## 5.1 Flexibility and Revisions to Growth Projections

## Source details

**Canonical URL:** [5.1 Flexibility and Revisions to Growth Projections](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/np/pp/eng/2012/_061812b.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/np/pp/eng/2012/_061812b.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/np/pp/eng/2012/_061812b.pdf.json)

---

### Overview
- Analysis period: between 2006 and September 2011, with some attention to 2002-05.
- Methodology: descriptive and regression analysis of programs’ macroeconomic adjustment and access, complemented by case studies, surveys, and program document review.
- Focus: whether program design and conditionality were consistent and even-handed; responsive to country-specific needs; based on reasonably good macroeconomic projections; and flexible to evolving country circumstances.

### Main findings on comparability and even-handedness
- Program design generally appeared tailored to country needs, even-handed, and flexible across the sample.
- Regressions indicate a limited number of initial conditions can explain programs’ macroeconomic adjustment and access, supporting a conclusion of even-handedness.
- Fit of regressions:
  - Generally good for GRA programs.
  - Weaker for PRGT programs, reflecting greater heterogeneity rather than clear evidence of inconsistent design.
- Specific empirical points:
  - Bayesian Model Averaging (BMA) methodology used for model selection.
  - Euro Area (EA) dummy was a powerful explanatory variable for access.
  - Stakeholder surveys: few respondents disagreed that programs have equivalent conditionality; around half of country authorities and donors responded that they did not know.
- Interpretation: an average fit of 0.5 is considered good for this method; lower R² for PRGT regressions likely reflects heterogeneity of PRGT countries, programs, and objectives.

### Macroeconomic projections and projection accuracy
- Overall: projections underpinning program design in the 2006-11 sample were generally unbiased and did not exhibit a systematic optimistic projection bias, contrary to studies based on earlier periods.
- Projection errors have grown for GRA cases; shocks from the global financial crisis help explain this and suggest room for improvement.
- Detailed patterns:
  - GRA: some evidence of pessimistic projections for reserves and perhaps current account balance; MEs rarely significantly different from zero for growth, government balance, and inflation.
  - PRGT: weak evidence for optimistic projections for growth and pessimistic projections for current account balance and reserves; reserves MEs often negative and significantly different from zero.
  - Programs did not appear more optimistic than non-program forecasts.

### Flexibility, revisions, and implementation
- Flexibility increased compared to the past, especially in:
  - Modifications of conditionality.
  - Augmentations of access.
- Increased flexibility helped maintain high implementation rates despite the global recession; conditionality implementation rate remained around 90 percent.
- Flexibility mechanisms and effects:
  - Adjustors built into program design used in more than half of QPCs on average; higher use in low-income country programs.
  - Modifications of QPCs and augmentations of access occurred when downward revisions to growth were largest (about 4 percentage points on average).
  - Application of adjustors increased implementation rates from 70 percent to 90 percent compared to a situation without adjustors.
  - Combining reviews, rephasing disbursements, and program extensions provided operational flexibility; extensions were more common in fragile states.

### Program design challenges, risks, and caveats
- "Wave 2" crisis programs (approved after August 2009) faced notable design difficulties:
  - Narrowed policy space increased the trade-off between adjustment and financing.
  - Exceptional challenges: high public debt; sustained loss of market access; low growth; competitiveness issues; currency union membership; macro-financial linkages; spillovers; systemic risk.
- Financial and program outcome considerations:
  - These relatively few programs represent a large proportion of access to Fund financing and create financial risks to the institution.
  - Too early to assess final outcomes for many programs; the first Greece program already encountered significant difficulties.
- Specific caveats and potential re-examination areas:
  - Fiscal adjustment in fragile states tended to be as large as or larger than other PRGT cases, controlling for other factors.
  - In some capital account crisis cases, there may be scope to relax initial fiscal targets since medium-term stabilization was better than expected despite narrowly missing initial fiscal targets.

### Euro Area membership, access, and adjustment
- Membership in the Euro Area was a significant variable explaining the level of access to Fund financing in GRA programs; its inclusion explained much of the variation represented by high access levels.
- Euro Area membership was not statistically significant in explaining adjustment, although “wave 2” countries (which include Euro Area programs) were associated with somewhat higher fiscal adjustment.
- Exceptional access examples at approval:
  - Ireland: 19.5 billion SDRs or 2,322 percent of quota at program approval (Ireland’s quota increased substantially in March, 2011, reducing total access to 1,548 percent of quota).
  - Greece: 26.4 billion SDRs or 3,212 percent of quota at program approval.
  - Portugal: 23.7 billion SDRs or 2,306 percent of quota.
- Program documents cited spillovers and systemic risks as justifications for higher access; general cross-border bank claims and IIP proxies did not capture contagion risks adequately.

### Fiscal adjustment — magnitude, composition, timing, and outturns
- General assessments:
  - Fiscal adjustment generally appropriate given initial conditions, program characteristics, and objectives.
  - More flexible in magnitude and timing than in the past; magnitude generally restrained because of concerns about output during the global crisis.
  - PRGT: fiscal adjustment larger for countries with stronger institutions (rule of law), though fragile states showed offsetting results.
- Timing:
  - Adjustment often back-loaded to accommodate growth concerns.
  - GRA programs: overall balance programmed to improve by just 1 percent on average in the first two program years.
  - PRGT cases: expenditure planned to increase early, supported by grants and debt relief, and remain above initial GDP ratio until the third year after initiation.
- Composition:
  - PRGT high-deficit cases: adjustment relied partly on expenditure restraint; other PRGT cases relied more on revenue and grants.
  - GRA high-deficit cases: adjustment fell mainly on spending.
  - Social spending on health and education and capital expenditures were fairly well protected; capital expenditures increased in PRGT programs both as a share of GDP and overall spending.
- Outturns and performance:
  - GRA programs initially adjusted faster than projected and about as much as programmed toward the end of the projection horizon; capital account crises missed targets throughout.
  - Regression: programs requiring larger fiscal adjustment in a given year were less likely to meet that year’s QPCs.
  - Projected adjustment, not initial fiscal balance, determines program performance; GRA programs with large initial deficits performed better according to regressions.

### Regression findings explaining programmed fiscal adjustment (Table 3.1 summary)
- Statistically significant explanatory variables (significance at the 10 percent level or higher):
  - Macro conditions: Fiscal balance at t-1 (-), Public debt at t-1 (+), IIP liabilities at t-1 (+), Change in current account balance from t-2 to t-1 (-)
  - Country characteristics: Rule of law (-) for GRA; Fragile states dummy (+) and Rule of law (+) for PRGT
  - Program characteristics: Aid dependence (-), Wave 2 dummy (+)
- Adj. R-squared:
  - GRA: 0.807
  - PRGT: 0.500

### Key descriptive and regression statistics (selected)
- GRA sample descriptive statistics:
  - Access (multiple of quota): Mean 4.57, Std. Dev. 6.21, Min 0.16, Max 32.12
  - Programmed fiscal balance adj. (percent of GDP): Mean 2.39, Std. Dev. 5.22, Min -12.24, Max 28.22
  - Public Debt/GDP (percent of GDP): Mean 51.30, Std. Dev. 38.08, Min 7.76, Max 164.97
  - IIP liabilities (US$ billions): Mean 147.63, Std. Dev. 484.30, Min 0.54, Max 3,544.04
  - Total cross-border bank claims (US$ billions): Mean 37.46, Std. Dev. 95.41, Min 0.14, Max 651.37
  - R-squared adjusted reported in selected regressions: 0.807, 0.955, 0.777, 0.494, 0.874, 0.828, 0.564
- Notable regression coefficients (GRA Appendix Table I.1 highlights):
  - Euro Area dummy: 19.68***
  - Currency Union dummy: 3.68***, 3.97***
  - Public debt (% of GDP): 0.03**
  - Total cross-border bank claims: 0.06***
  - Fiscal balance (percent of GDP): -0.45*** and 0.15***
  - Precautionary program dummy: -2.03***
  - Wave 1 dummy: 3.04***
  - Dummy for post-2006 programs: -3.61***

### Conditionality on expenditure measures and protecting the poor (Box 2)
- Scope and common measures:
  - Most common spending measures required: adjustment of utility tariffs and increases of domestic petroleum prices.
  - Some price increases preceded/accompanied by automatic price adjustment mechanisms to smooth adjustments and depoliticize increases.
- Incidence (2006–2010 case studies):
  - Nine of 18 case-study countries subject to price-affecting conditionality: examples include Uganda (2006), Pakistan (2006), Ghana (2009), Moldova (2006 and 2010), Sierra Leone (2006), The Gambia (2007), Togo (2008), Dominican Republic (2009).
  - Automatic fuel pricing conditionality present in The Gambia, Ghana, Togo.
- Analysis and TA:
  - In five of nine countries, price adjustment conditionality accompanied by distributive impact analysis.
  - Since 2002, Fiscal Affairs Department carried out 15 TA missions including Ghana and Moldova.
  - Where distributive analysis absent, staff sought World Bank support and included conditionality to strengthen social protection (e.g., Pakistan).
  - Where no TA, conditionality on price increases generally accompanied by a benchmark requiring expansion of social safety nets.

