## _wp1036

## Source details

**Canonical URL:** [_wp1036](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1036.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1036.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1036.pdf.json)

---

### I. Introduction
- Purpose: Examine monetary policy transmission in Mauritius using a VAR framework.
- Approach:
  - Benchmark VAR includes headline consumer price index (CPI).
  - Alternative VAR replaces headline CPI with core CPI (nets out administrative prices and energy prices).
  - Policy instruments examined: official Bank of Mauritius (BOM) interest rate (repo rate), nominal effective exchange rate (NEER), and money supply (M2).
  - Two VAR identification methods compared: recursive (Cholesky) and structural.
- Main high-level finding: Overall monetary transmission is weak, particularly for output. Differences exist between headline and core CPI models.

### II. Background and stylized facts
- Institutional and policy context:
  - BoM statutory objective: “to maintain price stability and to promote orderly and balanced economic development” (BoM Act 2004).
  - Evolution: direct monetary controls → reserve money target → Lombard rate in 1999 → repo rate introduced in December 2006 (repo replaced Lombard rate).
  - Monetary Policy Committee (MPC) formulates policy since April 2007.
- Recent developments and stylized facts:
  - Inflation: eased from an average of 8 percent in the 1990s to about 5 percent in the last five years; core inflation volatility has significantly declined.
  - Real GDP growth: averaged more than 5 percent in 1996-2008.
  - Policy actions: BoM reduced its policy rate by 250 basis points to 5.75.
  - Empirical staff estimate of policy reaction function includes a tolerance level of inflation estimated at 4.7 percent a year.
  - Exchange rate: reclassified from managed float to free float in AREAR classification; nominal effective exchange rate (NEER) depreciated about 4 percent since January 2009 in nominal effective terms while remaining relatively stable against the U.S. dollar.
- Behavioral patterns:
  - Looser monetary stance tends to be followed by higher inflation for most of the sample period (exceptions: 2003, 2006-07 and the recent period).
  - Core inflation does not seem associated with changes to the monetary policy stance.
  - Exchange rate depreciations are associated with higher core and headline inflation.
  - Growth outcomes depend heavily on the global economy (notably the EU); no clear strong relationship between short-term interest rates and real output over the sample.

### III. Empirical approach
- Data and variables:
  - Monthly data between 1999 Q1 and 2009 Q3.
  - Endogenous variables (tY): Real GDP (lgdp), Consumer price index (lcpi) [headline or core], Money stock (lm2), Repo rate (repo), Nominal effective exchange rate (lneer).
  - Exogenous variables (tX): U.S. Federal Funds rate (ffr), U.S. real GDP (lgdp_us).
  - All series in logarithms and seasonally adjusted except the repo rate.
- Unit roots and cointegration:
  - Unit root tests indicate all series are I(1); differences are I(0).
  - Standard cointegration tests indicate evidence of at least two cointegrating relationships.
  - Given short sample and to avoid imposing theoretical long-run restrictions, VARs are estimated in levels (no cointegration rank imposed).
- VAR specification and estimation:
  - VARs estimated with constant and no trend.
  - Two models estimated: (i) benchmark with headline CPI; (ii) alternative with core CPI.
  - Lag length criteria indicate the use of 2 lags for both models.
  - Residual diagnostic tests suggest well-behaved residuals.
  - Recursive estimation and Chow tests indicate VAR stability at the 1 percent significance level.
- Identification methods:
  - I - Recursive VAR (Cholesky): lower-triangular contemporaneous matrix; ordering matters and implies specific contemporaneous causality assumptions.
  - II - Structural VAR: relaxes recursive assumptions and imposes contemporaneous restrictions motivated by money demand and policy reaction considerations (permits monetary policy to respond contemporaneously to money demand and NEER but not to contemporaneous output and price shocks).

