## _wp1142

## Source details

**Canonical URL:** [_wp1142](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1142.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1142.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1142.pdf.json)

---

### Background and research question
- Traditional arguments emphasize expenditure switching under flexible exchange rates (Friedman(1953), Mundell(1961), Fleming(1962)).
- Theoretical literature is inconclusive when foreign currency debt and limited exchange rate pass-through are present.
- Central empirical question: whether a country with high foreign currency debt can better insulate itself from a negative external shock under a floating or a fixed exchange rate, and how import structure affects this insulation.

### Empirical approach and data
- Sample size: 101 countries (yearly data 1974–2007).
- Method: Interacted Panel Vector Autoregression (IPVAR) — Panel VAR augmented with interaction terms allowing VAR coefficients to vary with foreign currency debt (ξ) and import structure (ω).
- Interaction specification for coefficients:
  - αjkl,it = βjkl,1 + βjkl,2·PEGit + βjkl,3·FCDit + βjkl,4·RAWit + βjkl,5·FCDit·PEGit + βjkl,6·RAWit·PEGit
  - PEGit: exchange rate regime dummy; FCDit: foreign currency debt measure; RAWit: share of raw materials in imports.
- Key empirical measures:
  - Foreign Debt: primary measure short term external debt over GNI (WDI); transformations: log(1+debt).
  - Import Structure: share of primary commodities in total imports (agricultural goods, fuels, ores and metals); transformation: log(1+imp).
  - Exchange Rate Regime: de facto Levy-Yeyati and Sturzenegger (LYS) classification (1974 onward); also compared to IMF de jure classification.

### Main empirical findings (overview)
- The insulating properties of flexible exchange rate regimes are strong when:
  - the import share of high pass-through (homogeneous) goods is large (high RAW / high ω), and
  - foreign currency debt is low (low FCD / low ξ).
- When the share of homogeneous imports is small and foreign currency debt is high:
  - fixed exchange rates display better stabilization properties.
  - limited pass-through hinders relative price adjustment under a float and contractionary balance sheet effects dominate.
- Results contrast with earlier unconditional-volatility studies that found little difference between fixed and flexible regimes; accounting for real vs nominal shocks and country characteristics matters.
- No prior empirical study jointly analyzes buffer properties across varying ξ and ω; this study fills that gap.

### Theoretical framework (model summary)
- Model: Microfounded two-period (1 and 2) small open economy with sticky prices and wages, and a Bernanke–Gertler–Gilchrist (1998) type financial accelerator; extension of Céspedes, Chang, and Velasco (2003).
- Two key country-specific parameters:
  - ω: share of homogeneous goods in total imports (affects exchange rate pass-through).
  - ξ: share of foreign currency debt.
- Agents: workers and entrepreneurs. Entrepreneurs finance capital Q1K1 with net worth P1N1 and debt; external financing premium 1+η1 = (Q1K1 / P1N1)^μ, with μ capturing financial imperfections.
- Consumption aggregator: elasticity of substitution φ and weights γ and 1−γ; foreign composite: ω homogeneous and 1−ω heterogeneous.
- Pricing:
  - Homogeneous import goods priced on world market in foreign currency.
  - Heterogeneous import goods priced to domestic market (pricing-to-market).

### Reduced-form IS-LM-BP representation (log-linear)
- IS equation:
  - A y1 = λ i1 + (1−λ) x1 + [ (1−λ) + ω·λ g(φ) ] s1
  - Definitions:
    - A: positive constant.
    - λ < 1: steady state share of investment demand in domestic output net of consumption.
    - (1−λ): share of exports.
    - g(φ): increasing in substitution elasticity φ between domestic and foreign goods.
    - If ω = 0, expenditure switching channel is lost.
- BP (investment demand) equation:
  - Γ i1 = −ρ + μ(1+ψ) δy y1 − μ [ (1−γ′1) ω + ψ ξ ] s1 + [ 1 − (1−γ′1) ω ] s1
  - Definitions and interpretation:
    - Γ, δy, γ′1: positive coefficients.
    - ρ: world interest rate.
    - μ: degree of financial imperfections.
    - ψ: ratio of total debt over net worth (leverage).
    - Term [ 1 − (1−γ′1) ω ] s1: expansionary real risk-free rate effect of depreciation when μ = 0.
    - Term −μ [ (1−γ′1) ω + ψ ξ ] s1: contractionary balance sheet effects (investment cost and foreign debt valuation).
    - A depreciation can be contractionary when μ, ψ, and ξ are large.
- Monetary policy regimes:
  - Peg: s1 = 0.
  - Float (reduced form): s1 = −κ y1, where κ captures exchange rate response to output contractions.

### The Peg — mechanism and theoretical results
- Under a peg the exchange rate does not move; output and investment responses to external demand shocks are independent of ω and ξ.
- Result 1: ∂y/∂x|_{PEG} and ∂i/∂x|_{PEG} independent of ω and ξ because s = 0.
- Result 2: Leverage ψ amplifies balance-sheet effects; investment response increases with ψ and μ. If μ = 0, investment does not respond.
- Key statistics referenced for peg analysis:
  - Leverage parameter example calibration: ψ = 10.6.
  - Under peg column in Table 1: ∂y/∂x∂ξ = 0; ∂i/∂x∂ξ = 0; ∂y/∂x∂ω = 0; ∂i/∂x∂ω = 0.

### The Float — decomposition and theoretical implications
- Output response to an external shock decomposes into multiple effects:
  - First: identical to peg.
  - Second: expenditure switching effect — expansionary and increasing in φ and ω.
  - Third: export value effect — increasing in export share (1−λ).
  - Fourth and fifth: opposing effects on financing costs via exchange rate changes:
    - Depreciation lowers domestic real risk-free rate (reducing investment cost).
    - Depreciation can amplify balance-sheet effects by increasing investment price and the domestic-currency value of foreign debt.
- Result 3: Float's ability to stabilize output diminishes as ω declines and ξ increases.
- Result 4: Threshold foreign-currency debt share ξ_0 at which float becomes more destabilizing than peg falls when:
  - exchange rate pass-through is low (lower ω),
  - leverage ψ is strong,
  - financial imperfections μ are strong.
- Investment under float:
  - Effect 1: net worth and external finance premium channel via drop in output.
  - Effect 2: exchange rate increases risk premium via higher investment price and increased domestic value of foreign debt.
  - Effect 3: lowered domestic real risk-free rate due to depreciation reduces investment cost.
- Result 5: Higher ξ increases investment response.
- Result 6: Effect of ω on investment is ambiguous; limited pass-through moderates price effects but can amplify output effects; with high leverage output effect may dominate.

