## _wp0762

## Source details

**Canonical URL:** [_wp0762](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0762.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0762.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0762.pdf.json)

---

### Introduction: purpose and empirical focus
- Objective: assess how the quantity of government debt affects the level of the short-term interest rate consistent with the central bank’s monetary policy objectives, and whether the speed of adjustment to the targeted inflation rate is altered when government debt (relative to GDP) exceeds a threshold.
- Empirical focus: Canada during its 1990s inflation-targeting regime.
- Methodology overview:
  - Extend Clarida, Gali and Gertler (1999) (CGG) framework to incorporate a debt-constraint.
  - Derive an augmented forward-looking interest rate rule under discretion and compare baseline and debt-constrained rules.
  - Estimate a threshold debt-GDP level beyond which debt acts as a constraint on policy.

### Key assumptions and transmission mechanism
- Central bank minimizes quadratic loss function in expected inflation and output variability (CGG framework); inflation target normalized to zero; α denotes relative weight on output.
- Crucial fiscal assumption change: when government debt reaches a certain level, fiscal authorities are expected to reduce government spending (rather than raise net taxes) to meet debt service costs.
- No seigniorage revenues play a role.
- Mechanism: the interest-elasticity of the output gap (φ, slope of IS curve) interacts with lagged debt; under high debt the IS curve is altered, changing the responsiveness and speed of monetary policy adjustment to expected inflation and to expected future real interest rates.

### Theoretical results (debt-constrained CGG extension)
- Baseline discretionary optimal rule (CGG): it+1 = βπ E_t π_{t+1} + ... with β_π > 1 under baseline (equation (4) in source).
- Debt-constrained IS introduces additional terms for lagged debt, future expected debt servicing costs, and interactions between debt and output growth (equation (7) in source).
- Debt-constrained optimal interest-rate rule (equation (8) in source) features:
  - Dependence on expected inflation and lagged inflation (more gradual adjustment).
  - New dependence on expected future short-term real interest rate r^e_{t+1}.
  - Coefficients that vary with present or past debt level; if b ≤ b* rule reduces to baseline.
- Interest-rate response to expected inflation under high debt:
  - If b > b*, the response to expected inflation can be smaller than baseline (i.e., 'π_β can be lower than π_β), especially when the central bank places a low relative weight on output (inflation-targeting, low α).
  - High debt can thus produce weaker nominal-rate responses to a given expected inflation shock; policy must move more gradually because the IS constraint is altered.
- Responses to expected real rates and lagged inflation:
  - Sensitivity to expected future real interest rates is positive under high debt: higher r^e_{t+1} calls for higher target nominal rates.
  - Central bank pays attention to lagged inflation beyond forecasting role when debts are high.

### Empirical strategy (Canada case)
- Estimation approach:
  - Threshold regression estimated by iterative dummy-variable GMM: DUM = 1 if debt-GDP ratio > estimated threshold b*, 0 otherwise.
  - Grid search over debt-GDP ratio from 46 percent to 52 percent in 0.1 percentage point steps; threshold chosen to minimize GMM objective (lowest J-statistic).
  - Hansen’s J-statistic used to test over-identifying restrictions; orthogonality E[w_t v_t] = 0 exploited.
- Data (Canada):
  - Monthly sample: 1991:11 to 2000:12.
  - Key series: Overnight Rate (policy instrument i), CPI inflation π, core inflation (instrument), output gap x (linear detrend of log IIP), total interest-bearing marketable federal debt/GDP (for DUM), long-term indexed bond rate (proxy for expected future short-term real interest rate), exchange rate (C$/USD) annual percent change.
  - Expected inflation horizon chosen: n = 12 months (12-month ahead).