### Exchange rate flexibility and Fund-supported programs (Box 5)
- Key findings (2006–11):
  - Countries increasingly moved toward more flexible exchange rate regimes; about half of program countries considered exchange rate policies a strategy for program objectives.
  - About a quarter moved to more flexible arrangements, notably during the global crisis.
  - Explicit conditionality related to exchange rate policies and restrictions existed in 16 programs (8 GRA and 8 PRGT) launched during 2006-11.
  - Regression analysis: program countries did not move to more flexible regimes more often than non-program peers.
- Factors influencing moves to flexibility:
  - Initial level of reserves and regime type were important:
    - Low reserves × Fixed: 9 programs (move to greater flexibility)
    - Low reserves × Flexible: 36 programs (enhance flexibility to reduce reserve loss)
    - High reserves × Fixed: 18 programs (draw on reserve buffers)
    - High reserves × Flexible: 8 programs (use reserves to avoid large exchange rate movements)
- Institutional reforms supporting flexibility: rules-based FX intervention, central bank liquidity management, stronger bank supervision.

### The Vienna Initiative (Box 7)
- Context: Fall 2008 systemic concerns for CESE due to large Western bank exposures (US$450 billion) and overheating imbalances.
- Launch and components:
  - Informal discussions Nov. 2008; inaugural meeting Jan. 23, 2009.
  - Joint IFI Initiative launched Feb. 27, 2009; EBRD, World Bank, and EIB disbursed €33 billion over next two years.
- Commitments and outcomes:
  - Parent banks committed to maintain exposure and recapitalize subsidiaries; IFIs pledged financial assistance.
  - Feared meltdown avoided; all fixed exchange rate regimes held up; only Latvia experienced a full-fledged banking crisis (onset predated the Initiative).
  - Western bank exposure to CESE declined little, reflecting effective private sector "bail-in" under BCI arrangements.

### Projection bias and robustness (Section 42 & Section 52)
- Projection bias:
  - Evidence pointed to an overly pessimistic bias for some variables, not an overly optimistic bias.
  - No evidence that Fund program projections were more optimistic than non-program forecasts.
- Robustness checks:
  - Including public rollover needs in GRA regressions did not explain higher Euro Area access or divergent programmed paths; rollover needs selected once by BMA for structural conditionality.
  - Political economy variables added to regressions generally provided no consistent explanatory power; where significant they often carried counterintuitive signs and suggested over-fitting.
  - Iterative BMA limited model size to 15 regressors; baseline BMA-selected models re-estimated by OLS for presentation.
- Political economy and budget support findings:
  - Political economy specifications occasionally statistically preferred but gains limited and likely over-fitting.
  - Budget support programs: programmed fiscal adjustments larger by 1¾ percent of GDP in GRA cases and by 2½ percent of GDP in PRGT cases; GRA budget support programs include on average 1.3 more structural conditions per review; PRGT budget support programs envisage 2 percentage points of GDP less inflation reduction.

### Monetary policy, PRGT countries, and performance determinants
- PRGT monetary policy implementation often constrained by weak transmission, inefficient frameworks, and lack of market prerequisites; coordination of monetary and fiscal operations, debt management, and government cash management critical (examples: Uganda, Sierra Leone, Moldova).
- Case-study reforms emphasized monetary anchors, moves toward flexible exchange rates and inflation targeting in some crisis-era programs, and central bank role clarification.
- Determinants of QPC performance (selected regression results from Table III.1):
  - Projected fiscal deficit in t and projected fiscal adjustment in t+1 generally associated with lower share of QPCs met.
  - Number of prior actions associated with lower QPC implementation (coefficients e.g., -0.0430**, -0.0914*).
  - Sample sizes: N up to 3,621 in columns; R-sq range reported (example R-sq 0.625, adj. R-sq 0.531).

### Financial sector conditionality and classifications
- New classification categories and incidence (selected percentages):
  - PRGT before crisis: Regulation and supervision 38%; Bank restructuring and resolution 18%; Other 19%.
  - GRA before crisis: Regulation and supervision 48%; Bank restructuring and resolution 19%; Other 14%.
  - Capital account crises: Regulation and supervision 26%; Bank restructuring and resolution 37%; Liquidity and solvency support 10%.
  - PRGT crisis: Financial stability assessment/contingency 23%; Regulation and supervision 39%; Bank restructuring and resolution 16%.
  - Fragile states: Regulation and supervision 42%; Bank restructuring and resolution 24%; Other 14%.

### Euro Area program projections (Appendix IV highlights)
- Downward revisions in GDP growth forecasts present in all three Euro Area programs (Ireland, Greece, Portugal).
- Fiscal deficits in 2011 larger than initially forecasted for Greece and Portugal; public debt larger than forecast in 2011 for Greece.
- Projection revisions and review counts: Ireland (latest review 4th), Greece (5th), Portugal (2nd); some projections past 2011 are revisions to previous board approval projections.

*Source: _061812b - 5.1 Flexibility and Revisions to Growth Projections (excerpt).*

### 5.1 Flexibility and Revisions to Growth Projections ...............................................................48

### 5.1 Flexibility and Revisions to Growth Projections

### Overview
- Paper examines whether program design and conditionality were: consistent and even-handed; responsive to country-specific needs; based on reasonably good macroeconomic projections; and flexible to evolving country circumstances.
- Analysis focuses on the period between 2006 and September 2011, with some attention to the 2002-05 period.
- Methodology: descriptive and regression analysis of programs’ macroeconomic adjustment and access, complemented by case studies, surveys, and program document review.

### Main findings on comparability and even-handedness
- Program design generally appeared tailored to country needs, even-handed, and flexible across the sample.
- Regressions indicate a limited number of initial conditions can explain programs’ macroeconomic adjustment and access, supporting a conclusion of even-handedness.
- Fit of regressions:
  - Generally good for GRA programs.
  - Weaker for PRGT programs, reflecting greater heterogeneity rather than clear evidence of inconsistent design.
- Specific empirical points:
  - The Bayesian Model Averaging (BMA) methodology was used for model selection.
  - The Euro Area (EA) dummy was a powerful explanatory variable for access and warrants further examination.
  - Stakeholder surveys: few respondents disagreed that programs have equivalent conditionality; around half of country authorities and donors responded that they did not know.
- Regression fit interpretation:
  - An average fit of 0.5 is considered good for this type of method.
  - Lower R² for PRGT regressions likely reflects heterogeneity of PRGT countries, programs, and objectives rather than inconsistent design.

### Macroeconomic projections and projection accuracy
- Macroeconomic projections underpinning program design in the 2006-11 sample were generally unbiased and did not exhibit a systematic optimistic projection bias, contrary to studies based on earlier periods.
- However, projection errors have grown for GRA cases; the shocks associated with the global financial crisis help explain this outcome, and room for improvement exists.

### Flexibility, revisions, and implementation
- Flexibility increased compared to the past, especially in:
  - Modifications of conditionality.
  - Augmentations of access.
- Increased flexibility helped maintain high implementation rates despite the global recession.
- The analysis focuses mainly on initial program design because:
  - Initial design sets broad program outlines.
  - Numerous subsequent revisions make comprehensive treatment impractical.

