## _wp0667

## Source details

**Canonical URL:** [_wp0667](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0667.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0667.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0667.pdf.json)

---

### Introduction — solvency, sustainability, and the limits of deterministic DSA
- Solvency requires that the present value of future revenues net of non-interest spending (the primary balances) at least cover existing public debt; solvency is necessary but not sufficient for sustainability.
- Key conceptual points:
  - Sustainability assessment requires judgment about acceptable government strategies to satisfy the intertemporal budget constraint (Mendoza and Oviedo, 2004).
  - Solvency typically excludes outright default or forced restructuring and often excludes the inflation tax from acceptable strategies, limiting adjustments to the primary balance.
- Technical challenges with deterministic DSA:
  - Future paths of primary surpluses, interest rates, inflation, and productivity over horizons like “5 to 10 years” are uncertain.
  - Deterministic “bound tests” simulate alternative debt paths under isolated adverse shocks; calibration often uses fractions or multiples of the unconditional variance (e.g., “half a standard deviation”, “a quarter of a standard deviation”).
  - Limitations: ignore correlations among shocks, assume no fiscal reaction, and each bound test has near-zero formal probability — making meaningful risk quantification impossible.

### Proposed stochastic DSA methodology — purpose and advantages
- Purpose:
  - Improve understanding of risks surrounding debt dynamics and explicitly acknowledge the probabilistic nature of DSA.
  - Produce a more nuanced and credible assessment of long-term sustainability.
- Core idea:
  - Replace isolated deterministic bound tests with stochastic simulations that randomly generate large samples of shock combinations, yielding frequency distributions of the debt ratio for each projection year and depicting uncertainty with “fan charts.”
- Advantages over deterministic DSA:
  - Captures comovements among determinants of debt dynamics (GDP growth, interest rates, exchange rates).
  - Produces probabilistic assessments of debt outcomes rather than a few stylized scenarios.
  - Allows endogenous fiscal responses to shocks via an explicit fiscal reaction function.

### Algorithm architecture — three building blocks
- Block 1: Joint distribution of shocks
  - Calibrated using unrestricted VAR models fitted to historical data (quarterly estimation; quarterly projections annualized).
  - VARs describe comovements among determinants: GDP growth (g), real domestic interest rate (r), real effective exchange rate (z), foreign interest rate (us), etc.; ξt ~ N(0, Ω).
- Block 2: Fiscal reaction function
  - Characterizes fiscal behavior: primary balance equation pit = αi + ρ·di,t−1 + γ·ygapit + β·Xit + ηi + εit.
  - Calibrated using panel estimates for a sample of “34” emerging market economies during “1990-2004”; allows country-specific adjustments when data permit.
  - Rationale: fiscal policy tends to respond systematically to public debt and the business cycle; endogenous reaction improves risk analysis.
- Block 3: Simulation of public debt paths
  - Combines simulated economic scenarios from Block 1 with the fiscal policy process from Block 2.
  - Primary surplus in each simulation includes a fiscal policy shock φti ~ N(0, σφi^2), where σφi^2 is country-specific variance of reaction-function residuals.
  - Repeated simulations (e.g., “1000” or “10,000”) yield frequency distributions, fan charts, and probabilistic sustainability indicators.

### Implementation and illustrative application
- Countries illustrated: “Argentina, Brazil, Mexico, South Africa, and Turkey.”
- Data and timing:
  - Panel: “34” countries, maximum of “15 years (1990-2004).”
  - Simulations performed for “2004 = t” and τ = “5”, except Argentina where “2005 = t.”
- Relation to prior work:
  - Integrates fiscal reaction function approach (Abiad and Ostry, 2005) with stochastic debt analysis (Garcia and Rigobon, 2005; Penalver and Thwaites, 2004).
  - Builds on IMF (2003) concept of “overborrowing.”

### Fiscal reaction function — estimation and diagnostics
- Model and endogeneity:
  - Primary balance: pit = αi + ρ·di,t−1 + γ·ygapit + β·Xit + ηi + εit; endogeneity addressed via instrumentation.
  - Five specifications estimated: LIML (difference), System GMM (SGMM), GMM-DV, and two nonlinear specifications (splines and asymmetric gap responses).
  - Instruments include lags of U.S. one-year bond rates, changes in real oil prices, lagged fiscal costs of banking crises, import demand in industrial-country trading partners, and interactions for nonlinear terms.
- Key coefficient findings (selected estimates as reported):
  - Lagged Debt: 0.039 [0.032]; 0.030*** [0.007]; 0.046*** [0.008]; 0.121 [0.172]; 0.097*** [0.036]
  - Output Gap: 0.104 [0.109]; 0.217*** [0.072]; 0.328*** [0.113]
  - Real Oil Price: 0.481*** [0.072]; 0.084** [0.030]; 0.354*** [0.082]; 0.487*** [0.112]; 0.361*** [0.086]
  - IMF Program: 0.765** [0.347]; 1.121 [0.689]; 1.108*** [0.328]; 0.777** [0.344]; 0.939** [0.328]
  - Default: 0.870** [0.351]; 0.884 [0.813]; 1.190*** [0.401]; 0.749*** [0.297]; 1.077*** [0.368]
- Diagnostics (selected):
  - Observations: 349; 399; 418; 368; 418 (columns 1–5 respectively).
  - Hansen test (P-value): 0.84; 1.00; 0.45; - ; 0.03
  - AR(1) test (P-value): 0.05 (SGMM); AR(2) test (P-value): 0.09 (SGMM).
- Nonlinear findings:
  - Kink at “50 percent of GDP”: fiscal response to debt stronger when debt is below “50 percent of GDP”; above “50 percent of GDP” ρ declines to 0.01–0.03.
  - Output-gap asymmetry: larger deterioration in primary balances during contractions than improvements during booms.

### Non-fiscal VAR and stochastic simulation framework — technical design
- Unrestricted VAR for non-fiscal determinants: Yt = γ0 + Σk γk Yt−k + ξt, with ξt ~ N(0, Ω).
- Shock generation: ξ̂τ = Wυτ with υτ ~ N(1,0) and WW′ = Ω (Choleski factorization) to produce joint dynamic responses and VAR forecasts.
- Annualization: quarterly VAR projections annualized; simulated annual output gap = predicted GDP growth − annualized steady-state growth from the VAR (shocks to output gap zero on average).
- Debt recursion using stock-flow identity (symbols preserved): d̄t ≡ d˜t + d*t = d˜t−1 + s t + [r t − g t]d˜t−1 + ...
- Simulation output: fan charts (deciles; dark grey = “20 percent” confidence interval around median; colored cone = “80 percent” confidence interval).

