## wpiea2025215-source-pdf

## Source details

**Canonical URL:** [wpiea2025215-source-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025215-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025215-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025215-source-pdf.pdf.json)

---

### New monetary and operational policy framework & FPAS integration
- Bank of Mauritius (BOM) adopted a flexible inflation targeting (IT) regime (January 2023) with:
  - inflation target range of 2 to 5 percent and medium-term aim at the mid-point of 3.5 percent;
  - model-based inflation forecasts as intermediate targets and the Key Rate to signal stance;
  - floating exchange rate with smoothing of excessive volatility and forward-looking communications.
- Operational framework features:
  - Key Rate, interest rate corridor defined by Overnight Lending Facility (OLF) and Overnight Deposit Facility (ODF);
  - overnight interbank rate as operational target; open market operations to steer it close to the Key Rate;
  - 7-day BOM bills as main policy instrument issued at a fixed rate on full allotment basis;
  - additional instruments: fine-tuning operations, longer-term operations, reserve requirements (on rupee and foreign currency).
- FPAS development and role:
  - Forecasting and Policy Analysis System (FPAS) centralizes model-based forecasting and policy analysis; core tool is Mauritius Quarterly Projection Model (QPM).
  - FPAS supports quarterly forecasting rounds following an internal forecast calendar and includes nowcasting/near-term forecasting (NTF), climate-risk spreadsheet tool, medium-term steady-state analysis, and satellite/judgement-based analyses.

### Mauritius QPM: design, structure, and calibration
- Model characterization:
  - New Keynesian semi-structural QPM for a medium-size small open economy; decomposes real variables into gap and trend components; monetary policy via nominal interest rate changes to stabilize inflation at 3.5 percent mid-point over the medium-term (2-to-3 years).
  - Rational (forward-looking) expectations; calibrated (not estimated) to reflect small open economy features of Mauritius.
- Main blocks and linkages:
  - Aggregate demand: output gap driven by RMCI (real interest rate gap r̂t and REER gap ẑt), foreign output gap, fiscal impulse, and real wage income effect.
  - Prices: CPI decomposed into core2 (pc), energy (pe), and food (pf) subcomponents.
  - Labor market: nominal wage Phillips curve and Okun-type relation for unemployment gap.
  - External sector: modified UIP linking USD–rupee exchange rate to expectations and risk-adjusted interest differential.
  - Monetary policy rule: Taylor-type rule reacting to expected deviations of inflation and current output gap; neutral rate it neutral = r̅t + π̅t.
  - Fiscal block: parsimonious decomposition deft = deft str + deft gap + deft disc; fiscal impulse fiscimp t identified with discretionary component and structural changes; Special Funds adjustment εt sf.
  - Exogenous determinants: foreign prices, demand, interest rates, world oil and food prices, structural shocks.
- Calibration details and selected parameter values:
  - Calibration sample and steady states: 2001Q1:2025Q1 sample averages used for steady states.
  - Core inflation: b11 = 0.5; b12 = 0.6; b13 = 0.5; b14 = 0.3.
  - Energy inflation: b21 = 0.2; b22 = 0.5; b23 = 0.1; b24 = 0.8.
  - Food inflation: b31 = 0.2; b32 = 0.5; b33 = 0.1; b34 = 0.8.
  - Output/labor/exchange: a1 = 0.6; a2 = 0.25; e1 = 0.6; w1 = 0.4; w2 = 1.
  - Additional parameters (from interest-rate/fiscal lists): 푤 = 3; 1.5 푎4; 0.5 푐1; 0.8 푟1; 0.8 푎5; 0.1 푐2; 2 푠푠푤푒푑푔푒; ‒0.5 푎6; 0.33 푐3; 0.3 푢1; 0.6 푢2; 0.15; 푢3 = 0.1; 푓1 = 0.85; 푑1 = 0.95; 푓2 = 0.75; 푑2 = 0.025; 푓3 = 1.

### Stylized macroeconomic facts informing the model
- Long-run growth and structural change:
  - Economy expanded at an annual average of 4.8 percent over the last three decades.
- COVID-19 and recovery:
  - Nationwide lockdown in early 2020; full reopening of borders in October 2021.
  - Real GDP growth: 5.0 percent in 2023 and 4.9 percent in 2024.
- Fiscal dynamics:
  - Budget deficit surged to 13.6 percent in fiscal year 2019-20.
  - Public sector debt expanded to around 95 percent of GDP in fiscal year 2020-21.
  - Fiscal deficit reached nearly 10 percent in fiscal year 2024-25.
- Labor market and wages:
  - Labor force expanded to almost 600,000 by end-2024.
  - Unemployment rate stood at 6.0 percent in 2024.
  - Nominal wage growth was 1.0 percent in 2020.
- Inflation and CPI composition:
  - CPI fell to 2.5 percent in March 2025.
  - Prices increased by around 4 percent annually on average over 2014 to early 2025.
  - CPI basket shares: core2 accounts for 47 percent; core1 accounts for 61 percent; administered items around 14 percent; food and beverages 39 percent; micro-analysis food items 35.6 percent; imported items direct weight 42 percent.
- Exchange rate and MPF evolution:
  - Flexible IT introduced 16 January 2023; symmetric interest rate corridor of 200 basis points around Key Rate.
  - Reserve requirement ratio set at 9.0 percent of deposit liabilities on both rupee and foreign currency deposits; maintenance period lengthened from 14 to 28 days.
  - Policy rate history: Key Repo Rate slashed to 1.85 percent in early-2020; raised to 4.50 percent by end-2022; lowered to 4.00 percent in September 2024; raised back to 4.50 percent in February 2025.

