## _wp11144

## Source details

**Canonical URL:** [_wp11144](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11144.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11144.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11144.pdf.json)

---

### Introduction and recent trends
- International reserve accumulation reached 13 percent of global GDP in 2009, which represents a three-fold increase over ten years.
- Emerging market holdings rose to 32 percent of GDP.
- Footnote clarification: 26 percent if China is excluded.
- Paper focus: small non-dollarized Central American economies—Costa Rica, the Dominican Republic, Honduras, Guatemala and Nicaragua.

### Rationale for holding international reserves
- Long-standing view: reserves insure against balance of payments shocks.
- Common rules-of-thumb for reserve adequacy:
  - Cover some external commitments (e.g., three months of imports for countries with limited access to capital markets).
  - Measures of potential capital flight, e.g., short-term external debt at residual maturity and current account deficits.
- Alternative approach: buffer stock model treating reserves as a resource for smoothing consumption in the face of sudden stops (Frenkel and Jovanovic, 1981).
- Empirical determinants identified: reserves positively related to income volatility, openness and financial depth.
- Trade-off in optimal reserves: mitigate falls in domestic consumption during balance of payments crises versus costs of holding reserves (interest rate differential, quasi-fiscal costs).

### Rules-of-thumb for reserve cover (Central America comparisons)
- Central America increased gross international reserves by over 50 percent during 2000–2008.
- Larger Latin American economies: a three-fold increase during 2000–2008.
- A larger global sample of emerging markets: a six-fold increase during 2000–2008.
- Benchmarks and Central America:
  - Three months of imports rule: Central American economies barely reach this level; larger Latin American economies and the full sample of emerging markets exceed it by a wide margin.
  - Greenspan-Guidotti rule: Central America falls short despite modest external debt; other emerging markets meet or exceed it.
  - Reserves-to-broad-money (M2) rule: Historically 5–20 percent; more recently up to 50% suggested. Central America is below comparators.
- Financial intermediation context:
  - Broad money to GDP in Central America ≈ 40 percent.
  - Broad money to GDP in the larger sample of emerging economies ≈ 70 percent.
- Overall assessment: Central America’s reserve cover is low both in absolute terms and relative to emerging market peers.

### Reserve demand regressions — methodology
- Panel: N = 52 emerging economies, period 1993–2008 (T = 16).
- Dependent variable: reserves-to-GDP ratio (GIR/GDP).
- Model includes: vector of economic variables Xit, Central America dummy, crisis dummy (reversal of current account to GDP by more than five percentage points), and peer “Jones” variables (lagged).
- “Jones” variables: sum of reserve-to-GDP ratios of either other small emerging markets in the region or the five largest economies (Brazil, Chile, Colombia, Mexico, Peru).

### Reserve demand regressions — main findings (sign, magnitude, significance preserved)
- Reserve demand is positively related to openness and GDP volatility, and negatively related to exchange rate flexibility.
- Exchange rate flexibility has the largest effect: raising the period standard deviation of the nominal NCU/dollar exchange rate by one unit is associated with a decreased reserve/GDP ratio of about 25 percentage points.
- A one percentage point increase in openness or in the FDI/GDP ratio has effects on the reserve-to-GDP ratio close to the size of the exchange rate flexibility effect (direction: positive for openness and FDI).
- No significant effect of financial depth (M2/GDP) on reserve demand was found.
- Total external debt level and crisis dummies have expected positive signs but are not consistently significant.
- Short-term debt/GDP increases reserve-to-GDP by around 3 percentage points.
- Peer (Jones) effects: only reserve levels in the large emerging markets matter; an increase in the reserve-to-GDP ratio of the five largest economies in Latin America increases the reserve-to-GDP ratio of a Central American economy by 0.5 percentage points the following year. No significant effect from reserve accumulation in other Central American economies.
- Central America dummy: statistically significant negative coefficient interpreted as Central America having on average about two percentage points lower reserve-to-GDP ratio after controlling for other factors.