### Program design challenges, risks, and caveats
- "Wave 2" crisis programs (approved after August 2009) faced notable design difficulties:
  - Narrowed policy space in many countries increased the trade-off between adjustment and financing.
  - Exceptional design challenges included: high public debt; sustained loss of market access; low growth; competitiveness issues; currency union membership; macro-financial linkages; spillovers; and systemic risk.
  - Membership in the Euro Area was associated with higher access, and while large systemic risks justified access levels, the reference to systemic risk could have benefitted from more in-depth analysis at program inception.
- Financial and program outcome considerations:
  - These relatively few programs represent a large proportion of access to Fund financing and create financial risks to the institution.
  - It is too early to assess final outcomes for many programs in the sample; the first Greece program already encountered significant difficulties.
- Other caveats:
  - Findings are founded on aggregate overview; certain programs exhibited specific design flaws as indicated by EPAs and EPEs, though no systematic design flaws were apparent.
  - Two findings suggest scope to re-examine the rigor of initial fiscal adjustment in some programs:
    - Fiscal adjustment in programs in fragile states tended to be as large as or larger than other PRGT cases, controlling for other factors.
    - In capital account crisis cases, while design drew on past lessons, there may be room in certain cases for further relaxing initial fiscal targets, since stabilization in the medium term was better than expected despite narrowly missing fiscal targets in the initial program period.

*Italic: Source: _061812b - 5.1 Flexibility and Revisions to Growth Projections (excerpt).*

### 11. Membership in the Euro Area was a significant variable in explaining the level

### 11. Membership in the Euro Area was a significant variable in explaining the level of access to Fund financing in GRA programs

### Euro Area membership, access, and adjustment
- Membership in the Euro Area was a significant variable in explaining the level of access to Fund financing in GRA programs; its inclusion explained much of the variation represented by the high levels of access for these programs.
- Membership in the Euro Area was not statistically significant in explaining adjustment, although the “wave 2” countries (which include the Euro Area programs) were associated with somewhat higher fiscal adjustment.
- The extraordinarily high access in the Euro Area programs was viewed as necessary given the systemic risks from their crises and large debt rollover needs.

### Spillovers, systemic risk, and program design
- Program documents cited spillovers and systemic risks as specific justifications for higher access, emphasizing risks to European and global financial systems.
- General financial interconnectedness (proxied by total cross-border bank claims and liabilities from the international investment position) was not found to be an important explanatory variable for access, and these proxies do not adequately capture the contagion risks the programs were designed to contain.
- Euro Area program design faced additional constraints: membership in a currency union (eliminating nominal exchange rate devaluation), competitiveness problems, sustained total loss of market access, and complexity from large-scale co-financing from the European Union.
- The analysis and presentation of systemic risks in program documents expanded over time as risks evolved and became more prominent during implementation; the text recommends emulating this increased coverage in future exceptional access cases at program inception.

### Country cases and precedents
- The 2008 Hungary SBA explicitly cited reducing regional spillovers as an important objective because of large exposure of the largest Hungarian bank to Central, Eastern, and Southeastern Europe (CESE) and exposures of several Euro Area banks to Hungary via subsidiaries; this helped set a precedent for cross-border banking supervision and resolution frameworks at the European Union level.
- Exceptional access examples and proposed access levels at program approval:
  - Ireland: 19.5 billion SDRs or 2,322 percent of quota at program approval (Ireland’s quota increased substantially in March, 2011, reducing total access to 1,548 percent of quota).
  - Greece: 26.4 billion SDRs or 3,212 percent of quota at program approval.
  - Portugal: 23.7 billion SDRs or 2,306 percent of quota.
- These proposed exceptional access levels entailed substantial risks to the Fund in terms of outstanding credit stock and projected debt service over an extended period, but were justified by potential rapid spread of crises through financial exposures to and from peripheral Europe.
- Initial Greece program documentation lacked fully developed or quantified systemic risk analysis; deeper analysis in subsequent Euro Area programs improved decision-making.

### III. Adjustment and access in the design of Fund-supported programs — overview
- Program design aimed to achieve economic stabilization through a combination of adjustment and financing, including Fund access; promoting growth and poverty reduction was a goal in nearly all PRGT programs and about half of GRA cases.
- The section uses descriptive and regression analysis to examine programmed adjustment in fiscal, monetary, and external sectors (magnitude, composition, timing, outturns), factors explaining levels of access, extent of structural conditionality, and assessment of macroeconomic policy mixes for different policy challenges.
- Definition note: “adjustment” refers to planned or projected adjustment at program outset; projection horizon is t-1 to t+3 (five years from one period before initiation to three periods after initiation).
- Sample basis for averages: 36 GRA and 57 PRGT programs initiated between 2006 and 2011; 25 GRA and 30 PRGT programs initiated between 2002 and 2005 (unless noted otherwise).

### A. Fiscal adjustment — magnitude and flexibility
- Fiscal adjustment in Fund-supported programs generally appeared appropriate given initial conditions, program characteristics, and objectives.
- Fiscal adjustment was more flexible in magnitude and timing than in the past.
- The magnitude of fiscal adjustment was generally restrained, largely out of concern for output effects during the global economic crisis.
- In PRGT cases, fiscal adjustment was larger for countries with stronger institutions (rule of law), though fragile states showed offsetting results (regression results indicate fragile states had higher fiscal adjustment controlling for other factors, but these results should be interpreted carefully).
- Initial fiscal and debt positions were generally strong enough to allow limited adjustment while preserving sustainability and generating resources for priority spending.
- Fiscal adjustment was smaller during 2006-11 for PRGT programs and more back loaded for GRA cases compared to previous periods; magnitude of adjustment picked up for wave 2 programs.

### Regression findings explaining programmed fiscal adjustment (Table 3.1 summary)
- Statistically significant explanatory variables for programmed fiscal adjustment (all variables significant at the 10 percent level or higher):
  - Macro conditions: Fiscal balance at t-1 (-), Public debt at t-1 (+), IIP liabilities at t-1 (+), Change in current account balance from t-2 to t-1 (-)
  - Country characteristics: Rule of law (-) for GRA; Fragile states dummy (+) and Rule of law (+) for PRGT
  - Program characteristics: Aid dependence (-), Wave 2 dummy (+)
- R-squared (Adj.) reported: 0.807 for GRA; 0.500 for PRGT.
- Notes: See Appendix Tables I.1-3 for variable descriptions.

### Fiscal adjustment and debt dynamics
- The size of fiscal adjustment generally remained sufficient to at least stabilize public debt at sustainable levels, improving debt dynamics for most PRGT and GRA cases.
- Exceptions: programs with very high initial debt in the Euro Area and Caribbean.
- High debt programs involved relatively more fiscal adjustment than other programs, although sustainability concerns may still arise in some cases.
- Regression analysis shows fiscal adjustment magnitude is largely explained by high initial fiscal deficits in high-debt GRA cases in year t-1; similar but weaker pattern for PRGT programs with high debt ratios.
- In many high-debt PRGT cases, fiscal adjustment coupled with debt relief under HIPC and MDRI contributed to reductions in debt ratios; some PRGT countries (Guinea, Mauritania, Togo, Uganda) received significant HIPC/MDRI debt relief and reduced debt burdens abruptly.
- Greece (an example of a high-debt GRA case) saw debt ratios remain at high levels though on a downward trend after 2013, even with macroeconomic assumptions that proved more positive than outturns.

### Composition of fiscal adjustment
- Composition generally appeared tailored to country conditions and usually safeguarded priority spending.
- PRGT programs with high initial fiscal deficits: adjustment relied partly on expenditure restraint.
- Other PRGT cases: fiscal balances and expenditures were stable in GDP terms, with increases in revenue and grants playing a larger role initially.
- GRA programs with high initial fiscal deficits: most of the fiscal adjustment fell on spending, as spending levels were relatively high.
- Both social spending on health and education and capital expenditures were fairly well protected in both program types; capital expenditures actually increased in PRGT programs both as a share of GDP and overall spending.
- Programs with less fiscal adjustment tended to emphasize deeper structural reform agendas; extensive structural reform agendas were associated with less initial fiscal adjustment (suggesting program design accounted for political constraints by avoiding both large fiscal adjustment and extensive structural conditionality). In certain cases, needed structural reforms to support fiscal consolidation were found lacking (example: EPE for Ukraine).

