## _wp09263

## Source details

**Canonical URL:** [_wp09263](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09263.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09263.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09263.pdf.json)

---

### Context and objective
- Since late 2007, LICs faced exogenous shocks: surge in food and fuel prices in 2007−08 followed by the global financial crisis, weakening external stability.
- Demand for Fund financing for policy and exogenous shocks rose sharply in 2008:
  - Only twenty-seven such arrangements were approved over 2000–07; seventeen in 2008 alone.
- Objective: provide a systematic framework to identify vulnerable LICs likely to need Fund assistance and to assess LIC resilience to common and country-specific exogenous shocks.

### Scope, event definition, and identification refinements
- Event set (included):
  - SBAs, PRGF augmentations, and CFFs addressing immediate balance of payments (BOP) needs.
  - Some SAF/ESAF/PRGF arrangements included only if addressing immediate BOP needs.
- Exclusions:
  - Precautionary SBAs, SBAs/PRGF augmentations addressing natural disasters, ENDA (for main event sets).
- Supply-side exclusions from “normal” episodes:
  - (i) overdue financial obligations to the Fund;
  - (ii) interruption in current program;
  - (iii) lack of technical capacity/willingness to implement Upper Credit Tranche quality programs;
  - (iv) significant arrears to bilateral and multilateral creditors without cooperative framework.
- Sample period and counts:
  - Sample period: 1980–2004.
  - Model I (policy and/or exogenous shocks): 235 financing events identified initially; estimation sample details: # of observations 532; # of countries 55; sample probability 0.44.
  - Model II (exogenous shocks only): 91 financing events identified initially; estimation sample details: # of observations 388; # of countries 53; sample probability 0.23.

### Methodology
- Single-equation binary response framework estimating approval probability: panel probit family (pooled probit, RE probit, FE probit, correlated RE via Chamberlain-Mundlak).
- General form: P(y_it = 1 | x_it, c_i) = Φ(c_i + x_it' β).
- Explanatory variables (except exogenous shocks) lagged one period to mitigate endogeneity.
- Robustness checks include logit versus probit, alternative estimators, sub-samples (high non-oil export concentration), and out-of-sample prediction (1980–2000 estimated, 2001–04 predicted).

### Composite macroeconomic stability indicator (mitot)
- Rationale: single variables provide partial information; composite indicator aggregates complementary signals (inflation, nominal depreciation, reserve changes, imports, government balance, GDP, black market premium) with weights inverse to standard deviations.
- Formula (as presented in the source): grantportprofitfdishortIPDnrt
- Components and definitions:
  - nrt = net resource transfers
  - D = disbursements on long-term debt and IMF purchases/loans
  - P = repayments on long-term debt and IMF repurchases
  - I = total interest payments
  - short = change in short-term debt
  - fdi = foreign direct investment
  - profit = fdi profit remittances
  - port = portfolio investment flows
  - grant = official grants excluding technical cooperation
- Data coverage: net resource transfers series available only from 1985 onwards.

### Key empirical findings (benchmark estimation results)
- Models and estimation:
  - Model I: correlated random effects probit; # observations 532; # countries 55; sample probability 0.44; reported Pseudo R2 values include 0.58 and 0.57 for two specifications.
  - Model II: pooled probit with robust SEs; # observations 388; # countries 53; sample probability 0.23; reported Pseudo R2 values include 0.62 and 0.64 for two specifications.
- Variables found significant determinants (selected coefficients and t-statistics preserved exactly as reported):
  - Current account balance to GDP: -0.076 *** (-4.61); -0.076 *** (-4.51); -0.073 *** (-4.7); -0.073 *** (-4.13)
  - Reserve coverage in months of imports (CFA): -0.478 *** (-6.08); -0.492 *** (-6.15); -0.466 *** (-4.12); -0.448 *** (-3.96)
  - Reserve coverage in months of imports (non-CFA): -0.769 *** (-8.71); -0.804 *** (-8.82); -0.589 *** (-6.05); -0.598 *** (-6.13)
  - Macroeconomic stability indicator: 0.068 *** (2.89); 0.072 ** (2.21)
  - Depreciation: 0.016 *** (3.11); 0.020 *** (2.99)
  - Government balance to GDP: -0.047 (-1.6); -0.060 ** (-2.18)
  - Real GDP growth: -0.113 *** (-4.24); -0.127 *** (-4.69); -0.167 *** (-5.71); -0.180 *** (-5.85)
  - Official Grants to GDP: -0.046 (-1.64)
  - Total debt service to exports: 0.027 *** (3.03); 0.030 *** (3.15)
  - Change in terms of trade in previous year: -0.022 *** (-2.8); -0.019 ** (-2.49); -0.028 *** (-4.19); -0.028 *** (-4.26)
  - Change in real oil prices in previous two years: 0.009 *** (2.85); 0.010 *** (3.57); 0.012 *** (3.96); 0.013 *** (4.44)
  - Real world trade, cyclical component: -0.099 ** (-2.53); -0.087 ** (-2.28); -0.099 *** (-2.77); -0.109 *** (-2.71)
  - Change in real non-oil commodity prices: -0.020 (-1.58); -0.031 ** (-2.12); -0.022 (-1.45)
  - Real growth of goods exports in previous year: -0.009 * (-1.79)
  - Real GDP per capita: -0.281 (-1.37); -0.439 ** (-2.08)
  - Paris Club dummy: 0.774 *** (3.24); 0.685 *** (2.95); 0.525 * (1.67); 0.587 * (1.72)
  - Country specific averages: Total debt service to exports: 0.044 *** (2.63); 0.044 *** (2.64)
  - Country specific averages: FDI to GDP: -0.105 * (-1.76); -0.106 * (-1.77)
- Goodness of fit and model performance:
  - Reported Pseudo R2 values include 0.57, 0.58, 0.62, 0.64 across specifications.
  - In-sample predicted probabilities (Model I): median predicted probability for financing events = 0.96 vs 0.04 for normal episodes; 90 percent of financing events have probabilities above 0.41; 75 percent of normal episodes below 0.21.
  - In-sample predicted probabilities (Model II): 90 percent of financing events have predicted probabilities above 0.22; 90 percent of normal episodes below 0.29.
  - Out-of-sample predictions (1980–2000 estimated; 2001–04 predicted): proportion of financing events declines from about 54 percent to 12 percent in Model I out-of-sample; from 30 percent to 6 percent in Model II out-of-sample.

### Threshold probability analysis and operational classification
- Loss-function approach (Demirgüc-Kunt and Detragiache (1999)) used to derive thresholds that minimize expected weighted loss of type I and type II errors.
- Thresholds found:
  - Model I:
    - Asymmetrically weighted (3:1) loss minimized at threshold probability = 0.21 → ~24 percent wrong calls vs ~5 percent missed financing events.
    - Equally weighted loss threshold = 0.50 → ~12 percent wrong calls vs ~12 percent missed financing events.
    - Higher threshold 0.59 reduces wrong calls to <10 percent of normal episodes.
  - Model II:
    - Asymmetrically weighted (3:1) loss minimized at threshold probability = 0.14 → ~5 percent missed financing events, ~18 percent wrong calls.
    - Equally weighted loss threshold = 0.28 → ~10 percent wrong calls vs ~11 percent missed financing events.
    - Predicted probability above 0.28 reduces wrong calls to <10 percent.
- Operational use:
  - Predicted probabilities compared to thresholds provide a vulnerability classification.
  - “Vulnerable” if predicted probability exceeds basic threshold; “highly vulnerable” using higher thresholds (e.g., 0.59 in Model I; 0.28 in Model II).