### Baseline policy scenarios and simulation design
- Fiscal reaction used in simulations: p̂τ+ti = ρ̂τ·d̂τ+ti−1 + γ̂τ·ygap̂τ+ti + Λ̂τ+ti + φτ+ti.
  - Λ̂τ+ti captures predetermined, country-specific determinants; by default κ̂+τ,ti = 0 so Λ̂τ+ti is a country-specific constant anchor.
  - φτ+ti ~ N(0, σφi^2).
- Two baseline scenarios for Λ̂:
  - “Constant” policy scenario: deviation from predicted primary surplus in t persists (κ̂+τ,ti = 0).
  - “Predicted” policy scenario: recent deviations are temporary, simulated surplus paths follow reaction-function prediction (κ̂+τ,ti = −ε̂t).
- Simulations draw “1000” realizations to generate fan charts.

### Fan chart and risk findings — illustrative results for five emerging markets
- Countries and timing: Argentina (t=2005), Brazil (t=2004), Mexico (t=2004), South Africa (t=2004), Turkey (t=2004).
- Key insights:
  - Deterministic bound tests align roughly with the “40 percent” confidence interval of stochastic simulations — deterministic tests understate uncertainty.
  - Median debt paths generally stable/declining, but upper deciles show growing debt; in many cases at least a “30 percent” chance that adverse shocks lead to sustainability concerns.
  - Cross-country heterogeneity: volatile economies (Argentina, Turkey) have wider confidence intervals; South Africa narrower.
  - Large recent residuals materially affect fan charts (e.g., Mexico negative 2004 residual; Turkey large positive 2003–2004 residuals).
  - Upside skewness in distributions (Argentina, Turkey) suggests average fiscal response may be insufficiently stabilizing; stronger surplus response may be warranted.

### Probabilistic sustainability indicator — definition and reported values
- Definition: SI = Pr(d t+τ < d t) * (1 − Pr(d t+τ > d t + x)), with x = “0.10” (10 percent of GDP) in reported table.
- Table 3 (under the “Constant” Policy Scenario) — probabilities across horizons t+1 to t+5:
  - Pr(debt ratio lower than in t):
    - Argentina (t=2005): 0.83, 0.85, 0.82, 0.76, 0.73
    - Brazil (t=2004): 0.59, 0.63, 0.64, 0.63, 0.64
    - Mexico (t=2004): 0.51, 0.54, 0.53, 0.55, 0.56
    - South Africa (t=2004): 0.60, 0.64, 0.69, 0.70, 0.71
    - Turkey (t=2004): 0.40, 0.45, 0.46, 0.47, 0.47
  - Pr(debt ratio more than 10 percent of GDP higher than in t):
    - Argentina (t=2005): 0.06, 0.06, 0.10, 0.15, 0.20
    - Brazil (t=2004): 0.00, 0.05, 0.10, 0.15, 0.19
    - Mexico (t=2004): 0.03, 0.08, 0.11, 0.16, 0.20
    - South Africa (t=2004): 0.00, 0.00, 0.00, 0.03, 0.04
    - Turkey (t=2004): 0.21, 0.28, 0.34, 0.35, 0.38
  - Sustainability index (SI) values:
    - Argentina (t=2005): 0.78, 0.79, 0.74, 0.65, 0.59
    - Brazil (t=2004): 0.59, 0.60, 0.57, 0.54, 0.52
    - Mexico (t=2004): 0.50, 0.50, 0.47, 0.46, 0.45
    - South Africa (t=2004): 0.60, 0.64, 0.69, 0.68, 0.68
    - Turkey (t=2004): 0.32, 0.32, 0.30, 0.30, 0.29
- Interpretive guidance:
  - SI combines probability of decline with containment of upside risk; policymakers can set a critical threshold (example used in text: “0.4”) to flag problematic cases.

### Probabilistic debt-sustainability assessment — early-warning performance and robustness
- Early-warning retrospective checks:
  - Argentina at end-2000: indicator “shining red”; fan charts pointed to large upside/explosive debt risks.
  - Turkey at end-1999: indicator “shining red.”
  - Brazil at end-2000: indicator “yellow” — under control but with significant upside risks.
  - South Africa at end-2000: model predicts observed declining debt path; very high sustainability indicator.
- Robustness and sensitivity (selected from Table 5; “Constant” Policy Scenario; values for 2009 1/ unless noted):
  - Baseline — Argentina: 0.73; Brazil: 0.64; Turkey: 0.47
  - System GMM calibration — Argentina: 0.83; Brazil: 0.67; Turkey: 0.43
  - Non-linear reaction function — Argentina: 0.79; Brazil: 0.62; Turkey: 0.44
  - Deterministic fiscal policy — Argentina: 0.98; Brazil: 0.72; Turkey: 0.51
  - US junk bond interest rates — Argentina: 0.64; Brazil: 0.45; Turkey: 0.36
  - Footnote: “1/ Values for 2010 in the case of Argentina.”
- Main robustness conclusions:
  - System-GMM and nonlinear specifications do not materially change risk conclusions because weaker debt responsiveness is offset by higher predetermined fiscal components.
  - Deterministic fiscal paths substantially understate risk and raise sustainability indices.
  - Using higher-volatility foreign interest rates (U.S. junk bond yields) widens fan charts and lowers SI values markedly.

### Limitations and caveats
- Short time series and regime shifts in emerging markets reduce reliability of econometric forecasts.
- Extreme crisis events remain low-probability in this framework; fan charts detect risk of undesirable “non-crisis” situations more readily than full-blown crises.
- Annualization and combining quarterly VARs with annual fiscal data can shut off feedback from fiscal policy to economic variables (e.g., credibility effects on interest rates); methodology can accommodate imposed effects if warranted.
- Fiscal-policy volatility (reaction-function residuals) is treated as a source of risk via σφi^2.

### Policy implications and use cases
- The stochastic algorithm supplies:
  - Graphical fan charts summarizing distributions of public debt paths.
  - Numerical sustainability indicators combining probability of debt decline with containment of upside risks.
- Use cases:
  - Early-warning: would have signaled Argentina (2000), Turkey (1999), and warned on Brazil (2000).
  - Policy evaluation: quantify impact of delaying fiscal adjustment or of fiscal expansions on median debt paths and upside risks.
  - Consolidation strategy design: compare alternative adjustment plans by their effects on trend and upside risk.
- Advantages over deterministic templates:
  - Accounts for joint distributions and covariances of shocks.
  - Allows fiscal policy to adjust endogenously via estimated reaction functions.
  - Treats fiscal policy volatility as a risk factor; deterministic primary-balance assumptions can significantly underestimate risks.