### Model properties, impulse responses, and historical decompositions
- Impulse response functions (IRFs):
  - Monetary policy shock (one-unit interest rate shock): higher nominal rate ⇒ higher real rates, REER appreciation (negative REER gap), negative output gap, lower inflation with later easing restoring output and depreciating exchange rate to return inflation to target.
  - Aggregate demand shock: output gap rises by nearly 1 percentage point on impact; central bank raises nominal rate; UIP induces appreciation and closure of output gap.
  - Core2 supply shock: raises core2 and CPI above target; central bank raises nominal rate; exchange rate appreciation generates disinflationary forces; trade-off for central bank between price stability and activity.
  - UIP depreciation shock: depreciates nominal and real exchange rate ⇒ inflationary pressures and positive output gap; central bank raises interest rates, later restoring output and inflation.
  - Wage growth shock (10 percent quarterly annualized nominal wage shock): boosts incomes and output gap, triggers higher nominal rate response; exchange rate appreciates; RIR gap initially negative due to inflation expectations.
  - Structural fiscal deficit shock (1 percent of GDP increase in structural deficit; multiplier a4 = 0.5): stimulative fiscal impulse widens output gap, raises inflation and wages; central bank tightens policy generating positive RIR gap and negative REER gap.
- Historical decompositions and narratives (Kalman filter uses 2001Q1:2025Q1):
  - Fiscal impulse: large positive in 2020; negative in late-2021/early-2022; positive again aligning with fiscal year 2024-25 deficit.
  - RMCI decomposition: 2010s mainly tight due to REER; COVID-19 loosened both channels; from late-2022 RIR gap became more restrictive as policy tightened.
  - Output gap: deep negative in 2020; recovery after October 2021; post-2022 economy close to or above potential with stimulative fiscal policy and restrictive real monetary conditions.
  - RMC and core2 decomposition: external factors (REER, oil, food) important; mid-2023 RMC receded as global food and energy prices declined and policy tightening helped steer core2 toward target.

### Labor market block, specification variants, and policy relevance
- Labor market extension:
  - Nominal wage Phillips curve and unemployment block enable structural determination of wages and unemployment, embedding bargaining power, pricing power, productivity, and labor income effects on demand.
  - Inclusion reduces implied inflationary impact of demand shocks relative to a model without labor market (No LM), moderating optimal policy responses.
- Alternative calibration and robustness:
  - Alternative specification sets a6 = 0; b13 = 0.7; b23 = b33 = 0.2 to match initial QPM calibration.
  - Comparative IRFs: baseline (with labor market) shows more moderate central bank tightening; No LM requires stronger immediate tightening and yields larger estimated deflationary pressures during COVID-19.
  - Lower-right panel reference: "0.5 percentage points (lower-right panel)" in comparisons.

### In-sample forecast performance and institutional implications
- In-sample recursive eight-quarter-ahead forecast RMSFE ratios to random walk (QPM RMSFE / random walk RMSFE):
  - CPI (%, qoq ann.): 1Q 0.98; 2Q 1.02; 3Q 1.01; 4Q 0.84; 5Q 0.90; 6Q 0.81; 7Q 0.72; 8Q 0.68.
  - GDP (%, yoy): 1Q 0.67; 2Q 0.64; 3Q 0.63; 4Q 0.56; 5Q 0.60; 6Q 0.55; 7Q 0.57; 8Q 0.52.
  - Nominal depreciation (%, qoq ann.): 1Q 1.34; 2Q 0.72; 3Q 0.65; 4Q 0.67; 5Q 0.69; 6Q 0.71; 7Q 0.64; 8Q 0.58.
  - Interest rate (%): 1Q 2.09; 2Q 1.94; 3Q 1.73; 4Q 1.55; 5Q 1.44; 6Q 1.35; 7Q 1.25; 8Q 1.11.
- Interpretation:
  - QPM provides more accurate inflation, GDP growth, and exchange-rate forecasts than random walk benchmarks, especially at medium-term horizons.
  - Interest-rate forecasts are less accurate historically due to earlier monetary regimes differing from the current flexible IT framework; forecast accuracy is expected to improve as QPM informs MPC decisions.
- Institutional conclusions:
  - FPAS with Mauritius QPM aligns with the new inflation-targeting framework and strengthens BOM’s model-based forecasting, policy evaluation, and communications.
  - QPM supports baseline projections, counterfactuals, scenarios, and MPC deliberations while satellite tools and judgement inputs complement model outputs.
  - Planned extensions include disaggregated sectoral output, richer fiscal and external blocks, and multiple monetary instruments while preserving operational usability.

*Source: IMF Working Paper — Mauritius QPM: A Quarterly Projection Model for the Bank of Mauritius (Working Paper No. WP/2025/215).*

### 1. Introduction

### 1. Introduction

### New monetary and operational policy framework (BOM, January 2023)
- Bank of Mauritius (BOM) adopted a flexible inflation targeting (IT) regime with the following key elements:
  - explicit price stability commitment: inflation target ranging between 2 to 5 percent, with the aim of achieving the mid-point of 3.5 percent over the medium-term;
  - model-based inflation forecasts as intermediate targets;
  - using the key rate to signal the monetary policy stance;
  - a floating exchange rate regime, subject to market forces but with the Bank aiming to smooth out excessive volatility;
  - forward-looking external communications emphasizing medium-term prospects.
- Operational framework revisions:
  - introduced the Key Rate, alongside the interest rate corridor defined by the Overnight Lending Facility (OLF) and the Overnight Deposit Facility (ODF);
  - adopted the overnight interbank rate as its operational target, with open market operations aiming to steer it close to the Key Rate;
  - designated 7-day BOM bills as the main policy instrument, issued at a fixed rate and on full allotment basis;
  - identified additional instruments, including fine-tuning operations, longer-term operations, reserve requirements (on rupee and foreign currency).

### FPAS development and integration with policy processes
- BOM developed, with IMF Technical Assistance, a Forecasting and Policy Analysis System (FPAS) that centralizes model-based forecasting and policy analysis and is integrated into monetary policy processes and communications.
- The core analytical tool is the Mauritius Quarterly Projection Model (QPM), a semi-structural framework embedding key policy transmission channels and a flexible IT regime.
- The FPAS supports quarterly forecasting rounds following a defined internal forecast calendar specifying timing of meetings, deadlines, responsibilities, and deliverables to coordinate technical teams and decision-makers.
- Complementary analytical tools within the FPAS include:
  - rigorous and timely data summaries;
  - nowcasting and near-term forecasting (NTF) models;
  - a spreadsheet tool for the analysis of climate risks;
  - analytical formulation of medium-term steady states of key macroeconomic variables;
  - various satellite and judgement-based analyses.