### Counterfactual simulations: single-variable effects and reserve coverage example
- Simulation setup:
  - Two initial-value sets for other variables: (i) sample median; (ii) 75th (or 25th) percentile if coefficient positive (negative).
  - Global variables set to zero in simulations to isolate single-variable effects.
- Key counterfactual outcomes (Table 3 and text preserved):
  - Predicted probability for Fund financing:
    - Model I: 0.0 (Median), 0.84 (75th/weakly positioned country) with 95 percent CI [0.69 0.94].
    - Model II: 0.0 (Median), 0.59 (75th/weakly positioned country) with 95 percent CI [0.39 0.77].
  - Reserve coverage thresholds (Model I; non-CFA vs CFA):
    - For a non-CFA weakly positioned country: predicted probability falls below the threshold at 3.6 months of imports; the 95 percent confidence interval falls below the threshold if reserve coverage is above 4.4 months of imports.
    - For a similarly situated CFA country: corresponding figures are 5.1 and 7.3 months of imports respectively.
  - Median country requires extreme shocks (other variables fixed) to cross threshold:
    - Reserve coverage less than 1.4 months of imports for non-CFA countries.
    - Current account deficit to GDP exceeding 18¾ percent.
    - GDP contraction exceeding 4¾ percent.
    - Severe adverse terms of trade shocks.
    - Real oil prices increasing more than twofold.
- Other simulation findings:
  - For Model II (exogenous shocks):
    - Non-CFA country: predicted probability falls below threshold at 3.5 months of imports; 95 percent CI below threshold above 5 months.
    - CFA country: corresponding figures are 4.1 and 8 months of imports respectively.
  - GDP growth and terms of trade shocks swing predicted probability and confidence interval more dramatically for weakly positioned countries; a sharp adverse shock to real non-oil commodity prices could bring predicted probability over the threshold for a median country.

### Robustness, heterogeneity, and sub-sample findings
- Robustness:
  - Logit versus probit predicted values correlation = 0.99; similar log likelihoods.
  - Alternative estimators produce similar signs, significance, and broadly comparable magnitudes and average partial effects.
- Heterogeneity:
  - Country-specific averages matter in Model I: total debt service to exports, net present value of debt to exports (average), FDI to GDP, and net resource transfers to GDP capture persistent divergences explaining frequent Fund use by some countries.
  - No significant country heterogeneity detected for exogenous-shock-only approvals (Model II).
- CFA versus non-CFA:
  - Reserve coverage has a significantly larger effect for non-CFA countries versus CFA countries.
  - CFA dummy negative but insignificant (t-statistic ≈ -1.1; standard error 0.46).
  - Re-estimation for non-CFA sample: benchmark results qualitatively similar; change in real non-oil commodity prices becomes higher and significant at 10 percent in reduced sample.
- Export concentration (non-oil commodities):
  - Sub-sample estimations for LICs with high non-oil export concentration yield broadly similar significant determinants (reserve coverage; current account; real GDP growth; macro stability; terms of trade; oil prices; world trade cyclical component).
  - Observations by export share thresholds (benchmark >45%): # observations 532; sample probability 0.44; # countries 55. Additional thresholds and counts reported for >50%, >60%, >70% with corresponding # observations and sample probabilities.

### Findings on variables often insignificant
- Insignificant or weakly significant in core specifications:
  - Real effective exchange rate (REER).
  - Net resource transfers and many of its components (though country-specific average of net resource transfers significant in Model I).
  - U.S. real interest rate.
  - Election dummies (lagged and upcoming).
  - Alternative scalings for terms of trade shocks and oil-import-to-GDP scaling.

### Natural disasters and dedicated estimation
- Financing for natural disasters (ENDA and PRGF augmentations) estimated separately (Table 4):
  - Share of affected population: 0.041 ** (t = 2.47)
  - Reserve coverage in months of imports (lagged 1 period): -0.412 ** (t = -2.31)
  - Real GDP growth (lagged 1 period): -0.144 * (t = -1.79)
  - Constant: -1.233 (t = -1.39)
  - Pseudo R2: 0.24
  - # of observations: 312
  - Financing events: 17; Normal episodes: 295; Sample probability: 0.05; # of countries: 50
- Conclusion: financing for natural disasters is rare in sample and findings are tentative; conceptual justification for excluding natural disasters from the main event set is supported.

### Policy implications and applications
- Practical application:
  - Framework can be used with IMF’s semi-annual WEO forecasts to identify LICs with greater likelihood of requiring Fund financing.
  - Suggested extension: incorporate complementary “stress testing” approach to assess potential demand under adverse common and country-specific shocks and policy deteriorations.
- Main policy-relevant conclusions:
  - Demand for Fund resources by LICs is likely to be cyclical in response to global conditions; intensity determined by severity and persistence of adverse exogenous shocks.
  - Adverse global environment (oil price increases, declines in world trade, adverse commodity price shocks) can tip weakly positioned countries into higher probability of needing Fund financing.
  - Reserve coverage, current account balance, real GDP growth, macroeconomic instability, and terms of trade shocks are central indicators for vulnerability assessment.
  - Nonlinearities imply severe cumulative deterioration can increase probability of financing more than proportionally.

### Data sources and sample list
- Key data sources:
  - Present value of debt: World Bank internal study by Dikhanov (2004)
  - Paris Club rescheduling: Paris Club webpage
  - Principal arrears on long-term debt: WB Global Development Finance (GDF)
  - Gross international reserves: IMF WEO
  - Real oil prices and real non-oil commodity prices: IMF WEO (oil prices deflated by US CPI)
  - World trade index: IMF WEO (index number 2000=100)
  - Fund arrangements and program indicators: IMF databases (Fund Accounts, IFS, Fund Arrangements database)
- Country list for estimation sample (Table A1): 55 countries (examples include Albania, Bangladesh, Ethiopia, Ghana, Madagascar, Mozambique, Pakistan, Rwanda, Tanzania, Uganda, Vietnam, Zambia, Zimbabwe).

*Source: _wp09263 (IMF Working Paper content provided).*

### References .............................................................................................................

### _wp09263 - References .............................................................................................................

### Figures
- 1. Fund Financing to LICs for Policy and/or Exogenous Shocks (1980–2008) ..................... 11
- 2. Low Income Countries: Selected Economic Indicators (1980–2007) ................................ 14
- 3. Distribution of Reserve Coverage in the Sample Prior to Financing Events versus Normal Episodes: CFA versus Non-CFA Countries ....................................................................... 26
- 4. Threshold Probability Analysis .......................................................................................... 32
- 5. Model I: Counterfactual Simulations—Effects of Explanatory Variables on the Probability of Fund Financing .............................................................................................................. 36
- 6. Model II: Counterfactual Simulations—Effects of Explanatory Variables on the Probability of Fund Financing ............................................................................................ 37
- 7. Comparison of Predicted Probabilities: Probit versus Logit Estimates .............................. 38

### Tables
- 1. Estimation Results: Demand for the Fund Financing in Response to Shocks .................... 24
- 2. Predicted Probabilities (Percentiles) ................................................................................... 30
- 3. Values of Explanatory Variables in Counterfactual Simulations ....................................... 34
- 4. Financing for Natural Disasters: Estimation Results .......................................................... 40
- 5. The Effect of Export Concentration: Non-oil Commodities............................................... 41
- 6. Model I: The Fund Financing to Non-CFA Countries ....................................................... 42

### Boxes
- Box 1. Why A Composite Indicator for Macroeconomic Stability? ...................................... 18

### Appendices and Appendix Tables
- Appendices I. Data .................................................................................................................. 46
- Appendix Tables
  - A1. Country List: Estimation Sample ..................................................................................... 47
  - A2. Model I Policy and Exogenous Shocks: Estimation Results for Alternative Specifications .......................................................................................................................................... 48
  - A3. Model II Exogenous Shocks: Estimation Results for Alternative Specifications ........... 49
  - A4. Model I Policy and Exogenous Shocks: Comparison of Alternative Estimators and Average Partial Effects .................................................................................................... 50
  - A5. Model II Exogenous Shocks: Comparison of Alternative Estimators and Average Partial Effects .............................................................................................................................. 51
  - A6. Effect of Previous Fund Engagement .............................................................................. 52

* _wp09263 - References ............................................................................................................. *

### INTRODUCTION

### _wp09263 - INTRODUCTION

### Context and motivation
- Since late 2007, low-income countries (LICs) have been buffeted by a set of exogenous shocks: a surge in food and fuel prices in 2007−08 followed by the global financial crisis, affecting LICs’ external stability.
- Demand for Fund financing in response to policy and exogenous shocks rose sharply in 2008 after a prolonged period of subdued demand:
  - Only twenty-seven such arrangements were approved over 2000–07, and the number shot up to seventeen in 2008 alone.
  - Many arrangements were already approved or in the pipeline for 2009 (as noted in the source).
- Objective: provide a systematic framework to identify vulnerable LICs likely to need Fund assistance and to assess LIC resilience to common and country-specific exogenous shocks.