- Table 1. Reserve Demand Regressions. Dependent Variable: GIR/GDP (selected coefficients)
  - Log (Population): -0.008*    T-stat. -1.97
  - Openness /1: 0.117***    T-stat. 9.82
  - Exchange rate flexibility/2: -0.241*    T-stat. -1.87
  - GDP volatility/3: 0.101*    T-stat. 1.37
  - Central America dummy: -0.024*    T-stat. -1.40
  - FDI/GDP: 0.191***    T-stat. 2.90
  - Short term debt/GDP: 0.031*    T-stat. 1.71
  - Jones2*CA dummy: 0.004***    T-stat. 3.55
  - Notes:
    - 1/ Defined as (exports + imports)/GDP.
    - 2/ Defined as the period standard deviation of the NCU/dollar exchange rate, over the period average.
    - 3/ Defined as the period standard deviation of real GDP, over the period average.
    - 4/ Defined as M2/GDP.
    - ***, **, * denote results significant on the 1, 5 and 10% significance levels (one-sided t-tests).

### Models of optimal reserves — frameworks and calibration
- Frameworks used:
  - Jeanne and Rancière (2006) closed-form optimal reserves fraction of output ρ.
  - Gonçalves (2007) extension including dollar deposit withdrawals and government deposit coverage.
- Key intuition:
  - Reserves smooth domestic absorption against sudden stops (declines in the capital account and potentially output).
  - Approximation (setting δ+π = r – g = 0) implies ρ ≈ (λ + γ) / (1 + δ/π) adjusted by risk aversion σ; if δ = 0 then ρ = λ + γ; if γ = 0 then ρ = λ.
- Calibration (Non-country specific parameters)
  - Term premium, δ: 1.5%
  - Risk-free rate, r: 5%
  - Coefficient of relative risk aversion, σ: 2
  - Deposit coverage, α: 30-100%
  - Baseline crisis probability π: 10.2 percent (Jeanne and Rancière estimate); for countries with observed crises in the sample period, implied probabilities (e.g., one crisis in two decades → 5%) were used.

### Models of optimal reserves — comparative results (time averages 2003–08; selected table entries exact)
- Sample: Armenia, Bolivia, Bulgaria, Costa Rica, Dominican Republic, Guatemala, Honduras, Nicaragua, Paraguay, Romania, Uruguay.
- General patterns:
  - Optimal reserves per Jeanne and Rancière vary widely across time and countries.
  - Actual reserve levels are often close to, but typically exceeding, the Jeanne and Rancière optimal levels—especially for financially integrated emerging markets (Costa Rica, Dominican Republic, Bulgaria, Romania, Uruguay).
  - For countries with low short-term external debt (Guatemala, Honduras, Bolivia), the model prescribes low optimal reserves; Bolivia’s actual high reserves reflect terms-of-trade/commodity boom considerations not emphasized in the model.
  - Including dollar deposits (Gonçalves model) with α = 30% raises model-optimal reserves; under full (100%) deposit insurance, optimal reserves are higher still.
  - With Gonçalves calibration, actual reserves are lower than optimal for the majority of Central American countries; under full deposit insurance, Honduras is the only Central American country with actual reserves higher than the model-prescribed optimal level.

