37; t+2 = 66.93; t+3 = 110.05; t+4 = 111.83*; t+5 = 92.88
  - Guinea 2012 (A.4.1): t = -3.17; t+1 = -51.38*; t+2 = -62.94*; t+3 = -86.73*; t+4 = -66.05*; t+5 = -95.67*
- Selected robustness highlights (Table A.4.2):
  - Cameroon 2005 (robust): t = 19.27; t+1 = 67.17*; t+2 = 90.71*; t+3 = 117.15*; t+4 = 120.36*; t+5 = 157.45*
  - Chad 2005 (robust): t = 51.27; t+1 = 85.68; t+2 = 187.72; t+3 = 272.57; t+4 = 242.86; t+5 = 287.30
  - Sierra Leone 2006 (robust): t = 64.54; t+1 = -44.42; t+2 = -114.07; t+3 = -292.11; t+4 = -362.29; t+5 = -295.40
  - Central African Republic 2012 (robust): t = 17.94; t+1 = -15.75; t+2 = 62.74; t+3 = 73.59; t+4 = 116.63*; t+5 = 136.41*
  - Gambia 2012 (robust): t = 48.41; t+1 = 16.66; t+2 = 60.04; t+3 = 36.88; t+4 = 78.21; t+5 = 70.10
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

### A.5 Primary Enrollment Gender Parity Index (Table A.5.1 and Robustness Table A.5.2)
- Definition: A positive gender impact gap implies a higher ratio of girls to boys enrolled in public and private primary schools in IMF-supported programs compared to the control group.
- Predictors (A.5.1): Primary Enrollment Gender Parity (lagged), Reserve Coverage (lagged).
- Predictors (A.5.2 robustness): Primary Enrollment Gender Parity (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (Table A.5.1):
  - Georgia 2001: t = -0.01; t+1 = 0.06*; t+2 = 0.08; t+3 = 0.07; t+4 = 0.04; t+5 = 0.06
  - Burundi 2004: t = 0.00; t+1 = 0.01; t+2 = 0.03**; t+3 = 0.07*; t+4 = 0.05*; t+5 = 0.09*
  - Nepal 2003: t = -0.02; t+1 = -0.01; t+2 = -0.01; t+3 = 0.02; t+4 = 0.05; t+5 = 0.09
  - Morocco 2012 (robust in A.5.2): t = 0.00; t+1 = 0.01**; t+2 = 0.01**; t+3 = 0.00**; t+4 = -0.01**; t+5 = -0.01**
  - Note: selected robustness entries are in Table A.5.2.
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

### A.6 Secondary Enrollment Gender Parity Index (Table A.6.1 and Robustness Table A.6.2)
- Definition: A positive gender impact gap implies a higher ratio of girls to boys enrolled in public and private secondary schools in IMF-supported programs compared to the control group.
- Predictors (A.6.1): Secondary Enrollment Gender Parity (lagged), Reserve Coverage (lagged).
- Predictors (A.6.2 robustness): Secondary Enrollment Gender Parity (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (Table A.6.1):
  - Mozambique 2009: t = 0.00; t+1 = 0.02**; t+2 = 0.03**; t+3 = 0.06**; t+4 = 0.06**; t+5 = 0.05**
  - Cameroon 2017: t = 0.00; t+1 = 0.01*; t+2 = 0.08*; t+3 = 0.11*; t+4 = 0.19*; t+5 = 0.48*
  - Saint Kitts and Nevis 2011: t = 0.00; t+1 = 0.09*; t+2 = 0.05*; t+3 = 0.16*; t+4 = 0.00*; t+5 = 0.03*
  - Hungary 2008: t = 0.00; t+1 = -0.02**; t+2 = -0.01**; t+3 = -0.02**; t+4 = -0.02**; t+5 = -0.01**
- Selected robustness highlights (Table A.6.2):
  - Mozambique 2015 (robust): t = 0.00; t+1 = 0.12; t+2 = 0.15; t+3 = 0.03; t+4 = 0.08; t+5 = 0.12
  - Pakistan 2013 (robust): t = -0.07; t+1 = -0.07; t+2 = -0.02; t+3 = 0.01; t+4 = 0.05; t+5 = 0.08
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