### Mauritius QPM: design, structure, and role
- Characterization:
  - medium-size small open economy New Keynesian semi-structural model extending the canonical four-equation framework.
  - decomposes real variables into gap and trend components.
  - monetary policy conducted via changes in nominal interest rates aiming to stabilize inflation at the target by impacting the business cycle amid real and nominal rigidities.
  - allows for rational (forward-looking) expectations and is calibrated to represent reality features of a small open economy such as Mauritius.
- Main blocks and links:
  - aggregate demand: output gap driven by monetary conditions—real interest rate and real effective exchange rate—capturing two key monetary transmission channels;
  - prices: decomposed into core, food, and fuel components to reflect sectoral heterogeneity in price dynamics;
  - labor market: wage inflation Phillips curve and Okun-type relation between the unemployment rate gap and cyclical demand;
  - external sector: modified uncovered interest parity (UIP) linking the US dollar–rupee exchange rate to expectations and risk-adjusted interest rate differential;
  - monetary policy rule: policy reaction function specifies interbank rate dynamics in response to expected deviations of inflation from the target and current-period output gap;
  - fiscal block: parsimonious fiscal block models the fiscal impulse based on the cyclically adjusted fiscal balance;
  - exogenous determinants: foreign economy prices, demand, and interest rates; world oil and food prices; various structural shocks.
- Endogenous policy determination:
  - QPM delivers the interbank rate trajectory consistent with CPI inflation converging to the 3.5 percent mid-point of the target range over the medium-term (2-to-3 years), conditional on current situation and exogenous trajectories.
- Calibration:
  - parametrization reflects an iterative process and sequential calibrations to ensure theoretical and empirical coherency.
  - model was calibrated rather than estimated, consistent with adopted practice in many central banks in developing economies.

### Empirical properties, fit, and policy use
- The model exhibits:
  - economically intuitive propagation of shocks and monetary policy responses;
  - in-sample forecasting accuracy that captures dynamics of important economic variables;
  - gaps and structural shocks that align with historical narratives for Mauritius.
- Practical uses:
  - baseline projections, counterfactual simulations, alternative scenarios;
  - QPM-based results are a key input into BOM’s Monetary Policy Committee (MPC) deliberations and feature in forward-looking external communications (examples referenced to BOM releases in 2025).
- Similarity to other central bank QPM-type approaches:
  - comparable to frameworks adopted in Colombia, Ghana, India, Philippines, Rwanda, and others, while embedding Mauritius-specific mechanisms and parametrization.

### Limitations and planned extensions
- Notable limitations:
  - fiscal policy representation is abridged to aggregate fiscal impulse; does not model multidimensional public debt dynamics or spending/revenue composition;
  - balance of payments flows are not explicitly modeled; external balance reduced to cyclical position of the real effective exchange rate;
  - sectoral granularity limited: economic output proxied by real GDP rather than disaggregated sectors (e.g., agriculture, tourism);
  - monetary policy represented only by the interest rate, while in practice additional tools (reserve requirements, macroprudential measures, foreign exchange interventions) are used;
  - passthrough from policy rate to interbank rates and further to retail loan and deposit rates is not explicitly modeled.
- Ongoing and potential extensions:
  - work underway to model output in a disaggregated manner to capture heterogeneities across sectors and to enrich analysis of climate-related shocks and tourism flows;
  - future extensions could incorporate richer fiscal blocks, external and internal balances, and multiple monetary instruments, taking care to balance complexity and real-time operational usability.

### Paper organization
- Section 2 documents key stylized facts of the Mauritian economy that informed model specification.
- Section 3 details the structure of Mauritius QPM.
- Section 4 presents model-based results, including impulse response functions, equation decompositions, and forecast accuracy of in-sample simulations.
- Section 5 concludes.

*Source: IMF Working Paper — Mauritius QPM: A Quarterly Projection Model for the Bank of Mauritius, 1. Introduction (excerpt).*

### 2. Stylized Facts

### 2. Stylized Facts

### 2.1. Output and prices
- Long-term performance and structural transformation
  - Economy expanded at an annual average of 4.8 percent over the last three decades.
  - Transition from monoculture to diversified, export-oriented manufacturing, tourism, and developing financial services.
  - Described as the “Mauritian miracle” and the “success of Africa”.

- COVID-19 shock and recovery
  - Nationwide lockdown in early 2020 halted the island’s longest period of sustained economic growth.
  - Key sectors most affected: tourism, construction, and manufacturing.
  - Policy response: fiscal, monetary and financial support measures to safeguard jobs, alleviate cash-flow constraints, and ensure business continuity.
  - Second national lockdown in early 2021; full reopening of borders in October 2021.
  - Real GDP growth: 5.0 percent in 2023 and 4.9 percent in 2024.

- Fiscal developments
  - Budget deficit surged to 13.6 percent in fiscal year 2019-20.
  - Public sector debt expanded to around 95 percent of GDP in fiscal year 2020-21.
  - Fiscal deficit reached nearly 10 percent in fiscal year 2024-25.
  - Authorities expected to turn towards fiscal consolidation as growth normalizes.

- Labor market and wages
  - Labor force expanded to almost 600,000 by end-2024.
  - Unemployment rate stood at 6.0 percent in 2024 (a two-decade low).
  - Nominal wage growth was 1.0 percent in 2020; real wages declined in 2020 but recovered thereafter.
  - Wage formation features:
    - Public-sector wage adjustments guided by Pay Research Bureau (PRB) recommendations (collective bargaining every four to five years).
    - Private-sector wages often adjusted following public-sector revisions.
    - National Minimum Wage introduced January 2018.