### Scope and focus of the paper
- Focuses on a specific subset of Fund arrangements addressing immediate balance of payments needs of LICs in response to shocks:
  - SBAs, PRGF augmentations, and CFFs are the primary event set.
  - Precautionary SBAs and SBAs/PRGF augmentations addressing natural disasters are excluded.
  - Some SAF/ESAF/PRGF arrangements are included only if they address immediate balance of payments needs.
- Innovations relative to prior literature:
  - Examines a more homogeneous subset of arrangements (LICs + immediate BOP needs) to improve model specification.
  - Uses observable “supply constraints” (factors affecting the Fund’s evaluation of capacity to repay) to refine identification of normal episodes (periods without Fund financing for shocks).
  - Explicitly examines effects of global economic conditions (real oil and non-oil commodity prices, cyclical component of world trade) on demand for Fund arrangements by LICs.

### Key empirical findings (summary as stated)
- The econometric model fits well and remains resilient through robustness checks.
- A wider range of economic variables are significant determinants of approval of Fund arrangements with LICs, including:
  - reserve coverage,
  - current account balance to GDP,
  - real GDP growth,
  - macroeconomic stability indicator,
  - terms of trade shocks.
- Global conditions matter: changes in real oil and non-oil commodity prices, and the cyclical component of world trade, are significant determinants.
- Conclusion drawn: demand for Fund resources by LICs is likely to be cyclical in response to global conditions, with intensity depending on severity and persistence of adverse exogenous shocks.

### Rationale for LIC-only, shock-focused specification
- LICs share characteristics that make their balance of payments needs more homogeneous:
  - Narrow export base concentrated in primary commodities, making them more exposed to terms of trade shocks.
  - Historically prone to macroeconomic policy slippages.
  - Limited/no market access, greater dependence on official financing and foreign aid.
  - Often require Fund program “seal of approval” to access foreign aid and debt rescheduling.
  - Do not pose systemic or contagion risk like emerging markets (much smaller weight in global GDP and limited financial flows).
- Prior LIC-only work (Bird and Rowlands, 2009) had limited success; this study argues that further disaggregation by type of balance of payments need (immediate vs. protracted/structural) matters for model improvement.

### Supply-side (Fund) constraints and their treatment
- Observable circumstances under which access to Fund resources may be constrained are identified and used to exclude episodes from “normal” classifications:
  - (i) incurred overdue financial obligations to the Fund;
  - (ii) had interruption in its current program;
  - (iii) lacked the technical capacity/willingness to implement Upper Credit Tranche quality adjustment programs;
  - (iv) incurred significant arrears to bilateral and multilateral creditors and lacked a cooperative framework to address these arrears.
- Removing episodes with these characteristics aims to better distinguish the effects of economic factors on demand for Fund arrangements.

### Literature context (highlights)
- Existing literature on determinants of IMF arrangements is extensive but inconclusive; surveys by Steinwand and Stone (2008) and Bird (2007) conclude that models are far from definitive and key variables may be omitted.
- Earlier empirical consensus was limited: reserve holdings, previous participation in Fund programs, and low income sometimes significant; other variables (external debt burdens, terms of trade shocks, current account balance, fiscal deficits, monetary expansion, inflation) show mixed evidence.
- Political/supply-side variables (quota, UN voting, U.S./European influence measures, military aid etc.) yield mixed and sometimes limited improvements in predictive power.
- Few studies disaggregate by country group; where they do, results differ between LICs and MICs, and LIC specifications have often been weaker.

### Stylized facts on LIC landscape and Fund financing (selected points)
- Time variation in arrangements addressing immediate balance of payments needs:
  - Early 1980s: financing to LICs peaked at about twenty-five arrangements annually; 40 percent of LIC membership requested Fund assistance.
  - Mid-1990s: financing reached about 30 percent of membership (partly due to transition economies).
  - 1996–2007: subdued demand — nearly six arrangements per year during 1996–2002 and two arrangements per year during 2003–07.
  - 2008: picked up sharply to seventeen arrangements, reaching one fifth the LIC membership.
- Historical context for high-demand episodes:
  - Early 1980s episode related to the LDC debt crisis, elevated oil prices (1973–74 first oil shock; 1979 second oil shock), decline in non-oil commodity prices, poor macroeconomic management, and extensive debt reschedulings and concessional lending.
  - 2008 episode coincided with the food and fuel price shock followed by the global economic crisis; IMF Board papers (2009a and 2009b) note impacts on LICs via reduced export demand, lower FDI, and reduced remittances.
  - Result noted: average LIC growth weakened sharply from pre-crisis rates of 5–7 percent to 2.4 percent forecasted for 2009.
- Improvements in LIC macroeconomic indicators since 1980:
  - Growth picked up (median about 5½ percent), though worst quartile growth remained about 3 percent.
  - Inflation came down to single digits.
  - Fiscal deficits were sharply reduced — a correction reaching 6 percent of GDP for the worst quartile.
  - Reserve coverage improved, though worst quartile remained below three months of imports.
  - Total debt service to exports ratio fell sharply (median close to less than 10 percent from close to 30 percent previously), with an even steeper decline for the worst quartile.
- Remaining vulnerabilities:
  - Current account balances stayed volatile; worst quartile deficit remained about 10 percent of GDP.
  - Terms of trade volatility remained high; worst three-year cumulative terms of trade loss close to 30 percent for the worst quartile and a median loss of about 10 percent in the worst episodes.

### Potential economic determinants examined (variables and expected relationships)
- Standard variables (with prior mixed empirical evidence) reviewed and used in analysis:
  - International reserve coverage (months of imports): lower coverage increases likelihood of requesting Fund credit and is central to Fund’s assessment of need.
  - Real GDP growth: weaker growth increases likelihood of seeking Fund credit; prior evidence mixed.
  - Debt service to exports: heavy debt burden increases need but may reduce Fund willingness to lend; mixed results.
  - Current account balance to GDP: high deficits could increase requests, but previous results mixed—possible reasons include sample heterogeneity and event misclassification.
  - Income per capita: lower income associated with higher dependence on Fund and official flows; most studies find significant negative effect.
  - Change in terms of trade: negative commodity shocks deteriorate BOP; results previously mixed.
  - U.S. real interest rate (global liquidity indicator): higher real rates tighten global liquidity and increase debt servicing costs; mixed evidence in prior work.
  - Paris Club dummy: PC rescheduling in current or following year used to capture pre-conditions for rescheduling; previous work found significant positive effect from imminent rescheduling.
  - Election dummy: tests for effects of elections (lagged and lead versions); some studies find significance.
- New/alternative indicators used in this study:
  - Composite macroeconomic stability indicator (mitot) adapted from Jaramillo and Sancak (2009):
    - mitot combines inflation (cpi), nominal exchange rate depreciation (xr), change in reserves (res), imports of goods and services (mgs), government balance (gbal), GDP (gdp), and black market premium (blackpr), with weights inverse to standard deviations.
    - Higher mitot indicates increased macroeconomic instability.
  - Net resource transfers to GDP: captures creditor/donor willingness to finance current account deficits; important for LICs where donor assistance matters.

### Methodological approach (high-level)
- Estimates a single-equation probability model as a reduced-form estimate reflecting both the member’s decision to request and the Fund’s decision to approve, following common practice because identifying restrictions for separate modeling are hard to find.
- Improves event identification by:
  - Focusing on arrangements addressing immediate BOP needs (SBAs, PRGF augmentations, CFFs) and excluding precautionary/natural-disaster cases.
  - Excluding episodes where observable supply constraints suggest the Fund would preclude access.
- Tests robustness across sub-samples, including arrangements explicitly addressing exogenous shocks and LICs with high export concentration in non-oil commodities.