### Simulation of responses under Peg and Float (calibration & comparative statics)
- Calibration and parameter choices:
  - γ = 0.6
  - θ = 11 (mark-up for differentiated products → 10%)
  - φ = 2
  - μ = 0.2
  - κ = 2
  - β = 0.96
  - ψ = 10.6 chosen so that external debt to GDP = 36 %
  - Baseline normalization: ̄K1, ̄X1, ̄X2 chosen so output growth ̄Y2/̄Y1 and real exchange rates ̄S1/̄P1 and ̄S2/̄P2 are one.
- Comparative statics performed:
  - ω varied from close to zero up to 50%.
  - ξ varied from zero up to 36% (holding overall debt constant).
- Simulation finding: peg performs better (smaller fall in output) when foreign currency debt is high and the share of homogeneous goods is low; Figure 1 illustrates (∂y/∂x|FL − ∂y/∂x|PEG) across grid.

### Data, empirical design, estimation and inference
- Sample restrictions:
  - Exclude G7 countries.
  - Require > 5 data points per country.
  - Drop countries with PPP adjusted GDP per capita < 1000 dollars in 2007.
  - Drop small countries with population < 1 million.
  - Drop observations where annual change in real GDP > twenty percent.
  - Consider observations where inflation < fifty percent.
  - Restrict regime to be in place at least one period to avoid transition contamination.
- Data sources: IMF IFS, World Bank WDI, BIS.
- Empirical model:
  - Recursive Interacted Panel VAR with variables EXVit (log ToT or foreign real interest rate), GDPit (log real GDP), INVit (log real investment).
  - External shocks identified by small open economy assumption (strict exogeneity of external variables).
- Estimation:
  - Panel VAR estimated by OLS with country fixed effects; two lags by Schwartz Criterion.
  - Interaction terms allow slope heterogeneity.
- Inference:
  - Bootstrapped standard errors and confidence intervals adapted from Runkle (1987) for panel interaction context.
  - Simulation/bootstrap steps repeated 500 times to form 90 % confidence intervals.
  - Wald tests for joint significance of interaction terms and empirical distribution checks for IRF differences.

### Empirical results — Floats versus Pegs (selected numeric findings)
- Terms of trade shock: initial negative 10% ToT shock, LYS classification — cumulative responses (selected horizons):
  - Output (cumulative)
    - 1st year: PEG −0.39 ∗∗∗ ; FLOAT −0.25 ∗ ; Difference −0.14
    - 2nd year: PEG −0.93 ∗∗∗ ; FLOAT −0.49 ∗∗ ; Difference −0.44 ∗
    - 5th year: PEG −1.10 ∗∗∗ ; FLOAT −0.60 ∗∗ ; Difference −0.50 ∗
  - Investment (cumulative)
    - 1st year: PEG −1.25 ∗∗∗ ; FLOAT −1.24 ∗∗∗ ; Difference −0.01
    - 2nd year: PEG −2.63 ∗∗∗ ; FLOAT −2.50 ∗∗∗ ; Difference −0.14
    - 5th year: PEG −3.10 ∗∗∗ ; FLOAT −2.92 ∗∗∗ ; Difference −0.18
  - Wald Test p-value: 0.03
- Results are qualitatively similar using IMF de jure classification and for foreign interest rate shocks (100 bps) with analogous relative patterns.

### Empirical results — Role of foreign currency debt (FCD)
- Key empirical patterns (LYS, negative 10% ToT shock):
  - Low FCD:
    - 2nd year output: FLOAT response insignificant (near zero); PEG output declined > 1 percent.
    - Investment: declines ~1.7% under peg and ~1.0% under float within two years.
  - High FCD:
    - 2nd year output: PEG response similar to low-debt peg; FLOAT response substantially stronger — output declines > 1 percent.
    - Investment: declines ~2.8% under peg and ~4.5% under float within two years.
- Simulation threshold result:
  - Output responds less under a float up to a short term external debt to GNI ratio of 10 percent; roughly 25 percent of observations lie above this threshold.
- Robustness:
  - Similar results using total external debt, BIS claims, or IMF de jure classification (with higher parameter uncertainty).
  - Wald tests find interaction terms involving FCD jointly significant; double interaction FCD*PEG less clear in some specifications.
- Selected numeric examples (Table 3, LYS, short term external debt):
  - High FCD, 2nd year output: PEG −0.78 ∗∗∗ ; FLOAT −1.12 ∗∗∗ ; Difference 0.34
  - Low FCD, 2nd year output: PEG −1.23 ∗∗∗ ; FLOAT 0.14 ; Difference −1.37 ∗∗∗
  - High minus Low diff (2nd year): 1.71 ∗∗

### Empirical results — Role of import structure (RAW)
- Key empirical patterns (negative 10% ToT shock, LYS):
  - High RAW (high pass-through):
    - 2nd year output: PEG −1.36 ∗∗∗ ; FLOAT −0.56 ∗ ; Difference −0.80 ∗∗
    - Investment tends to fall more under peg than float in some horizons.
  - Low RAW (low pass-through):
    - 2nd year output: PEG −0.42 ∗ ; FLOAT −1.21 ∗∗∗ ; Difference 0.79 ∗∗
    - Investment falls more under float: 2nd year example FLOAT −4.59 ∗∗∗ ; PEG −1.77 ∗
  - Threshold:
    - Peg and float responses intersect at a raw material share of 14 %; roughly 55 % of observations lie below the threshold.
- Robustness:
  - De jure classification yields similar patterns but smaller and often not statistically significant pairwise differences.
  - Wald tests find joint significance of interaction terms involving RAW.