*Italic: Source — _wp0667 - References..............................................................................................................*

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

### _wp0667 - References..............................................................................................................

### H3: Introduction — solvency, sustainability, and the limits of deterministic DSA
- Solvency requires that the present value of future revenues net of non-interest spending (the primary balances) at least cover existing public debt; solvency is necessary but not sufficient for sustainability.
- Key conceptual issues:
  - Sustainability assessment requires judgment about acceptable government strategies to satisfy the intertemporal budget constraint (Mendoza and Oviedo, 2004).
  - Solvency excludes outright default or forced restructuring; most analysts also exclude the inflation tax from acceptable strategies, limiting adjustments to the primary balance.
- Technical challenges:
  - Future paths of primary surpluses, interest rates, inflation, and productivity over horizons like 5, 10, or 20 years are uncertain.
  - Standard medium-term DSA (generally 5 to 10 years) interprets a declining debt-to-GDP trend as signaling sustainable policies, whereas a rising trend or stabilization at a high level raises concerns.
- Deterministic “bound tests” in standard DSA:
  - Simulate alternative debt paths under adverse shocks to single variables (lower growth, higher interest rates, lower primary balance, exogenous debt increases such as exchange rate depreciation or recognition of off-budget obligations).
  - Shock calibration often uses a multiple or fraction of the unconditional variance of the underlying series.
- Limitations of deterministic bound tests:
  - Ignore correlations among shocks and joint dynamic responses of relevant variables.
  - Assume fiscal policy does not react to simulated economic developments.
  - Each individual bound test formally has a near-zero probability of occurrence, making meaningful quantification of risk impossible.

### H3: Proposed stochastic DSA methodology — purpose and advantages
- Purpose:
  - Improve understanding of risks surrounding debt dynamics and explicitly acknowledge the probabilistic nature of DSA.
  - Produce a more nuanced and credible assessment of long-term sustainability.
- Core idea:
  - Replace isolated deterministic bound tests with stochastic simulations that randomly generate a large sample of shock combinations, yielding frequency distributions of the debt ratio for each projection year.
  - Depict uncertainty using “fan charts” that show confidence bands around the median debt projection.
- Advantages over deterministic DSA:
  - Captures comovements among determinants of debt dynamics (GDP growth, interest rates, exchange rates).
  - Produces probabilistic assessments of debt outcomes rather than relying on a few stylized scenarios.
  - Allows endogenous fiscal responses to shocks via an explicit fiscal reaction function.

### H3: Algorithm architecture — three building blocks
- Block 1: Joint distribution of shocks
  - Calibrated to fit statistical properties of historical data using unrestricted VAR models.
  - VARs:
    - Describe comovements among determinants of debt dynamics (GDP growth, interest rates, exchange rates).
    - Provide estimates of conditional variances and covariances of shocks.
    - Generate consistent projections for determinants of debt dynamics.
  - Estimation requires quarterly data; quarterly projections are annualized for the stock-flow debt identity.
- Block 2: Fiscal reaction function
  - Characterizes fiscal behavior through an explicit reaction function.
  - Calibrated using panel estimates for a sample of 34 emerging market economies during 1990-2004.
  - Allows country-specific adjustments if sufficient time-series data are available.
  - Rationale: fiscal policy tends to systematically respond to variations in public debt and the business cycle; allowing endogenous fiscal reaction improves risk analysis.
- Block 3: Simulation of public debt paths
  - Combines simulated economic scenarios from Block 1 with the fiscal policy process from Block 2.
  - Produces annual public debt paths that reflect plausible constellations of shocks and consistent projections for growth, interest rates, exchange rates, and fiscal policy.
  - Repeated simulations yield a large sample of public debt projections and associated frequency distributions for each year.

### H3: Implementation and illustrative application
- Countries analyzed (illustration of versatility): Argentina, Brazil, Mexico, South Africa, and Turkey.
- Empirical inputs and considerations:
  - VAR estimation requires quarterly data; concern expressed about the high noise-to-signal ratio in quarterly fiscal data.
  - Simulated primary balances and public debt are determined recursively to account for fiscal policy responses to debt developments.
- Relation to prior work:
  - Marries fiscal reaction function approach in Abiad and Ostry (2005) with stochastic debt analysis in Garcia and Rigobon (2005) and Penalver and Thwaites (2004).
  - Builds on IMF (2003) concept of “overborrowing” (excess of current public debt over the annuity value of future primary surpluses) and refines estimates of fiscal reaction functions using an expanded dataset.

### H3: Comparative diagnostics — deterministic versus stochastic insights
- Deterministic DSA (example: South Africa, 2005-2010 template) uses standardized bound tests including:
  - Small but permanent adverse shocks (half a standard deviation) to real GDP growth, the real interest rate, and the primary balance.
  - A combination of these three shocks (quarter of a standard deviation).
  - Two large temporary disturbances: a 30 percent real depreciation and a contingent-liabilities shock equivalent to 10 percent of GDP.
- Critiques of deterministic scenarios:
  - Scenarios may not reflect shock patterns observed in the economy (ignore comovements).
  - Degree of realism varies across countries due to standardization.
  - Likelihood of resulting debt paths cannot be precisely assessed; plausibility judged on core assumptions rather than probabilistic merits.
- Stochastic approach provides:
  - Realistic constellations of shocks based on estimated joint distributions.
  - Country-specific frequency distributions of debt that enable quantitative risk assessment and can refine fiscal policy recommendations.

- Key numeric values preserved from source text (examples cited in the methodology and illustrations):
  - Projection horizon commonly “5 to 10 years.”
  - VAR estimation sample period for panel fiscal reaction function: “1990-2004.”
  - Illustrative large temporary disturbances: “30 percent real depreciation” and contingent-liabilities shock “equivalent to 10 percent of GDP.”
  - Deterministic small permanent shocks calibrated as “half a standard deviation” and combinations at “a quarter of a standard deviation.”
  - Five emerging market economies used to illustrate the algorithm: “Argentina, Brazil, Mexico, South Africa, and Turkey.”
  - South Africa: DSA example covering “2005-2010” (and figures labeled “2004-2010” appear in source material).