*Source: _wp09263 - INTRODUCTION (IMF working paper content provided).*

### Box 1. Why A Composite Indicator for Macroeconomic Stability?

### Box 1. Why A Composite Indicator for Macroeconomic Stability?

### Definition and Formula
- The composite indicator incorporates resource transfers, as defined in the World Bank’s Global Development Finance (GDF) database, to capture the effect of overall country specific resource availability.
- Formula as presented in the source:
  grantportprofitfdishortIPDnrt

### Variable Definitions
- nrt is net resource transfers
- D is disbursements on long-term debt and IMF purchases/loans
- P is repayments on long-term debt and IMF repurchases
- I is total interest payments
- short is the change in short-term debt
- fdi is the foreign direct investment
- profit is fdi profit remittances
- port is portfolio investment flows
- grant is official grants excluding technical cooperation

### Data Coverage
- This series is available in the World Bank’s GDF only 1985 onwards.29

*Source: _wp09263 - Box 1. Why A Composite Indicator for Macroeconomic Stability?*

### Box 1. Why A Composite Indicator for Macroeconomic Stability?

### Box 1. Why A Composite Indicator for Macroeconomic Stability?

### Rationale for a composite indicator
- Any single variable provides only partial information; combined indicator advocated by Jaramillo and Sancak (2009), Fischer (1993), and Sahay and Goyal (2006).
- Inflation: in principle the most obvious indicator of macroeconomic instability, but
  - controlled and/or fixed prices and regulated utilities in many developing countries limit usefulness;
  - Fischer (1993): low and stable inflation can be maintained for long periods by policies that are not ultimately sustainable.
- Fiscal deficit: a good indicator of unsustainable policies (direct price controls lead to higher fiscal deficits because of subsidies), but measurement problems in low income countries (inadequate coverage of public sector and quasi-fiscal deficits) may reduce usefulness.
- Exchange rate developments and reserve losses: more likely to reflect underlying macroeconomic problems despite measurement issues; fixed exchange rates defended by reserve loss.
- Conclusion: composite indicator captures complementary signals that single variables may miss.

### Modifications introduced in this study to Jaramillo and Sancak (2009)
- Addition of black market premium as a separate component, inspired by Fischer (1993).
  - Rationale: black market premium indicates distorted or controlled exchange market and often reflects underlying fiscal position.
  - Satyanath and Subramanian (2004): change in nominal parallel market exchange rate is a better measure of nominal macroeconomic instability than inflation.
  - Limitations of relying exclusively on black market premium:
    - short dataset would constrain estimation and forecasting framework;
    - absence of a premium does not guarantee sustainable policies — forecaster must also assess other macroeconomic variables.
- Reserves scaling: reserves are scaled by lagged value of imports of goods and services (instead of by base money as in the original index) to assure comparability across countries.

### Additional country-specific and global variables considered
- Variables reflecting country-specific external vulnerability and financing capacity:
  - Combined indicator reflecting debt service burden, creditors’ willingness to finance balance of payments need, and donor inflows that may offset financial outflows (individual effects of components also tested).
  - Real export growth: LICs with narrow export base are exposed to frequent real quantity shocks that may pressure balance of payments sustainability.
- Global variables examined to capture time-specific common external shocks:
  - Global demand: cyclical components of World GDP and real World trade.
  - Change in real non-oil commodity prices: relevant because LICs are major exporters of non-oil primary commodities.
  - Change in real oil prices: interacted with dummy for oil importers; country specific time series of oil imports to GDP also used.
  - World interest rates and global business cycle considered (Elekdag (2008) referenced as an example of prior work).

### Identification of dependent variable (approval of IMF arrangements)
- Dependent variable: panel dummy equal to 1 if a new Fund arrangement is approved, 0 otherwise.
- Included arrangements addressing immediate balance of payments need: SBAs, SAF/ESAF/PRGF augmentations, CFFs, and some SAF/ESAF/PRGF arrangements addressing immediate needs.
- Sample period: 1980–2004, identifying 235 financing events and 297 normal episodes.
- Variant dependent variable for arrangements explicitly addressing exogenous shocks: only SAF/ESAF/PRGF augmentations and CFFs; 91 such financing events identified in estimation sample.
- Refinements to identify financing events:
  - (i) precautionary SBAs and SBAs/PRGF augmentations addressing natural disasters excluded;
  - (ii) some SAF/ESAF/PRGF arrangements added if addressing immediate balance of payments needs arising from policy shocks (systematically determined using program interruptions).
- Normal episodes defined as initial year of two successive years with no Fund financing for shocks when member eligible to access Fund resources. Observations excluded if member had overdue obligations, program interruptions, SMPs, EPCAs, three years leading up to EPCAs, or arrears to other creditors.

### Econometric specification
- Binary response model for panel data: panel probit framework with alternatives for pooled probit, random effects (RE) probit, fixed effects (FE) probit, and correlated RE (Chamberlain-Mundlak device).
- General form: P(y_it = 1 | x_it, c_i) = Φ(c_i + x_it' β) with Φ the normal c.d.f.
- Explanatory variables (except exogenous shocks) are lagged one period for Fund financing events to avoid endogeneity.
- Estimation approaches:
  - Pooled probit with cluster-robust standard errors (assumes independence over t and i).
  - RE probit treats c_i as random with specified distribution and assumes independence between c_i and x_i.
  - Correlated RE probit relaxes independence between c_i and x_i using time averages.
- Interpretation note: coefficients β̂ do not equal marginal effects; marginal effect depends on β̂ and x_it via normal density φ.

### Key estimation results (benchmark specifications; Table 1 summary)
- Two models estimated:
  - Model I: All shocks (policy and/or exogenous shocks). Estimated by correlated random effects probit.
  - Model II: Exogenous shocks only. Estimated by pooled probit with robust standard errors.
- Sample and modeling details:
  - Fifty-five countries included in sample over 1980–2004.
  - # of observations: Model I columns (1),(2): 532; Model II columns (3),(4): 388.
  - # of countries: 55 (Model I), 53 (Model II).
  - Sample probability: 0.44 (Model I), 0.23 (Model II).
- Selected estimated coefficients (coefficient (t statistic)) — preserve exactly as in Table 1:
  - Current account balance to GDP: -0.076 *** (-4.61); -0.076 *** (-4.51); -0.073 *** (-4.7); -0.073 *** (-4.13)
  - Reserve coverage in months of imports (CFA): -0.478 *** (-6.08); -0.492 *** (-6.15); -0.466 *** (-4.12); -0.448 *** (-3.96)
  - Reserve coverage in months of imports (non-CFA): -0.769 *** (-8.71); -0.804 *** (-8.82); -0.589 *** (-6.05); -0.598 *** (-6.13)
  - Macroeconomic stability indicator: 0.068 *** (2.89); 0.072 ** (2.21)
  - Depreciation: 0.016 *** (3.11); 0.020 *** (2.99)
  - Government balance to GDP: -0.047 (-1.6); -0.060 ** (-2.18)
  - Real GDP growth: -0.113 *** (-4.24); -0.127 *** (-4.69); -0.167 *** (-5.71); -0.180 *** (-5.85)
  - Official Grants to GDP: -0.046 (-1.64)
  - Total debt service to exports: 0.027 *** (3.03); 0.030 *** (3.15)
  - Change in terms of trade in previous year: -0.022 *** (-2.8); -0.019 ** (-2.49); -0.028 *** (-4.19); -0.028 *** (-4.26)
  - Change in real oil prices in previous two years: 0.009 *** (2.85); 0.010 *** (3.57); 0.012 *** (3.96); 0.013 *** (4.44)
  - Real world trade, cyclical component: -0.099 ** (-2.53); -0.087 ** (-2.28); -0.099 *** (-2.77); -0.109 *** (-2.71)
  - Change in real non-oil commodity prices: -0.020 (-1.58); -0.031 ** (-2.12); -0.022 (-1.45)
  - Real growth of goods exports in previous year: -0.009 * (-1.79)
  - Real GDP per capita: -0.281 (-1.37); -0.439 ** (-2.08)
  - Paris Club dummy: 0.774 *** (3.24); 0.685 *** (2.95); 0.525 * (1.67); 0.587 * (1.72)
  - Country specific averages: Total debt service to exports: 0.044 *** (2.63); 0.044 *** (2.64)
  - Country specific averages: FDI to GDP: -0.105 * (-1.76); -0.106 * (-1.77)
- Model performance metrics:
  - Pseudo R2: 0.58; 0.57; 0.62; 0.64 (for the four benchmark specifications reported)
  - # of observations repeated: 532; 532; 388; 388
  - Wald test (prob) values shown in table footnotes.
- Main empirical conclusions from Table 1 and robustness checks:
  - Economic variables significantly determine approval of Fund arrangements.
  - Time-varying country-specific variables significant across models: reserve coverage, current account balance to GDP, real GDP growth, macroeconomic stability indicator, and terms of trade shocks.
  - Among time-specific global shocks, growth in real oil prices and cyclical component of world trade significant in both models.
  - Paris Club dummy significant at 1 percent in Model I and 10 percent in Model II (38 out of 55 countries had Paris Club debt relief in estimation sample).
  - Reserve coverage has larger impact on demand for Fund financing for non-CFA countries versus CFA countries; LR test rejects equality of coefficients for reserve coverage at 5 percent in Model I.
  - Nominal depreciation highly significant in both models when included; inflation and nominal depreciation are highly correlated and entered one at a time.
  - Macroeconomic stability indicator significant in both models; specifications using the composite indicator preferred for robustness.