### Main empirical results and key statistics
- Threshold estimate:
  - Threshold debt-GDP ratio = 50.2 percent.
  - Debt was above this level from June 1993 through May 1997 in the sample.
- GMM and specification diagnostics:
  - J-statistic: baseline = 0.18; debt-constrained = 0.17.
  - p-values for J-statistic: baseline = 0.96; debt-constrained = 0.99 (cannot reject validity of instruments).
- Smoothing / adjustment:
  - Smoothing parameter estimate = 0.93.
- Expected-inflation coefficients:
  - Baseline πβ = 0.76.
  - Debt-constrained πβ = 1.69.
- Debt impact on inflation response:
  - Estimated 1θ = -0.99 (i.e., for every percentage point increase in expected inflation, the debt-constrained interest-rate response is 0.99 percentage point less than during normal times).
  - Overall debt-constrained effect reported as 0.70 (= πβ - 1θ) percentage point increase in the overnight rate.
  - Baseline overall effect: 0.76 percentage point.
- Interaction with indexed bond rate:
  - A simultaneous 1 percentage point increase in 12-month ahead inflation expectation and a 50-basis point increase in the indexed bond rate produces an additional high-debt effect of 13(/2)θθ+ = 0.63 percentage point below normal (as reported).
- Selected parameter estimates (from Table 1 as presented):
  - Baseline: α = 2.81 (0.75); πβ = 0.76 (0.21); μ = 0.07 (0.02); 1ψ = 1.24 (0.04); 2ψ = -0.31 (0.04).
  - Debt-constrained: α = 5.44 (3.42); πβ = 1.69 (0.28); 1θ = -0.99 (0.45); γ = -0.40 (0.33); 2θ = 0.23 (0.43); σ = -1.13 (0.78); 3θ = 0.72 (0.28); μ = 0.05 (0.02); 1ψ = 0.96 (0.02); 2ψ = -0.06 (0.02).
  - Note: Standard errors in parentheses; statistically significant estimates at 5% are indicated in the source table.

### Empirical interpretation and substantive findings
- Evidence of a debt-constraint:
  - Significant evidence that monetary policy faced a debt-constraint when debt > 50.2 percent of GDP.
  - During low-debt periods the Bank of Canada followed a strong “lean against the wind” strategy (πβ estimate substantially above 1 in debt-constrained specification).
  - During high-debt periods, the interest-rate response to expected inflation was constrained: for each 1 percentage point increase in expected inflation, the effective interest-rate response was 0.99 percentage point below normal.
- Policy smoothing:
  - Nominal interest-rate adjustment is highly smoothed (smoothing parameter = 0.93).
  - Baseline πβ = 0.76 (< 1) could be interpreted as accommodative, but the threshold-GMM results indicate this may instead reflect parameter shifts when debt is high.
- Role of other variables:
  - No evidence that past inflation (beyond its forecasting role) was used in policy formulation in the sample.
  - Long-term indexed bond rate (proxy for expected future short-term real rate) positively affects the overnight rate.
  - No separate response to U.S. Federal Funds Rate and output gap beyond their roles in predicting inflation.
- Cycle dependence:
  - Constraining effects of high debt differ across business cycles: larger constraint during recessions than booms, and depend on the level of the equilibrium real interest rate (proxied by expected future short-term rate).
- Historical implication for Canada:
  - During the high-debt period in the 1990s, inflation and the output gap were low (calling for loosening), but debt-constraints prevented sufficient loosening—policy was less loose than inflation indicators called for.

### Policy implications
- Fiscal policy that raises public debt can “constrain” monetary authorities by altering the IS curve and weakening the nominal-rate response to expected inflation.
- This constraint is relevant for developing and emerging economies adopting inflation-targeting without concomitant public debt reduction, and can also affect industrialized countries if public debt crosses a critical threshold.
- Central banks may need to account for interactions with fiscal expectations (e.g., expected government spending adjustments) when setting policy rules.

### Limitations and suggested extensions
- The paper modifies a key CGG assumption but does not propose a fully-specified fiscal-monetary model that checks Blanchard-Kahn conditions for a unique saddle-path equilibrium.
- The threshold estimate is based on minimizing the GMM J-statistic; proper confidence intervals for the threshold were not computed and would strengthen inference.
- Additional empirical work using historical episodes from other countries with inflation targeting and high debt (e.g., United Kingdom) could bolster the findings.

*Source: _wp0762 - Box 1 and selected sections (IMF working paper content provided).*

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

### _wp0762 - References..............................................................................................................

### Introduction: purpose and questions
- Objective: assess how the quantity of government debt affects the level of the short-term interest rate consistent with the central bank’s monetary policy objectives, and whether the speed of adjustment to the targeted inflation rate is altered when government debt (relative to GDP) exceeds a threshold.
- Empirical focus: Canada during its 1990s inflation-targeting regime.
- Methodology overview:
  - Extend Clarida, Gali and Gertler (1999) (CGG) framework to incorporate a debt-constraint.
  - Derive an augmented forward-looking interest rate rule under discretion, comparing a baseline rule and a rule when a debt-constraint operates (two-step procedure).
  - Estimate a threshold level of debt beyond which debt acts as a constraint on policy.