### Timing and outturns of fiscal adjustment
- Timing: fiscal adjustment was often back-loaded to accommodate growth concerns in initial program years.
  - In GRA programs (many initiated after the global crisis), the overall balance was programmed to improve by just 1 percent on average in the first two program years.
  - In PRGT cases, expenditure was planned to increase early in programs, supported by grants and debt relief, and to remain above its initial ratio to GDP in all but the third year after program initiation.
- Outturns:
  - Strong outturns compared to targets in early program years indicated most fiscal adjustment targets were not over-ambitious; partly driven by faster than expected economic growth.
  - GRA programs initially adjusted faster than projected and about as much as programmed toward the end of the projection horizon; capital account crises missed targets throughout the program horizon.
  - Regression analysis suggests programs requiring larger fiscal adjustment in a given year were less likely to meet that year’s quantitative performance criteria (QPCs); large required adjustment may be detrimental to program performance.
  - PRGT programs also adjusted faster than programmed in initial years (except Policy Support Instrument arrangements) but adjusted less than initially programmed subsequently.
- Notable projected vs. performance finding: projected adjustment, not the initial fiscal balance, determines program performance; GRA programs with large initial deficits performed better according to the regression findings.

### Boxes and comparative findings (2002-05 vs 2006-11)
- Box 1: Flexibility in Fiscal Adjustment
  - Compared to 2002-05, GRA programs in 2006-11 planned more fiscal adjustment in the face of higher deficits and inflation rates; adjustment was more back loaded in 2006-11 due to larger growth declines during the global crisis.
  - PRGT programs in 2006-11 targeted less adjustment and higher deficits and inflation than 2002-05; initial periods relied solely on revenue and grants for adjustment while expenditure remained expansionary to allow counter-cyclical policies.
  - Monetary policy in 2006-11 PRGT programs targeted inflation rates of about 4 percent, slightly higher than in 2002-05.
- Box Figures summarized: programmed growth, fiscal balance, current account, and inflation comparisons between 2002-05 and 2006-11 samples for GRA and PRGT programs; composition of fiscal adjustment (revenue and grants vs. expenditure) differences across periods.

*Source: IMF staff analysis in the referenced chapter/section.*

### Box 2. Conditionality on Expenditure Measures and Protecting the Poor

### Box 2. Conditionality on Expenditure Measures and Protecting the Poor

### Scope and common measures
- Expenditure policy measures potentially affecting the poor focused mainly on price increases.
- The most common spending measures required:
  - adjustment of utility tariffs, or
  - increases of domestic petroleum prices.
- Rationale for price increases included:
  - avoid excessive volatility of taxes and revenues,
  - curtail untargeted subsidies,
  - eliminate quasi-fiscal deficits,
  - put loss-making public enterprises on a sounder footing.
- Some price increases were preceded or accompanied by automatic price adjustment mechanisms intended to:
  - smooth out periodic adjustments,
  - dampen price volatility,
  - depoliticize price increases.

### Incidence across case-study countries (2006–2010)
- Between 2006 and 2010, nine of the 18 countries in the case study sample were subject to program conditionality affecting prices of products consumed by the poor.
- Programs predicating specific tariff increases: Uganda (2006), Pakistan (2006), Ghana (2009).
- Programs implementing measures related to utilities tariff adjustment mechanisms or institutional responsibilities for price setting: Moldova (2006 and 2010), Sierra Leone (2006), The Gambia (2007), Togo (2008), Dominican Republic (2009).
- Programs including conditionality related to automatic fuel pricing: The Gambia, Ghana, Togo.

### Analysis of distributional impacts and technical assistance (TA)
- In five out of these nine countries, price adjustment conditionality was accompanied by an analysis of their distributive impact.
- Technical assistance explicitly considered adverse impacts on the poor and mitigation options.
- Since 2002, the IMF‘s Fiscal Affairs Department carried out 15 missions, including missions in Ghana and Moldova.
- In the Gambia, Sierra Leone, and Togo, TA focused on the design and implementation of a price smoothing mechanism that explicitly accounted for effects on the poor and provided recommendations for offsetting measures.

### Use of external support and social protection conditionality
- Where distributive impact analysis was not available, Fund staff sought support from the World Bank and included conditionality aimed at strengthening social protection.
- In cases without distributive impact analysis or relevant TA, conditionality on price increases was generally accompanied by a benchmark requiring expansion of existing social safety nets and social benefits to the poor.
  - Example: Pakistan’s measures occurred with World Bank support.
- Uganda: price increases were not explicit conditionality; authorities expressed desire to increase electricity tariffs in the Memorandum of Economic and Financial Policies, coupled with a floor on expenditures from the Poverty Action Fund as an indicative target in the program.

*Source: IMF staff summary from Box 2. Conditionality on Expenditure Measures and Protecting the Poor*

### Box 5. Exchange Rate Flexibility and Fund-Supported Programs

### Box 5. Exchange Rate Flexibility and Fund-Supported Programs

### Key findings on exchange rate regime shifts (2006–11)
- During 2006-11, countries—both with and without programs—increasingly moved toward more flexible exchange rate regimes.
- Among program countries:
  - About half considered exchange rate policies as a strategy for achieving program objectives.
  - About a quarter moved to more flexible arrangements, notably during the global crisis.
  - There was explicit conditionality related to exchange rate policies and restrictions in 16 programs (8 GRA and 8 PRGT) launched during 2006-11 (including removing multiple currency practices, abolishing market rationing and removal of current account restrictions).
  - Countries with currency boards (both with and without programs) did not change their exchange rate regimes.
  - Regression analysis concluded that program countries have not moved to more flexible exchange rate regimes more often than their non-program peers (Appendix III Table III.2). This result did not support a commonly held view that Fund-supported programs influenced countries‘ decisions to move toward more flexible exchange rate regimes.
  - The shift towards greater exchange rate flexibility under programs was supported by reforms in the institutional framework and market infrastructure, including rules-based and transparent foreign exchange intervention policies; enhancing central bank liquidity management; and stronger supervision of commercial banks‘ risk management.

### Factors influencing decisions to increase exchange rate flexibility
- The analysis addresses program documents that had discussions on exchange rate policies.
- Initial level of reserves and competing macroeconomic objectives were important factors:
  - In countries with high international reserves, program design tended to envisage a drawdown in reserves to limit exchange rate movements (in both fixed and flexible exchange rate regimes), with goals of dampening inflationary pressures, calming market sentiment, and preserving financial stability.
  - In countries with fixed exchange rate regimes and low reserves, programs envisioned a move towards more flexibility, particularly when multiple macroeconomic and monetary policy objectives were becoming increasingly incompatible.
  - Countries with floating exchange rates and low reserves relied more heavily on exchange rate movements in order to avoid reserve losses and absorb large exogenous shocks.

- Box Table 1. Factors in Moving to More Flexible Exchange Rate Regimes, 2006-2011 (initial level of international reserves × type of exchange rate regime):
  - Low reserves × Fixed: 9 programs (move to greater flexibility)
  - Low reserves × Flexible: 36 programs (enhance flexibility to reduce loss)
  - High reserves × Fixed: 18 programs (initially draw on reserve buffers)
  - High reserves × Flexible: 8 programs (consider using the reserves to avoid large exchange rate movements)

### Regression and econometric findings
- Regression analysis did not find that Fund-supported programs led countries to adopt more flexible exchange rate regimes more often than non-program peers (Appendix III Table III.2).
- The institutional and market infrastructure reforms accompanying shifts to greater flexibility included:
  - Rules-based and transparent foreign exchange intervention policies.
  - Enhancements to central bank liquidity management.
  - Stronger supervision of commercial banks‘ risk management.

### Policy implications and operational observations
- Program design considered initial reserve levels and regime type when deciding whether to allow or encourage greater exchange rate flexibility, balancing objectives of inflation control, market calming, and financial stability.
- Where reserves were high, programs often allowed reserve drawdowns to stabilize exchange rates; where reserves were low and regimes rigid, programs more often envisaged greater flexibility.
- Institutional reforms (intervention rules, liquidity management, bank supervision) supported the credibility and effectiveness of moves toward flexibility, especially during crisis periods.

*Source: Box 5. Exchange Rate Flexibility and Fund-Supported Programs (extracted content).*

### Box 7. The Vienna Initiative

### Box 7. The Vienna Initiative

### Background and systemic risks in CESE
- In the fall of 2008, concerns ran high that certain economies of central, eastern and southeastern Europe (CESE) would suffer a contagious financial meltdown.
- Most economies in the region were seriously overheated with wide current account deficits, high external debt, and many years of extremely rapid credit growth.
- Much of bank credit was foreign-currency denominated and financed from abroad.
- The US$450 billion exposure of western banks to CESE mostly took the form of loans by western parent banks to their local affiliates, with exposure easily exceeding 50 percent of GDP in many countries.
- Vulnerabilities raised the specter of reversing capital flows, undermining banking systems and triggering downward exchange rate spirals across the region.