*Italic: Source — _wp0667 - References..............................................................................................................*

### Section III). Also, the estimated residuals of the reaction functions provide information on

### _wp0667 - Section III). Also, the estimated residuals of the reaction functions provide information on

### Fiscal behavior, reaction functions, and implications for risk analysis
- Estimated residuals of the reaction functions provide information on the stability of fiscal behavior, with erratic policy impulses being another independent source of uncertainty.
- The passive policy assumption fails to capture the stabilizing response of the primary surplus to the debt itself, thereby increasing estimated risks to debt.
- Accounting for systematic features of the policy process:
  - Improves reliability of the risk analysis.
  - Should lead to more focus of policy advice on fiscal effort (measured as the departure from the estimated reaction function) rather than on the unconditional change in the fiscal balance.
- Deviations of actual policies from the benchmark provided by the estimated reaction function may prove useful in assessing such effort, and thereby the room for fiscal adjustment (Abiad and Ostry, 2005 and Debrun, 2005b).
- Table 1 summarizes some key difference between the usual DSA framework and the extended DSA proposed in this paper.

### Overview of the simulation algorithm — objectives
- The simulation algorithm seeks to satisfy three main objectives:
  1. Provide a sensible way to account for fiscal behavior and simulate changes to it;  
  2. Provide a tool easily applicable to different emerging market countries while referring to a common benchmark for policy assessment; and  
  3. Keep data requirements close to those in the standard DSA.
- Implications of the objectives:
  - The first objective puts the fiscal reaction function as an essential building block.
  - The second objective points to panel data techniques to estimate an average behavior across a group of countries that could serve as a “positive” benchmark for each individual member of the group though, as suggested earlier, if sufficient country-specific data are available, it would in principle be possible to estimate a country-specific fiscal reaction function.
  - The third objective suggests relying as much as possible on annual data.
- Footnote: "This being said, our simulation tool can accommodate any normative policy scenario, including the constant policy assumption."8

### Key contrasts between deterministic (commonly used DSA) and the paper's probabilistic approach
- Diagnostic based on:
  - Deterministic (commonly used DSA): "...a few stylized, isolated shocks; exogenous policies."
  - Probabilistic approach (in this paper): "...a large number of random shock constellations drawn from an estimated joint distribution; endogenous fiscal policy."
- Calibration of shocks:
  - Deterministic: "Fraction or multiple of historical standard deviations of underlying variables. Calibration based on the likelihood of the resulting debt path for a sample of emerging market economies."
  - Probabilistic: "Directly based on the estimated joint distribution of disturbances (country specific)."
- Output:
  - Deterministic: "Large temporary shocks provide a probabilistic upper bound to the debt ratio; small permanent shocks delineate interval of most probable outcomes."
  - Probabilistic: "Frequency distributions of the debt ratio over time, “fan charts.”"
- Main advantages:
  - Deterministic: "Amenable to standardized bound tests across countries; low data requirement."
  - Probabilistic: "Better reflection of country specificity (in terms of shocks and fiscal policy behavior); explicitly probabilistic output."
- Sources: IMF, and the authors.

### Data requirements and estimation strategy — reasons for algorithm design
- The objectives impose a significant departure from existing algorithms.9
- Need to initially separate the estimation of the fiscal reaction function from that of the other economic relationships before merging them again when simulating the behavior of the public debt ratio. Two compelling reasons:
  1. The estimation of the variance-covariance matrix of shocks inevitably relies on time-series techniques (an unrestricted VAR model) that demand higher-than-annual frequency data. However, in contrast to financial variables and GDP, budgetary data at such frequency are often either unavailable or unreliable for the purpose of policy evaluation.10
  2. The VAR framework restricts the specification of the reaction function in undesirable ways: e.g., the primary surplus responds to contemporaneous variations in the output gap, not lagged ones, as imposed by the VAR framework.
- The approach tries to better internalize data constraints, focusing on a shorter time span (during which regime shifts and structural changes are less likely) while keeping annual fiscal data at the center of the analysis. This potentially increases the number of countries to which the algorithm could be applied.

### Algorithm structure
- The algorithm comprises three building blocks:
  - The first building block is discussed more fully in Section III (below in the source).
  - The other two building blocks are discussed in the Appendix.

*Source: _wp0667 - Section III). Also, the estimated residuals of the reaction functions provide information on*

### 1. A fiscal reaction function—in the positive sense of a description of average fiscal policy

### 1. A fiscal reaction function—in the positive sense of a description of average fiscal policy

### Model specification and identification
- Primary balance equation (panel): pit = αi + ρ·di,t−1 + γ·ygapit + β·Xit + ηi + εit (equation (1) as described).
  - pit is the ratio of the primary balance to GDP in country i and year t.
  - di,t−1 is the public debt-to-GDP ratio observed at the end of period t-1.
  - ygapit is the output gap.
  - ηi is an unobserved country fixed-effect.
  - Xit is a vector of control variables.
- Endogeneity concerns:
  - Output gap endogenous with contemporaneous fiscal shocks εit.
  - Lagged debt correlated with country fixed effects ηi.
  - Persistence in idiosyncratic errors εit makes lagged debt endogenous via past surpluses.
- Instrumentation strategy:
  - Five specifications estimated: LIML (difference) instrumenting lagged change in debt and contemporaneous change in output gap; System GMM (SGMM); GMM with country dummies instrumenting only the output gap (GMM-DV); and two nonlinear specifications (splines and asymmetric gap responses).
  - Instruments used (LIML/SGMM/LIML nonlinear): lags of U.S. one-year bond rates, changes in real oil prices, lagged fiscal costs of banking crises, import demand in industrial-country trading partners, and interactions for nonlinear terms.

### Non-fiscal VAR and stochastic simulation framework
- For each country an unrestricted VAR for non-fiscal determinants: Yt = γ0 + Σk γk Yt−k + ξt, where Yt = (z, g, r, us, ...)′ with:
  - r = real domestic interest rate
  - us r = real foreign interest rate
  - g = real GDP growth rate
  - z = (log of the) real effective exchange rate
  - ξt ~ N(0, Ω)
- Simulations:
  - Generate sequences of shocks ξ̂t,...,ξ̂T drawn so that ξ̂τ = Wυτ with υτ ~ N(1,0) and WW′ = Ω (Choleski factorization), producing joint dynamic responses and VAR forecasts consistent with shocks.
  - Quarterly VAR projections annualized; simulated annual output gap = predicted GDP growth − annualized steady-state growth from the VAR (ensures shocks to the output gap are zero on average).
  - Debt path computed recursively using the reaction function and the stock-flow identity:
    - d̄t ≡ d˜t + d*t = d˜t−1 + s t + [r t − g t]d˜t−1 + ... (stock-flow identity notation as in text, preserving symbols t, d*, d˜, s).
  - Primary surplus in each simulation incorporates a fiscal policy shock φti ~ N(0, σφi^2) where σφi^2 is country-specific variance of reaction-function residuals.
  - Algorithm can produce arbitrarily many debt paths (e.g., 1000 or 10,000) to obtain frequency distributions, fan charts, and probabilistic sustainability indicators.