### Key assumptions and mechanisms
- Central bank loss function minimized subject to expectation IS and Aggregate Supply functions (CGG framework).
- No seigniorage revenues play a role in the analysis.
- Crucial assumption change: expected path of government spending is altered so that when government debt reaches a certain level, fiscal authorities are expected to reduce government spending (rather than raise net taxes) in line with debt service costs.
- Transmission mechanism implication: the interest-elasticity of the output gap (the slope of the IS curve) changes under the debt-constraint, affecting the speed of adjustment and the responsiveness of the interest rate to expected inflation.
- Policy coordination context referenced: active monetary/passive fiscal versus passive monetary/active fiscal (Leeper, 1991) and the forward-looking Taylor rule from CGG which assumes fiscal behavior that supports prevailing monetary policy.

### Literature and motivating evidence
- Seigniorage channel: historically relevant in extreme cases (example: post-war Germany hyperinflation used to wipe out debt); generally negligible in highly industrialized countries.
- Interest-rate channel: higher interest rates increase debt service; example cited (Dornbusch, 1998): a percentage point increase in the interest rate amounts to roughly $30 billion in extra deficit, or half a percentage point of GDP.
- Historical episode: mid-1930s to mid-1950s bond yields were kept at or below 2.5 percent to accommodate high debts.
- Policy thresholds mentioned:
  - Canada’s federal debt rose to above 50 percent of GDP in the early 1990s (context for Bank of Canada concerns cited from Bernanke, Mishkin et al (1999), pg.137; Annual Report, 1993 released March 1994).
  - Stability and Growth Pact reference: concern when debt beyond 60 percent of GDP for euro-area countries.
- Fiscal theory of price level determination references (Leeper, 1989; Sims, 1994; Woodford, 1998, 2001): quantity of debt can affect the price level via wealth effects when primary surplus is exogenous; maturity structure matters (shorter maturity implies larger inflation needed to reduce real debt value).

### Main findings (as stated)
- A high government debt level could constrain monetary policy if government spending—rather than taxes—is expected to adjust in the future in line with debt service costs. The constraint operates through changes in the policy transmission mechanism that alter the nature of the IS curve.
- The effects of the debt-constraint on monetary policy differ across the business cycle: the constraint is larger during a recession than during a boom.
- Empirical estimates indicate that Canadian monetary policy was constrained by a high government debt-GDP ratio during the 1990s: policy was less loose than inflation indicators called for.

### Empirical approach highlights
- Formulate an econometric model based on the modified IS/Aggregate Supply implications to estimate the threshold debt level.
- Use monthly Canadian data (1991:11 to 2000:12) in the empirical application (as indicated by figures and tables headings).
- J-Statistics and GMM estimates are used for choosing the threshold level of debt and estimating parameters of baseline and debt-constrained models (table and figure references in the source).

*Source: _wp0762 - References..............................................................................................................*

### Box 1. The optimal interest rate rule under discretionary monetary policy (Clarida, Gali, Gertler, 1999)

### Box 1. The optimal interest rate rule under discretionary monetary policy (Clarida, Gali, Gertler, 1999)

### Central bank objective and baseline setup
- Central bank objective: minimize quadratic loss function increasing in present and future expected inflation and output variability (equation (1)).
- Loss function (as written): 22 1 (1)max[] 2 tt xαπ−+ .
- Variables and parameters:
  - Output gap: x ≡ y − z (log of output y from trend z).
  - Inflation target normalized to zero.
  - α: relative weight on output in the objective function; weight on inflation normalized to 1.

### Structural model (state of the economy)
- Expectational IS curve (equation (2)):
  - Relates current output gap, t x, to expected real interest rate and future expected output gap.
  - Negative relationship between real interest rate and output gap; consumption smoothing links future expected output to current consumption.
- Phillips curve with Calvo pricing (equation (3)):
  - Current inflation depends on current and future expected output and inflation.
  - Cost-push disturbance u_t is autoregressive; û_t and ε̂_t are i.i.d. zero-mean disturbances.

### Baseline optimal interest rate rule (discretionary policy)
- CGG two-step solution: maximize (1) subject to (3) and substitute FOCs into IS (2) to derive target interest rate rule.
- Baseline (target) interest rate rule (equation (4)):
  - (4) 1 1 ttt t iE π β πε φ + = + 
  - β_π > 1 (noted): an increase in expected inflation calls for short-term nominal rates to be moved up more than proportionately.