### IV. Estimation results and key findings
- General inference setup:
  - Impulse responses computed for a one-standard-deviation monetary policy shock (unexpected temporary rise in the repo rate) on output, prices, repo rate, money supply, and NEER.
  - Forecast error variance decompositions used to assess the relative importance of monetary policy shocks for fluctuations in each variable.
  - Both benchmark (headline CPI) and alternative (core CPI) models estimated under both identification schemes.
- Principal findings:
  - Overall conclusion: transmission channel is weak, particularly for output.
  - Headline CPI model:
    - Changes in the repo rate result in small and short-lived (albeit statistically significant) responses of the headline CPI and to a lesser extent output.
    - Exchange rate shocks transmit to the headline CPI and output in a statistically significant way and tend to last longer than repo rate shocks.
    - Some evidence of a very short-lived transmission of money supply shocks to output (but not CPI).
  - Core CPI model:
    - No transmission of repo rate shocks to output or core CPI.
    - Core CPI exhibits a quick response to exchange rate shocks.
    - Evidence of a statistically significant transmission of money supply shocks to both output and inflation (and to the exchange rate).
  - Identification sensitivity:
    - Results depend on identification; both recursive and structural identifications are examined to assess robustness.
- Policy implications preserved from the source:
  - Different monetary policy “rules” could be appropriate depending on which CPI is targeted:
    - For headline CPI, where the interest rate channel is stronger, “Taylor-type” rules may be more applicable.
    - For core CPI, where money supply effects are more important, “McCallum-type” rules that target money supply could be more appropriate.
  - Weak transmission to output suggests addressing structural rigidities in the financial system and regulatory framework that may hamper transmission to the real sector.
  - Further work recommended: improve model specifications for the Mauritian economy, consider alternative structural identifications, and account for data limitations given the short analysis period.

### V. Impulse responses and variance decomposition (headline CPI model)
- Repo-rate (monetary policy) shock (one-standard-deviation, positive):
  - Lowers output and inflation; effects on inflation are more persistent and more statistically significant.
  - Output returns to pre-shock levels by the end of the 6th period; inflation remains below pre-shock value for more than 10 periods (with some effects marginally insignificant).
  - Structural identification indicates faster transmission: statistically significant effects on output and inflation appear almost immediately and disappear after period 3.
  - Variance decomposition: a repo-rate shock accounts for about 2-8 percent (on average) of fluctuations in both prices and output.
- Unexpected nominal appreciation (exchange rate shock, positive):
  - Lowers both output and inflation; effects on inflation remain statistically significant for more than 10 quarters.
  - Both effects appear about 4 periods after the shock; output returns to pre-shock levels almost immediately.
  - Variance decomposition: exchange rate shocks account for a small proportion of fluctuations in output and prices (about 12 percent of price fluctuations on average in the core-CPI model; for headline CPI the exchange-rate share is described as small).
  - Patterns suggest a slow and persistent pass-through of exchange rate changes to prices.
- Money supply (M2) shock (positive):
  - Translates into a persistent increase in inflation lasting for more than 10 periods (effects not always statistically significant).
  - Produces a short-lived increase in output which disappears by period 3.
  - No statistically significant change in NEER following the shock to M2 in the headline-CPI recursive specification; in the core-CPI model, a monetary shock elicits a statistically significant response of NEER.
  - Variance decomposition (core-CPI model): a money supply shock accounts for about 10-11 percent of fluctuations in output and prices, and about 7 percent of fluctuations in the nominal exchange rate.
- Overall interpretation:
  - Shocks to the repo rate, exchange rate, and money supply have small and statistically significant transmission to prices and output, particularly for inflation.
  - Transmission is stronger for nominal variables (inflation) than for real variables (output), consistent with patterns in transition economies.
  - Weak transmission to real variables may reflect financial market bottlenecks and structural problems (financial frictions, rigidities) rather than pure long-run money neutrality.

### VI. Model diagnostics and forecasting (headline CPI model)
- Model fit and stability:
  - Dynamic within- and out-of-sample forecasts for year-on-year inflation perform quite well.
  - Chow tests indicate model stability and parameter constancy at the 1 percent significance level.
- Out-of-sample forecast performance:
  - The model performs extremely well up to 2008Q4, but overestimates inflation for the three quarters of 2009 by about 1.5-2 percent.
  - Dynamic out-of-sample forecast estimated using a VAR from 1999 Q1 to 2008Q1. Static one-step-ahead forecasting yields an even closer fit.
- Forecast for 2009 Q4 to 2010 Q3:
  - After a small further drop in inflation in 2009 Q4, inflation resumes to the range of 3 to 4 percent in 2010.

### VII. Alternative model (core CPI)
- Impulse responses:
  - A monetary policy shock has no statistically significant effect on output or inflation; both fluctuate around pre-shock values.
  - A positive shock to NEER has a small, statistically significant negative effect on core CPI that lasts two periods after the shock; transmission is shorter and not persistent compared to headline CPI.
  - No statistically significant effect of an exchange rate shock on output.
  - A positive money supply shock increases both output and prices; effects persist (though not always statistically significant), and NEER responds statistically significantly to the monetary shock.
  - Structural-model impulse responses give similar conclusions.
- Variance decomposition (core CPI):
  - Repo shock accounts for a negligible percentage (less than 2 percent, on average) of fluctuations in prices and output.
  - Exchange-rate shock accounts on average for about 12 percent of price fluctuations.
  - Money-supply shock accounts for about 10-11 percent of fluctuations in output and prices, and about 7 percent of fluctuations in the nominal exchange rate.
- Forecasts:
  - Dynamic within- and out-of-sample forecasts perform reasonably well but perhaps not as well as headline case.
  - Forecast for 2009 Q4 to 2010 Q3: after a small further drop in 2009 Q4, core inflation resumes to an average of 2.5 percent in 2010.