- Table 3. Actual vs. Optimal Reserves (% of GDP), Time Averages 2003–08 (selected entries)
  - Costa Rica: Actual 10.2; Jeanne and Rancière 8.2; Gonçalves (30% guarantee) 14.8; Gonçalves (Full guarantee) 21.6
  - Dom. Republic: Actual 3.4; Jeanne and Rancière 4.9; Gonçalves (30% guarantee) 5.8; Gonçalves (Full guarantee) 9.6
  - Guatemala: Actual 11.0; Jeanne and Rancière 6.4; Gonçalves (30% guarantee) 8.7; Gonçalves (Full guarantee) 15.8
  - Honduras: Actual 19.1; Jeanne and Rancière 7.3; Gonçalves (30% guarantee) 7.9; Gonçalves (Full guarantee) 12.7
  - Nicaragua: Actual 13.2; Jeanne and Rancière 17.7; Gonçalves (30% guarantee) 21.7; Gonçalves (Full guarantee) 44.4
  - Armenia: Actual 6.0; Jeanne and Rancière 13.8; Gonçalves (30% guarantee) 14.4; Gonçalves (Full guarantee) 15.6
  - Bolivia: Actual 16; Jeanne and Rancière 2.5; Gonçalves (30% guarantee) 2.7; Gonçalves (Full guarantee) 4.3
  - Bulgaria: Actual 28.9; Jeanne and Rancière 24.9
  - Paraguay: Actual 14.4; Jeanne and Rancière 11.3; Gonçalves (30% guarantee) 11.7; Gonçalves (Full guarantee) 12.3
  - Romania: Actual 15.2; Jeanne and Rancière 16.7; Gonçalves (30% guarantee) 19.0; Gonçalves (Full guarantee) 24.4
  - Uruguay: Actual 13.6; Jeanne and Rancière 11.3; Gonçalves (30% guarantee) 12.8; Gonçalves (Full guarantee) 17.7

### Sensitivity analysis and scenario results (selected quantitative anchors preserved)
- Risk aversion σ:
  - Increasing σ from 2 to 5 increases average optimal reserve-to-GDP by about 4 percent for the full sample; for Central American economies the increase is about 5 percent on average.
- Coverage parameter change:
  - Changing the coverage parameter from 1.5 to 0 raises the average optimal reserve-to-GDP ratio by almost 7 percent; for the Central American economies the increase is over 8 percent.
- Term premium δ and risk-free rate r:
  - A lower term premium decreases the opportunity cost of holding reserves and thus raises the optimal reserve-to-GDP ratio.
- Crisis probability π:
  - Baseline π = 10.2 percent.
  - If π falls by half to 5 percent, the average reserves-to-GDP ratio would decline by close to 11 percent.
  - For Central American economies, lowering π to 5 percent would reduce optimal reserve-to-GDP by around 8 percent.
- Size of output loss γ:
  - Mexico 1994 output loss: -6.5 percent → average optimal reserve-to-GDP ratio would increase by almost 70 percent for Central American economies versus 40 percent for the full sample.
  - Uruguay 2001 output loss: -14 percent → optimal reserve-to-GDP ratio would increase by 104 percent for Central American economies and by 68 percent for the full sample.
- Numerical note in text: "1.5 to 0, the average optimal reserve-to-GDP ratio would increase by almost 7 percent."

### Main findings on reserve adequacy in Central America
- Central America scores relatively low on traditional reserve benchmarks (reserves to months of imports, short-term external debt and current account positions).
- After controlling for determinants of reserve demand, Central America has significantly lower reserve-to-GDP ratios (Central America dummy ≈ -0.024*).
- Evidence of “keeping up with the Joneses” effects: Central American countries respond to reserve accumulation in the five largest Latin American economies (Jones2*CA dummy: 0.004***).
- Using Jeanne and Rancière (2006) and Gonçalves (2007) frameworks, actual reserve levels in Central America are often lower than model-prescribed optimal insurance against sudden stops and partial withdrawals of dollar deposits.

### Policy implications and recommendations
- Further reserve accumulation is an important policy option for Central American economies given model indications that many actual levels are below optimal insurance needs.
- Accumulation costs: relatively low return on international reserves, quasi-fiscal costs from sterilization operations, and potential valuation losses from exchange rate movements.
- Accumulation should not compromise inflation-targeting credibility where applicable.
- Policy alternatives to reduce reserve needs:
  - Allow higher exchange rate flexibility.
  - Reduce short-term external debt.
  - Use multilateral liquidity sources: several Central American countries accessed IMF liquidity during the crisis; multilateral reserve pools remain a cheap and attractive means to obtain international reserves given increasing flexibility of IMF lending facilities.