### Empirical results — Joint role of FCD and RAW
- Full specification with PEG, FCD, and RAW:
  - Simulated cumulated output response within two years across grid of FCD and RAW.
  - Difference = Float response minus Peg response (negative implies float performs better).
  - Main empirical pattern:
    - Float insulates output better when RAW is high and FCD is low.
    - Float is relatively worse when RAW is low or FCD is high.
  - Empirical difference surface resembles theoretical model predictions: expenditure switching (via ω) versus contractionary balance sheet effects (via ξ) interact to determine regime preference for output stability.

### Policy-relevant conclusions and implications
- Main conclusions:
  - Flexible exchange rates do not universally shield output better from real external shocks.
  - The insulating advantage of floats depends critically on:
    - Foreign currency debt (higher FCD weakens float insulation; can make floats more destabilizing).
    - Import structure/pass-through (higher RAW increases buffering role of floats).
  - Balance sheet effects and limited pass-through can outweigh expenditure switching benefits of depreciation under a float.
- Policy implication (ceteris paribus):
  - The case for a floating exchange rate weakens if a country has high foreign currency debt and imports mainly low pass-through goods because of increased output volatility.
- Caveats:
  - Policy makers should consider additional dimensions (inflation, trade volumes, long term growth, likelihood of financial crises); the framework is silent on these dimensions.

### Appendices and supporting materials (contents summarized)
- Appendix A: Model — two-period small open economy with financial accelerator, pricing assumptions, and monetary regimes.
- Appendix B: Derivation — linearization and IS, LM, BP relations; definitions of parameters and intermediate expressions.
- Appendix C: Solving the system — comparative analysis under peg, float, pure float; derivation of threshold ξ_0 and signs of comparative statics.
  - Example closed-form peg responses:
    - ∂y/∂x |_{PEG} = (1−λ) Γ / [ A Γ − λ μ (1+ψ) δ_y ].
    - ∂i/∂x |_{PEG} = (1−λ) μ (1+ψ) δ_y / [ A Γ − λ μ (1+ψ) δ_y ].
  - Threshold foreign-currency debt share ξ_0 expression (equation (36)) and interpretation: lower ω_0, higher ψ, higher μ reduce ξ_0.
- Appendix D: Data sources and variable construction (detailed definitions, harmonization of AREAER taxonomies, ToT construction, PASS measure, country sample list).

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1142.pdf*

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

### _wp1142 - References .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .

### Background and research question
- Traditional arguments (Friedman(1953), Mundell(1961), Fleming(1962)) emphasize expenditure switching under flexible exchange rates.
- Theoretical literature is inconclusive when foreign currency debt and limited exchange rate pass-through are present.
- Key empirical question addressed: whether a country with high foreign currency debt can better insulate itself from a negative external shock under a floating or a fixed exchange rate, and how import structure affects this insulation.

### Empirical approach and data
- Sample size: 101 countries.
- Method: Interacted Panel Vector Autoregression (IPVAR) — a Panel VAR augmented with interaction terms allowing VAR coefficients to vary with foreign currency debt (ξ) and import structure (ω).
- Motivations for interaction terms: allow coefficients to vary deterministically across time and countries; alternative to stochastically time-varying coefficient frameworks.

### Main empirical findings
- The insulating properties of flexible exchange rate regimes are strong when:
  - the import share of high pass-through (homogeneous) goods is large, and
  - foreign currency debt is low.
- When the share of homogeneous imports is small and foreign currency debt is high:
  - fixed exchange rates display better stabilization properties.
  - limited pass-through hinders relative price adjustment under a float and contractionary balance sheet effects dominate.
- Results contrast with earlier unconditional-volatility studies that found little difference between fixed and flexible regimes; accounting for real vs nominal shocks and country characteristics matters.
- No prior empirical study was identified that jointly analyzes buffer properties of exchange rate regimes across varying levels of foreign currency debt and import structure; this study fills that gap.

### Theoretical framework overview
- Model type: Microfounded two-period (1 and 2) small open economy with sticky prices and wages, and a financial accelerator mechanism as in Bernanke, Gertler, and Gilchrist (1998). Extension of Céspedes, Chang, and Velasco (2003).
- Two key country-specific parameters:
  - ω: share of homogeneous goods in total imports (affects exchange rate pass-through),
  - ξ: share of foreign currency debt.
- Agents: workers and entrepreneurs. Entrepreneurs finance capital Q1K1 with net worth P1N1 and debt; risk premium η1 increases with the share of debt via 1+η1 = (Q1K1 / P1N1)^μ, where μ captures financial imperfections.
- Consumption bundle: consumers choose between home and foreign goods with elasticity φ and weights γ and 1−γ. Foreign goods are a Cobb-Douglas bundle with share ω homogeneous, 1−ω heterogeneous.
- Pricing assumptions:
  - Homogeneous import goods priced on the world market in foreign currency.
  - Heterogeneous import goods priced to the domestic market (pricing-to-market).

### Reduced-form IS-LM-BP representation (log-linear)
- IS equation (output y1 depends on investment i1, foreign demand x1, exchange rate s1):
  - A y1 = λ i1 + (1−λ) x1 + [ (1−λ) + ω·λ g(φ) ] s1
  - Parameters and interpretations:
    - A: positive constant.
    - λ < 1: steady state share of investment demand in domestic output net of consumption.
    - (1−λ): share of exports.
    - g(φ): increasing in substitution elasticity φ between domestic and foreign goods.
    - If ω = 0, the expenditure switching channel is lost (no effect of s on relative prices).
- BP (investment demand) equation (Γ i1 as function of foreign interest rate ρ, output y1, exchange rate s1):
  - Γ i1 = −ρ + μ(1+ψ) δy y1 − μ [ (1−γ′1) ω + ψ ξ ] s1 + [ 1 − (1−γ′1) ω ] s1
  - Parameters and interpretations:
    - Γ, δy, γ′1: positive coefficients.
    - ρ: world interest rate; higher ρ depresses investment.
    - μ: degree of financial imperfections.
    - ψ: ratio of total debt over net worth (leverage); higher ψ amplifies output effects on investment.
    - The term [ 1 − (1−γ′1) ω ] s1 captures the expansionary real risk-free rate effect of a depreciation when μ = 0.
    - The term −μ [ (1−γ′1) ω + ψ ξ ] s1 captures contractionary balance sheet effects (investment cost effect and foreign debt valuation effect).
    - A depreciation can be contractionary when μ, ψ, and ξ are large.
- Monetary policy regimes:
  - Peg: s1 = 0.
  - Float (reduced-form): s1 = −κ y1, where κ captures how the exchange rate responds to output contractions (this can represent a broad class of policies that let the exchange rate depreciate in response to adverse shocks).