### A.7 Tertiary Enrollment Gender Parity Index (Table A.7.1 and Robustness Table A.7.2)
- Definition: A positive gender impact gap implies a higher ratio of females to males enrolled in public and private post-secondary education in IMF-supported programs compared to the control group.
- Predictors (A.7.1): Tertiary Enrollment Gender Parity (lagged), Reserve Coverage (lagged).
- Predictors (A.7.2 robustness): Tertiary Enrollment Gender Parity (lagged), Reserve Coverage (lagged), Government Consumption, Reserves (Percent of External Debt).
- Selected country-year impact gaps (Table A.7.1):
  - Sri Lanka 2016: t = 0.04; t+1 = 0.16; t+2 = 0.26; t+3 = 0.28; t+4 = 0.35; t+5 = 0.39
  - Moldova 2016: t = 0.00; t+1 = 0.06; t+2 = 0.07; t+3 = 0.10; t+4 = 0.13*; t+5 = 0.17*
  - Rwanda 2016: t = 0.00; t+1 = -0.07*; t+2 = -0.00*; t+3 = -0.12*; t+4 = -0.14*; t+5 = -0.23**
  - Mozambique 2009: t = -0.30; t+1 = -0.05; t+2 = -0.02; t+3 = 0.16; t+4 = 0.04; t+5 = 0.05
- Selected robustness highlights (Table A.7.2):
  - Moldova 2016 (robust): t = -0.03; t+1 = 0.04; t+2 = 0.02; t+3 = 0.07; t+4 = 0.12; t+5 = 0.17
  - Mozambique 2009 (robust): t = -0.22; t+1 = -0.09; t+2 = 0.03; t+3 = 0.08; t+4 = 0.10; t+5 = 0.11
- Note: *, **, and *** represent significance at the 0.10, 0.05, and 0.01 levels, respectively.

*Annex A tables report country-level synthetic control gender impact gaps for multiple gender-relevant indicators (adolescent fertility, labor force participation, maternal mortality, and enrollment gender parity indices) with the predictors used and robustness specifications shown in the tables.*

### Annex B. Data Sources

### Annex B. Data Sources

### Databases and key variables (gender indicators)
- WDI DATABASE (GENDER INDICATORS) — World Bank World Development Indicator (WB 2023a)
  - Adolescent Fertility Rate
    - WB: SP.ADO.TFRT
  - Female/Male Labor Force Participation
    - WB: SL.TLF.CACT.FM.ZS
  - Female Labor Force Participation
    - WB: SL.TLF.CACT.ZS
  - Maternal Mortality Rate
    - WB: SH.STA.MMRT
  - Primary Enrollment GPI
    - WB: SE.ENR.PRIM.FM.ZS
  - Secondary Enrollment GPI
    - WB: SE.ENR.SECO.FM.ZS
  - Tertiary Enrollment GPI
    - WB: SE.ENR.TERT.FM.ZS

### Databases and key variables (control variables)
- WDI DATABASE (CONTROL VARS) — World Bank World Development Indicator (WB 2023a)
  - Total reserves in months of imports
    - WB: FI.RES.TOTL.MO
  - Total reserves (% of total external debt)
    - WB: FI.RES.TOTL.DT.ZS
  - Government final consumption expenditure (% of GDP)
    - WB: NE.CON.GOVT.ZS

### IMF and World Bank program and classification sources
- IMF LENDING DATABASE
  - IMF commitment database — IMF-supported Programs
  - URL: https://www.imf.org/external/np/fin/tad/query.aspx
- WORLD BANK INCOME CLASSIFICATION DATABASE — World Bank (WB 2023b)
  - World Bank Income Classification
  - Data download reference: http://databank.worldbank.org/data/download/site-content/OGHIST.xls

### Additional data references cited in the annex
- World Bank, 2024. World Development Indicators Data Files.
  - URL: https://databank.worldbank.org/source/world-development-indicators
- International Monetary Fund. 2024c. “Monitoring of Fund Arrangements Database (MONA).”
  - URL: https://www.imf.org/external/np/pdr/mona/index.aspx

### Notes on content scope
- The annex lists databases, variables, and the exact database codenames used for the analysis.
- The annex also provides external URLs for IMF and World Bank query or download pages referenced in the data compilation.

*Source: Annex B. Data Sources, wpiea2024247-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024247-print-pdf.pdf_