### VIII. Conclusion and policy considerations
- Summary:
  - The overall monetary policy transmission of an unexpected temporary increase in the repo rate is weak for both headline and core CPI models, particularly for output.
  - Repo-rate shocks, exchange-rate shocks, and money-supply shocks produce statistically significant changes in headline CPI; transmission to output is not consistently significant.
  - With core CPI, exchange-rate and money-supply shocks transmit to prices, while repo-rate shocks do not; transmission of money-supply shocks is stronger in the core-CPI model than in the headline-CPI model.
- Policy implications:
  - Headline CPI (interest-rate channel relatively stronger): “Taylor-type” rules may be more applicable.
  - Core CPI (money-supply effects more pronounced): “McCallum-type” rules that target money supply could be more appropriate.
  - Weak transmission to output suggests the need to address structural rigidities in the financial system and regulatory framework.
- Further work recommended:
  - Improve model specifications to better reflect characteristics of the Mauritian economy.
  - Consider alternative structural identifications beyond the one used here.
  - Be mindful of possible data limitations given the short analysis period.

*Source: _wp1036 - Appendix A: VAR Modeling and Diagnostics (selected content from the provided PDF).*

### Appendix A: VAR Modeling and Diagnostics ........................................................................21

### Appendix A: VAR Modeling and Diagnostics

### Appendix title and location
- Appendix A: VAR Modeling and Diagnostics ........................................................................21

### Adjacent appendix
- Appendix B: Additional Impulse Responses and Variance Decompositions ..........................27

*Source: _wp1036 - Appendix A: VAR Modeling and Diagnostics (page listing from provided PDF content).*

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

### _wp1036 - References  ...............................................................................................................................32

### I. Introduction
- Purpose: Examine monetary policy transmission in Mauritius using a VAR framework.
- Approach:
  - Benchmark VAR includes headline consumer price index (CPI).
  - Alternative VAR replaces headline CPI with core CPI (nets out administrative prices and energy prices).
  - Policy instruments examined: official Bank of Mauritius (BOM) interest rate (repo rate), nominal effective exchange rate (NEER), and money supply (M2).
  - Two VAR identification methods compared: recursive (Cholesky) and structural.
- Main high-level finding: Overall monetary transmission is weak, particularly for output. Differences exist between headline and core CPI models.

### II. Background and stylized facts
- Institutional and policy context:
  - BoM statutory objective: “to maintain price stability and to promote orderly and balanced economic development” (BoM Act 2004).
  - Evolution: direct monetary controls → reserve money target → Lombard rate in 1999 → repo rate introduced in December 2006 (repo replaced Lombard rate).
  - Monetary Policy Committee (MPC) formulates policy since April 2007.
- Recent developments and stylized facts (selected exact figures and facts from the source):
  - Inflation: eased from an average of 8 percent in the 1990s to about 5 percent in the last five years; core inflation volatility has significantly declined.
  - Real GDP growth: averaged more than 5 percent in 1996-2008.
  - Policy actions: BoM reduced its policy rate by 250 basis points to 5.75 (recent easing referenced in the text).
  - Empirical staff estimate of policy reaction function includes a tolerance level of inflation estimated at 4.7 percent a year.
  - Exchange rate: reclassified from managed float to free float in AREAR classification; nominal effective exchange rate (NEER) depreciated about 4 percent since January 2009 in nominal effective terms while remaining relatively stable against the U.S. dollar.
- Behavioral patterns:
  - Looser monetary stance tends to be followed by higher inflation for most of the sample period (exceptions noted: 2003, 2006-07 and the recent period).
  - Core inflation does not seem associated with changes to the monetary policy stance.
  - Exchange rate depreciations are associated with higher core and headline inflation.
  - Growth outcomes depend heavily on the global economy (notably the EU); no clear strong relationship between short-term interest rates and real output over the sample.