### Data and sample
- Panel: 34 countries, maximum of 15 years (1990-2004).
- Annual frequency used; fiscal data from IMF desk economists for largest available coverage of the fiscal sector.
- Simulations performed for 2004 = t and τ = 5, except Argentina where 2005 = t (to account for the debt exchange operation).

### Estimation results — linear and nonlinear reaction functions
- General findings:
  - All three linear estimations in Table 2 indicate a positive response of primary surpluses to public debt.
  - LIML and SGMM estimates of ρ imply responses of 0.030-0.039; the GMM-DV regression gives 0.046.
  - Positive estimated coefficient on the output gap → primary balances estimated to be countercyclical, driven more by worsening balances in recessions than improvements in booms.
  - Countries with IMF-supported programs run higher surpluses (IMF Program coefficients significant).
  - Countries in default run larger primary balances (Default dummy positive and significant in several specifications).
- Selected coefficient estimates from Table 2 (standard errors in brackets; significance indicated exactly as in table):
  - Lagged Debt: 0.039 [0.032]; 0.030*** [0.007]; 0.046*** [0.008]; 0.121 [0.172]; 0.097*** [0.036]
  - Output Gap: 0.104 [0.109]; 0.217*** [0.072]; 0.328*** [0.113]
  - Real Oil Price: 0.481*** [0.072]; 0.084** [0.030]; 0.354*** [0.082]; 0.487*** [0.112]; 0.361*** [0.086]
  - Institutions: 0.374 [0.484]; −0.219 [0.322]; −0.675*** [0.258]; 0.463 [0.445]; −0.380 [0.256]
  - IMF Program: 0.765** [0.347]; 1.121 [0.689]; 1.108*** [0.328]; 0.777** [0.344]; 0.939** [0.328]
  - Default: 0.870** [0.351]; 0.884 [0.813]; 1.190*** [0.401]; 0.749*** [0.297]; 1.077*** [0.368]
- Tests and diagnostics (as reported in Table 2):
  - Observations: 349; 399; 418; 368; 418 (columns 1–5 respectively).
  - Hansen test (P-value): 0.84; 1.00; 0.45; - ; 0.03
  - AR(1) test (P-value): 0.05 (as reported for SGMM)
  - AR(2) test (P-value): 0.09 (as reported for SGMM)
  - Cragg-Donald Stat.: 7.23; 19.63; 1.96 (for specified columns)
- Nonlinear (kinked at 50 percent of GDP and asymmetric output-gap response):
  - Specification: pit = ηi + ρ·di,t−1 + γ·ygapit + β·Xit + ρ50·(di,t−1·D50) + γP·(ygapit·P) + γN·(ygapit·N) + εit (notation as in text).
  - Findings: fiscal response to debt stronger when debt is below 50 percent of GDP; once debt exceeds 50 percent of GDP, ρ declines to 0.01-0.03 (slightly lower than linear specification).
  - Output-gap asymmetry: deterioration in primary balances during contractions larger than improvements during booms.

### Simulation design and baseline policy scenarios
- Fiscal reaction function used in simulations (equation (3) notation preserved):
  - p̂τ+ti = ρ̂τ·d̂τ+ti−1 + γ̂τ·ygap̂τ+ti + Λ̂τ+ti + φτ+ti
  - Λ̂τ+ti captures predetermined, country-specific determinants (institutional quality, IMF program, default, oil-price effects); by default κ̂+τ,ti = 0 so Λ̂τ+ti is a country-specific constant anchor.
  - φτ+ti is a policy shock drawn from mean-zero normal distribution with variance equal to country-specific variance of reaction-function residuals.
- Two baseline scenarios for non-debt discretionary component Λ̂:
  - “Constant” policy scenario: any deviation from predicted primary surplus in t is assumed to persist over the simulation horizon (κ̂+τ,ti = 0 over horizon).
  - “Predicted” policy scenario: recent deviations are temporary, simulated surplus paths follow the reaction-function prediction (κ̂+τ,ti = −ε̂t over horizon).
- Simulations draw 1000 realizations to generate fan charts (deciles; dark grey = 20 percent confidence interval around median; colored cone = 80 percent confidence interval).

### Fan chart and risk findings (five illustrative emerging markets)
- Countries analyzed: Argentina (t=2005), Brazil (t=2004), Mexico (t=2004), South Africa (t=2004), Turkey (t=2004).
- Main insights:
  - Deterministic bound tests correspond to roughly the 40 percent confidence interval of stochastic simulations — deterministic tests understate uncertainty.
  - Median debt paths generally stable/declining (suggesting average fiscal response sufficient for median sustainability), but upper deciles show growing debt: in many cases at least a 30 percent chance that adverse shocks lead to sustainability concerns.
  - Cross-country heterogeneity: more volatile economies (e.g., Argentina, Turkey) have wider confidence intervals; less volatile (South Africa) narrower cones.
  - Large last-year residuals materially affect fan charts (Mexico had a highly negative 2004 residual; Turkey had large positive 2003–2004 residuals).
  - Upside skewness in distributions (Argentina, Turkey) indicates standard average fiscal response may be insufficiently stabilizing given typical shock configurations — stronger surplus response to debt may be warranted.

### Probabilistic sustainability indicator (definition and results)
- Sustainability indicator defined as: SI = Pr(d t+τ < d t) * (1 − Pr(d t+τ > d t + x)), where x is a positive mark-up over initial public debt (x = 0.10 = 10 percent of GDP in reported table).
- For x = 0.10, Table 3 (under the “Constant” Policy Scenario) reports:
  - Pr(debt ratio lower than in t) across horizons t+1 to t+5:
    - Argentina (t=2005): 0.83, 0.85, 0.82, 0.76, 0.73
    - Brazil (t=2004): 0.59, 0.63, 0.64, 0.63, 0.64
    - Mexico (t=2004): 0.51, 0.54, 0.53, 0.55, 0.56
    - South Africa (t=2004): 0.60, 0.64, 0.69, 0.70, 0.71
    - Turkey (t=2004): 0.40, 0.45, 0.46, 0.47, 0.47
  - Pr(debt ratio more than 10 percent of GDP higher than in t) across horizons t+1 to t+5:
    - Argentina (t=2005): 0.06, 0.06, 0.10, 0.15, 0.20
    - Brazil (t=2004): 0.00, 0.05, 0.10, 0.15, 0.19
    - Mexico (t=2004): 0.03, 0.08, 0.11, 0.16, 0.20
    - South Africa (t=2004): 0.00, 0.00, 0.00, 0.03, 0.04
    - Turkey (t=2004): 0.21, 0.28, 0.34, 0.35, 0.38
  - Sustainability index (SI = product) across horizons t+1 to t+5:
    - Argentina (t=2005): 0.78, 0.79, 0.74, 0.65, 0.59
    - Brazil (t=2004): 0.59, 0.60, 0.57, 0.54, 0.52
    - Mexico (t=2004): 0.50, 0.50, 0.47, 0.46, 0.45
    - South Africa (t=2004): 0.60, 0.64, 0.69, 0.68, 0.68
    - Turkey (t=2004): 0.32, 0.32, 0.30, 0.30, 0.29
- Interpretation and use:
  - The SI combines the probability of a declining debt ratio with containment of upside risk (no rise above d t + 10 percentage points).
  - Policymakers can set a critical threshold for SI (example threshold used in text: 0.4) to flag problematic cases; choice of threshold depends on risk aversion and desired reduction in debt.