### Methodological notes and quantitative anchors
- Panel regression framework: time period 1993-2008 and a panel of 52 emerging markets.
- Jeanne and Rancière (2006) approximation cited:
  - ρ = λ + γ (1 − q^(1/σ)) where ρ is optimal reserves-to-GDP, λ is short-term private external debt as share of output, γ is the output loss fraction in a sudden stop, q is the marginal rate of substitution term (denoted ϕ in the source derivation), and σ is CRRA risk aversion.
  - If term premium δ = 0 then ϕ = 1 and ρ = λ + γ.
- Numerical benchmarks cited:
  - Baseline π = 10.2 percent; alternative π = 5 percent scenario.
  - Mexico 1994 output loss: -6.5 percent.
  - Uruguay 2001 output loss: -14 percent.

### Concluding synthesis
- Multiple methods—benchmark ratios, panel regressions, and optimal-reserve models—converge to indicate that Central American economies generally hold reserves below levels implied by precautionary insurance against sudden stops.
- Policy choices should weigh insurance benefits of additional reserves against accumulation costs and consider alternatives (exchange-rate flexibility, reducing short-term external debt, multilateral liquidity facilities).

*Source: _wp11144 - References (excerpt) — IMF PDF chapter/section*

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

### _wp11144 - References

### Introduction and recent trends
- International reserve accumulation reached 13 percent of global GDP in 2009, which represents a three-fold increase over ten years.
- Emerging market holdings rose to 32 percent of GDP.
- Footnote clarification: 26 percent if China is excluded.

### Rationale for holding international reserves
- Long-standing view: reserves insure against balance of payments shocks.
- Common rules-of-thumb for reserve adequacy:
  - Cover some external commitments (e.g., three months of imports for countries with limited access to capital markets).
  - Measures of potential capital flight, e.g., short-term external debt at residual maturity and current account deficits.
- Alternative approach: buffer stock model treating reserves as a resource for smoothing consumption in the face of sudden stops of external credit and associated output falls (Frenkel and Jovanovic, 1981).
- Empirical determinants identified in the literature: reserve holdings are positively related to income volatility, openness and financial depth (Edwards, 1983; Flood and Marion, 2002; Obstfeld et al., 2008).
- Work on optimal reserves weighs benefits of mitigating falls in domestic consumption during balance of payments crises against costs of holding reserves, such as the interest rate differential between long-term debt issued to finance reserves and the return on reserves (Jeanne and Rancière, 2006; Gonçalves, 2007; Valencia, 2010).

### Existing findings and open questions
- Several authors conclude recent reserve holdings of most emerging markets are hard to justify on economic grounds; alternative explanations include export-oriented growth strategies (Dooley et al., 2003).
- Evidence from Asia and Latin America suggests over-accumulation in one country may prompt others to follow through an attempt to "keep up with the Joneses" (Cheung and Qian, 2009; Cheung and Sengupta, 2011).

### Study focus and motivation
- This paper focuses on small non-dollarized economies in Central America: Costa Rica, the Dominican Republic, Honduras, Guatemala and Nicaragua.
- Rationale for focus:
  - Lack of studies on reserve coverage for the region and smaller emerging markets.
  - Assess appropriateness of commonly used methods for reserve adequacy for smaller and less financially integrated emerging markets, including the possible presence of "keeping up with the Joneses" effects.
- Regional characteristics relevant for reserve adequacy:
  - Relatively little short-term external debt compared with larger emerging market peers, suggesting a smaller vulnerability along that dimension.
  - Share of dollar-denominated deposits to total deposits is about 45 percent on average in the region, well above levels of dollarization in the five largest economies in Latin America and other emerging markets—raising risks of large deposit withdrawals during crises that need to be taken into account when assessing reserve adequacy.
- Countries excluded from the sample and reasons:
  - El Salvador and Panama were fully dollarized for the period studied; the notion of international reserves is not readily applicable to them.
  - Belize was excluded due to its smaller size than the other economies in the sample.