### Noteworthy substantive findings and interpretations
- Reserve coverage:
  - Some CFA countries did not request Fund financing despite exceptionally low reserves, possibly due to reserve pooling arrangement and French Treasury convertibility guarantee.
  - CFA countries sometimes obtained Fund financing with comfortable reserve levels, especially for exogenous shocks.
  - Non-CFA countries rarely had Fund financing with reserve coverage higher than four months of imports (except for exogenous shocks).
- Commodity prices and exogenous shocks:
  - Increase in real non-oil commodity prices tends to reduce likelihood of Fund financing in Model II (exogenous shocks), though not significant across all specifications; suggests "bunching" of Fund financing in response to adverse common commodity price shocks.
  - Growth in real oil prices has a positive and significant effect on probability of financing events.
- Heterogeneity and persistent factors:
  - Country-specific averages of external debt service to exports and FDI to GDP significant in Model I, suggesting persistent divergences in debt burden and net resource inflows explain frequent use of Fund resources by some countries.
  - Net present value of debt to exports not significant contemporaneously, but country-specific average of net present value of debt to exports is an important source of country heterogeneity.
- Variables tested that do not work or are insignificant:
  - Real effective exchange rate (REER) not significant (sign ambiguous in literature and equilibrium REER movements may explain insignificance).
  - Alternative measures of terms of trade shocks scaled by GDP and oil imports to GDP not significant.
  - Net resource transfers and its components largely not significant, though country specific average of net resource transfers to GDP is significant in Model I.
  - U.S. real interest rate not significant (likely because LICs lack market access and rely on concessional financing).
  - Election dummies (lagged and upcoming) not significant.

### Goodness of fit and predictive performance
- Pseudo-R2 indicates about 57 to 64 percent of the uncertainty in the data is explained by the empirical models.
- Likelihood ratio test strongly rejects null that coefficients are jointly zero for both models.
- In-sample predicted probabilities (percentiles) show good separation between financing events and normal episodes:
  - Model I (in sample): median predicted probability for financing events = 0.96 vs 0.04 for normal episodes; 90 percent of financing events have probabilities above 0.41; 75 percent of normal episodes below 0.21.
  - Model II (in sample): 90 percent of financing events have predicted probabilities above 0.22; 90 percent of normal episodes below 0.29.
- Out-of-sample predictions (model estimated for 1980–2000, predicted for 2001–04) remain distinct despite rarity of financing events in out-of-sample period:
  - Proportion of financing events declines from about 54 percent to 12 percent in Model I out-of-sample; from 30 percent to 6 percent in Model II out-of-sample.
- Table 2 reported detailed predicted probability percentiles for financing events and normal episodes (in sample and out of sample).

*Source: _wp09263 - Box 1. Why A Composite Indicator for Macroeconomic Stability?*

### 0.73 for all 15 financing events, while 75 percent of normal episodes have predicted

### _wp09263 - 0.73 for all 15 financing events, while 75 percent of normal episodes have predicted

### Threshold Probability Analysis: method and rationale
- The study applies the loss function approach suggested by Demirgüc-Kunt and Detragiache (1999) to determine a threshold probability: a predicted probability exceeding this threshold signals a financing event.
- The expected loss function is the weighted average of type I and type II errors, with weights reflecting costs attached to each type of error. Minimization of the loss function yields the threshold probability.
- Threshold probabilities are calculated for:
  - equal weights (type I and type II errors equally weighted), and
  - a 3:1 ratio weighting missed financing events (type I error) as three times more costly than wrong calls (type II error).
- The 3:1 ratio is used to reduce the risk of missing financing events (make missed financing events more costly) and thus capture countries that might need Fund financing at an early stage.

### Threshold results and error trade-offs
- Baseline descriptive observation:
  - "0.73 for all 15 financing events, while 75 percent of normal episodes have predicted probabilities lower than 0.55."
- Model I (first model):
  - Asymmetrically weighted (3:1) loss function minimized at threshold probability of 0.21.
    - Corresponds to about 24 percent wrong calls versus 5 percent missed financing events in the sample.
  - Equally weighted loss function yields threshold probability of 0.50.
    - Corresponds to 12 percent wrong calls of normal episodes and 12 percent missed financing events.
  - A higher threshold probability of 0.59 in Model I reduces wrong calls to less than 10 percent of normal episodes.
- Model II (second model):
  - Asymmetrically weighted (3:1) loss function minimized at threshold probability of 0.14.
    - Limits missed financing events to only 5 percent, with wrong calls at 18 percent.
  - Equally weighted loss function yields threshold probability of 0.28.
    - Corresponds to about 10 percent wrong calls versus 11 percent missed financing events.
  - In Model II, a predicted probability above 0.28 reduces wrong calls to less than 10 percent (same as equally weighted threshold).

### Operational forecasting framework and vulnerability classification
- The probabilities predicted by both models compared to their corresponding threshold probabilities constitute an operational forecasting framework.
- Thresholds are chosen to minimize the risk of missing a financing event.
- Classification implications:
  - Countries with predicted probabilities above the basic thresholds can be considered “vulnerable,” indicating a possible need for Fund financing.
  - Within this group, a set of “highly vulnerable” countries can be identified using higher threshold probabilities associated with fewer wrong calls (e.g., 0.59 in Model I; 0.28 in Model II).

### Effects of explanatory variables on the probability of financing events
- Because the model is nonlinear, the effect of a variable on the probability of a financing event depends on:
  - its coefficient, and
  - the values of other explanatory variables and their coefficients.
- Estimated coefficients therefore have no direct interpretation; marginal effects are typically reported at preset values of other variables. This study uses counterfactual simulations instead.
- Simulation design:
  - Other explanatory variables are fixed at two initial-value sets:
    - (i) sample median, and
    - (ii) 75th (25th) percentile if the estimated coefficient is positive (negative).
  - Changes to global variables are set to zero in these simulations to highlight the effect of each variable on:
    - a “median PRGF country,” and
    - a weakly positioned PRGF country,
    - in the absence of common shocks.
- Key simulated outcomes:
  - For both models, the predicted probability for Fund financing is zero for the median country.
  - For the weakly positioned country, the predicted probability increases to:
    - 0.84 in the first model, and
    - 0.59 in the second model.
  - Relative to the thresholds above, the median country is predicted not to request Fund financing, while the weakly positioned country is predicted to request Fund financing in both models.

### Counterfactual simulations and reserve coverage example
- Figure 5 (Model I, all shocks) shows that an isolated change in a single explanatory variable across its sample range has a much larger effect on the weakly positioned country than on the median country.
- Reserve coverage is highlighted as particularly influential:
  - Both the predicted probability and the upper bound of its confidence interval cross the threshold probability within the sample range for the weakly positioned country.
  - Example implication: a financing event is predicted for a weakly positioned non-CFA country at reserve coverage less than (text truncated in source).

*Source: Author's calculations from _wp09263 - 0.73 for all 15 financing events, while 75 percent of normal episodes have predicted probabilities lower than 0.55.*

### 3.6 months of imports. The 95 percent confidence interval of the predicted probability falls

### _wp09263 - 3.6 months of imports. The 95 percent confidence interval of the predicted probability falls

### Counterfactual simulations: reserve coverage, current account and predicted probability
- For a non-CFA country: predicted probability falls below the threshold at 3.6 months of imports. The 95 percent confidence interval of the predicted probability falls below the threshold if reserve coverage is above 4.4 months of imports.  
- For a similarly situated CFA country: corresponding figures are 5.1 and 7.3 months of imports respectively.  
- Only a current account surplus above 15 percent of GDP would bring the predicted probability below the threshold for the weakly positioned country.  
- Simulations are counterfactual: all other variables remain unchanged while a single variable is allowed to change over its entire sample range. Some time-specific variables (real oil prices, real commodity prices, cyclical component of world trade) are extended beyond their actual sample range. Dummy variable is set to zero.