### Limitations highlighted
- Short time series and regime shifts in many emerging markets reduce reliability of econometric estimates for forecasting.
- Extreme events (crises) remain low-probability in this framework; fan charts may at best detect risk of undesirable “non-crisis” situations.
- Annualization (combining annual and quarterly data) shuts off feedback from fiscal policy to economic variables (e.g., credibility effects on interest rates, fiscal reforms affecting future growth and spreads). The methodology can accommodate imposed effects (overriding VAR steady states) if warranted.

*Source: _wp0667 - 1. A fiscal reaction function—in the positive sense of a description of average fiscal policy; canonical URL: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0667.pdf*

### 0.50 X (1 - 0.20)]; alternatively, of course, probabilities of decline of less (more) than 50

### _wp0667 - 0.50 X (1 - 0.20)]; alternatively, of course, probabilities of decline of less (more) than 50

### Probabilistic debt-sustainability assessment: key findings and early-warning performance
- Probabilistic indicators (fan charts + a sustainability index) highlight upside risks to public debt driven by shocks identified via country-specific VARs—particularly domestic real interest rates and the exchange rate for Turkey.
- The sustainability indicator is defined as [Pr( debt < 2000 level )*(1-Pr(debt > (2000 level + 10 percent)] (except for Turkey, where 1999 replaces 2000).
- Eve-of-crisis checks (information available to policymakers at the time):
  - Argentina at end-2000: indicator “shining red”; fan charts point to particularly large upside/explosive debt risks.
  - Turkey at end-1999: indicator “shining red”; debt dynamics looked unmanageable under most circumstances.
  - Brazil at end-2000: indicator “yellow”; situation broadly under control but with significant upside risks and rapid deterioration signaled over the simulation horizon.
  - South Africa at end-2000 (retrospective control): model predicts the declining debt path effectively observed; South Africa exhibits a very high sustainability indicator.
- A policymaker more averse to upside risk (reducing acceptable probability from 20 percent to 10 percent) would shade more cells: “More cells would be shaded if the policymaker wanted to reduce the probability of upside risk to say 10 percent from 20 percent... In the limit... all cells below 0.50 would be shaded.”

### Alternative calibrations and sensitivity to assumptions
- Variation in the fiscal reaction function specification:
  - System-GMM estimates of the linear reaction function (to account for possible upward bias in responsiveness) and a nonlinear reaction function (debt spline and differentiated responses to the output gap) do not materially affect the risk analysis.
  - Reason: weaker responsiveness to public debt under system-GMM is compensated by a higher pre-determined component of fiscal policy (τ+Λti,).
- Deterministic primary-surplus paths (from IMF staff reports) versus endogenous fiscal policy:
  - Assuming deterministic primary surpluses substantially reduces perceived risk by ignoring fiscal response to output gap variations and fiscal volatility.
  - Deterministic policy assumptions lead to underestimating risks to public debt sustainability and produce an upward bias in the sustainability index (Table 5).
- Choice of foreign interest rate (10-year treasury vs U.S. junk bond rate):
  - Using U.S. junk bond yields widens fan-chart confidence intervals and dramatically lowers sustainability indices, confirming vulnerability to higher and more volatile interest rates.
- Robustness summary (selected entries from Table 5; "Constant" Policy Scenario unless otherwise indicated, values for 2009 1/):
  - Baseline — Argentina: 0.73; Brazil: 0.64; Turkey: 0.47
  - System GMM calibration — Argentina: 0.83; Brazil: 0.67; Turkey: 0.43
  - Non-linear reaction function — Argentina: 0.79; Brazil: 0.62; Turkey: 0.44
  - Deterministic fiscal policy — Argentina: 0.98; Brazil: 0.72; Turkey: 0.51
  - US junk bond interest rates — Argentina: 0.64; Brazil: 0.45; Turkey: 0.36
  - (Table 5 also reports probabilities of debt being lower than in 2004 and probabilities of debt being more than 10 percent of GDP higher than in 2004; see source.)
- Footnotes and calibration notes:
  - 1/ Values for 2010 in the case of Argentina.
  - 2/ The sustainability index is defined as [Pr( debt < 2004 level )*(1-Pr(debt > (2004 level + 10 percent)] for Table 5.

### Implications for policy and risk management
- The probabilistic algorithm provides policymakers:
  - Graphical fan charts summarizing the distribution of public debt paths.
  - Numerical sustainability indicators that combine probability of debt decline with containment of upside risks.
- Use cases:
  - Early-warning: indicators and fan charts would have signaled impending troubles for Argentina (2000), Turkey (1999), and warned for Brazil (2000).
  - Policy evaluation: quantify the effect of delaying fiscal adjustment or undertaking fiscal expansions on both median debt paths and upside risks.
  - Consolidation strategy design: evaluate alternative adjustment plans by their impact on trend and upside risk of debt paths.
- Advantages over deterministic DSA templates:
  - Accounts for joint probability distributions of shocks and their covariances.
  - Allows fiscal policy to adjust endogenously (via estimated fiscal reaction functions).
  - Treats fiscal policy as a potential source of risk (volatility of primary balance).
  - Deterministic primary-balance assumptions can produce significant underestimation of risks; probabilistic approach represents a considerable improvement.