### Launch, governance, and components
- Informal discussions began in November 2008; inaugural Vienna Initiative meeting held in Vienna, Austria on January 23, 2009.
- The Vienna Initiative brought together western parent bank groups, home and host-country authorities (financial supervisors, finance ministries, and central banks), and multilateral organizations (IMF, European Bank for Reconstruction and Development (EBRD), European Commission (EC), European Investment Bank (EIB), and the World Bank).
- Joint IFI (International Financial Institutions) Initiative launched on February 27, 2009 as the financial assistance arm of the Vienna Initiative.
- Over the next two years, the EBRD, World Bank, and EIB would disburse €33 billion to strengthen banks in the region.

### Structure, participants, and launches
- Vienna Initiative (VI)
  - Launched: Jan. 09
- Joint IFI Initiative
  - Participants: EBRD, World Bank, EIB (IMF as observer)
  - Launched: Feb. 09
  - Objectives: Financial assistance to strengthen banks and support lending to the real economy (€33 billion); engage other stakeholders; facilitate coordination
- European Bank Coordination Initiative (EBCI) — Country Meetings
  - Participants: EBRD, IMF, EC, EIB, World Bank; 4-10 EU parent banks; supervisors, MoFs, and CBs from relevant home countries and the host country (ECB as observer)
  - Launched: Romania (Mar. 09); Serbia (Mar. 09); Hungary (May 09); BiH (Jun. 09); Latvia (Sept. 09)
  - Objectives: Private Sector Involvement (PSI); coordination of exposure maintenance and capitalization of subsidiaries for financial stability
- Full Forum Meetings
  - Participants: EBRD, IMF, EC, EIB, World Bank; 15 EU parent banks; supervisors, MoFs, and CBs from 7 home and 5-6 host countries (ECB as observer)
  - Meetings: Sept. 09; Mar. 10; Mar. 11
  - Objectives: Policy discussion on regional issues and medium-term challenges; stocktaking

### Commitments by participant groups
- Parent banks: committed to maintain exposure to certain CESE countries and recapitalize their regional subsidiaries as needed.
- International Financial Institutions (IFIs): pledged financial assistance under adjustment programs and the Joint IFI Initiative.
- Home-country authorities: agreed that any public support for parent banks would not discriminate between the groups’ domestic and foreign operations.
- Host-country authorities: committed to implement programs as agreed and not to discriminate between domestic and foreign banks.

### Country-specific agreements and variation
- Banks’ commitments were stronger in Romania, Serbia, and Bosnia-Herzegovina, where agreements were integral to adjustment program discussions.
- Commitments were weaker in Hungary and Latvia, where agreements were concluded only after macro-adjustment programs were already in place.

### Outcomes and impact
- The feared financial meltdown in CESE was successfully avoided.
- All fixed exchange rate regimes held up.
- Any excessive depreciation of flexible exchange rates that had occurred earlier corrected quickly.
- Only Latvia experienced a full-fledged banking crisis; its onset predated the Vienna Initiative and mainly concerned Parex Bank—a domestic bank without a western parent.
- Banks largely complied with their EBCI commitments.
- Overall exposure of western banks to CESE declined little and far less than in other regions—in effect, the private sector was "bailed in" under the BCI arrangements.
- The success of the BCI in CESE underscored the potential value of co-financing with regional financing arrangements.

*Source: _061812b - Box 7. The Vienna Initiative*

### 42. There was no evidence of an overly optimistic bias in 2006-11 Fund program

### 42. There was no evidence of an overly optimistic bias in 2006-11 Fund program

### Summary finding on projection bias
- Analysis examined mean error (ME) and counts of negative and positive errors to assess bias.
- The evidence pointed to a presence of an overly pessimistic bias in projections for some variables, not an overly optimistic bias.
- Program projections did not appear to be more optimistic than non-program forecasts (Table 4.2).
- This lack of evidence for an optimistic bias in Fund program projections contrasts with findings in the academic literature and the previous RoC.
- The paper could not test whether this finding held for larger, more recent programs because the number of outturns under those programs was too limited; preliminary indications suggest some concerns.

### GRA (General Resources Account) program projection patterns
- Some evidence for pessimistic projections for reserves and perhaps the current account balance, but mean errors (MEs) were rarely significantly different from zero.
- Growth, government balance, and inflation:
  - The picture was mixed in terms of counts of negative and positive errors.
  - MEs were never significant.
- Reserves and current account balance:
  - Projections were more often pessimistic.
  - MEs were not significant.
- Programs initiated between 2002 and 2005:
  - Some evidence for an over-optimistic inflation bias (negative errors).
  - Over-pessimistic biases in fiscal balance, growth, and reserves.

### PRGT (Poverty Reduction and Growth Trust) program projection patterns
- Some weak evidence for optimistic projections for growth and for pessimistic projections for the current account balance and reserves.
- Growth:
  - Projections were more often optimistic, but MEs were not significant—suggesting a weak optimistic bias.
  - Evidence was weaker when considering just the crisis period.
- Government balance and inflation:
  - No indication of bias.
- Current account balance:
  - Projections were more often pessimistic, but MEs were not significant.
  - Evidence of a pessimistic bias was stronger when considering just the crisis period.
- Reserves:
  - MEs were often negative and significantly different from zero, suggesting a pessimistic bias in projections.
- Programs initiated between 2002 and 2005:
  - Some weak evidence for pessimistic projections for the government balance and optimistic projections for inflation.

### Flexibility of Fund-supported programs (design and implementation)
- Increased flexibility during design and implementation contributed to program success in a dynamic environment.
- Programs made active use of flexibility to accommodate changing conditions; active use of flexibility helped maintain the implementation rate of conditionality at around 90 percent despite the global recession.
- Flexibility did not appear to come at the expense of program success in meeting objectives.
- Flexibility in program design included:
  - Revisions to program conditionality through adjustors built into program design.
  - Modifications of QPCs (quantitative performance criteria) and augmentations during program implementation.
  - Rephasing of disbursements, combining reviews, and program extensions.

### Specific findings on adjustors, augmentations, and program adjustments
- Revisions in macroeconomic policy frameworks, especially for growth, consistently led to modifications of QPCs and augmentations of access.
- The number of modifications and augmentations increased when downward revisions to growth projections were largest (about 4 percentage points on average).
- Downward revisions in growth projections were associated with the size of augmentation, although less strongly.
- Augmentations of access levels were requested and approved even more than once in a single program in some cases (examples cited in source).
- Combining reviews and program extensions provided flexibility for countries needing more time to implement programs; this sometimes helped manage difficult domestic social or political constraints.
- Programs in fragile states faced difficult challenges; higher rate of program extensions suggested some programs for fragile states might be overly ambitious in timing of reaching objectives.
- Adjustors:
  - Provided flexibility so countries were not penalized for foreseeable shocks beyond their control (common example: change in level of external assistance).
  - Use of adjustors has remained fairly stable over time and depends on the type of QPCs and arrangements.
  - On average, adjustors are used in more than half of QPCs.
  - Use of adjustors tends to be higher in Fund-supported programs for low-income countries.
- Effect on implementation:
  - Program flexibility in a review clearly helped to improve the implementation rate in the subsequent review (with the exception of combined reviews).
  - The application of adjustors led to an increase in implementation rates from 70 percent to 90 percent, compared to a situation without adjustors.

### Methods and robustness notes (regression appendix)
- Estimation used iterative Bayesian Model Averaging (BMA) to select best available models among candidate regressors, integrating model and parameter uncertainty.
- To avoid overfitting, (i) model size limited to a maximum of 15 regressors and (ii) robustness checks undertaken separately after BMA estimation.
- Reported tables present BMA-selected best models re-estimated by frequentist OLS for ease of presentation; adjusted R squared figures are based on these OLS regressions.
- Three additional robustness checks were performed; baseline regressions were highly robust to these variations.
- First robustness check split intercept by area department dummies; results indicate program design was evenhanded across area departments after accounting for macroeconomic and other characteristics, with limited exceptions in GRA current account and reserves adjustment (but sample sizes were small for those cases).