### Median and percentile values used in simulations (Table 3)
- Current account balance to GDP (%): -4.9 (Median), -8.9 (75th/25th percentile entry shown as a combined line in source)  
- Reserve coverage in months of imports 1/: 3.1 (Median)  
- Macroeconomic stability indicator: 1.9 (Median), 4.5 (75th)  
- Real GDP growth (%): 4.3 (Median), 1.7 (75th/25th)  
- Total debt service to exports (%): 16.1 (Median), 26.1 (75th)  
- Change in terms of trade in previous year (%): 0 (Median), -8.4 (75th/25th)  
- Real GDP per capita 2/: 3272 (Median), 21 (75th/25th) — (In constant 2000 US$ per source formatting)  
- Country Specific Averages: FDI to GDP (%): 1.1 (Median), 0.4 (75th/25th)  
- Predicted probability 3/: Model I: 0.0 (Median), 0.84 (75th); Model II: 0.0 (Median), 0.59 (75th). The 95 percent confidence interval for the predicted probability is in brackets in the source (e.g., Model I [0.0 0.0] [0.69 0.94]; Model II [0.0 0.0] [0.39 0.77]).

### Key counterfactual findings and thresholds for median country
- An isolated change in most other explanatory variables is unlikely to reduce predicted probability below the threshold for the weakly positioned country.  
- Predicted probability approaches 1 with significant isolated deterioration in any explanatory variable, increasing classification as “highly vulnerable.”  
- More plausible combined deterioration (e.g., adverse terms of trade shock leading to wider current account deficit and lower reserve coverage) would accelerate convergence of predicted probability toward 1.  
- Median country does not need Fund financing except in the following extreme cases (with other variables held fixed):  
  - Reserve coverage less than 1.4 months of imports for non-CFA countries.  
  - Current account deficit to GDP exceeding 18¾ percent.  
  - GDP contraction exceeding 4¾ percent.  
  - Severe adverse terms of trade shocks.  
  - Real oil prices increasing more than twofold.  
- The threshold probability falls within the confidence interval of predicted probabilities when any of the above hold while other variables are fixed; however, the counterfactual nature implies the threshold would likely be exceeded at less extreme values when other variables also move.

### Model II (exogenous shocks) specific simulations (Figure 6 summary)
- For a non-CFA country: predicted probability falls below the threshold at 3.5 months of imports; its 95 percent confidence interval also falls below the threshold above 5 months of imports.  
- For a CFA country: corresponding figures are 4.1 and 8 months of imports respectively.  
- GDP growth and terms of trade shocks have a more profound effect and are estimated with higher precision, swinging predicted probability and confidence interval from below the threshold to 1 for a weakly positioned country.  
- A sharp adverse shock to real non-oil commodity prices could bring the predicted probability over the threshold for a median country.  
- Real GDP per capita has only a limited effect in this specification.

### Robustness checks: Logit vs Probit and alternative estimators
- Logit versus probit (both as correlated random effects estimators) yield very similar predicted values; correlation of predicted values is 0.99. Log likelihoods are also quite close.  
- Alternative probit estimators (including linear probability model with cluster robust SEs, pooled probit with cluster robust SEs) show similar sign and significance of covariates; magnitudes of coefficients and average partial effects (APEs) are broadly comparable. This suggests correlation between unobserved heterogeneity and covariates may not be strong.  
- In model I, FDI to GDP and total debt to exports become insignificant in some estimators, while their country specific averages are significant in the correlated RE model.

### Persistence of Fund engagement and previous programs
- A dummy variable equals one if the member had a Fund program (SAF/ESAF/PRGF or EFF, including shock financing) in the previous five years; zero otherwise.  
- For both models, other explanatory variables remain broadly the same when this dummy is included (Table A6).  
- The dummy for Fund engagement is significantly negative for all shocks and insignificant for exogenous shocks: previous Fund engagement seems to significantly reduce the likelihood of financing in response to policy shocks.  
- This negative correlation contrasts with earlier research finding positive association between previous and current participation in mixed samples; possible explanation relates to distinguishing immediate balance of payments needs from protracted adjustment programs.

### Natural disasters and financing events (ENDA and PRGF augmentations)
- Emergency Natural Disaster Assistance (ENDA) and PRGF augmentations for natural disasters are deliberately excluded from the financing events in both models because such events are unpredictable conditional on explanatory variables used for policy and exogenous shocks.  
- Estimation for financing for natural disasters (final specification reported in Table 4):  
  - Share of affected population: 0.041 ** (t = 2.47)  
  - Reserve coverage in months of imports (lagged 1 period): -0.412 ** (t = -2.31)  
  - Real GDP growth (lagged 1 period): -0.144 * (t = -1.79)  
  - Constant: -1.233 (t = -1.39)  
  - Pseudo R2: 0.24  
  - # of observations: 312  
  - Financing events: 17  
  - Normal episodes: 295  
  - Sample probability: 0.05  
  - # of countries: 50  
- Financing for natural disasters is rare in the sample (17 events vs 295 normal episodes), making results tentative; empirical evidence supports conceptual argument to exclude natural disasters from main event set.

### Export concentration: non-oil commodities (sub-sample results, Table 5)
- Sub-sample estimation for LICs with high concentration of non-oil commodity exports (sample averages 1980–2004) yields:  
  - Current account balance to GDP 2/: coefficients around -0.076 *** to -0.085 *** across thresholds.  
  - Reserve coverage in months of imports (CFA) 2/: coefficients around -0.478 *** to -0.518 ***.  
  - Reserve coverage in months of imports (non-CFA) 2/: coefficients around -0.769 *** to -0.963 ***.  
  - Macroeconomic stability indicator 2/: positive and significant coefficients (e.g., 0.068 *** to 0.086 **).  
  - Real GDP growth 2/: negative and significant coefficients around -0.113 *** to -0.094 ***.  
  - Change in terms of trade in previous year: negative and often significant (e.g., -0.022*** to -0.016).  
  - Change in real oil prices in previous two years: positive and often significant (e.g., 0.009*** to 0.006).  
  - Real world trade, cyclical component: negative and often significant (e.g., -0.099** to -0.124**).  
  - Change in real non-oil commodity prices: mixed significance with some negative coefficients (e.g., -0.020 to -0.033*).  
  - Paris Club dummy: positive and significant across specifications (e.g., 0.774 *** to 0.685 **).  
- Observations and sample probabilities by threshold of non-oil commodity export share:  
  - Benchmark (>45%): # of observations 532, Sample probability 0.44, # of countries 55.  
  - >50%: # of observations 405, Sample probability 0.48, # of countries 42.  
  - >60%: # of observations 381, Sample probability 0.49, # of countries 40.  
  - >70%: # of observations 347, Sample probability 0.48, # of countries 36.  
  - (Additional column) # of observations 321, Sample probability 0.46, # of countries 33.  
- Quick observation from source: for countries with less than 45 percent of their exports in non-oil commodities, the sample probability of Fund financing drops to about 0.31 (text ends mid-sentence in supplied content).

*Source: _wp09263 (excerpt provided).*

### 0.48 for those above that threshold. Nevertheless, the benchmark specification holds up

### _wp09263 - 0.48 for those above that threshold. Nevertheless, the benchmark specification holds up

### Resilience of the benchmark specification
- The benchmark specification "holds up remarkably well" through significant step-wise reductions in the sample; results are qualitatively similar and sizes of coefficients are broadly comparable.
- After a certain point, "small sample problems" appear to kick in.
- Estimated coefficients for:
  - Change in real non-oil commodity prices: "much higher and become significant in reduced samples."
  - Cyclical movements in real world trade: "considerably higher."
- Working with sub-groups of LICs with high concentration in non-oil commodities "improves the specification for common shocks."
- For other variables, "the benchmark model remains resilient."