### III. Empirical approach
- Data and variables:
  - Monthly data between 1999 Q1 and 2009 Q3.
  - Endogenous variables (tY): Real GDP (lgdp), Consumer price index (lcpi) [headline or core], Money stock (lm2), Repo rate (repo), Nominal effective exchange rate (lneer).
  - Exogenous variables (tX): U.S. Federal Funds rate (ffr), U.S. real GDP (lgdp_us).
  - All series in logarithms and seasonally adjusted except the repo rate.
- Unit roots and cointegration:
  - Unit root tests indicate all series are I(1); differences are I(0).
  - Standard cointegration tests indicate evidence of at least two cointegrating relationships.
  - Given short sample and to avoid imposing theoretical long-run restrictions, VARs are estimated in levels (no cointegration rank imposed).
- VAR specification and estimation:
  - VARs estimated with constant and no trend.
  - Two models estimated: (i) benchmark with headline CPI; (ii) alternative with core CPI.
  - Lag length criteria indicate the use of 2 lags for both models.
  - Residual diagnostic tests suggest well-behaved residuals.
  - Recursive estimation and Chow tests indicate VAR stability at the 1 percent significance level.
- Identification methods:
  - I - Recursive VAR (Cholesky): assumes lower-triangular contemporaneous matrix (ordering matters); implies specific contemporaneous causality assumptions (e.g., prices may affect output contemporaneously; money stock has no immediate effect on prices; monetary policy shock has no immediate effect on money stock; NEER has no immediate effect on monetary policy).
  - II - Structural VAR: relaxes recursive assumptions and imposes contemporaneous restrictions motivated by money demand and policy reaction considerations (permits monetary policy to respond contemporaneously to money demand and NEER but not to contemporaneous output and price shocks).

### IV. Estimation results and key findings
- General estimation setup for inference:
  - Impulse responses computed for a one-standard-deviation monetary policy shock (unexpected temporary rise in the repo rate) on output, prices, repo rate, money supply, and NEER.
  - Forecast error variance decompositions used to assess the relative importance of monetary policy shocks for fluctuations in each variable.
  - Both benchmark (headline CPI) and alternative (core CPI) models estimated under both identification schemes.
- Summary of principal findings (preserving source language and distinctions):
  - Overall conclusion: transmission channel is weak, particularly for output.
  - Headline CPI model:
    - Changes in the repo rate result in small and short-lived (albeit statistically significant) responses of the headline CPI and to a lesser extent output.
    - Exchange rate shocks transmit to the headline CPI and output in a statistically significant way and tend to last longer than repo rate shocks.
    - Some evidence of a very short-lived transmission of money supply shocks to output (but not CPI).
  - Core CPI model:
    - No transmission of repo rate shocks to output or core CPI.
    - Core CPI exhibits a quick response to exchange rate shocks.
    - Evidence of a statistically significant transmission of money supply shocks to both output and inflation (and to the exchange rate).
  - Identification sensitivity:
    - Results depend on identification; both recursive and structural identifications are examined to assess robustness.
- Policy implications and recommendations preserved from the source:
  - Different monetary policy “rules” could be appropriate depending on which CPI is targeted:
    - For headline CPI, where the interest rate channel is stronger, “Taylor-type” rules may be more applicable.
    - For core CPI, where money supply effects are more important, “McCallum-type” rules that target money supply could be more appropriate.
- Model performance checks:
  - Dynamic in-sample forecasting and out-of-sample predictive performance evaluated for inflation forecasting (models’ forecasting performance assessed and compared with actual values).

*Source: _wp1036 - References  ...............................................................................................................................32*

### 0.2 percent, respectively. The effect is more persistent (and more statistically

### _wp1036 - 0.2 percent, respectively. The effect is more persistent (and more statistically

### Impulse responses and variance decomposition (headline CPI model)
- A positive repo-rate (monetary policy) shock:
  - Lowers output and inflation; effects on inflation are more persistent and more statistically significant.
  - Output returns to pre-shock levels by the end of the 6th period; inflation remains below pre-shock value for more than 10 periods (with some effects marginally insignificant).
  - Structural identification (Figure 3b) indicates faster transmission: statistically significant effects on output and inflation appear almost immediately and disappear after period 3.
  - Variance decomposition: a repo-rate shock accounts for about 2-8 percent (on average) of fluctuations in both prices and output.
- A positive unexpected nominal appreciation (exchange rate shock):
  - Lowers both output and inflation; effects on inflation remain statistically significant for more than 10 quarters.
  - Both effects appear about 4 periods after the shock; output returns to pre-shock levels almost immediately.
  - Variance decomposition: exchange rate shocks account for a small proportion of fluctuations in output and prices (about 12 percent of price fluctuations on average in the core-CPI model; for headline CPI the exchange-rate share is described as small).
  - These patterns suggest a slow and persistent pass-through of exchange rate changes to prices.
- A positive shock to money supply (M2):
  - Translates into a persistent increase in inflation lasting for more than 10 periods (effects not always statistically significant).
  - Produces a short-lived increase in output which disappears by period 3.
  - There is no statistically significant change in the nominal effective exchange rate following the shock to M2 in the headline-CPI recursive specification; in the core-CPI model, a monetary shock elicits a statistically significant response of the nominal effective exchange rate.
  - Variance decomposition (core-CPI model): a money supply shock accounts for about 10-11 percent of fluctuations in output and prices, and about 7 percent of fluctuations in the nominal exchange rate.
- Overall interpretation:
  - Shocks to the repo rate, exchange rate, and money supply have small and statistically significant transmission to prices and output, particularly for inflation.
  - Transmission is stronger for nominal variables (inflation) than for real variables (output), consistent with patterns in transition economies.
  - Weak transmission to real variables may reflect financial market bottlenecks and structural problems (financial frictions, rigidities) rather than pure long-run money neutrality.