- Inflation measurement and drivers
  - Inflation aggregates used: headline CPI, core1, and core2.
    - core2 accounts for 47 percent of the CPI basket (excludes volatile food and energy and administered-price items).
    - core1 accounts for 61 percent of the CPI basket (includes fuel and administered items).
    - Administered items account for around 14 percent of the CPI basket.
    - Food and beverages account for 39 percent of the CPI basket.
  - Food items account for 35.6 percent of the basket (micro analysis).
  - Direct weight of imported items: 42 percent of the overall CPI basket.
  - Inflation episodes:
    - Double-digit inflation in 2021-22 after imported inflation and second-round effects.
    - CPI fell to 2.5 percent in March 2025.
    - Prices increased by around 4 percent annually on average over the period 2014 to early 2025.
  - Short-term CPI fluctuations largely driven by food (notably locally produced fresh vegetables); energy and administered prices show more delayed pass-through.
  - Wage-related transmission channel:
    - Non-performance-based end-of-year bonuses and increases in minimum wages can raise production costs and affect retail prices, particularly core2 prices (market services).

- Model implications
  - BOM QPM includes a dedicated labor-market block linking unemployment, wages, output gap, and inflation, allowing potential central bank response to labor-market shocks.

### 2.2. Exchange rate and interest rate
- Exchange rate regime and trends
  - Historical evolution: from currency board in early 1990s to floating exchange rate (IMF de facto classification).
  - Rupee depreciation trend since 2018 driven by US dollar strength and domestic demand-supply conditions.
  - BOM intervenes to mitigate excessive volatility and avoid disorderly market conditions.
  - Depreciation pressures intensified after COVID-19 and the Russia-Ukraine war; Bank sold FX to maintain supply.

- Monetary policy framework evolution
  - Transitioned from priority lending and reserve money targeting to interest rate-based policy (since 2006).
  - New flexible inflation-targeting (IT) regime introduced 16 January 2023.
  - Flexible IT operational and strategic features:
    - Flexible inflation target range of 2-5 percent with the aim of attaining 3.5 percent over the medium-term.
    - Continued reliance on interest rates; operational target: overnight rate; intermediate target: inflation forecasts.
    - Greater clarity on role and nature of FX interventions under flexible exchange rate.
    - Symmetric interest rate corridor of 200 basis points around the Key Rate.
    - Main instrument: 7-day BOM bill issued at a fixed rate equal to the Key Rate on full allotment basis (this arrangement in place from January to July 2023).
    - Fine-tuning operations and longer-term BOM instruments via competitive bidding.

- Reserve and liquidity rules under new MPF
  - Maintenance period for holding Cash Reserve Ratio (CRR) on both rupee and foreign currency deposits lengthened from 14 to 28 days.
  - Reserve requirement ratio set at 9.0 percent of deposit liabilities on both rupee and foreign currency deposits.
  - Daily reserve requirements scrapped; foreign currency reserve requirements no longer remunerated.

- Interest rate history and recent moves
  - Policy rate cuts during pandemic: Key Repo Rate slashed to 1.85 percent in early-2020.
  - Tightening cycle: Key Repo Rate raised to 4.50 percent by end-2022.
  - Policy rate lowered to 4.00 percent in September 2024.
  - Key Repo Rate raised back to 4.50 percent in February 2025 due to excess liquidity, negative interest rate differentials, and sustained rupee depreciation.
  - Short-term money market rates generally moved in tandem with the policy rate, influenced by liquidity management and the new MPF.

- Model alignment
  - Most elements of the new MPF were incorporated in Mauritius QPM as of late-2022 to anticipate the formal shift to a flexible IT framework.

*Source: wpiea2025215-source-pdf - 2. Stylized Facts.*

### 2023. In the first half of 2023, the overnight interbank rate and the 91-day bill yield moved up to the range

### wpiea2025215-source-pdf - 2023. In the first half of 2023, the overnight interbank rate and the 91-day bill yield moved up to the range

### Liquidity operations and interest rate corridor (2023–2024)
- In the first half of 2023, the overnight interbank rate and the 91-day bill yield moved up to the range of 4.25 to 4.75 percent, reflecting open market operations being conducted through the 7-day BOM bills at the policy rate of 4.50 percent.
- In July 2023, a review in liquidity management operations capped the 7-day BOM auctions at 4.50 percent; following this cap, banks started depositing excess funds in the Overnight Deposit Facility.
- After banks deposited excess funds in the Overnight Deposit Facility, the overnight interbank rate and the 91-day bill yield dropped down to the lower bound of the corridor.

### Developments in late 2024 and policy-rate tracking
- Reflecting renewed open market operations in the latter part of 2024, the overnight interbank rate has since increased to around 3.5 percent.
- The 91-day bill yield has also been tracking the policy rate, hovering around 4.50 percent.

*See IMF (2025a) for an overview of the latest economic developments and policy-relevant discussions in Mauritius.*

### 3. Model Structure

### 3. Model Structure

### Major building blocks
- Mauritius QPM consists of four key blocks: aggregate demand, aggregate supply (Phillips curves), uncovered interest parity (UIP), and monetary policy reaction function.
- Extensions relative to a canonical QPM:
  - Decomposition of headline inflation into core, food, and energy subcomponents.
  - Parsimonious fiscal block with structural, cyclical, and discretionary elements.
  - Labor market block with a nominal wage Phillips curve and an Okun-type relation for the unemployment rate gap.
  - Real Monetary Conditions Index (RMCI) combining real interest rate and real effective exchange rate gaps.

### Aggregate demand (IS-type specification)
- Output gap equation (gaps denoted with a “hat”):
  - ŷt = a1 ŷt−1 − a2 rmci t + a3 ŷt* + a4 fiscimp t + a5 (rŵt − unemp̂t) + εt y
  - rmci t = a6 r̂t + (1−a6)(−ẑt)
- Key determinants:
  - Persistence: a1 ŷt−1
  - Real monetary conditions (RMCI): rmci t (weighted average of real interest rate gap r̂t and REER gap ẑt)
  - Foreign (euro area) output gap: a3 ŷt*
  - Fiscal impulse: a4 fiscimp t
  - Real wage income effect: a5(rŵt − unemp̂t)
  - Aggregate demand shocks: εt y
- Notes:
  - RMCI interpretation: higher values imply tighter monetary conditions.
  - Habit formation included to match GDP persistence.
  - Fiscal impulse captures discretionary fiscal actions and medium-term structural changes.