### Analytical framework and technical highlights (Appendix extracts)
- Government flow budget identity (nominal form):
  - 1 (1) t t t t t DiDBP − = + + − .
- Real representation (deflated by inflation) given as equation (A.1):
  - 1 1 , 1 t t t t t r d d b p g − ⎛⎞ + = + − ⎜⎟ + ⎝⎠
- Backward-solution for debt (equation (A.2)):
  - 0 1 (). 1 j t t t j i t j j r d b y g a p γ η ε ∞ − − = ⎛⎞ + = − − − − ⎜⎟ + ⎝⎠ ∑
- Necessary (stability) condition stated verbatim:
  - "Assuming that the output gap and the shocks to the primary balance are stationary, equation (A.2) implies that a necessary condition for the path of debt to be non explosive—that is, for the debt to GDP ratio to converge to a finite level—is that 1 1 1 r g ρ ⎛⎞ + −< ⎜⎟ + ⎝⎠ ."
- Role of asymmetric output-gap response:
  - If primary deficits worsen in recessions and do not improve in booms, debt will tend to rise unless the response of primary surpluses to debt (ρ) offsets the deficit bias.

### Econometric estimation of fiscal reaction functions (appendix summary)
- Sources of endogeneity addressed:
  - Output-gap endogeneity to contemporaneous policy shocks.
  - Lagged debt endogeneity due to correlation with country fixed effects (ηi) and persistence in policy shocks (εit).
- Estimation methods employed and rationale:
  - OLS and LSDV (benchmarks).
  - GMM-DV (GMM with country dummies and instrumented output gap).
  - Difference GMM and LIML for equation in differences (B.1), using lags of interest rates, real oil prices, fiscal banking-crisis cost proxy, and trade-partners import demand as instruments.
  - Blundell and Bond (1998) system-GMM (SGMM) for efficiency and precision gains.
- Baseline empirical findings (Table B.1, 1990-2004; dependent variable: level or change in the primary fiscal balance — standard errors in brackets):
  - Lagged Debt coefficients:
    - OLS: 0.023*** [0.005]
    - LSDV: 0.062*** [0.008]
    - GMM with DV: 0.069*** [0.008]
    - GMM (Difference): 0.070* [0.042]
    - LIML (Difference): 0.055 [0.037]
    - System GMM variants: 0.025* [0.014], 0.031* [0.017], 0.034** [0.014], 0.037** [0.016]
  - Output Gap coefficients:
    - OLS: 0.047 [0.058]
    - LSDV: 0.181*** [0.050]
    - GMM with DV: 0.3 4*** [0.133]
    - GMM (Difference): 0.279** [0.109]
    - LIML (Difference): 0.312** [0.121]
    - System GMM variants: 0.138* [0.070], 0.155 [0.101], 0.243*** [0.059], 0.251*** [0.078]
  - Constant terms and country-dummies usage noted in table.
- Diagnostic/statistics (selected):
  - Observations: typically 330 (except LSDV with 33).
  - Cragg-Donald statistic reported for first stage: e.g., 13.29; 7.58.
  - Hansen test p-values (varied across specs, reported in table).
  - AR(1) and AR(2) tests reported (p-values in table).

### Data, VARs, and calibrations (selected)
- Dataset covers: Argentina, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Côte d’Ivoire, Croatia, Ecuador, Egypt, Hungary, India, Indonesia, Israel, Jordan, Korea, Lebanon, Malaysia, Mexico, Morocco, Nigeria, Pakistan, Panama, Peru, the Philippines, Poland, Russia, South Africa, Thailand, Turkey, Ukraine, Uruguay, and Venezuela.
- Key series definitions and sources (as listed):
  - Primary fiscal surpluses and public debt: IMF desk and fiscal economists (percent of GDP).
  - Output gap: percent deviation of real GDP from HP-filtered trend; real GDP from IMF IFS or WEO.
  - Institutional quality: average of five ICRG indicators (range 0-6).
  - IMF program indicator, Default Indicator, Fiscal Costs of Banking Crises (Honohan et al. 2005), U.S. one-year government bond rates (FRED), real oil prices (Brent/Dubai/Texas deflated by U.S. CPI).
  - Real interest rate formula (quarterly, equation (C.1)):
    - 0.25 t t+1 (1+(int. rate /100)) 1 1+CPI inflation −,  where tint. rate is domestic nominal interest at quarter t and t+1 CPI inflation is CPI inflation for quarter t+1.
- VAR structure and selected implied steady-state values (annualized percent):
  - Argentina (foreign interest rate: 10-year US government bond yield)
    - Implied steady state values: Real foreign interest rate 2.33456; Real domestic interest rate 9.76359; GDP growth 6.49988.
  - Argentina (foreign interest rate: US junk bond yield)
    - Implied steady state values: Real foreign interest rate 10.36704; Real domestic interest rate -0.19766; GDP growth 0.16626 (table entries reflect VAR coefficients and residual correlations).
  - Brazil (foreign interest rate: 10-year US government bond yield)
    - Implied steady state values: Real foreign interest rate 2.51465; Real domestic interest rate 9.79475; GDP growth 2.44727.
  - Brazil (US junk bond yield)
    - Implied steady state values: Real foreign interest rate 7.26897; Real domestic interest rate 9.34110; GDP growth 6.07628.
  - Turkey (10-year US government bond yield)
    - Implied steady state values: Real foreign interest rate 3.60884; Real domestic interest rate 16.97227; GDP growth 3.18259.
  - Turkey (US junk bond yield)
    - Implied steady state values: Real foreign interest rate 8.23170; Real domestic interest rate 18.65478; GDP growth 3.86027.
- Baseline parameters used in simulations (examples shown for Argentina, Brazil, Mexico, South Africa, Turkey across different estimation strategies: Fixed effect-IV for output gap only; system GMM no fixed effect; non-linear model). Example baseline parameter set (system GMM, no fixed effect):
  - Starting debt level (end-2004) for Argentina: 0.84142; Brazil: 0.5454; Mexico: 0.465; South Africa: 0.357; Turkey: 0.635.
  - Fiscal behavior (system-GMM non-linear): Constant -0.03628; Positive Output gap coefficient 0.18100; Negative Output gap coefficient 0.26800; Debt coefficient 0.09700; Spline coefficient -0.06200.
  - Conditional variance of primary surplus examples: Argentina 0.00010; Brazil 0.00017; Mexico 0.00090; South Africa 0.00025; Turkey 0.00094.
- The appendix includes a full table describing the sustainability-indicator value surface (probability of decline in column vs absence of upside risk in row), with shaded entries below 0.4.