### Theoretical implications and comparative statics
- Expenditure switching increases with ω and g(φ); higher ω (more homogeneous imports) strengthens the beneficial effect of depreciation on net exports.
- Limited exchange rate pass-through (lower ω) reduces IS expenditure switching but can partially offset contractionary BP effects by limiting the increase in domestic prices (reducing the investment cost effect).
- When foreign currency debt ξ is large, balance sheet effects dominate and depreciations are more likely to be contractionary.
- Theoretical exercises illustrate adjustment to an external demand shock x1; analogous results hold for world interest rate shocks ρ.
- Appendices provide full derivations, solutions under the Peg, the Float, and Pure Float, and evaluations of cross-partial effects (e.g., ∂^2 y / ∂ x ∂ ω).

### Policy-relevant conclusions
- Exchange rate regime choice interacts with country characteristics:
  - Countries with low ξ and high ω are better insulated by floating exchange rates.
  - Countries with high ξ and low ω may prefer fixed exchange rates to avoid contractionary balance sheet effects and limited expenditure switching.
- Monetary policy design and exchange rate flexibility should consider import composition (ω) and foreign currency debt exposure (ξ) jointly.
- The IPVAR empirical strategy allows direct assessment of how the responses of output and investment to external shocks vary with ξ, ω, and the exchange rate regime.

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1142.pdf*

### 1.   The Peg

### 1.   The Peg

### The Peg: mechanism and theoretical results
- Under a peg the exchange rate does not move; combine equations (1), (2) and (3) and take the derivative of output with respect to external demand to obtain the effects.
- Result 1: The response of output and investment to an external demand shock is independent of the import share of differentiated goods and the level of foreign currency debt, since the exchange rate does not move under a peg.
- Leverage ψ amplifies the balance sheet effect related to a fall in output and thus increases the strength of the output response.
- Result 2: The investment response is a multiple of the output response and increases with leverage and the level of financial imperfection. With no imperfection investment does not respond, since financing costs do not move.

### Key statistics and parameters referenced for the peg analysis
- Leverage parameter: ψ (example calibration used below: ψ=10.6).
- Financial imperfection parameter: μ.
- Under the peg column in Table 1: ∂y/∂x∂ξ = 0; ∂i/∂x∂ξ = 0; ∂y/∂x∂ω = 0; ∂i/∂x∂ω = 0.

---

### 2.   The Float

### The Float: decomposition of effects on output and investment
- Combine IS and BP equations (1),(2) with the float rule (4). The effect of the external shock on output decomposes into five components:
  - First effect: identical to the peg.
  - Second effect: expenditure switching effect — expansionary and increasing in the elasticity of substitution.
  - Third effect: export value effect — increasing in the export share.
  - Fourth and fifth effects: opposing effects on financing costs via exchange rate changes:
    - Depreciation lowers the domestic real risk free rate (reducing investment cost).
    - Depreciation can amplify balance sheet effects in the presence of financial imperfections by increasing the ratio of investment over net worth (raising investment cost) through higher investment price and reduced net worth due to increased debt burden; this effect is related to the share of foreign debt.
- Result 3: The ability of a float to stabilize output diminishes as the share of homogeneous import goods declines and the level of foreign currency debt increases.
- Result 4: An economy reaches the threshold foreign currency debt share at which a float becomes more destabilizing than a peg more quickly if:
  - exchange rate pass-through is low (expenditure switching effects are small),
  - leverage is strong,
  - financial imperfections are strong.
  - Thus interactions between pass-through, leverage, and financial frictions affect the buffer properties of a float.

### Investment under the float
- The investment response decomposes into three main effects:
  - Effect 1: impact on net worth and the external finance premium caused by a drop in output — lower output reduces net worth and raises risk premium, reducing investment demand.
  - Effect 2: additional effects from exchange rate changes — depreciation increases the risk premium by increasing the price of investment and the domestic currency value of foreign debt.
  - Effect 3: positive effect from a lowered domestic real risk free rate due to depreciation — reduces investment cost independent of financial frictions.
- Result 5: A higher share of foreign currency debt increases the response of investment.
- Result 6: The effect of the import structure is ambiguous: limited pass-through moderates the effect on prices but increases the effect on output; with high leverage the effect on output dominates.

### Float versus Peg: sign summary (Table 1)
- Peg: ∂y/∂x∂ξ = 0; ∂i/∂x∂ξ = 0; ∂y/∂x∂ω = 0; ∂i/∂x∂ω = 0.
- Float: ∂y/∂x∂ξ > 0; ∂i/∂x∂ξ > 0; ∂y/∂x∂ω < 0; ∂i/∂x∂ω =<?>0 (ambiguous in sign depending on parameters).

---

### 3.   Simulation of Responses under Peg and Float

### Calibration and parameter choices
- Consumption home-goods share: γ = 0.6
- Mark-up for differentiated products: 10% (θ = 11)
- Elasticity of substitution between foreign and home consumption goods: φ = 2
- Capital market imperfection: μ = 0.2
- κ = 2
- β = 0.96
- ψ = 10.6 chosen so that the ratio of external debt to GDP equals 36 % (midpoint between average short term external debt and total external debt in the sample).
- Baseline normalization: ̄K1, ̄X1, ̄X2 chosen such that output growth ̄Y2/̄Y1 and the real exchange rate in both periods ̄S1/̄P1 and ̄S2/̄P2 are one.