### Fiscal block (parsimonious)
- Fiscal deficit decomposition:
  - deft = deft str + deft gap + deft disc
- Components:
  - Structural deficit: deft str follows a univariate mean-reverting process:
    - deft str = f1 deft−1 str + (1−f1) deft ss str + εt deft str
  - Cyclical deficit (automatic stabilizers):
    - deft gap = −f2 ŷt
  - Fiscal impulse identification:
    - fiscimp t = deft disc + f3(deft str − deft−1 str) + εt sf
- Special Funds adjustment term: εt sf captures ad-hoc government expenditures (e.g., post-pandemic infrastructure, social housing, hospitals) tuned by expert judgment.

### Labor market block
- Nominal wage Phillips curve:
  - Δwt = (1−w1) Et Δwt+1 + w1 Δwt−1 − w2 ût − w3 rŵt + εt w
  - Where Δwt is qoq annualized change in log nominal wages; Et denotes rational expectations.
- Real wage definition and trend/gap decomposition:
  - rŵt = wt − pt
  - rŵt = rŵ̅t + rŵ̂t
- Trend real wage growth:
  - Δrŵ̅t = r1 Δrŵ̅t−1 + (1−r1)(Δy̅t + ss_wedge) + εt rŵ
  - ss_wedge set equivalent to observed discrepancy between average real wage growth and output growth (around 0.5 percentage point).
- Unemployment block (Okun-type and trend/NAIRU dynamics):
  - ut = u̅t + ût
  - ût = u1 ût−1 − u2 ŷt + u3 rŵ̂t + εt u
  - u̅t = d1 u̅t−1 + (1−d1) u̅ss − d2 (Δy̅t − Δy̅ss) + εt u̅
- Interpretation:
  - Unemployment gap negatively correlated with output gap (Okun’s law).
  - NAIRU (trend unemployment) mean-reverting and linked to trend output growth deviations.

### Aggregate supply and inflation (three subcomponents)
- CPI decomposition:
  - pt = wc pct c + we pt e + (1−wc − we) pt f + εt p
  - pt: CPI; pt c: core2; pt e: energy; pt f: food. εt p captures non-additivity from constant weights and log-approximations.
- Two exchange rate transmission channels:
  - Direct: exchange rate enters each Phillips curve (high imported share of goods).
  - Indirect: RMCI (real exchange rate component) captures expenditure-switching effects via net exports.

Core inflation (core2)
- Phillips curve:
  - πt c = b11 πt−1 c + (1−b11) Et πt+1 c + b12 rmc t c + εt π c
- Real marginal cost for core producers:
  - rmc t c = b13 ŷt + b14 (ẑt − rp̂t c) + (1−b13 − b14) rŵ̂t
- Relative core price definitions:
  - rp t c = pt c − pt
  - rp̂t c = rp t c − rp̅t c
- Notes: rmc t c combines output gap, real exchange rate adjusted for relative core prices, and real wage gap.

Energy (fuel) inflation
- Phillips curve:
  - πt e = b21 πt−1 e + (1−b21) Et πt+1 e + b22 rmc t e + εt π e
- Real marginal cost for energy:
  - rmc t e = b23 ŷt + b24 (rp̂t woil + ẑt − rp̂t e) + (1−b23 − b24) rŵ̂t
- Relative world energy price:
  - rp t woil = p t woil − cpi t us
  - rp̂t woil = rp t woil − rp̅t woil
- Notes: RMC component (rp̂t woil + ẑt − rp̂t e) defines energy import price in domestic currency; energy and world oil price trends follow autoregressive processes.

Food inflation
- Phillips curve:
  - πt f = b31 πt−1 f + (1−b31) Et πt+1 f + b32 rmc t f + εt π f
- Real marginal cost for food:
  - rmc t f = b33 ŷt + b34 (rp̂t wfood + ẑt − rp̂t f) + (1−b33 − b34) rŵ̂t
- Relative world food price:
  - rp t wfood = p t wfood − cpi t us
  - rp̂t wfood = rp t wfood − rp̅t wfood
- Notes:
  - Food component includes food, beverages and mortgage interest on housing loans (mortgage interest included because its share is 3.1 percent as of 2025 and modeling it separately would complicate sign differences under monetary policy).
  - Food supply shock εt π f can capture weather-related disturbances in reduced form.

Calibration and cross-component differences
- Calibration uses 2001Q1:2025Q1 sample averages for steady states, behavioral parameter analogues from other QPMs, impulse response analysis, data filtration, and in-sample simulations.
- Persistence differences:
  - Energy and food inflation persistence lower than core; higher volatility due to supply-side disturbances.
  - Weights in RMC definitions differ: output gap more prominent for core; world energy/food price gaps and real exchange rate gap higher for energy and food sectors.

### Exchange rate (modified UIP)
- Nominal exchange rate equation:
  - st = st+1 exp −(it − it* − prem t)/4 + εt s
  - st+1 exp = e1 Et st+1 + (1−e1)[st−1 + 2/4(π̅t − π* + Δz̅t)]
- Variables:
  - st: log nominal exchange rate (domestic currency per US dollar); increase denotes depreciation.
  - it: domestic nominal money market interest rate; it*: foreign nominal interest rate; prem t: time-varying sovereign risk premium.
  - εt s: exchange rate (UIP) shock.
- Expectations mixing:
  - e1 captures forward-looking degree; blending Et st+1 and lagged st−1 adjusted by trend depreciation (inflation target differential π̅t − π* and trend REER depreciation Δz̅t) to match observed persistency.
- Real effective exchange rate (z t) definition:
  - zt = [wus st + (1−wus) st eur] + [wus p t us + (1−wus) p t ea] − pt
  - st eur computed via rupee-dollar and dollar-euro rates; wus is weight of US dollar; p t us, p t ea, pt are price levels for US, euro area, and Mauritius.