### Model diagnostics and forecasting (headline CPI model)
- Model fit and stability:
  - Dynamic within- and out-of-sample forecasts for year-on-year inflation perform quite well (Figures 4 and 5).
  - Chow tests indicate model stability and parameter constancy.
- Out-of-sample forecast performance:
  - The model performs extremely well up to 2008Q4, but overestimates inflation for the three quarters of 2009 by about 1.5-2 percent (2009 being the global recession year, not explicitly modeled).
  - The dynamic out-of-sample forecast is estimated using a VAR from 1999 Q1 to 2008Q1. Using static one-step-ahead forecasting yields an even closer fit.
- Forecast for 2009 Q4 to 2010 Q3:
  - After a small further drop in inflation in 2009 Q4, inflation resumes to the range of 3 to 4 percent in 2010.

### Alternative model (core CPI)
- Impulse responses (core CPI replacing headline CPI):
  - A monetary policy shock has no statistically significant effect on output or inflation; both fluctuate around pre-shock values.
  - A positive shock to the nominal effective exchange rate has a small, statistically significant negative effect on core CPI that lasts two periods after the shock; transmission is shorter and not persistent compared to headline CPI, implying lower exchange-rate pass-through once volatile energy and administered prices are excluded.
  - There is no statistically significant effect of an exchange rate shock on output.
  - A positive money supply shock increases both output and prices; effects persist (though not always statistically significant), and the nominal effective exchange rate responds statistically significantly to the monetary shock.
  - Structural-model impulse responses give similar conclusions: no effect of monetary-policy shock on output or inflation; a small statistically significant effect of exchange-rate shock on inflation.
- Variance decomposition (core CPI):
  - Repo shock accounts for a negligible percentage (less than 2 percent, on average) of fluctuations in prices and output.
  - Exchange-rate shock accounts on average for about 12 percent of price fluctuations.
  - Money-supply shock accounts for about 10-11 percent of fluctuations in output and prices, and about 7 percent of fluctuations in the nominal exchange rate.
- Forecasts (core CPI model):
  - Dynamic within- and out-of-sample forecasts perform reasonably well but perhaps not as well as headline case (Figures 7 and 8).
  - Forecast for 2009 Q4 to 2010 Q3: after a small further drop in 2009 Q4, core inflation resumes to an average of 2.5 percent in 2010.

### Conclusion and policy considerations
- Summary of findings:
  - The overall monetary policy transmission of an unexpected temporary increase in the repo rate is weak for both headline and core CPI models, particularly for output.
  - Repo-rate shocks, exchange-rate shocks, and money-supply shocks produce statistically significant changes in headline CPI; transmission to output is not consistently significant.
  - With core CPI, exchange-rate and money-supply shocks transmit to prices, while repo-rate shocks do not; transmission of money-supply shocks is stronger in the core-CPI model than in the headline-CPI model.
- Policy implications:
  - Different monetary policy rules might be considered depending on the target:
    - Headline CPI (where the interest-rate channel is relatively stronger): “Taylor-type” rules may be more applicable.
    - Core CPI (where money-supply effects are more pronounced): alternative “McCallum-type” rules that target money supply could be more appropriate.
  - The weak transmission to output suggests the need to address structural rigidities in the financial system and regulatory framework that may hamper monetary transmission to the real sector.
  - Further work recommended:
    - Improve model specifications to better reflect characteristics of the Mauritian economy.
    - Consider alternative structural identifications beyond the one used here.
    - Be mindful of possible data limitations given the short analysis period.

*Source: _wp1036 - 0.2 percent, respectively. The effect is more persistent (and more statistically*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp1036.pdf_