### Comparative statics performed
- Vary the share of homogeneous good imports (parameter ω) from close to zero up to 50%.
- Vary the foreign currency debt to GDP ratio (parameter ξ) from zero up to 36%, holding overall debt constant.
- Figure 1 displays the joint role of foreign currency debt and import structure as the difference in the fall of output to a negative external demand shock between a float and a peg: (∂y/∂x|FL − ∂y/∂x|PEG).
- Finding illustrated: the response under a peg is relatively smaller (peg performs better) if foreign currency debt is high and the share of homogeneous goods is low.

---

### 4.   Data and empirical design

### Sample and sample restrictions
- Sample coverage: yearly data for 101 countries from 1974-2007.
- Exclusions and restrictions:
  - Exclude G7 countries.
  - Require more than five data points per country.
  - Drop countries with PPP adjusted GDP per capita less than 1000 dollars in 2007.
  - Drop small countries with population less than 1 million.
  - Drop observations where annual change in real GDP exceeds twenty percent.
  - Consider observations where inflation was below fifty percent.
  - Restrict exchange rate regime to be in place at least one period to avoid contamination from regime transitions.

### Data sources
- IMF’s International Financial Statistics (IFS)
- World Bank’s World Development Indicators (WDI)
- Bank of International Settlements (BIS)

### Key empirical measures
- (a) Foreign Debt
  - Primary measure: short term external debt over GNI (WDI).
  - Robustness: total external debt.
  - Alternative: claims on the domestic economy by foreign banks scaled by GDP (BIS) — starts in 1983 and covers reporting-country banks only.
  - Transformation used to reduce sensitivity to outliers: log(1+debt), where debt is the corresponding debt measure expressed in percentage points.
  - Note: Lane and Shambaugh (2010) dataset indicates almost hundred percent of external debt is in foreign currency.
- (b) Import Structure
  - Proxy for high pass-through goods: share of primary commodities in total imports (sum of agricultural goods, fuels, ores and metals over total imports) from WDI.
  - Transformation used: log(1+imp), where imp is the import share of raw materials in percentage points.
  - Cross-sectional mean of this measure varies between 23% in 1981 and 11% in 1999.
- (c) Exchange Rate Regime
  - Preferred classification: de facto Levy-Yeyati and Sturzenegger (LYS) classification (covers 1974 onward), to capture actual conduct of exchange rate policy.

*Source: _wp1142 - 1.   The Peg (PDF chapter/section).*

### 2004. The authors use cluster analysis on exchange rate data and official reserves to infer the

### _wp1142 - 2004. The authors use cluster analysis on exchange rate data and official reserves to infer the

### Data and key measures
- Exchange rate regime
  - Exchange rate dummy: one for a peg and zero for non peg.
  - Classification based on cluster analysis on exchange rate data and official reserves to infer actual exchange rate policy; comparison also made with the IMF’s de jure classification (1974-2007).
- Terms of Trade
  - Terms of trade measure derived from various sources: IFS for most developed countries; UNCTAD for others; if unavailable, constructed from constant and current export and import values from the WDI.
- Foreign interest rate
  - Real foreign interest rate measured as the short term real interest rate of the reference country; nominal short term rate from money market or treasury bill rate minus CPI inflation in the reference country.
- National accounts
  - Real GDP and investment in local currency taken from the WDI.

*For a detailed description and the respective measures employed see the appendix (as indicated in source).*

### Empirical model and identification
- Model specification
  - A recursive Interacted Panel VAR is estimated with variables:
    - EXVit: external variable (log terms of trade or foreign real interest rate)
    - GDPit: log real GDP
    - INVit: log real investment
  - Equation (5) structure: deterministic country fixed effects μi, L lags, and deterministically varying coefficients αjkl,it.
- Identification
  - External shocks identified under the small open economy assumption implying strict exogeneity of external variables (α12l,it = α13l,it = 0 for all l).
  - Focus on the response of output and investment to real external shocks; domestic real shocks not identified.

### Interaction terms and heterogeneity
- Coefficients allowed to vary with country characteristics:
  - αjkl,it = βjkl,1 + βjkl,2·PEGit + βjkl,3·FCDit + βjkl,4·RAWit + βjkl,5·FCDit·PEGit + βjkl,6·RAWit·PEGit
  - PEGit: exchange rate regime dummy
  - FCDit: foreign currency debt measure (continuous)
  - RAWit: share of raw materials (continuous)
- Empirical approach
  - Benchmark specifications set some β coefficients to zero to examine separate effects.
  - External dynamics restricted to be independent of country characteristics: α11l = β11l,1 for all l (Wald test does not reject this).
  - Continuous variables evaluated at 20th and 80th percentiles for comparisons; PEG evaluated at 1 (peg) and 0 (float).

### Estimation and inference
- Estimation
  - Panel VAR estimated by OLS with country fixed effects; two lags chosen by the Schwartz Criterion.
  - Interaction terms used to allow slope heterogeneity and mitigate fixed-slope bias.
- Inference
  - Bootstrapped standard errors and confidence intervals used (procedure adapted from Runkle (1987) for panel and interaction context).
  - Simulation/bootstrap steps summarized:
    1. Estimate (5) by OLS.
    2. Draw errors ε̂i,~t from N(0, Σ̂).
    3. Simulate recursively using ε̂i,~t and initial observations with α̂jkl,it to generate artificial Ŷi,1… and iterate.
    4. Repeat draws and simulation across periods, accounting for interaction terms.
    5. Re-estimate coefficients and IRFs on artificial sample.
    6. Repeat steps 2–5 500 times; 90 % confidence interval from simulated estimates.
  - Tests employed:
    - Wald tests for joint significance of interaction terms (all, those involving FCD or RAW, and double interactions with PEG).
    - Direct assessment of empirical distribution of impulse response differences (fraction above zero) using bootstrap.