### Interest rate and monetary policy (Taylor-type rule)
- Short-term nominal interest rate (proxy: weighted average OIR) equation:
  - it = c1 it−1 + (1−c1)[it neutral + c2 πt exp,dev + c3 ŷt] + εt i
  - it neutral = r̅t + π̅t
  - πt exp,dev = Et π4 t+3 − π̅t+3
- Interpretation:
  - Lagged interest rate it−1 captures policy inertia; high c1 yields smoother interest-rate path.
  - Central bank reacts to three-quarters-ahead expected annual inflation deviation from target and to output gap with parameters c2 and c3.
  - Neutral nominal rate equals trend real rate r̅t plus inflation target π̅t; trend real rate linked to trend foreign real interest rate due to Mauritius’ international financial center status.
  - εt i captures monetary policy shock (non-systematic interest rate movements).

### Calibration: selected parameter values (Table 1)
- Core inflation parameters:
  - b11 = 0.5
  - b12 = 0.6
  - b13 = 0.5
  - b14 = 0.3
- Energy inflation parameters:
  - b21 = 0.2
  - b22 = 0.5
  - b23 = 0.1
  - b24 = 0.8
- Food inflation parameters:
  - b31 = 0.2
  - b32 = 0.5
  - b33 = 0.1
  - b34 = 0.8
- Output gap / exchange rate / labor market parameters (partial list shown):
  - a1 = 0.6
  - a2 = 0.25
  - a3 (listed in table but value truncated in source)
  - e1 = 0.6
  - w1 = 0.4
  - w2 = 1

*Source: IMF staff summary of “3. Model Structure” from the Mauritius QPM chapter.*

### 0.4 Interest rate

### 0.4 Interest rate

### Parameter values and coefficients
- 푤: 3
- 1.5 푎4
- 0.5 푐1
- 0.8 푟1
- 0.8 푎5
- 0.1 푐2
- 2 푠푠푤푒푑푔푒
- ‒0.5 푎6
- 0.33 푐3
- 0.3 푢1
- 0.6 푢2
- 0.15

### Fiscal block
- 푢3: 0.1
- 푓1: 0.85
- 푑1: 0.95
- 푓2: 0.75
- 푑2: 0.025
- 푓3: 1

_Source: Authors’ representation based on an iterative process._

### 4. Model Properties and Results

### 4. Model Properties and Results

### 4.1. Impulse response functions
- The section presents IRFs of key macroeconomic variables to a one-unit monetary policy shock and other shocks, illustrating conventional monetary transmission channels embedded in Mauritius QPM.
- Monetary policy (one-unit interest rate shock):
  - Higher nominal interest rate ⇒ higher real rates ⇒ positive real interest rate (RIR) gap.
  - Real effective exchange rate (REER) gap turns negative (overvaluation) owing to a stronger domestic currency in both nominal and real terms.
  - Contractionary influences on economic activity through interest rate and exchange rate channels ⇒ negative output gap.
  - Weak activity and overvalued REER drive down inflation, which partly helps stimulate real wages; easing later restores output and depreciates exchange rate, pushing inflation back to target.
- Aggregate demand shock (Figure 11):
  - Output gap increases by nearly 1 percentage point on impact, with effects dissipating thereafter, partly due to monetary policy response.
  - Higher domestic demand ⇒ build-up of inflationary pressures through higher real marginal costs, particularly for core inflation, and upward pressure on nominal wages (positive real wage gap).
  - Central bank reacts to above-target inflation and positive output gap by gradually increasing the nominal interest rate; no trade-off for the central bank in this scenario.
  - Through UIP, higher interest rate differential (Mauritius vis-à-vis the US) ⇒ nominal and real appreciation ⇒ negative REER gap, tighter monetary conditions, gradual closing of output gap.
- Core2 (supply) inflation shock (Figure 12):
  - Unexpected increase in core2 raises core2 inflation and causes CPI inflation to overshoot target.
  - Central bank raises nominal interest rate; UIP induces nominal exchange rate appreciation and a negative REER gap (overvalued currency), creating disinflationary pressures.
  - RIR gap moves negative initially (driven by higher inflation expectations), then rises as inflation and expectations wane.
  - Above-target inflation dampens real wages (negative income effect), constraining aggregate demand and creating a negative output gap.
  - Supply shocks move prices and activity in opposite directions, imposing a trade-off on the central bank; in Mauritius QPM the central bank emphasizes price stability and raises rates in response.
- UIP (exchange rate depreciation) shock (Figure 13):
  - Depreciation ⇒ nominal and real exchange rate depreciation ⇒ inflationary pressures via higher import prices and a positive output gap (stimulating net exports and aggregate demand).
  - With inflation overshooting and excess demand, central bank raises interest rates; initial RIR gap remains stimulative (negative) due to high inflation expectations.
  - Gradual rate increases (and UIP feedback) strengthen currency, tighten real monetary conditions, counteract demand pressures, and restore output; overvaluation and falling domestic cost pressures bring inflation back to target.
  - Note: model’s depreciation shock is expansionary (aggregate demand-type), but the text notes that in practice depreciation may be contractionary if triggered by capital outflows or risk-off; model adjustments to capture such effects are left for future research.
- Wage growth shock (10 percent quarterly annualized nominal wage shock, Figure 14):
  - Wage shock lifts real labor incomes ⇒ real wage gap opens ⇒ consumption spending and output rise (positive output gap).
  - Higher wages increase labor costs and may push firms to lay off workers ⇒ unemployment up.
  - Central bank raises policy rate responding to inflation and widening output gap; RIR gap turns negative initially as inflationary effects dominate.
  - Exchange rate appreciates nominally and real terms due to positive nominal interest rate differentials; monetary conditions become more restrictive (higher weight on exchange rate), allowing gradual narrowing of output gap and inflation convergence to target.
- Structural fiscal deficit shock (1 percent increase in structural fiscal deficit to GDP ratio, Figure 15):
  - The 1 percent shock amounts to a stimulative fiscal impulse; contemporaneous multiplier a4 = 0.5.
  - Positive fiscal shock supports domestic activity ⇒ output gap widens and triggers demand-side inflationary pressures.
  - Stronger output ⇒ unemployment declines ⇒ nominal and real wages increase.
  - Central bank raises interest rate ⇒ monetary policy becomes contractionary (positive RIR gap); exchange rate appreciates nominally and in real terms ⇒ REER gap negative.
  - Tight RIR and REER ⇒ tight monetary conditions ⇒ eventual cooling of activity and narrowing of output gap; pressures on inflation are alleviated, allowing interest rate cuts and exchange rate depreciation later.