### Fund financing: CFA versus non-CFA countries (Table 6 summary)
- Earlier results: reserve coverage has a "significantly larger effect on demand for Fund financing for non-CFA countries vis-à-vis CFA countries."
- Question: Do CFA members need Fund financing less due to reserve pooling arrangement backed by the French Treasury?
- Selected coefficient estimates from Table 6 (Model I; columns reported):
  - Current account balance to GDP 2/: -0.076 ***, -0.075 ***, -0.074 *** 
  - Reserve coverage in months of imports (non-CFA) 2/: -0.769 ***, -0.797 ***, -0.774 ***
  - Macroeconomic stability indicator 2/: 0.068 ***, 0.062 ***, 0.053 ***
  - Real GDP growth 2/: -0.113 ***, -0.114 ***, -0.114 ***
  - Change in terms of trade in previous year: -0.022 ***, -0.021 ***, -0.020 **
  - Change in real oil prices in previous two years: 0.009 ***, 0.008 ***, 0.009 ***
  - Real world trade, cyclical component: -0.099 **, -0.096 **, -0.057
  - Change in real non-oil commodity prices: -0.020; -0.021*; -0.027*
  - Real growth of goods exports in previous year: -0.009 *, -0.009 *, -0.007
  - Paris Club dummy: 0.774 ***, 0.801 ***, 0.769 ***
  - CFA Dummy: -0.506
  - Constant: 0.551, 0.666, 0.569
- Country specific averages:
  - Total debt service to exports: 0.044 ***, 0.045 ***, 0.048 ***
  - FDI to GDP: -0.105 *, -0.104 *, -0.108 **
- Sample and estimation details (from Table 6):
  - # of observations: 532, 532, 412
  - Sample probability: 0.44, 0.44, 0.42
  - # of countries: 55, 54, 45
  - Note: 1/ Fund financing for policy and/or exogenous shocks excluding natural disasters. Estimated by a correlated random effects probit model. Significant at 10 percent:*; 5 percent:**; and 1 percent:***. 2/ These variables are lagged one period for Fund financing events in order to avoid endogeneity problem.
- Interpretation:
  - Adding a CFA dummy: dummy is "negative but insignificant" (standard error of estimated coefficient for CFA dummy is 0.46 corresponding to a t-statistic of -1.1).
  - Re-estimating the benchmark only for non-CFA countries: results "qualitatively similar" and benchmark "remains valid."
  - In non-CFA sample, change in real non-oil commodity prices: coefficient "becomes higher and significant at 10 percent level."
  - In non-CFA sample, cyclical movements in real world trade: coefficient becomes lower and "insignificant."

### Key empirical findings across models and robustness checks
- Determinants found very significant in both models:
  - Reserve coverage
  - Current account balance to GDP
  - Real GDP growth
  - Macroeconomic stability indicator
  - Terms of trade shocks
- Global economic conditions significantly affecting demand for Fund financing:
  - Change in real oil prices
  - Change in real non-oil commodity prices
  - Cyclical component of world trade
- Consequence: "An adverse global environment may tip the balance for weakly positioned countries, thus increasing the demand for Fund financing."
- Country-specific heterogeneity:
  - Country specific averages of external debt burden indicators and resource inflows are significantly associated with unobserved country heterogeneity in the model for all shocks:
    - Total debt service to exports
    - Net present value of debt to exports
    - FDI to GDP
    - Net resource transfers to GDP
  - No significant country heterogeneity detected for approval of arrangements addressing exogenous shocks.
- Income per capita:
  - Significance of income per capita "disappears in the model for all shocks," suggesting previous findings of significance may be driven by mixed samples of LICs and MICs.
  - Some weak evidence income per capita may be significant for Fund financing addressing exogenous shocks.
- Persistence and previous Fund engagement:
  - "No persistence of Fund programs could be detected."
  - Previous Fund engagement is found to "reduce the likelihood of financing in response to policy shocks."
  - This contrasts with findings for PRGF-eligible countries where prior participation increases duration of program spells.
  - Distinguishing immediate balance of payments needs from protracted needs may explain this difference.

### Quantitative magnitudes and model performance (selected table highlights)
- Model performance indicators (from various tables):
  - Pseudo R2 values reported across specifications:
    - Example values: 0.57, 0.56, 0.58, 0.59, 0.61, 0.63 (varying by model and specification)
  - Sample probabilities:
    - Model I (all shocks): 0.44
    - Model II (exogenous shocks): 0.23
- Average Partial Effects (selected APEs from Table A4 and A5):
  - Model I APE examples:
    - Current account balance to GDP: -0.009 *** (Linear model); correlated RE probit: -0.070 ***
    - Reserve coverage (non-CFA): -0.101 *** (Linear); correlated RE probit APE: -0.638 ***
    - Real GDP growth: -0.014 *** (Linear); correlated RE probit APE: -0.107 ***
    - Change in terms of trade: -0.002 ** (Linear); correlated RE probit APE: -0.016 ***
    - Change in real oil prices: 0.002 *** (Linear); correlated RE probit APE: 0.010 ***
  - Model II APE examples:
    - Current account balance to GDP: -0.011 *** (Linear); correlated RE probit: -0.073 ***
    - Reserve coverage (non-CFA): -0.065 *** (Linear); correlated RE probit APE: -0.589 ***
    - Real GDP growth: -0.024 *** (Linear); correlated RE probit APE: -0.167 ***
    - Change in terms of trade: -0.003 *** (Linear); correlated RE probit APE: -0.028 ***
    - Change in real oil prices: 0.002 *** (Linear); correlated RE probit APE: 0.012 ***

### Concluding remarks and policy implications
- The study identifies economic factors affecting LIC demand for Fund financing in response to policy and exogenous shocks, filling a gap by focusing on this specific subset of Fund arrangements.
- Two key contributions:
  - Examination of a more homogenous subset: Fund financing addressing immediate balance of payments needs of LICs in response to shocks, with further differentiation for exogenous-shock arrangements.
  - Use of observable "supply constraints" to identify normal episodes and distinguish supply from demand effects.
- Practical application:
  - Framework can be used with IMF’s semi-annual World Economic Outlook (WEO) forecasts to help identify LICs with greater likelihood of requiring Fund financing.
  - Suggested extension: incorporate a complementary "stress testing" approach to gauge potential demand for Fund financing under a range of adverse common and country-specific shocks and policy deteriorations.
- Overall outlook:
  - "Recent adverse common shocks impacting LICs and weakening their external positions are likely to significantly increase the number of LICs requiring Fund financing."
  - Due to nonlinearities, "a severe enough cumulative deterioration" could increase probability of Fund financing more than proportionally.
  - Future demand for Fund resources "is likely to be characterized by cycles in response to global conditions" with intensity determined by severity and persistence of exogenous shocks.

### Data and sample (Appendix I and Tables A1)
- Key variables and sources (selected):
  - Present value of debt: World Bank internal study by Dikhanov (2004)
  - Paris Club debt rescheduling and cancellations: Paris Club webpage
  - Principal arrears on long-term debt outstanding: WB Global Development Finance (GDF) dataset
  - Gross international reserves at year end: IMF WEO
  - Real Oil prices (ASPS): Oil prices deflated by US CPI (WEO)
  - Real non-oil commodity prices: Non-oil commodity prices deflated by US CPI (WEO)
  - World trade (index number 2000=100): IMF WEO
  - SBA, CFF, PRGF/SAF/ESAF arrangements, ENDA, EPCA, SMP: IMF databases (Fund Accounts, IFS, Fund Arrangements database)
- Country list for estimation sample (Table A1): 55 countries enumerated (examples include Albania, Bangladesh, Ethiopia, Ghana, Madagascar, Mozambique, Pakistan, Rwanda, Tanzania, Uganda, Vietnam, Zambia, Zimbabwe, etc.).
- Estimation samples:
  - Model I (policy and/or exogenous shocks excluding natural disasters): # of observations typically reported as 532; # of countries: 55.
  - Model II (exogenous shocks excluding natural disasters): # of observations typically reported as 388; # of countries: 53.