### Paper organization (contents)
- Section II: investigates international reserve coverage using traditional rules-of-thumb.
- Section III: presents results from reserve demand regressions.
- Section IV: contains a model of optimal reserve coverage and its predictions.

*Source: _wp11144 - References (excerpt) — IMF PDF chapter/section*

### Section V concludes.

### Section V concludes.

### II. RULES-OF-THUMB FOR RESERVE COVER
- Central America increased gross international reserves by over 50 percent during 2000–2008, while the five largest Latin American economies experienced a three-fold increase and a larger global sample of emerging markets a six-fold increase during the same period.
- Traditional benchmarks and how Central America compares:
  - Three months of imports rule: Central American economies barely reach this level; larger Latin American economies and the full sample of emerging markets exceed it by a wide margin.
  - Greenspan-Guidotti rule (gross international reserves should cover short-term external debt on a residual maturity basis): Central America falls short of this benchmark despite relatively modest external debt levels; other emerging markets meet or exceed it.
  - Reserves-to-broad-money (M2) rule: Historically 5–20 percent of broad money considered adequate; more recently up to 50% suggested. Central America is below comparators on this dimension.
- Financial intermediation context:
  - Broad money to GDP in Central America ≈ 40 percent.
  - Broad money to GDP in the larger sample of emerging economies ≈ 70 percent.
- Overall assessment: Central America’s reserve cover is low both in absolute terms and relative to emerging market peers.

### III. RESERVE DEMAND REGRESSIONS
- Methodology:
  - Panel of N = 52 emerging economies, period 1993–2008 (T = 16).
  - Dependent variable: reserves-to-GDP ratio (GIR/GDP).
  - Model includes: vector of economic variables Xit, Central America dummy, crisis dummy (reversal of current account to GDP by more than five percentage points), and peer “Jones” variables (lagged).
  - “Jones” variables constructed by summing reserve-to-GDP ratios of either other small emerging markets in the region or the five largest economies (Brazil, Chile, Colombia, Mexico, Peru).
- Main regression findings (sign, magnitude, and significance reported):
  - Reserve demand is positively related to openness and GDP volatility, and negatively related to exchange rate flexibility.
  - Exchange rate flexibility has the largest effect: raising the period standard deviation of the nominal NCU/dollar exchange rate by one unit is associated with a decreased reserve/GDP ratio of about 25 percentage points.
  - A one percentage point increase in openness (defined as (exports + imports)/GDP) or in the FDI/GDP ratio has effects on the reserve-to-GDP ratio close to the size of the exchange rate flexibility effect (in direction consistent with positive relationship for openness and FDI).
  - No significant effect of financial depth (M2/GDP) on reserve demand was found, contrary to some prior studies.
  - Total external debt level and crisis dummies have expected positive signs but are not consistently significant.
  - Short-term debt/GDP increases reserve-to-GDP by around 3 percentage points.
  - Peer (Jones) effects: only reserve levels in the large emerging markets matter; an increase in the reserve-to-GDP ratio of the five largest economies in Latin America increases the reserve-to-GDP ratio of a Central American economy by 0.5 percentage points the following year. No significant effect from reserve accumulation in other Central American economies.
  - Central America dummy: statistically significant negative coefficient interpreted as Central America having on average about two percentage points lower reserve-to-GDP ratio after controlling for other factors.
- Table 1. Reserve Demand Regressions. Dependent Variable: GIR/GDP
  - Log (Population): -0.008*    T-stat. -1.97
  - Openness /1: 0.117***    T-stat. 9.82
  - Exchange rate flexibility/2: -0.241*    T-stat. -1.87
  - GDP volatility/3: 0.101*    T-stat. 1.37
  - Central America dummy: -0.024*    T-stat. -1.40
  - FDI/GDP: 0.191***    T-stat. 2.90
  - Short term debt/GDP: 0.031*    T-stat. 1.71
  - Jones2*CA dummy: 0.004***    T-stat. 3.55
  - Notes:
    - 1/ Defined as (exports + imports)/GDP.
    - 2/ Defined as the period standard deviation of the NCU/dollar exchange rate, over the period average.
    - 3/ Defined as the period standard deviation of real GDP, over the period average.
    - 4/ Defined as M2/GDP.
    - ***, **, * denote results significant on the 1, 5 and 10% significance levels (one-sided t-tests).
  - All estimated coefficients, different specifications and robustness checks are in the Appendix.