### Results — Floats versus Pegs
- Terms of trade shock (initial negative 10% ToT shock), LYS classification — cumulative response:
  - Output
    - 1st year: PEG −0.39 ∗∗∗ ; FLOAT −0.25 ∗ ; Difference −0.14
    - 2nd year: PEG −0.93 ∗∗∗ ; FLOAT −0.49 ∗∗ ; Difference −0.44 ∗
    - 5th year: PEG −1.10 ∗∗∗ ; FLOAT −0.60 ∗∗ ; Difference −0.50 ∗
  - Investment
    - 1st year: PEG −1.25 ∗∗∗ ; FLOAT −1.24 ∗∗∗ ; Difference −0.01
    - 2nd year: PEG −2.63 ∗∗∗ ; FLOAT −2.50 ∗∗∗ ; Difference −0.14
    - 5th year: PEG −3.10 ∗∗∗ ; FLOAT −2.92 ∗∗∗ ; Difference −0.18
  - Wald Test p-value: 0.03
- Using IMF de jure classification yields qualitatively similar results (float exhibits smaller output fall); results for foreign interest rate shock (100 bps) similar:
  - 2nd year: PEG output falls by about −0.22 ∗ ; FLOAT remains virtually unchanged (example) as reported in text.

### Results — Role of foreign currency debt (FCD)
- Key patterns (LYS classification, negative 10% ToT shock):
  - With low FCD:
    - 2nd year output: FLOAT response insignificant (near zero), PEG output declined by more than one percent.
  - With high FCD:
    - PEG output response similar to low-debt peg; FLOAT response substantially stronger — output declines by more than one percent.
  - Investment differences drive much of the variation:
    - Low FCD: investment declines ~1.7% under peg and ~1.0% under float within two years.
    - High FCD: investment declines ~2.8% under peg and ~4.5% under float within two years.
- Simulation threshold result
  - Output responds less under a float up to a short term external debt to GNI ratio of 10 percent; roughly 25 percent of observations lie above this threshold.
- Robustness and alternative measures
  - Results similar using total external debt (higher parameter uncertainty), BIS claims, or IMF de jure classification.
  - Wald tests find interaction terms involving FCD jointly significant; double interaction FCD*PEG less clearly significant in some specifications.
- Selected numerical examples (Table 3, LYS, short term external debt)
  - High FCD, 2nd year output: PEG −0.78 ∗∗∗ ; FLOAT −1.12 ∗∗∗ ; Difference 0.34
  - Low FCD, 2nd year output: PEG −1.23 ∗∗∗ ; FLOAT 0.14 ; Difference −1.37 ∗∗∗
  - High minus Low diff (2nd year): 1.71 ∗∗

### Results — Role of import structure (RAW)
- Key patterns (negative 10% ToT shock, LYS classification):
  - High RAW (high share of raw materials in imports; high pass-through)
    - 2nd year output: PEG −1.36 ∗∗∗ ; FLOAT −0.56 ∗ ; Difference −0.80 ∗∗
    - Investment tends to fall more under peg than float in some horizons (see table).
  - Low RAW (low pass-through)
    - 2nd year output: PEG −0.42 ∗ ; FLOAT −1.21 ∗∗∗ ; Difference 0.79 ∗∗
    - Investment falls more under float: example 2nd year: FLOAT −4.59 ∗∗∗ ; PEG −1.77 ∗
  - Intersection/threshold
    - Peg and float responses intersect at a raw material share of 14 %; roughly 55 % of observations lie below the threshold.
- Robustness
  - Using de jure classification yields similar picture but smaller and often not statistically significant pairwise differences.
  - Wald tests find joint significance of interaction terms involving RAW.

### Results — Joint role of FCD and RAW
- Full specification (exchange rate regime, foreign currency debt, and import structure)
  - Simulated cumulated output response within two years across grid of FCD and RAW.
  - Difference = Float response minus Peg response (negative implies float performs better/insulates output more).
  - Main pattern:
    - Float insulates output better when RAW is high and FCD is low.
    - Float becomes relatively worse (less insulating / more destabilizing) when either RAW is low or FCD is high.
  - The shape of the empirical difference surface resembles theoretical model predictions: expenditure switching (via pass-through) vs. contractionary balance sheet effects (via foreign currency debt) interact to determine which regime is preferable for output stability.

### Conclusion and policy-relevant findings
- Main empirical conclusions
  - Flexible exchange rates do not universally shield output better from real external shocks.
  - The insulating advantage of floats depends critically on:
    - Foreign currency debt (higher FCD weakens the insulation of floats; can make floats more destabilizing).
    - Import structure/pass-through (higher raw material share increases the buffering role of floats).
  - Balance sheet effects and limited pass-through can outweigh expenditure switching benefits of depreciation under a float.
- Policy implication (ceteris paribus)
  - The case for a floating exchange rate weakens if a country has high foreign currency debt and imports mainly low pass-through goods because of increased output volatility.
- Caveats
  - Policy makers should also consider other dimensions (inflation, trade volumes, long term growth, likelihood of financial crises) beyond output response to external shocks; framework is silent on these dimensions.

*Source: Excerpt from _wp1142 (2004) — Interacted Panel VAR analysis of terms of trade and foreign interest rate shocks, with interaction terms for PEG, FCD, and RAW, and bootstrapped inference as described in the source document.*

### REFERENCES

### _wp1142 - REFERENCES

### Major components of the content unit
- References: comprehensive bibliography of works related to exchange rate regimes, balance-sheet effects, pass-through, terms-of-trade shocks, monetary policy in open economies, and financial accelerator mechanisms.
- Appendix A: Model — a two-period, small open economy microfounded IS-LM-BP model with sticky prices, wages, and a Bernanke–Gertler–Gilchrist type financial accelerator; limited exchange rate pass-through and non-unit elasticity of substitution between domestic and foreign goods.
- Appendix B: Derivation of the IS-LM-BP equations — linearization and characterization of IS, LM, and BP relations; definitions of parameters and intermediate expressions.
- Appendix C: Solving the system — comparative analysis of outcomes under a peg, a float, and a pure float; derivation of conditions under which a float is stabilizing or destabilizing.
- Appendix D: Data sources — definition and construction of key empirical variables and harmonization choices.