*Source: Author's calculations, from the content unit _wp09263 - 0.48 for those above that threshold. Nevertheless, the benchmark specification holds up*

### REFERENCES

### _wp09263 - REFERENCES

### Academic articles and journal papers
- Andersen, Thomas Barnebeck, Thomas Harr, and Finn Tarp, 2006, “On U.S. Politics and IMF Lending,” European Economic Review, Vol. 50, pp. 1843–62.
- Barro, Robert and Jong-Wha Lee, 2005, “IMF Programs: Who is Chosen and What are the Effects?” Journal of Monetary Economics, Vol. 52, No. 7, pp. 1245–69.
- Bird, Graham and Dane Rowlands, 2009, “A Disaggregated Empirical Analysis of the Determinants of IMF Arrangements: Does One Model Fit All?” Journal of International Development, Vol. 21, pp. 915–31.
- Bird, Graham and Dane Rowlands, 2007, “Should It be Curtains for Some of the IMF’s Lending Windows?” The Review of International Organizations, Vol. 2, pp. 281–99.
- Bird, Graham and Dane Rowlands, 2001, “IMF Lending: How is It Affected by Economic, Political and Institutional Factors?” Journal of Policy Reform, Vol. 4, pp. 243–70.
- Bird, Graham, 2007, “The IMF: A Bird’s Eye View of its Role and Operations,” Journal of Economic Surveys, Vol. 21, No. 4, pp. 683–745.
- Bird, Graham, Mumtaz Hussain, and Joseph P. Joyce, 2004, “Many Happy Returns? Recidivism and the IMF,” Journal of International Money and Finance, Vol. 23, No. 2, March, pp. 231–51.
- Cameron, A. Colin, Frank A.G. Windmeijer, 1997, “An R-Squared Measure of Goodness of Fit for some Common Nonlinear Regression Models,” Journal of Econometrics, Vol. 77, pp. 329–42.
- Chamberlain, Gary, 1982, “Multivariate Regression Models for Panel Data,” Journal of Econometrics, Vol. 1, pp. 5–46.
- Conway, Patrick, 1994, “IMF Lending Programs: Participation and Impact,” Journal of Development Economics, Vol. 45, pp. 365–91.
- Conway, Patrick, 2005, “The Revolving Door: Duration and Recidivism in IMF Programs,” The Review of Economics and Statistics, Vol. 89, No. 2, pp. 205–20.
- Dreher, Axel and Nathan M. Jensen, 2007, “Independent Actor or Agent? An Empirical Analysis of the Impact of U.S. Interests on IMF Conditions,” Journal of Law and Economics, Vol. 50, pp. 105–24.
- Elekdağ, Selim, 2006, “How Does the Global Economic Environment Influence the Demand for IMF Resources?” IMF Staff Papers, Vol. 55, No. 4, pp. 624–53.
- Fischer, Stanley, 1993, “The Role of Macroeconomic Factors in Growth,” NBER Working Paper No. 4565 (Cambridge, Massachusettes: National Bureau of Econimic Research).
- Ghosh, Atish, Manuela Goretti, Bikas Joshi, Alun Thomas, and Juan Zalduendo, 2007, “Modeling Aggregate Use of Fund Resources—Analytical Approaches and Medium-Term Projections,” IMF Working Paper 07/70 (Washington: International Monetary Fund).
- Hauser, J.A., 1978, “Testing the Accuracy, Usefulness, and Significance of Probabilistic Choice Models: An Information-Theoretic Approach,” Operations Research, Vol. 26, pp. 406–21.
- Hutchison, Michael M. and Ilan Noy, 2003, “Macroeconomic Effects of IMF-Sponsored Programs in Latin America: Output Costs, Program Recidivism and the Vicious Cycle of Failed Stabilizations,” Journal of International Money and Finance, Vol. 22, Issue 7, pp. 991–1014.
- Jaramillo, Laura and Cemile Sancak, 2009, “Why Has the Grass Been Greener on One Side of Hispaniola? A Comparative Growth Analysis of the Dominican Republic and Haiti,” IMF Staff Papers, Vol. 56, pp. 323–49.
- Joyce, Joseph P., 1992, “The Economic Characteristics of IMF Program Countries,” Economics Letters, Vol. 38, No. 2, pp. 237–42.
- Joyce, Joseph P., 2005, “Time Present and Time Past: a Duration Analysis of IMF Program Spells,” Review of International Economics, Vol. 13, pp. 283–97.
- Knight, M. and Julio A. Santaella, 1997, “Economic Determinants of IMF Financial Arrangements,” Journal of Development Economics, Vol. 54, pp. 405–36.
- Milesi-Ferretti, Gian Maria and Assaf Razin, 1996, “Current Account Sustainability: Selected East Asian and Latin American Experiences,” NBER Working Paper No. 5791 (Cambridge, Massachusettes: National Bureau of Economic Research).
- Mundlak, Yair, 1978, “On the Pooling of Time Series and Cross Section Data,” Econometrica, Vol. 46, pp. 69–85.
- Oatley, Thomas and Jackson Yackee, 2004, “American Interests and IMF Lending,” International Politics, Vol. 41, No. 3, pp. 415–29.
- Reinhart, C. M. and Kenneth Rogoff, 2004, “The Modern History of Exchange Rate Arrangements: A Reinterpretation,” The Quarterly Journal of Economics, Vol. CXIX, February 2004, Issue 1, pp. 1–48.
- Przeworski, Adam, James Raymond Vreeland, 2000, “The Effect of IMF Programs on Economic Growth,” Journal of Development Economics, Vol. 62, pp. 385–421.
- Santaella, Julio A., 1996, “Stylized Facts Before IMF-Supported Macroeconomic Adjustment,” IMF Staff Papers, Vol. 43, No. 3, September 1996, pp. 502–44.
- Steinwand, Martin and Randall Stone, 2008, “The Fund: A Review of the Recent Evidence,” The Review of International Organizations, Vol. 3, No. 2, pp. 123–49.
- Stone, Randall W., 2004, “The Political Economy of IMF Lending in Africa,” American Political Science Review, Vol. 98, pp. 577–92.
- Sturm, Jan-Egbert, Helge Berger, and Jakob de Haan, 2005, “Which Variables Explain Decisions on IMF Credit? An Extreme Bound Analysis,” Economics and Politics, Vol. 17, No. 2, pp. 117–213.
- Thorsten, Beck, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh, 2007, “New Tools and New Tests in Comparative Political Economy: The Database of Political Institutions,” The Development Research Group of the World Bank (Washington: World Bank).

### IMF working papers, staff papers, and institutional reports
- Cerutti, Eugenio, 2007, “IMF Drawing Programs: Participation Determinants and Forecasting,” IMF Working Paper 07/152 (Washington: International Monetary Fund).
- Demirgüc-Kunt, Asli and Enrica Detragiache, 1999, “Monitoring Banking Sector Fragility: A Multivariate Logit Approach,” IMF Working Paper 99/147 (Washington: International Monetary Fund).
- Dreher, Axel, Jan-Egbert Sturm, and James Raymond Vreeland, 2006, “Does Membership on the UN Security Council Influence IMF Decisions? Evidence from Panel Data,” Swiss Institute for Business Cycle Research No. 151 (Zurich: Institute for Business Cycle Research).
- Elekdağ, Selim, 2006, “How Does the Global Economic Environment Influence the Demand for IMF Resources?” IMF Staff Papers, Vol. 55, No. 4, pp. 624–53.
- Ghosh, Atish, Manuela Goretti, Bikas Joshi, Alun Thomas, and Juan Zalduendo, 2007, “Modeling Aggregate Use of Fund Resources—Analytical Approaches and Medium-Term Projections,” IMF Working Paper 07/70 (Washington: International Monetary Fund).
- IEO (Independent Evaluation Office, IMF), 2002, Evaluation of the Prolonged Use of IMF Resources (Washington: International Monetary Fund).
- Ivanova, Anna, Mayer, W., Mourmouras, A., and Anayiotas, G., 2003, “What determines the implementation of IMF-supported programs?” IMF Working Paper 03/8. Published in A. Mody and A. Rebucci (eds), 2006, IMF-Supported Programs: Recent Staff Research (Washington: International Monetary Fund).
- IMF, 2009a, “The Implications of the Global Crisis for Low-Income Countries,” March 2009, (Washington: International Monetary Fund).
- IMF, 2009b, “The Implications of the Global Financial Crisis for Low-Income Countries— An Update,” September 2009, (Washington: International Monetary Fund).
- Sahay, Ratna and Rishi Goyal, 2006, “Volatility and Growth in Latin America: An Episodic Approach,” IMF Working Paper 06/287 (Washington: International Monetary Fund).

### Books, chapters, and other publications
- Bird, Graham and Dane Rowlands, 2006, in Globalization and the Nation State: The Impact of the IMF and the World Bank, ed. by Gustav Ranis, James Raymond Vreeland, and Stephen Kosack (New York: Routledge), pp.231–62.
- Stone, Randall W., 2002, Lending Credibility: the International Monetary Fund and the Post-Communist Transition Princeton (New Jersey: Princeton University Press).
- Milesi-Ferretti, Gian Maria and Assaf Razin, 1996, “Current Account Sustainability: Selected East Asian and Latin American Experiences,” NBER Working Paper No. 5791 (Cambridge, Massachusettes: National Bureau of Economic Research).

*References as listed in _wp09263 - REFERENCES*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09263.pdf_