### 4.2. Model-based decompositions and historical narratives
- Estimation details:
  - Kalman filter results use data sample 2001Q1:2025Q1; observed data mostly to 2024Q4, with 2025Q1 as nowcast; figures focus on 2010Q1:2024Q4.
- Fiscal impulse (Figure 16):
  - Pre-pandemic, especially early-2010s, model estimates systematically positive fiscal impulses consistent with government investment to generate growth and jobs.
  - Large positive fiscal impulse estimated in 2020 due to pandemic support (less revenue, more recurrent spending).
  - Negative fiscal impulse in late-2021 and early-2022 aligning with post-pandemic fiscal consolidation.
  - More recent positive fiscal impulse aligns with high budget deficit for fiscal year 2024-25.
- Real monetary conditions index (RMCI) decomposition (Figure 17):
  - RMCI split into RIR gap and REER gap components; positive values denote tight conditions.
  - Pre-pandemic (2010-19) the exchange rate appreciated (positive contributions) and REER stance was overvalued, with exception of 2015-16 depreciation episode.
  - Composite monetary conditions in the 2010s were driven primarily by tight REER stance; interest rate contributions were less meaningful before flexible IT framework.
  - During COVID-19 both interest rate and exchange rate channels loosened conditions to support aggregate spending.
  - Starting 2021, exogenous supply-type shocks led to inflation becoming more deep-rooted; central bank raised interest rates and real effects were reinforced by decelerating inflation and expectations, making RIR gap more restrictive even with flat nominal rates; exchange rate firmed, ensuring tight real monetary conditions starting late-2022.
- Output gap decomposition (Figure 18):
  - Output gap fluctuated mildly around zero before the pandemic; positive values during 2010-15 and 2018-19 reflected stimulative fiscal conditions and favorable demand shocks, partly offset by tight monetary conditions.
  - 2020 saw unprecedented decline in output gap due to COVID-19 lockdowns and border closures; fiscal and monetary support restrained contraction (positive fiscal impulse in 2020Q3; stimulative RIR gap).
  - Additional contraction in 2021Q2 during second wave; recovery after full reopening in October 2021 with negative output gap closing by end-2022.
  - Post-2022 the output gap remained positive with economy operating close to or above potential; fiscal policy remained stimulative while real monetary conditions turned restrictive starting 2023.
- Nominal wage dynamics (Figure 19):
  - Nominal wage growth (quarterly annualized rate) driven primarily by expectations (backward- and forward-looking) and, to a lesser extent, by catching-up via real wage gap.
  - Short-term unemployment fluctuations were evident during the pandemic, exerting negative influence on wages.
  - Recovery in activity and employment gains supported nominal wage pick-up; shocks capture exogenous spikes from wage policies and allowances.
- CPI composite real marginal costs (RMC) decomposition (Figure 20):
  - RMC driven importantly by external factors given high import reliance; key components are gaps in REER, global oil prices, and global food prices.
  - Early-2010s: RMC strongly inflationary due to high global commodity prices, partly offset by overvalued exchange rate.
  - 2014-19: composite RMC mainly negative (disinflationary) with weak international food prices offset by above-trend oil prices and sequential REER under-/overvaluation.
  - COVID-19: RMC largely disinflationary due to subdued demand, negative real wage gap, and lower commodity prices.
  - Post-pandemic: RMC increased due to rise in global commodity prices and some depreciation; by mid-2023 RMC moved back into disinflationary region led by declining global food prices and real appreciation of the rupee, though positive output and global oil price gaps continued to exert upward pressure.
- Core2 inflation decomposition (Figure 21):
  - Pre-pandemic cost pressures were mostly benign; domestic cycle and REER movements produced counterbalancing effects, tilting realized core2 inflation to the downside.
  - Start of pandemic: negative output gap and falling commodity prices maintained disinflationary pressures, partly offsetting rupee depreciation.
  - Mid-2021 commodity price shocks reversed the situation with cost-push factors driving core2 increases; backward-looking inflation expectations and pass-through of commodity and freight costs into sectoral prices contributed.
  - From mid-2023, core2 RMC receded due to declining global food and energy prices and fading supply disruptions; policy tightening also helped steer core2 inflation back towards target.

### 4.3. Labor market block: additional considerations
- The labor market extension in Mauritius QPM structurally determines wage and unemployment dynamics, embedding bargaining power, pricing power, labor market tightness, productivity, cost linkages, and labor income effects on aggregate demand in the IS curve.
- Policy relevance:
  - Incorporation allows analysis of implications of wage and labor policy decisions (e.g., new legislation) on inflation and the necessity for monetary policy responses to avoid a wage-inflation spiral via de-anchored expectations.
  - Supports forward-looking analytical input to the BOM’s MPC process.
- Comparative IRF analysis (Figure 22) — baseline model vs. model without labor market ("No LM"):
  - "No LM" obtained by setting elasticity of aggregate demand to labor income = 0 and increasing output-gap weights in sectoral RMCs so real wage gap share = 0.
  - Both models: excess demand ⇒ higher inflation. Differences:
    - Baseline: labor is a production factor, so output gap’s importance in domestic costs is relatively lower; initial wage response muted ⇒ demand-side price pressures weaker.
    - No LM: wages exogenous ⇒ higher weight on output-gap-related costs ⇒ demand shock yields larger inflationary impact.
  - Central bank response is more moderate in baseline; RIR stance remains neutral in baseline for first year following shock, whereas alternative model elicits immediate tightening.
  - Initial adjustment in baseline occurs mainly through exchange rate channel.
- Kalman filter estimates comparison (Figure 23 and discussion):
  - With higher output gap weight in RMCs (No LM), deflationary pressures during COVID-19 lockdowns are estimated stronger.
  - Since actual inflation did not fall as much in 2020-21, the alternative model requires larger positive supply shocks and implies higher inflation expectations and a marginally looser RMCI over 2020-21.
  - The alternative specification implies a need for a tighter interest rate trajectory and a higher estimated nominal neutral rate (discussion notes the nominal neutral rate is higher by a maximum of about — text ends at this point and does not provide the exact completed value).