### IV. MODELS OF OPTIMAL RESERVES
- Purpose: Assess reserve optimality from a consumption insurance perspective, balancing benefits of holding reserves in sudden stops against quasi-fiscal costs.
- Frameworks used:
  - Jeanne and Rancière (2006) model (closed-form solution for optimal reserves as fraction of output ρ).
  - Gonçalves (2007) extension to include dollar deposit withdrawals (dollar deposit coverage by the government).
- Key model definitions and intuition:
  - Reserves smooth domestic absorption against sudden stops (declines in the capital account and potentially output).
  - Optimal reserves-to-output ratio ρ derived from first-order conditions; approximation (setting δ+π = r – g = 0) yields:
    - ρ ≈ (λ + γ) / (1 + δ/π) adjusted by risk aversion σ etc. — intuition: optimal reserves increase one-for-one with short-term external debt λ and output cost of a crisis γ; decline with opportunity cost (1+ δ/π).
  - If δ = 0, full insurance optimum ρ = λ + γ. If γ = 0, ρ = λ.
  - Gonçalves extension: include deposit withdrawal risk. Use α = 30% as a baseline government backing of dollar deposits (Gonçalves value from Uruguay 2001), and also present full (100%) backing as upper bound.
  - The model does not account for valuation effects from real exchange rate depreciation; Jeanne and Rancière show optimal reserves are not significantly affected by real exchange rate in aggregate.
- A. Parameters (Non-country specific model parameters)
  - Term premium, δ: 1.5%
  - Risk-free rate, r: 5%
  - Coefficient of relative risk aversion, σ: 2
  - Deposit coverage, α: 30-100%
  - Notes on parameters:
    - r = 5% corresponds roughly to average U.S. 3-month Treasury bill rate over the last decades.
    - δ = 1.5% approximates average difference between yield on 10-year U.S. treasury bills and the federal funds rate.
    - Jeanne and Rancière cross-country probit estimate for crisis probability π = 10.2% (average emerging market crisis every ten years); for countries with observed crises in the sample period, implied probabilities (e.g., one crisis in two decades → 5%) were used.
- B. Results (comparisons of actual reserves vs. model optimal reserves; time averages 2003–08)
  - Sample: Armenia, Bolivia, Bulgaria, Costa Rica, Dominican Republic, Guatemala, Honduras, Nicaragua, Paraguay, Romania, Uruguay.
  - General findings:
    - Optimal reserves per Jeanne and Rancière vary widely across time and countries.
    - Actual reserve levels are often close to, but typically exceeding, the Jeanne and Rancière optimal levels—especially for financially integrated emerging markets (Costa Rica, Dominican Republic, Bulgaria, Romania, Uruguay).