*Source: IMF content unit "_061812b - 42. There was no evidence of an overly optimistic bias in 2006-11 Fund program"*

### 52. A second robustness check includes public rollover needs in the GRA

### 52. A second robustness check includes public rollover needs in the GRA

### Robustness checks: rollover needs
- Objective: test whether including public rollover needs explains higher Euro Area access and whether programmed paths varied for high-rollover cases.
- Finding: evidence does not support either hypothesis.
- BMA selection: the rollover needs variable is selected by BMA once—as an effective regressor in the structural conditionality regression—where higher rollover is associated with more structural conditionality.
- Practical note: inclusion of rollover needs reduces sample size considerably; results not reported for space reasons.

### Robustness checks: political economy variables
- Hypothesis tested: powerful Fund stakeholders influence program design to channel more resources, potentially with less adjustment and conditionality, to politically proximate or business-linked countries (literature surveyed by Steinwand and Stone (2008)).
- Empirical result: adding a large set of political economy variables to baseline frequentist specifications yields no consistent patterns; vast majority of political economy variables are utterly insignificant in explaining program design.
- Where political economy variables enter significantly:
  - They generally carry counterintuitive signs, often suggesting harsher conditionality for proximate countries.
  - In a few cases the political economy specification is statistically preferred to the baseline, but this is attributed to over-fitting given low numbers of observations.
- Additional BMA check: when political economy variables are added to the BMA candidate set:
  - Most resulting best models include some political economy variable(s), while baseline variables remain very robust, particularly for GRAs.
  - Political economy variables that enter often carry unintuitive signs and tend to offset each other.
  - Gains in overall fit versus baseline are generally very limited; Vuong (1989) tests side with the baseline over the political economy specification for the vast majority of regressions.

### Key regression findings (Appendix Table I.1 — GRA regressions)
- Dependent variables covered include: Programmed inflation reduction (percent of GDP), Programmed current account adj. (percent of GDP), Programmed reserves adj. (months of imports), Access (multiple of quota), Number of Structural Conditions, Programmed fiscal balance adj. (percent of GDP).
- Selected coefficient highlights (initial macro conditions, country and program characteristics):
  - Public debt (% of GDP): 0.03**
  - IIP liabilities (US$ billions): -0.01**
  - Total cross-border bank claims (US$ billions): 0.06***
  - Inflation (percent): 0.91***
  - Fiscal balance (percent of GDP): -0.45*** and 0.15***
  - Reserves (months of imports): -0.14*
  - Growth (percent): -0.13**
  - Current account balance (percent of GDP): -0.73***, -0.12***, -0.19***
  - Reserves change (months of imports, t-2 vs t-1): -0.86***
  - Current account balance change (percent of GDP): -0.30***, -0.31***
  - Prior actions at board approval: 0.72***
  - Currency Union dummy: 3.68***, 3.97***
  - Euro Area dummy: 19.68***
  - Trade openness: -0.02**, -0.05***
  - Rule of law: -1.06**
  - Capital account crisis (narrow sample): 1.16**, 4.79***, 5.85***
  - Precautionary program dummy: -2.03***
  - Wave 1 dummy: 3.04***
  - Wave 2 dummy: 3.19***, -1.98***
  - Successor program dummy: 1.34**, -1.06**
  - Fund credit outstanding (Multiple of quota): 0.30**
  - Dummy for post-2006 programs: -3.61***
  - Intercepts (various regressions): -1.95***, -3.23***, -1.23**, 1.92***, 2.75***, 5.65***, 6.47***
  - R-squared adjusted (selected regressions): 0.807, 0.955, 0.777, 0.494, 0.874, 0.828, 0.564
- Notes: ***, **, * denote 1, 5, and 10 percent significance levels, respectively. iBMA used to select regressors. Number of observations is 56 for all GRA regressions.

### Key regression findings (Appendix Table I.2 — PRGT regressions)
- Dependent variables covered similarly for PRGT programs (number of observations generally 85; access regression excludes PSI programs and has 72 observations).
- Selected coefficient highlights:
  - Public debt (% of GDP): 0.02**, 0.03***
  - IIP liabilities (US$ billions): 0.22*** (multiple regressions), -0.02**
  - Total cross-border bank claims (US$ billions): 0.13***
  - Inflation (percent): 0.80***
  - Fiscal balance (percent of GDP): -0.64***
  - Growth (percent): 0.11**, -0.04***
  - Current account balance (percent of GDP): -0.39***, 0.09**
  - Transition country dummy: 0.35**
  - Fragile States dummy: 2.77***
  - Trade openness: 0.06***, 0.01**
  - Rule of law: 2.58***
  - Aid dependence (Aid in percent of program country GDP): -0.15***
  - Crisis program dummy: -1.21***
  - HIPC completion point dummy: 1.44***
  - Dummy for post-2006 programs: 0.56***, -1.26**
  - Intercepts (selected): -1.40*, -2.49***, -9.72***, -3.02***, 0.74***, 7.15***
  - R-squared adjusted (selected regressions): 0.500, 0.926, 0.534, 0.574, 0.626, 0.267
- Notes: ***, **, * denote 1, 5, and 10 percent significance levels, respectively. iBMA used to select regressors. Number of observations is 85 for PRGT/PSI regressions (72 for access regression).

### Descriptive statistics (Appendix Table I.3 — selected means and ranges)
- Access (multiple of quota) GRA sample: Mean 4.57, Std. Dev. 6.21, Min 0.16, Max 32.12
- Programmed fiscal balance adj. (percent of GDP) GRA sample: Mean 2.39, Std. Dev. 5.22, Min -12.24, Max 28.22
- Programmed current account adj. (percent of GDP) GRA sample: Mean 2.87, Std. Dev. 7.81, Min -10.29, Max 39.60
- Programmed inflation adj. (percentage points) GRA sample: Mean -4.84, Std. Dev. 8.06, Min -47.02, Max 7.80
- Programmed reserves adj. (months of imports) GRA sample: Mean 0.35, Std. Dev. 1.78, Min -3.28, Max 4.71
- Number of structural conditions per program review GRA sample: Mean 6.75, Std. Dev. 4.85, Min 0.80, Max 23.70
- Public Debt/GDP (percent of GDP) GRA sample: Mean 51.30, Std. Dev. 38.08, Min 7.76, Max 164.97
- IIP liabilities (US$ billions) GRA sample: Mean 147.63, Std. Dev. 484.30, Min 0.54, Max 3,544.04
- Total cross-border bank claims (US$ billions) GRA sample: Mean 37.46, Std. Dev. 95.41, Min 0.14, Max 651.37
- Inflation rate (percent) GRA sample: Mean 8.67, Std. Dev. 8.82, Min -2.19, Max 51.46
- Fiscal balance (percent of GDP) GRA sample: Mean -3.61, Std. Dev. 5.94, Min -22.07, Max 9.22
- Reserves (months of imports) GRA sample: Mean 4.33, Std. Dev. 2.84, Min 0.14, Max 11.05
- GDP growth (percent) GRA sample: Mean 3.22, Std. Dev. 5.53, Min -14.46, Max 13.82
- Current account balance (percent of GDP) GRA sample: Mean -5.29, Std. Dev. 8.40, Min -35.51, Max 15.56
- Political economy averages (PRGT and PSI sample examples):
  - UN voting with US, important votes (percent): Mean 51.02, Std. Dev. 20.85, Min 0.00, Max 88.90
  - UN voting with US (percent): Mean 22.23, Std. Dev. 11.27, Min 6.67, Max 45.10
  - IMF A-level economists (fraction): Mean 0.46, Std. Dev. 0.52, Min 0.00, Max 1.87
  - Aid from US (Fraction of US total aid): Mean 0.01, Std. Dev. 0.02, Min -0.01, Max 0.14
- Notes on samples: GRA sample has 56 observations; PRGT and PSI sample combined has 85 observations (72 and 13 obs).

### Implications and interpretation
- Including public rollover needs in GRA regressions provides limited explanatory power; not supported as a driver of higher Euro Area access or divergent programmed paths, except for a single structural conditionality link.
- Political economy variables generally do not explain program design consistently; where statistically significant their counterintuitive signs and tendency to offset suggest over-fitting and limited substantive influence relative to macroeconomic and program factors.
- Baseline macro, country, and program characteristics (e.g., prior actions, capital account crises, currency union, Euro Area dummy, trade openness, reserves, fiscal balance, inflation) are robust correlates of programmed macroeconomic adjustment and program design across GRA and PRGT samples.