### Incorporating government debt concerns: rationale and mechanism
- Motivation: high government debt may constrain central bank ability to use nominal interest rate to fight inflation.
- Government consumption G_t enters IS error term due to derivation from consumption Euler equation:
  - (5) { } 11 Ezg t tt t ε ≡ ∆ − ∆ ++  where g_t ≡ log(G_t−1 / Y_t−1).
- Government budget constraint (ratio to output, identity):
  - Relates growth of government debt to interest obligations and primary deficit; variables include debt-to-GDP ratio b_t, real interest rate r_t, growth rate of output ŷ, government purchases-to-GDP, and net tax (tx − tr)/Y_t.
- Fiscal regime with high debt:
  - When debt-GDP ratio above threshold b*, expected change in government consumption depends on debt level, growth, and servicing costs because net taxes are not expected to absorb debt shocks entirely.
  - Government spending expected to fall proportionally to expected debt-service payments; if future interest rates expected lower, spending need not fall as much.

### Debt-constrained IS and augmented dynamics
- New IS with debt-constraints (equation (7)):
  - (7) ' 1111112111 ttttttttttttttttt xbiEExEbiExbxbφ φπ φπφε −++ ++ +− = − + − ++ − − − ∆ −∆ + 
  - Features:
    - Real interest elasticity φ interacts with lagged debt 1−t b, reinforcing higher interest rate effect on output gap.
    - Additional terms: future expected debt servicing costs and interactions between debt and growth in output gap.
    - Error term 'ε is AR and assumed to share same AR coefficient as ε in (2).

### Debt-constrained optimal interest rate rule
- Re-solving with unchanged objective (1), PC (3), and new IS (7) yields (equation (8)):
  - (8) ' ' 111 111 111111 e ttt ttttt ttt bb iEr bbb π φφ ε λ βπ π φφαφφφφ − ++ − − − ⎛⎞ = + ++ ⎜⎟ +++ ⎝⎠
  - Auxiliary definitions included for composite coefficients (preserved as in source).
- Notable properties:
  - If b ≤ b* (1 0 φ = ), rule reduces to baseline CGG (equation (4)).
  - Policy depends on expected inflation and lagged inflation (central bank looks at past inflation beyond forecasting role—more gradual adjustment).
  - New independent variable: future expected short-term real interest rate r^e_{t+1}; a rise in r^e_{t+1}, ceteris paribus, calls for a rise in target nominal rate.
  - Coefficients vary with present or past debt level.

### Interest rate response to expected inflation under high debt
- For debts high and near maximum level b, equation (8) can be rewritten (equation (9)) and the interest rate response to E_{t+1} π is given by (10):
  - (10) ' 1 11 (() 2) 1 b bb πππ φφφραλλ βββ φφφφαρφ ⎛⎞ ⎛⎞ +− ≡ +− + ⎜⎟ ⎜⎟ ++ ⎝⎠ ⎝⎠
- Sign and magnitude considerations:
  - If b ≤ b*, no debt attention and baseline response applies.
  - If b > b*, then (1 1 b φ φφ − +) < 0 given 0 > φ, making the second term of (10) negative.
  - Third term sign ambiguous unless assumptions on α and λ:
    - If λ α ≤ (low weight on output in the objective, i.e., central bank tough on inflation), third term negative ⇒ overall 'π_β < π_β.
    - If λ α >, ambiguity remains.
  - 'π_β decreases if b increases and λ α ≤ (partial derivative analysis shown; condition leads to < 0).
- Policy interpretation:
  - In an inflation-targeting regime (low α), policy responds less to expected inflation when debts are high and this response falls as debts rise.
  - This does not necessarily mean policy is accommodating; rather, a given expected inflation requires a more gradual response due to extra IS-curve constraints—central bankers feel constrained by high and growing debts.

### Responses to expected real rates and lagged inflation under high debt
- Sensitivity to expected future real interest rates (second term in (9)) is positive:
  - If debts are high, an increase in r^e_{t+1} leads to raising the interest rate regardless of cycle.
  - High debts can constrain monetary policy when recessionary pressures call for rate cuts.
- Response to past inflation (third term in (9)) is positive when debts are high:
  - Central bank pays attention to lagged inflation beyond forecasting role, producing more gradual adjustment.