    - For countries with low short-term external debt (Guatemala, Honduras, Bolivia), the model prescribes low optimal reserves; Bolivia’s actual high reserves reflect terms-of-trade/commodity boom considerations not emphasized in the model.
    - When including dollar deposits (Gonçalves model) even with only 30% government backing, actual reserves are lower than optimal reserves for the majority of Central American countries. Under full deposit insurance, Honduras is the only Central American country with actual reserves higher than the model-prescribed optimal level.
  - Table 3. Actual vs. Optimal Reserves (% of GDP), Time Averages 2003–08 (selected entries shown exactly as reported)
    - Costa Rica: Actual 10.2; Jeanne and Rancière 8.2; Gonçalves (30% guarantee) 14.8; Gonçalves (Full guarantee) 21.6
    - Dom. Republic: Actual 3.4; Jeanne and Rancière 4.9; Gonçalves (30% guarantee) 5.8; Gonçalves (Full guarantee) 9.6
    - Guatemala: Actual 11.0; Jeanne and Rancière 6.4; Gonçalves (30% guarantee) 8.7; Gonçalves (Full guarantee) 15.8
    - Honduras: Actual 19.1; Jeanne and Rancière 7.3; Gonçalves (30% guarantee) 7.9; Gonçalves (Full guarantee) 12.7
    - Nicaragua: Actual 13.2; Jeanne and Rancière 17.7; Gonçalves (30% guarantee) 21.7; Gonçalves (Full guarantee) 44.4
    - Armenia: Actual 6.0; Jeanne and Rancière 13.8; Gonçalves (30% guarantee) 14.4; Gonçalves (Full guarantee) 15.6
    - Bolivia: Actual 16; Jeanne and Rancière 2.5; Gonçalves (30% guarantee) 2.7; Gonçalves (Full guarantee) 4.3
    - Bulgaria: Actual 28.9; Jeanne and Rancière 24.9
    - Paraguay: Actual 14.4; Jeanne and Rancière 11.3; Gonçalves (30% guarantee) 11.7; Gonçalves (Full guarantee) 12.3
    - Romania: Actual 15.2; Jeanne and Rancière 16.7; Gonçalves (30% guarantee) 19.0; Gonçalves (Full guarantee) 24.4
    - Uruguay: Actual 13.6; Jeanne and Rancière 11.3; Gonçalves (30% guarantee) 12.8; Gonçalves (Full guarantee) 17.7
  - Note: Results refer to time averages for 2003–2008; data limitations prevented covering 2009.
- C. Sensitivity analysis (illustrative results reported for Jeanne and Rancière model; example: Costa Rica, year 2008)
  - Risk aversion parameter σ:
    - Increasing σ from 2 to 5 increases average optimal reserve-to-GDP by about 4 percent for the full sample; for Central American economies the increase is about 5 percent on average.
  - Term premium δ and lower risk-free rate r:
    - A lower term premium decreases the opportunity cost of holding reserves and thus raises the optimal reserve-to-GDP ratio.
    - (Text truncated in the source at this point; sensitivity analysis continues in the Appendix.)