*Source: Authors’ simulations and estimations as presented in "4. Model Properties and Results" of the Mauritius QPM chapter.*

### 0.5 percentage points (lower-right panel).

### wpiea2025215-source-pdf - 0.5 percentage points (lower-right panel)

### Labor market extension and model specification
- Introducing the labor market block into Mauritius QPM:
  - Allows taking into account additional propagation channels and estimating unobserved variables, like the unemployment rate gap.
  - Provides more informed policy advice; during the COVID-19 episode the labor market extension—by explicitly taking into account the declines in real wages and increases in unemployment—would have allowed to avoid too tight policy recommendations, which would have delayed the post-pandemic recovery.
- Formal alternative calibration specified in the text:
  - Set 푎6 = 0 in equation (1).
  - Set 푏13 = 0.7, 푏23 = 푏33 = 0.2 in equations (9), (12), and (15).
  - This alternative version matches the model specification and calibration of the initial version of Mauritius QPM.
- Figures referenced:
  - Figure 23: Kalman filter estimates for several unobserved variables in the baseline model and in the alternative model without the labor market extension (figure depicts baseline vs. model without labor market).
  - Lower-right panel reference: "0.5 percentage points (lower-right panel)."

### In-sample simulations and conditioning
- Purpose and setup:
  - In-sample simulations generate plain recursive forecasts in each quarter assuming data is available only up to that quarter.
  - Simulations are conditioned on actual observations or full sample filtered estimates for foreign variables and trends (i.e., for each simulation, their future values are assumed to be known).
  - Foreign-sector variables (e.g., Fed Funds rate, euro area output and inflation, international oil and food prices) and domestic trends are typically provided to the model from outside sources or modeled as univariate autoregressive processes converging to steady state.
- Findings from in-sample simulations:
  - Figure 24 shows recursive in-sample eight-quarter ahead forecasts for selected macroeconomic variables.
  - QPM captures tendencies in actual data reasonably well for inflation, real GDP growth, and nominal exchange rate changes, including turning points.
  - Interest rate forecasts are less accurate due to:
    - The monetary policy regime in previous years being different from the current flexible inflation targeting regime.
    - QPM being specified and calibrated to approximate the newly introduced MPF, implying a rather active interest rate policy in the model that produces more reactive interest rate predictions for pre-2023 in-sample forecasts compared to historical values.
  - Expectation that interest rate forecast accuracy should improve going forward as QPM-based medium-term forecasts become a key input to monetary policy decisions within the new framework.

### Forecast accuracy: RMSFE relative to random walk benchmarks
- Methodology:
  - Compute root mean squared forecast errors (RMSFE) and compare to random walk model forecasts as a benchmark.
  - Random walk benchmark: assume next quarter value equals current quarter (useful for stationary variables).
- Table 2: QPM Mauritius RMSFE relative to random walk models (values shown are ratios of QPM RMSFE to random walk RMSFE)
  - CPI (%, qoq ann.): 1Q 0.98, 2Q 1.02, 3Q 1.01, 4Q 0.84, 5Q 0.90, 6Q 0.81, 7Q 0.72, 8Q 0.68
  - GDP (%, yoy): 1Q 0.67, 2Q 0.64, 3Q 0.63, 4Q 0.56, 5Q 0.60, 6Q 0.55, 7Q 0.57, 8Q 0.52
  - Nominal depreciation (%, qoq ann.): 1Q 1.34, 2Q 0.72, 3Q 0.65, 4Q 0.67, 5Q 0.69, 6Q 0.71, 7Q 0.64, 8Q 0.58
  - Interest rate (%): 1Q 2.09, 2Q 1.94, 3Q 1.73, 4Q 1.55, 5Q 1.44, 6Q 1.35, 7Q 1.25, 8Q 1.11
- Interpretation:
  - QPM-based inflation, GDP growth and exchange rate forecasts are significantly more accurate than the random walk benchmark, especially over the medium-term (relevant horizon for QPMs).
  - Interest rate forecasting accuracy is lower; the random walk is a reasonable approximation of historical interest rate behavior during the exchange rate-centric monetary policy framework.

### Conclusion and institutional implications
- FPAS and QPM role:
  - The Forecasting and Policy Analysis System (FPAS) implemented by the Bank of Mauritius (BOM) aligns well with the newly adopted inflation targeting monetary policy framework.
  - The upgraded FPAS enhances BOM’s institutional capabilities in model-based forecasting and policy evaluation, supports monetary policy decision-making, and strengthens communication.
  - The Quarterly Projection Model (QPM) is the core analytical tool for forecasts, policy-relevant scenarios, and assessing monetary policy effects in response to macroeconomic shocks.
- Model characteristics and calibration:
  - Mauritius QPM is a sophisticated extension of a canonical New Keynesian semi-structural macroeconomic model, customized for the Mauritian economy and current policy framework.
  - Calibration ensures theoretical coherency (intuitive shock propagation and policy response channels) and empirical fit, both necessary for credible analytical support.
- Practical benefits:
  - BOM’s framework enables exploration of a wide range of economic scenarios, evaluation of trade-offs during disturbances, and comparison of alternative policy paths to ensure price stability.
  - Additional satellite tools (near-term forecasting models and sector-specific analysis) provide judgment-based inputs.
  - BOM has strengthened forecasting capabilities and reorganized policy processes in line with best central bank FPAS practices; Mauritius QPM plays a pivotal role in reinforcing the Bank’s forward-looking monetary policy framework.

*Source: IMF Working Paper — Mauritius QPM: A Quarterly Projection Model for the Bank of Mauritius (Working Paper No. WP/2025/215).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025215-source-pdf.pdf_