### Empirical strategy: joint estimation of threshold and policy rule (Canada)
- Estimation approach:
  - Threshold regression model to choose threshold debt level and estimate debt-constrained rule jointly.
  - Iterative dummy-variable approach estimated by generalized method of moments (GMM).
- Econometric specification (as written, equations (11)–(14)):
  - Policy instrument: target short-term interest rate T_t i responds to n-period-ahead expected inflation, r^e_{t+1}, and other variables z in information set Ω_t.
  - Dummy DUM = 1 if debt-GDP ratio > estimated threshold b*, 0 otherwise.
  - Expected ranges for coefficients: 0,0,0,0,0,0,1 321 >><>==> θ θ θ μ σ γ β π .
  - Realized n-period-ahead inflation used to eliminate unobserved expectation term; target-to-actual interest rate adjustment modeled with smoothing parameters ψ1 and ψ2.
  - Orthogonality conditions E[w_t v_t] = 0 exploited for GMM; Hansen’s J-statistic used to test over-identifying restrictions.
- Threshold search:
  - Grid search over debt-GDP ratio from 46 percent to 52 percent, step size 0.1 percentage point.
  - Threshold chosen to minimize GMM objective (lowest J-statistic).

### Data used for Canada case
- Sample: monthly data from 1991:11 to 2000:12 (first decade of inflation targeting in Canada).
- Key series and sources:
  - Overnight Rate: short-term interest rate instrument (i).
  - Consumer Price Index (CPI): yearly rate of inflation π.
  - Core inflation rate: used as instrument to predict future inflation.
  - Output gap x: calculated by linearly de-trending log of Index of Industrial Production (IIP).
  - Total interest-bearing marketable federal debt divided by GDP → debt-GDP ratio (used to define DUM; not used directly in estimation).
  - Expected future short-term real interest rate proxied by long-term indexed bond rate.
  - Expected inflation series attempted via long-term nominal minus indexed bond yields—failed to provide convincing results; chosen horizon n = 12 (12 months) for expected inflation.
  - Exchange rate (C$/USD) annual percentage change used as element of z.
- Stationarity: monthly series for inflation rate, output gap, and expected real interest rate are stationary for the short sample.

### Estimation results (Canada)
- Grid search outcome:
  - J-statistic–minimizing threshold level of debt estimated to be 50.2 percent of GDP.
- Debt incidence in sample:
  - Debt was above this level from June 1993 through May

*Source: Box 1, "_wp0762 - Box 1. The optimal interest rate rule under discretionary monetary policy (Clarida, Gali, Gertler, 1999)" (PDF).*

### 1997.  The  standard  error  bands  around  a  Hodrick-Prescott  ‘trend’  is  also  shown  in  the

### _wp0762 - 1997.  The  standard  error  bands  around  a  Hodrick-Prescott  ‘trend’  is  also  shown  in  the

### Estimation approach and instruments
- GMM estimation used; parameter estimates reported in Table 1.
- Instruments (first through sixth, ninth and twelfth lags) for: overnight interest rate, the core inflation rate, interest rate on the long-term indexed bond, output gap, and the Fed Funds Rate.
- For the debt-constrained model the same set of instruments was used except for lags of the indexed bond rate.
- Over-identifying restrictions: p-value for the J-statistic = 0.96 for the baseline model and 0.99 for the debt-constrained model; cannot reject instrument validity.
- Note: The level of the nominal exchange rate is non-stationary.

### Key parameter estimates and statistics (GMM, Canada, monthly data, 1991:11 to 2000:12)
- Adjustment / smoothing:
  - Smoothing parameter estimate: 0.93.
- Expected-inflation coefficient:
  - Baseline πβ = 0.76.
  - Debt-constrained πβ = 1.69.
- Debt-threshold:
  - Threshold level of debt-GDP ratio = 50.2 percent.
- Debt impact on inflation response:
  - Estimated 1θ = -0.99 (for every percentage point increase in expected inflation, the debt-constrained interest rate response is 0.99 percentage point less than during normal times).
  - Overall effect during high debt levels: 0.70 (= πβ - 1θ) percentage point increase in the overnight rate for the debt-constrained model.
  - Baseline overall effect: 0.76 percentage point.
- Interaction with indexed bond rate:
  - If there is a simultaneous 1 percentage point increase in the 12-month ahead inflation expectation and a 50-basis point increase in the rate on indexed bond, the additional effect of high debts on the overnight rate is 13(/2)θθ+ 0.63 percentage point below normal.
- J-statistics and related figures:
  - J-statistic values: 0.18 (baseline), 0.17 (debt-constrained).
  - p-values for J-statistic: 0.96 (baseline), 0.99 (debt-constrained).
  - Figure note: The light broken lines are ±2*S.E. bands around a Hodrick-Prescott trend (dark broken line).