*Source: _wp11144 - Section V concludes.*

### 1.5 to 0, the average optimal reserve-to-GDP ratio would increase by almost 7 percent. Again,

### _wp11144 - 1.5 to 0, the average optimal reserve-to-GDP ratio would increase by almost 7 percent. Again,

### Main findings on reserve adequacy in Central America
- Central America scores relatively low on traditional reserve benchmarks such as reserves to months of imports or short-term external debt and current account positions.
- After controlling for determinants of reserve demand (openness, exchange rate flexibility, financial depth, external debt stocks) in a panel of emerging markets, Central America has significantly lower reserve-to-GDP ratios.
- Evidence of “keeping up with the Joneses” effects: Central American countries take into account reserve accumulation of large emerging markets in Latin America when setting policy.
- Using the Jeanne and Rancière (2006) and Gonçalves (2007) frameworks, actual reserve levels in Central America are often lower than model-prescribed optimal insurance against sudden stops and partial withdrawals of dollar deposits.

### Sensitivity analysis and scenario results
- Changing the coverage parameter from 1.5 to 0 raises the average optimal reserve-to-GDP ratio by almost 7 percent; for the Central American economies the increase is over 8 percent.
- Varying the risk-free rate produces similar effects on optimal reserves as adjusting coverage.
- Crisis probability:
  - Baseline crisis probability estimated by Jeanne and Rancière: 10.2 percent.
  - If the crisis probability falls by half to 5 percent, the average reserves-to-GDP ratio would decline by close to 11 percent.
  - For Central American economies (with relatively low benchmark crisis probabilities), lowering π to 5 percent would reduce optimal reserve-to-GDP by around 8 percent.
- Size of output loss (γ):
  - If countries experienced an output drop equal to Mexico 1994, -6.5 percent, the average optimal reserve-to-GDP ratio would increase by almost 70 percent for Central American economies versus 40 percent for the full sample.
  - If output loss equals Uruguay 2001, -14 percent, the optimal reserve-to-GDP ratio would increase by 104 percent (i.e., double) for Central American economies and by 68 percent for the full sample.
  - The large sensitivity for Central America reflects the absence of such dramatic output losses in the sample period.

### Policy implications and recommendations
- Further reserve accumulation is an important policy option for Central American economies, given model indications that many actual levels are below optimal insurance needs.
- Reserve accumulation comes with non-negligible costs: relatively low return on international reserves, quasi-fiscal costs from sterilization operations, and potential valuation losses from exchange rate movements.
- Accumulation should not compromise inflation-targeting credibility for countries committed to such regimes.
- Policy alternatives to reduce the need for reserves include:
  - Allowing higher exchange rate flexibility.
  - Reducing short-term external debt.
  - Using multilateral liquidity sources: several Central American countries accessed IMF liquidity during the crisis; with increasing flexibility of IMF lending facilities, multilateral reserve pools remain a cheap and attractive means to obtain international reserves.

### Key methodological notes and quantitative anchors
- Panel regression framework: time period 1993-2008 and a panel of 52 emerging markets.
- Jeanne and Rancière (2006) model summary and approximations:
  - Optimal-reserve-to-GDP ratio (exact closed-form given in the source) can be approximated, in the calibration range, by:
    - ρ = λ + γ (1 − q^(1/σ))
      - where ρ is optimal reserves-to-GDP, λ is short-term private external debt as share of output, γ is the output loss fraction in a sudden stop, q is the marginal rate of substitution term (denoted ϕ in the source derivation), and σ is the CRRA risk aversion parameter (see the source for full notation and derivation).
  - If the term premium δ = 0 then ϕ = 1 and the perfect-smoothing level is ρ = λ + γ.
  - The Greenspan-Guidotti rule is a special case with no output cost: ρ = λ.
- Model drivers emphasized:
  - Probability of a sudden stop, π.
  - Term premium, δ.
  - Output loss in a sudden stop, γ.
  - Risk aversion parameter, σ.
  - Short-term debt share, λ.
- Numerical benchmarks cited:
  - Baseline π = 10.2 percent (Jeanne and Rancière estimate).
  - Alternative π = 5 percent scenario.
  - Mexico 1994 output loss: -6.5 percent.
  - Uruguay 2001 output loss: -14 percent.

### Empirical regression highlights (select qualitative results)
- Openness is positively and significantly associated with higher reserves (coefficients significant at conventional levels in the regressions).
- Exchange rate volatility and financial depth enter with the expected signs in various specifications.
- Central America dummy is negative and in some specifications statistically significant, indicating lower reserves-to-GDP relative to the broad emerging-market sample.
- “Keeping up with the Joneses” (average of LA-5 economies) interactions are positive and significant in several specifications, consistent with peer effects in reserve accumulation.

### Concluding synthesis
- Multiple methods—benchmark ratios, panel regressions controlling for standard determinants, and optimal-reserve models—converge to indicate that Central American economies generally hold reserves below levels implied by precautionary insurance against sudden stops.
- Policy response should weigh the insurance benefits of additional reserves against accumulation costs and alternative instruments (exchange-rate flexibility, reducing short-term external debt, and multilateral liquidity facilities).

*Source: IMF working paper content (selected sections, figures, and appendices as provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11144.pdf_