### Appendix A — Model (structure and key definitions)
- Economy:
  - Two periods (1 and 2); agents: workers and entrepreneurs.
  - Adverse external demand shock occurs in period 1; no further shocks in period 2.
- Preferences and consumption aggregator:
  - Worker utility: U = sum_{t=1}^2 β^t [ log C_t − (σ−1)/σ (1/υ) L_t^υ + log(M_t/Q_t) ].
  - Consumption aggregator: C_t = υ [ γ C_{H,t}^{1−φ} + (1−γ) C_{F,t}^{1−φ} ]^{1/(1−φ)}, with φ the elasticity of substitution and υ = γ^{−γ}(1−γ)^{−(1−γ)}.
  - Foreign composite: C_{F,t} = (C_{HOM,F,t})^ω (C_{HET,F,t})^{1−ω}.
  - CES aggregates for varieties with elasticity θ.
- Pricing and production:
  - Domestic firm production: Y_{jt} = A K_{jt}^α L_{jt}^{1−α}.
  - Price-setting: P_{H,t} = θ/(θ−1) E_{t−1} MC_{H,t}.
  - Foreign homogeneous good price in domestic currency: P_{HOM,F,t} = S_t (nominal exchange rate).
  - Foreign differentiated goods: P_{HET,F,t} = θ/(θ−1) E_{t−1} S_t MC_{F,t}; limited exchange rate pass-through because differentiated goods are priced in domestic currency.
- Entrepreneurs and financing:
  - Entrepreneurs start with capital K_1 and external debt: foreign currency D^*_1 and domestic D_1.
  - Investment price Q_1; capital fully depreciates after one period.
  - Balance: P_1 N_1 + S_1 D^*_2 + D_2 = Q_1 I_1.
  - Net worth: P_1 N_1 = R_1 K_1 + Π_1 − S_1 D^*_1 − D_1 = P_1 Y_1 − W_1 L_1 − S_1 D^*_1 − D_1.
  - External financing premium: 1 + η_1 = (Q_1 I_1 / P_1 N_1)^μ, with μ measure of capital market imperfections.
  - Arbitrage condition: R_2 / Q_1 = (1+ρ)(1+η_1) S_2/S_1, where ρ is foreign interest rate.
- Monetary policy regimes:
  - Peg: s_t = 0.
  - Float (reduced form): m_1 = 0, p_2 = 0, leading to s_t = −κ y_t with κ the exchange-rate–output sensitivity parameter.
- Market clearing:
  - Period 1: Y_1 = γ (P_1/Q_1)^{−θ} (I_1 + C_1) + S_1/P_1 X_1.
  - Period 2: Y_2 = γ (P_2/Q_2)^{−θ} C_2 + S_2/P_2 X_2.
  - In period 2, y_2 = α i_1 (investment determines output when no further shocks).

### Appendix B — IS-LM-BP derivation (key relations and definitions)
- Period-2 output as function of real exchange rate:
  - y_2 = Ξ (s_2 − p_2), with Ξ = [ 1 + (φ−1)(1−γ'_2) κ / (1−κ) ] ≥ 1.
  - Price index: q_2 = (1−γ'_2) s_2 + γ'_2 p_2, where γ'_2 = 1 − (1−γ) (S̄_2/Q̄_2)^{(1−φ)}.
- Period-1 linearized market clearing:
  - τ y_1 + (1−τ)[ c_1 + φ q_1 ] = λ[ i_1 + φ q_1 ] + (1−λ)( s_1 + x_1 ).
  - τ = 1/(1−γ) (P̄_1/Q̄_1)^{−φ} (1−α)(1−θ^{−1}) > 1.
  - λ = γ (P̄_1/Q̄_1)^{−φ} Ī_1 / [ γ (P̄/Q̄)^{−φ} Ī_1 + S̄_1 P̄_1 X̄_1 ] < 1.
- IS curve (period 1):
  - A y_1 = λ i_1 + (1−λ) x_1 + [ (1−λ) + λ ω g(φ) ] s_1, where A = (1 − τ α/(1−α)) and g(φ) = (1−γ'_1)[ φ + (φ−1)(τ−1)/λ ].
- LM (float) money-demand relation:
  - m_1 = B y_1 − C i_1, with B = 1/[ ζ (1−α) ] and C = (1−ζ) α / ζ, ζ = 1 − β Q̄_1 C̄_1 / (Q̄_2 C̄_2) > 0.
- BP (entrepreneurs / external finance premium):
  - Arbitrage and risk-premium relations lead to:
    - Γ i_1 = −ρ + μ(1+ψ) δ_y y_1 − μ[(1−γ)ω + ψ ξ] s_1 + [1 − (1−γ) ω] s_1,
    - with Γ = (Φ^{−1} + μ), Φ = Ξ / [ Ξ(1−α) + α ] ≥ 1,
    - δ_y = θ^{−1} [ 1 − (1−α)(1−θ^{−1}) ]^{−1},
    - ψ = (S_1 D^*_1 + D_1) / P_1 N_1 (debt-to-net-worth ratio),
    - ξ = S̄_1 D̄^*_1 / (S̄_1 D̄^*_1 + D̄_1) (foreign-currency share in total debt).

- Calibration example (standard parametrization θ = 11 and α = 1/3):
  - For domestic debt to GDP ratio 12% and foreign debt to GDP ratio 24%: P̄_1 N̄_1 / P̄_1 Ȳ_1 = 0.04 and ψ = 9 and ξ = 2/3.