### Table 1 — Parameter estimates (selected, as presented)
- Baseline Estimates:
  - α = 2.81 (0.75)
  - πβ = 0.76 (0.21)
  - μ = 0.07 (0.02)
  - 1ψ = 1.24 (0.04)
  - 2ψ = -0.31 (0.04)
  - Jstatistic− = 0.18
  - p-value for J-statistic = 0.96
- Debt-constrained Estimates:
  - α = 5.44 (3.42)
  - πβ = 1.69 (0.28)
  - 1θ = -0.99 (0.45)
  - γ = -0.40 (0.33)
  - 2θ = 0.23 (0.43)
  - σ = -1.13 (0.78)
  - 3θ = 0.72 (0.28)
  - μ = 0.05 (0.02)
  - 1ψ = 0.96 (0.02)
  - 2ψ = -0.06 (0.02)
  - Jstatistic− = 0.17
  - p-value for J-statistic = 0.99
- Note from Table: Statistically significant (at 5% level) estimates are in bold. Standard errors (in parentheses) are adjusted from the ones derived from estimation of (9”) by the delta-method.

### Empirical findings and interpretation
- Presence of a debt-constraint:
  - Significant evidence of debt-constraint when debt > 50.2 percent of GDP.
  - During low-debt periods, the Bank of Canada followed a ‘lean against the wind’ strategy: estimate of πβ = 1.69, substantially above 1.
  - During high-debt periods, the interest-rate response to expected inflation is constrained: for every 1 percentage point increase in expected inflation the interest-rate response is 0.99 percentage point below normal.
- Policy smoothing and accommodation:
  - Nominal interest-rate adjustment is considerably smoothed (smoothing parameter = 0.93).
  - Baseline πβ = 0.76 (considerably less than 1) — could suggest accommodation of inflation, but alternative explanation: parameter break due to high debt levels (thus threshold-GMM used to test debt-constraint).
- Role of other variables:
  - No evidence that past inflation (beyond its role in predicting future inflation) was used in policy formulation.
  - Interest rate on long-term indexed bond (proxy for future expected short-term real rate) positively affects the overnight rate.
  - No separate response of the overnight rate to the U.S. Federal Funds Rate and the output gap, beyond their roles in predicting inflation.
- Historical context:
  - During the high-debt period in the 1990s, inflation and the output gap were low (calling for loosening), but debt-constraints prevented sufficient loosening.

### Main conclusions (V. SUMMARY AND CONCLUSIONS)
- Three main findings:
  1. A high government debt could constrain monetary policy if government spending is expected to adjust in future in line with debt service costs. The constraint operates through an altered policy transmission mechanism (IS curve changes). The policy responds less to expected inflation if debts are high; the same expected inflation requires a more gradual response due to IS-curve constraints.
  2. The constraining effects of high debt on monetary policy differ during booms and recessions; effects depend on the level of the equilibrium real interest rate (proxied by expected future short-term rate).
  3. Empirical estimates suggest Canadian monetary policy might have been constrained by a high government debt-GDP ratio during the 1990s; policy was less loose than indicated by inflation indicators.
- Threshold estimate:
  - Threshold debt-GDP ratio estimated to be 50.2 percent (slightly higher than 48 percent referenced by the Bank of Canada governor as a warning level).
- Policy implications:
  - Fiscal decisions leading to higher debts can “constrain” monetary authorities by changing the IS curve; relevant caution for developing and emerging economies moving to inflation-targeting without lowering public debt.
  - Debt-constraints can also affect industrialized countries; even the United States could face a debt-constraint if public debt crosses a certain threshold.

### Limitations and avenues for future work
- The paper modifies an assumption of the CGG model without proposing a fully-specified model or checking the Blanchard-Kahn condition for a unique saddle-path equilibrium.
- The threshold estimate is based on minimizing the J-statistic of the GMM regressions; proper confidence intervals for the threshold could not be computed and would enhance analysis quality.
- Historical episodes from other countries (e.g., the United Kingdom) with inflation targeting and high government debt could strengthen empirical results.

*Source: IMF working paper content (pages and figures as provided in the supplied content).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp0762.pdf_