*Source: IMF staff estimations, Appendix Tables I.1–I.3 (regressions and descriptive statistics).*

### Appendix Table I.4. Robustness for Regressions Explaining Planned Macroeconomic Adjustment in GRA Programs Including Reg

### _061812b - Appendix Table I.4. Robustness for Regressions Explaining Planned Macroeconomic Adjustment in GRA Programs Including Regional Variables

### Robustness results — GRA regressions (regional variables)
- Sample distribution: total 56 observations across area departments: African (3), Asia-Pacific (3), European (19), Middle East and Central Asia (5), Western Hemisphere (26).
- Area department dummies (coefficients shown by dependent variable):
  - African: -1.47*, -2.28***, -9.70***, -3.19***, 0.69***, 7.10***
  - Asia-Pacific: -0.84, -3.11***, -10.13***, -1.62, 0.84***, 8.80***
  - European: -0.17, -3.08***, -9.18***, -2.56**, 0.73**, 8.85***
  - Middle East and Central Asia: -1.85, -2.85***, -9.74***, -3.53***, 0.72***, 7.11***
  - Western Hemisphere: -1.08, -3.27***, -10.58***, -3.63***, 0.93***, 7.85***
- R-squared adjusted by dependent variable: 0.490, 0.936, 0.519, 0.566, 0.822, 0.880
- Area dept. dummies are jointly significant at 5 percent? No for all six dependent variables.
  - Likelihood ratio test result (Prob > chi2): 0.869, 0.470, 0.993, 0.484, 0.732, 0.544
- Selected initial condition and other regressors (coefficients shown where selected as part of iterative BMA):
  - Public debt (% of GDP): 0.02**, 0.03***
  - IIP liabilities: 0.22*** (repeated in multiple columns), -0.02**
  - Total cross-border bank claims (US$ billions): 0.13***
  - Inflation (percent): 0.81***
  - Fiscal balance (percent of GDP): -0.64***
  - Growth (percent): 0.11**, -0.04***
  - Current account balance (percent of GDP): -0.40***, 0.09**
  - Fragile States dummy: 2.78***
  - Prior actions at board approval: 0.38***
  - Trade openness: 0.06***, 0.01**
  - Rule of law: 2.57***
  - Aid dependence (Aid in percent of program country GDP): -0.16***
  - Crisis program dummy: -1.24***
  - HIPC completion point dummy: 1.52***
  - Dummy for post-2006 programs: 0.57***, -1.22**
- Notes:
  - ***, **, * denote 1, 5, and 10 percent significance levels.
  - Set of regressors determined by iterative Bayesian Model Averaging (iBMA).
  - Grey shading in original indicates variables excluded from candidate regressors in iBMA.
  - Excludes PSI programs for access regression (access in PSIs is zero by definition).

### Robustness results — PRGT regressions (regional variables)
- Sample distribution: total 85 observations across area departments: African (58), Asia-Pacific (3), European (4), Middle East and Central Asia (14), Western Hemisphere (6). For access regression, observations drop to 72 due to omission of PSI programs.
- R-squared adjusted by dependent variable: 0.793, 0.945, 0.770, 0.473, 0.858, 0.817, 0.559
- Political economy variables jointly significant at 5 percent? Yes for one dependent variable (last column); No for others.
  - Likelihood ratio test result (Prob > Chi2): 0.273, 0.724, 0.257, 0.179, 0.382, 0.034, 0.207
- Selected coefficients (examples from candidate regressors selected by iBMA):
  - Public debt (% of GDP): 0.04*
  - IIP liabilities: 0.04**
  - Inflation (percent): 0.92***
  - Fiscal balance (percent of GDP): -0.63***, 0.10
  - Current account balance (percent of GDP): -0.82***, -0.14**, -0.24***
  - Reserves (months of imports) change (t-2 vs t-1): -0.78**
  - Prior actions at board approval: 0.75***
  - Euro Area dummy: 20.88***
  - Trade openness: -0.02, -0.05***
  - Rule of law: -1.40*
  - Capital account crisis (narrow sample): 1.99***, 3.44**, 3.23*
  - Political-economy examples (selected coefficients):
    - UN voting with Japan (percent): -0.61**
    - Aid from US (Fraction of US total aid): 38.71 (in one column) and substantial negative coefficients in others (e.g., -52.74)
    - Aid from France (Fraction of French total aid): large positive coefficients in some columns (e.g., 410.82, 299.24)
    - Trade with UK (fraction of total trade): -22.85*** in one column
- Intercepts shown by column: -9.11, -8.98, -8.60, 0.40, -0.08, 4.01, 9.46
- Notes:
  - ***, **, * denote 1, 5, and 10 percent significance levels.
  - Political economy variables include UN voting alignments, IMF staff composition (A-level/B-level economists), aid shares from donors, trade shares with donors.
  - Source: IMF staff estimations.

### Political economy variables in GRA regressions (separate specification)
- Number of observations: 56 for all regressions.
- Political econ. variables jointly significant at 5 percent? Yes for some dependent variables (notably columns with likelihood ratio test results 0.020, 0.001, 0.033); No for others (results 0.066, 0.144, 0.215).
- Selected coefficients (examples):
  - Public debt (% of GDP): 0.01, 0.03***
  - IIP liabilities: 0.17, 0.12
  - Total cross-border bank claims (US$ billions): 0.14***
  - Inflation (percent): 0.84***
  - Fiscal balance (percent of GDP): -0.73***
  - Growth (percent): 0.11**, -0.06***
  - Fragile States dummy: 2.25**
  - Trade openness: 0.05**, 0.02***
  - Rule of law: 2.07**
  - HIPC completion point dummy: 1.43***
  - UN voting / aid / trade variables with selected coefficients:
    - UN voting with UK (percent): 0.09* in one column
    - UN voting with Japan (percent): 0.22*
    - Aid from Germany (Fraction of German total aid): 110.68** in one column
    - Trade with US (fraction of total trade): 27.99**, 42.50**, 24.98*** in selected columns
    - Trade with Japan (fraction of total trade): -42.14* in one column; 51.07 in another
- R-squared adjusted across columns: 0.556, 0.931, 0.549, 0.666, 0.662, 0.276
- Intercepts reported across columns: -8.88, -3.08, -25.99**, -6.62*, 0.04, 5.23
- Notes: political economy variables tested jointly versus baseline model without political economy variables.

### Impact of Budget Support (GRA and PRGT regressions)
- Quantitative program design finding (text summary):
  - Programs with direct budget support did not differ substantially in quantitative design from other programs, controlling for other factors.
  - Main exceptions:
    - Programmed fiscal adjustments larger by 1¾ percent of GDP in GRA cases.
    - Programmed fiscal adjustments larger by 2½ percent of GDP in PRGT cases.
    - GRA programs with budget support include on average 1.3 more structural conditions per review.
    - PRGT budget support programs envisage 2 percentage points of GDP less inflation reduction.
  - Coefficients on other variables remain robust over the 2006-11 sample period versus the baseline that excludes the budget support indicator.
- Candidate regressors and selected coefficients in BMA runs (combined summary shown in tables):
  - Baseline 2006-11 and Incl. Budget Support columns report coefficients such as:
    - Public debt (% of GDP): 0.02, 0.01
    - IIP liabilities: 0.15, 0.15
    - Inflation (percent): 0.71***, 0.77***
    - Fiscal balance (percent of GDP): -0.84***, -0.90***
    - Change in inflation (pp): 0.13**, 0.11**
    - Fragile States dummy: 1.68, 1.73
    - Rule of law: 2.13, 2.23
    - Aid dependence: -0.14*, -0.10
    - Budget Support dummy coefficients reported: 2.44*, 1.92*** (various columns and samples)
  - Baseline 2006-2011 R-squared adjusted reported: 0.5143, 0.5414, 0.9014, 0.9158 (by column)
  - Sample size in those BMA-based regressions: Number of observations 58, sample 2006-11.