### Appendix C — Solving the system: peg, float, pure float and comparative statics
- Under a peg (s = 0):
  - System reduces to:
    - A y = λ i + (1−λ) x
    - Γ i = −ρ + μ(1+ψ) δ_y y
  - Responses to an external demand shock x:
    - ∂y/∂x |_{PEG} = (1−λ) Γ / [ A Γ − λ μ (1+ψ) δ_y ].
    - ∂i/∂x |_{PEG} = ∂y/∂x |_{PEG} μ (1+ψ) δ_y / Γ = (1−λ) μ (1+ψ) δ_y / [ A Γ − λ μ (1+ψ) δ_y ].
  - Key results under peg:
    - Result 1: Responses independent of ω and ξ.
    - Result 2: Strength of response increases with leverage ψ; investment response is multiple of output response and increases with ψ and μ. If μ = 0, investment does not respond.
- Under a float (rule s_t = −κ y_t):
  - Output response decomposed into six components (see equation (33)), where κ captures exchange-rate countercyclicality.
  - Interpretations:
    - Terms that make depreciations expansionary: expenditure switching (depends on ω and g) and export value effect (depends on 1−λ).
    - Terms that make depreciations contractionary: balance-sheet effect via foreign debt share ξ and investment-cost effect on the risk premium.
  - Comparative derivatives:
    - ∂(∂y/∂x)/∂ξ = [ ∂y/∂x ]^{−2} λ (1−λ) μ κ ψ / Γ > 0.
      - Higher ξ makes a float less stabilizing (Result 3).
    - ∂(∂y/∂x)/∂ω = [ ∂y/∂x ]^{−2} [ (1−γ'_1)(1+μ) Γ − g ] λ κ / (1−λ) < 0 under φ > 1 (proof in C.4).
      - Higher pass-through ω makes a float more stabilizing (Result 3).
  - Threshold foreign-currency debt share ξ_0 beyond which float is more destabilizing than peg:
    - ξ_0 = [ (1−λ) Γ + λ ] / [ λ μ ψ ] + [ Γ λ g − λ (1+μ) (1−γ'_1) ] / [ λ μ ψ ] ω_0 (equation (36)).
    - Interpretation (Result 4): Lower ω_0, higher ψ, and higher μ reduce the threshold ξ_0 (float becomes destabilizing earlier).
  - Investment response under float:
    - ∂i/∂x |_{FL} given by expression combining Γ^{−1} μ(1+ψ) δ_y times output responses and additional κ-driven terms; higher ξ increases ∂i/∂x (Result 5).
    - Effect of ω on investment is ambiguous (Result 6): first term negative, third term positive, second ambiguous; high leverage makes output effect dominate.
- Pure float as reduced form:
  - Money-market equilibrium under pure float yields i = 1/[ (1−ζ)(1−α) α ] y.
  - Then s = −κ(ω,ξ) y with κ(ω,ξ) = B [ (1+μ)(1−γ') ω + μ ψ ξ − 1 ]^{−1} and B = (Φ^{−1} + μ) [ (1−ζ)(1−α) α ]^{−1} − μ (1+ψ) δ_y.
  - Pure float adjusts κ endogenously: ∂κ/∂ω < 0 implying lower pass-through leads to larger κ (stronger depreciation) to reestablish money market equilibrium.
  - Implication: pure float can insulate output better if foreign debt low and pass-through low; but can amplify contractionary effects if foreign debt high and differentiated imports share high.
- Proof that ∂^2 y / ∂x ∂ ω < 0 for μ ≥ 0 provided (sketch in C.4):
  - Show derivative negative at μ = 0 and decreasing in μ; thus derivative cannot be positive for μ ≥ 0. Conclusion used to sign comparative statics.

### Appendix D — Data sources and variable construction (definitions and harmonization)
- Exchange Rate Regimes:
  - De jure IMF AREAER classifications underwent four taxonomies (Taxonomy 1: 1975–1976; Taxonomy 2: 1977–1981; Taxonomy 3: 1982–1998; Taxonomy 4: 1998–today).
  - Authors harmonize into three categories: Peg, Interm (intermediate), Float using mapping from multiple taxonomy codes into unified classification.
- Import structure:
  - PASS = log( 1 + ( ∑ raw material imports / total imports ) * 100 ), where raw material imports = agricultural goods + fuels + ores and metals; data from WDI; shares rescaled to total imports of goods and services.
- Terms of Trade (ToT):
  - Proxy constructed when IFS, WDI net barter ToT, or UNCTAD not sufficient:
    - ToT = (EXPORTS^{cur} / EXPORTS^{con}) / (IMPORTS^{cur} / IMPORTS^{con}), scaled to base year 2000.
  - Final ToT measure uses IFS if possible, then UNCTAD, then ToT proxy from WDI; alternative used when series breaks or constant series for three consecutive years.
- Country sample list: explicit list of countries and notation indicating source of ToT measure (superscripts 1 for IFS, 2 for UNCTAD/WDI, none for WDI-derived proxy). (Table 5 referred.)
- Variable sources and descriptions (selected):
  - Exchange Rate Regime: LYS (de facto) — Levy-Yeyati and Sturzenegger (2005) — Broad 1 to 3 classification; IMF (de jure) provided by IMF staff — harmonized taxonomies.
  - Import Data – Raw Material Trade: World Development Indicators — share of agricultural, fuel, ores and metals in total merchandise imports.
  - Terms of Trade: IFS (Index, base year 2000 = 100); ToT own calc. based on WDI; net barter ToT (WDI); ToT UNCTAD.
  - National Accounts: GDP (WDI) — real, in local currency and current; Gross Fixed Capital Formation (WDI) — real, in local currency.
  - Foreign Currency Debt: BIS and own calculation — in percent of GDP; (Short Term) External Debt: WDI — in percent of GNI.
  - Interest Rates: Money Market or T-Bill — WDI and national sources — in percent.
  - Base Country data: Shambaugh supplemented with RR for recent period.

### Key analytical findings and comparative results (summary)
- Peg vs Float:
  - Under a peg, output and investment responses to an external demand shock do not depend on import-pass-through ω or the foreign-currency debt share ξ (Result 1).
  - Financial frictions (μ) and leverage (ψ) amplify shock transmission under both regimes; investment response requires μ > 0 (Result 2).
  - A floating exchange rate can be stabilizing via expenditure switching (higher ω increases effectiveness) and export valuation (higher 1−λ increases effectiveness), but can be destabilizing via balance-sheet channels when foreign-currency debt share ξ and leverage ψ are large (Results 3–5).
  - There exists a threshold ξ_0 for foreign-currency debt share beyond which a float becomes more destabilizing than a peg; ξ_0 decreases with lower pass-through ω, higher μ, and higher ψ (Result 4).
  - Effects of the import structure ω on investment and output under a float are nuanced and can be ambiguous; with μ > 0 some effects have opposite signs and net effect depends on parameter values (Result 6).
- Pure float representation:
  - A pure float that stabilizes short-run money supply and medium-run prices can be represented as s = −κ(ω,ξ) y with κ endogenously depending on ω and ξ; lower ω implies larger κ (stronger depreciation) to restore money market equilibrium.

*Source: _wp1142 - REFERENCES (PDF chapter/section and appended model and data appendices).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp1142.pdf_