*Italicized source attribution: Content unit from the IMF PDF chapter provided in the source content.*

### REFERENCES

### REFERENCES

### Journal articles, working papers, and chapters
- Abiad, Abdul, and Jonathan D. Ostry, 2005, “Primary Surpluses and Sustainable Debt Levels in Emerging Market Countries,” IMF Policy Discussion Paper 05/6 (Washington: International Monetary Fund).
- Anderson, Theodore W., and Cheng Hsiao, 1981, “Formulation and Estimation of Dynamic Models Using Panel Data,” Journal of Econometrics, Vol. 18, pp. 67–82.
- Anderson, Theodore W., and H. Rubin, 1949, “Estimators of the Parameters of a Single Equation in a Complete Set of Stochastic Equations,” Annals of Mathematical Statistics, Vol. 21, pp. 570–82.
- Arellano, Manuel, and Stephen R. Bond, 1991, “Some Specification Tests for Panel Data: Monte Carlo Evidence and an Application to Employment Equations,” Review of Economic Studies, Vol. 58, pp. 277–98.
- Balassone, Fabrizio, and Maura Francese, 2004, “Cyclical Asymmetry in Fiscal Policy, Debt Accumulation and the Treaty of Maastricht,” Temi di Discussione No. 531 (Rome: Banca d’Italia).
- Blundell, Richard, and Stephen R. Bond, 1998, “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models,” Journal of Econometrics, Vol. 87, pp. 115–43.
- Bohn, Henning, 1998, “The Behavior of U.S. Public Debt and Deficits,” Quarterly Journal of Economics, Vol. 113 (August), pp. 949–63.
- Bond, Steven, 2002, “Dynamic Panel Data Models: A Guide to Microdata Methods and Practice,” Cemmap Working Paper Series No. CWP09/02 (London: Institute of Fiscal Studies).
- Caprio, Gerard, Daniela Klingebiel, Luc Laeven, and Guillermo Noguera, 2005, "Banking Crisis Database" in Systemic Financial Distress: Containment and Resolution, ed. by Patrick Honohan and Luc Laeven (New York: Cambridge University Press).
- Celasun, Oya, and Joong Shik Kang, 2005, “Estimating Fiscal Reactions: A Monte Carlo Study” (unpublished; Washington: International Monetary Fund).
- Cragg, J.G., and S.G. Donald, 1993, “Testing Identifiability and Specification in Instrumental Variable Models,” Econometric Theory, Vol. 9, pp. 222–40.
- Debrun, Xavier, 2005a, “Public Debt in South Africa: A Risk Analysis,” in South Africa—Selected Issues, IMF Country Report 05/345 (Washington: International Monetary Fund).
- Debrun, Xavier, 2005b, “Lessons from an Empirical Model of Fiscal Policy in Emerging Markets” in Moghadam, Reza, and others, Turkey at the Crossroads—From Crisis Resolution to EU Accession, IMF Occasional Paper No. 242 (Washington: International Monetary Fund).
- Fatàs, Antonio, and Illian Mihov, 2003, “On Constraining Fiscal Policy Discretion in EMU,” Oxford Review of Economic Policy, Vol. 19, No. 1, pp. 1–28.
- Fabrizio, Stefania, and Ashok Mody, 2005, “Can Budget Institutionas Counteract Political Indiscipline?” (unpublished; Washington: International Monetary Fund).
- Favero, Carlo, 2002, “How Do European Monetary and Fiscal Authorities Behave?” CEPR Discussion Paper No. 3426 (London: Center for Economic Policy Research).
- Galí, Jordi, and Roberto Perotti, 2003, “Fiscal Policy and Monetary Integration in Europe,” Economic Policy, Vol. 37 (October), pp. 535–72.
- Garcia, Márcio, and Roberto Rigobon, 2005, “A Risk Management Approach to Emerging Markets’ Sovereign Debt Sustainability with an Application to Brazilian Data,” Chapter 5 in Inflation Targeting, Debt, and the Brazilian Experience, 1999 to 2003, ed. by Francesco Giavazzi, Ilan Goldfajn, and Santiago Herrera (Cambridge, Massachusetts: MIT Press).
- Goldfajn, Ilan, 2005, “Comment on Chapter 5,” in Inflation targeting, Debt, and the Brazilian Experience, 1999 to 2003, ed. by Francesco Giavazzi, Ilan Goldfajn, and Santiago Herrera (Cambridge, MA: MIT Press).
- International Monetary Fund, 2003, “Public Debt in Emerging Markets: Is it too High?” Chapter III in World Economic Outlook, September (Washington: International Monetary Fund).
- International Monetary Fund, 2004, “Has Fiscal Behavior Changed Under the European Economic and Monetary Union?” in Chapter II of World Economic Outlook, September (Washington: International Monetary Fund).
- Judson, Ruth A., and Ann L. Owen, 1999, “Estimating Panel Data Models: A Guide for Macroeconomists,” Economics Letters, Vol. 65, pp. 9–15.
- Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos Végh, 2004, “When It Rains, It Pours: Procyclical Macropolicies and Capital Flows, NBER Macroeconomics Annual 2004.
- Kiviet, Jan F., 1995, “On Bias, Inconsistency, and Efficiency of Various Estimators in Dynamic Panel Data Models,” Journal of Econometrics, Vol. 68, pp. 53–78.
- Mélitz, Jacques, 1997, “Some Cross-country Evidence About Debts, Deficits and the Behaviour of Monetary and Fiscal Authorities," CEPR Discussion Paper No. 1653 (London: Centre for Economic and Policy Research).
- Mendoza, Enrique G., and P. Marcelo Oviedo, 2004, “Public Debt, Fiscal Solvency and Macroeconomic Uncertainty in Latin America: The Cases of Brazil, Colombia, Costa Rica and Mexico,” NBER Working Paper, No. 10637 (Cambridge, Massachusetts: National Bureau of Economic Research).
- Nickell, Steven, 1981, “Biases in Dynamic Models with Fixed Effects,” Econometrica, Vol. 49, pp. 1417–26.
- Stock, James H., and Moto Yogo, 2005, “Testing for Weak Instruments in Linear IV Regressions,” Chapter 5 in Identification and Inference for Econometric Models: A Festschrift in Honor of Thomas Rothenberg ed. by J.H. Stock and D.W.K Andrews (Cambridge, United Kingdom: Cambridge University Press).
- Penalver, Adrian, and Gregory Thwaites, 2004, “Analysing Sovereign Debt Sustainability: A New Probabilistic Approach” (unpublished; London: Bank of England).
- Tanner, Evan and Issouf Samake, 2005, “The Shocking Facts: Public Debt Sustainability in Brazil, Mexico and Turkey” (unpublished; Washington: International Monetary Fund).
- Wyplosz, Charles, 2005, “Institutions for Debt Sustainability in Brazil,” Chapter 6 in Inflation Targeting, Debt, and the Brazilian Experience, 1999 to 2003, ed. by Francesco Giavazzi, Ilan Goldfajn, and Santiago Herrera (Cambridge, Massachusetts: MIT Press).

*Content unit: _wp0667 - REFERENCES (source PDF: _wp0667 - REFERENCES)*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0667.pdf_