- Appendix Table I.8 (GRA regression evaluating impact of budget support) — selected candidate-regressor coefficients:
  - Fiscal balance (percent of GDP): -0.66***, -0.61***
  - Current account balance (percent of GDP) change (t-2 vs t-1): -0.22***, -0.21***
  - Prior actions at board approval: 0.51***, 0.44**
  - Rule of law: -1.55***, -1.80***
  - Budget Support dummy: 2/1.77**, 1.26*
  - Wave 2 dummy: 2.36**, 2.70***
  - Intercepts shown: -2.24***, -3.13***, 3.27***, 2.89***
  - R-squared adjusted across these regressions: 0.868, 0.883, 0.187, 0.240
  - Number of observations: 34; Sample: 2006-11

- Appendix Table I.9 (PRGT regression evaluating impact of budget support) is indicated but specific coefficient table not reproduced in supplied content.

### Lessons from case studies of monetary policy adjustment (Appendix II)
- General findings:
  - Fund-supported programs incorporated monetary reforms to support monetary adjustment and program objectives; depth/type influenced by initial conditions (inflation, fiscal and current account balances, capital and remittances inflows, level of reserves) and structural agenda.
  - In many programs launched during the global financial crisis, establishing a clear monetary anchor was a priority (Dominican Republic, Ukraine, Seychelles, Sri Lanka, Armenia, Moldova). These programs tended to move toward flexible exchange rate arrangements and toward inflation targeting.
  - Clarification of central bank role in foreign exchange market was important in some cases (Costa Rica, Hungary), with recommendation to avoid frequent FX interventions except to stabilize volatile conditions.
  - In PRGT countries, monetary adjustment played a limited role due to weak transmission mechanisms, inefficient frameworks, lack of prerequisites for market operations, and lack of competition in money and FX markets (Sierra Leone, Moldova, Uganda). These countries continued targeting the monetary base as an anchor.
  - Both GRA and PRGT programs proposed major reforms in monetary policy forecasting and implementation: strengthen transmission mechanisms, detailed liquidity forecasting for central bank interventions, eliminate segmentation in money markets, central bank recapitalization reforms (Dominican Republic, Costa Rica, Pakistan) to strengthen credibility and independence and open market operation efficiency.
- Table II.1. Number of Monetary Policy Reforms in GRA and PRGT Programs (based on case studies) — counts:
  - Change in monetary regime: GRAs 5, PRGTs 5
  - Monetary and fiscal consolidation: GRAs 3, PRGTs 6
  - Central Bank recapitalization/independence: GRAs 3, PRGTs 4
  - Exchange rate regime: GRAs 8, PRGTs 6
  - Monetary policy instruments/liquidity management: GRAs 7, PRGTs 6
  - Other: GRAs 2, PRGTs - (dash indicates none or not applicable in supplied content)

*Source: IMF staff estimations.*

### 60. In PRGT countries, the efficiency of monetary adjustment often depended on

### _061812b - 60. In PRGT countries, the efficiency of monetary adjustment often depended on

### Monetary policy implementation in PRGT countries
- Efficiency of monetary adjustment often depended on coordination of monetary and fiscal operations, debt management, and government cash management.
- Country examples where improvement of coordination was seen as an important pre-requisite for enhancing the efficiency of monetary policy implementation, and the link between the money market and lending interest rates:
  - Uganda
  - Sierra Leone
  - Moldova

### Fiscal adjustment: patterns and mechanics
- Figure III.2 (Expenditure (% GDP), periods t-1, t, t+1, t+2, t+3):
  - In PRGT cases with high initial fiscal deficits, adjustment relied both on expenditure restraint and increased revenue and grants.
  - In GRA cases, the burden of adjustment fell mainly on spending, as initial spending levels were high.
- Time labels in figures preserved as shown: t-1 t t+1 t+2 t+3.
- Source cited for figures: WEO, MONA, and Staff calculations.

### Determinants of performance against QPCs (Table III.1)
- Dependent variable: share of QPCs met in year t+1.
- Selected estimated coefficients (as presented):
  - Projected* fiscal deficit in t: -0.0155, -0.0252*, -0.0092, -0.0177**, -0.0082, -0.0125*
    - t-statistics (in parentheses): (-1.51), (-2.38), (-1.26), (-3.38), (-1.89), (-2.91)
  - Projected* fiscal adjustment** in t+1: -0.0259, -0.0328*, -0.0171, -0.0240**, -0.0136*, -0.0179**
    - t-statistics: (-1.66), (-2.20), (-1.71), (-3.63), (-2.29), (-3.30)
  - Projected* inflation in t: -0.0166, -0.00721, -0.0132*, -0.0125**, -0.00539, -0.00802*
    - t-statistics: (-1.85), (-0.77), (-2.18), (-3.01), (-1.50), (-2.36)
  - Number of prior actions: -0.0430**, -0.0914*, -0.0381***, -0.0182, -0.0202**, -0.00748
    - t-statistics: (-3.00), (-2.89), (-3.74), (-1.48), (-3.34), (-0.74)
  - Constant terms: 1.010***, 0.929***, 1.036***, 0.970***, 0.960***, 0.955***
    - t-statistics: -12.65, -10.65, -17.01, -19.52, -26.5, -23.46
- Sample sizes and fit statistics (as presented):
  - N: 3217, 3621, 3621, 3621, 3621, 3621 (as arranged in table columns)
  - R-sq: 0.4610, 0.690, 0.465, 0.625, 0.438, 0.517
  - adj. R-sq: 0.3810, 0.5860, 0.3970, 0.5310, 0.3660, 0.396
- Notes from table:
  - * Projections are as of period t (e.g. fiscal adjustment in t+1 as projected in t)
  - ** Fiscal adjustment in t+1 is defined as the fiscal balance (% GDP) in t+1 minus the fiscal balance (% GDP) in t.
  - The table presents results of probit regressions; a positive coefficient suggests that an increase in the regressor is associated with increased probability of a higher than programmed fiscal deficit in period t+1.

### Financial sector conditionality: classification and incidence
- New classification categories (legend):
  - 1 Regulation and supervision
  - 2 Infrastructure and market development
  - 3 Bank restructuring and resolution
  - 4 Financial stability assessment/contingency
  - 5 Liquidity and solvency support
  - 6 Household and corporate debt restructuring
  - 7 Anti-money laundering/terrorism
  - 8 Other
- Selected percentage distributions by group (as presented):
  - PRGT before crisis:
    - 1: 38%
    - 2: 14%
    - 3: 18%
    - 4: 8%
    - 7: 3%
    - 8: 19%
  - GRA before crisis:
    - 1: 48%
    - 2: 9%
    - 3: 19%
    - 4: 10%
    - 8: 14%
  - Capital account crises:
    - 1: 26%
    - 3: 37%
    - 4: 14%
    - 5: 10%
    - 6: 2%
    - 8: 11%
  - PRGT crisis:
    - 1: 39%
    - 2: 7%
    - 3: 16%
    - 4: 23%
    - 5: 4%
    - 6: 1%
    - 7: 1%
    - 8: 9%
  - GRA crisis:
    - 1: 35%
    - 2: 1%
    - 3: 27%
    - 4: 16%
    - 5: 9%
    - 6: 2%
    - 7: 1%
    - 8: 9%
  - Transition countries:
    - 1: 12%
    - 3: 46%
    - 4: 24%
    - 5: 9%
    - 6: 3%
    - 8: 6%
  - Fragile states:
    - 1: 42%
    - 2: 6%
    - 3: 24%
    - 4: 12%
    - 5: 1%
    - 7: 1%
    - 8: 14%
- Source cited for classification and figures: WEO, MONA, and Staff calculations.

### Euro Area program projections (Appendix IV)
- General findings:
  - Downward revisions in GDP growth forecasts were present to differing degrees in all three Euro Area programs.
  - Fiscal deficits in 2011 were larger than initially forecasted for Greece and Portugal.
  - Public debt was larger than forecast in 2011 for Greece.
- Notes on the figure data:
  - Red denotes Ireland, Blue Greece and Green for Portugal.
  - The solid line represents their latest review (4th for Ireland, 5th for Greece and 2nd for Portugal).
  - All projections past 2011 are revisions to previous projections made at board approvals.
  - Note: Ireland's 2010 Actual Value of -32.6% is omitted.
- Time series labels in figures preserved as shown: 2010 2011 2012 2013 2014 2015.

*Source: IMF staff review material as presented in the provided content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/np/pp/eng/2012/_061812b.pdf_