## 1. Exchange Rates and Foreign Exchange Interventions

## Source details

**Canonical URL:** [1. Exchange Rates and Foreign Exchange Interventions](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020167-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020167-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020167-print-pdf.pdf.json)

---

### Introduction and recent evolution
- Before 2011, reserve money growth was mostly driven by accumulation of Net Foreign Assets (NFA).
- NFA increased by an annual average of 20 percent over 2001–10.
- Base money and broad money growth averaged 13 and 11 percent over 2001–10, respectively.
- Inflation averaged 3.4 percent over 2001–10.
- After the Revolution, NFA losses were offset by growing Net Domestic Assets (NDA).
- Monetary base remained stable until 2015; after the 2015 terrorist attacks the CBT’s balance sheet expanded due to accelerating demand for cash and foreign currency and substantial scaling-up of central bank refinancing operations.
- Inflation peaked in June 2018 at 7.7 percent as balance sheet expansion fueled broad money growth.

### Phases of post-2010 monetary history
- Three phases:
  - (1) Move to further exchange rate flexibility after the Arab Spring and emergence of banks’ structural liquidity deficit.
  - (2) Growing central bank refinancing after the 2015 terrorist attacks, loss of control over monetary aggregates, and ineffective rate hikes.
  - (3) More forceful policy tightening since 2018 to counter accelerating inflation.

### Monetary policy actions and outcomes
- Policy rate path and actions:
  - Policy rate increased in April 2017; policy rate rose from 5 percent in early 2018 to 7.75 percent in February 2019.
  - CBT increased policy rate several times to a maximum of 4.75 percent in June 2014; earlier decreased from 4.5 percent at end-2010 to 3.5 percent in September 2011.
  - CBT increased policy rate by 50 basis points in April 2017 (from 4.25 percent maintained since October 2015) and by 25 basis points in May 2017.
  - Cumulative policy rate increase since early 2018 amounted to 275 basis points.
  - Policy rate decreased by 100 basis points in March 2020 (Covid-19 shock).
- Liquidity injections and refinancing:
  - Total volume of liquidity injections (mainly 7-day MROs) grew from an average of TND 1.6 billion in January 2011 to TND 4.3 billion in mid-February 2012.
  - Total volume of CBT refinancing increased by 60 percent between July 2017 and July 2018 before stabilizing in August 2018.
- Transmission and trends:
  - Negative real interest rates and increasing refinancing accommodated money demand and demand for cash and foreign currency.
  - Monetary tightening beginning in 2017 contained inflation from July 2018.
  - Broad money growth halved from 2017 levels by mid-2018.
  - Credit growth to the economy decreased from 12.7 percent in 2017 to 8 percent in 2018 and 4.9 percent in August 2019.
  - Inflation declined from 7.7 percent in June 2018 to 5.8 percent in February 2020, then increased to 6.2 percent in March 2020.

### Structural liquidity, reserve money composition, and de-cashing
- Structural liquidity turned negative in 2011 due to FX outflows and FX interventions.
- Reserve requirement ratio reduced from 12.5 to 2 percent in May 2011, then to 1 percent in late 2013.
- Cap on outstanding 7-day MRO introduced in July 2017 widened spread between MMR and policy rate from an average 10 basis points (Jan 2015–Jul 2017) to an average 41 basis points after July 2017.
- Cap shifted refinancing into other instruments and increased TND liquidity injections through FX swaps.
- Reserve money composition (August 2019 snapshot):
  - Total reserve money TND 20.9 billion.
  - CIC (currency in circulation) TND 14.1 billion (68 percent).
  - Bank reserves in FX TND 6.5 billion (31 percent).
  - Bank reserves in dinar TND 0.3 billion (1 percent).
- FX accounts’ share in reserve money rose from 18 percent in 2011 to 35 percent in 2018.
- Authorities launched a de-cashing action plan in March 2018 to limit cash payments by the public administration and adopt a regulatory framework for electronic payments.

### Assessment of monetary policy frameworks and transmission
- Monetary targeting ineffective because reserve money is mainly composed of CIC and bank accounts in foreign currency.
- Money multiplier volatility increased after 2010 due to unstable money demand.
- Exchange rate targeting is no longer feasible given FX reserve levels, current account deficit, and inflation differentials with main trading partners.
- Progress toward Inflation Targeting (IT):
  - Implemented an interest rate mid-corridor system (200 basis points).
  - Developed macro-forecasting models including a medium-term Quarterly Projection Model (QPM) and short-term ARIMA models; DSGE development planned under 2019–21 strategic plan.
  - Introduced more competitive FX auctions since August 2018 and respected monthly net intervention limits under the EFF arrangement since December 2018; started to buy FX in March 2019.
  - CBT’s legal framework: cannot directly lend to the government (Article 25 of Law No. 2016-35 of April 25, 2016); implicit inflation target is four percent; O/N interbank rate (MMR) is the operational target per Circular No. 2017-02.
- Passthrough and transmission evidence:
  - Exchange rate passthrough long-term effect about 25-30 percent.
  - Share of FX accounts in reserve money rose from 18 percent (2011) to 35 percent (2018).
  - Long-run policy rate pass-through to money market rate:
    - January 10, 2015–July 11, 2017: long-run coefficient = 1.
    - July 12, 2017–April 30, 2019: long-run coefficient = 0.88.
  - Despite weakening to MMR, policy rate transmission to bank deposit and lending rates remained effective; deposit and lending rates co-move with CBT policy rate.

### Empirical evidence — VAR and passthrough estimates
- VAR setup: 2010Q1–2019Q3; variables: GR (real GDP growth), PI (inflation), USDTND (bilateral exchange rate to US dollar), BCT_RATE (CBT policy rate), M3 growth; oil price exogenous.
- Impulse response to a one standard deviation policy rate shock (+0.2 percentage points):
  - Inflation decreases by about 0.2 percentage points after six quarters.
  - Real GDP growth decreases by about 0.4 percentage points after six quarters.
- Variance decomposition (long-run):
  - Policy rate explains about 13 percent of inflation variance after 9 quarters.
  - Broad money growth explains about 13 percent of inflation variance.
  - Exchange rate fluctuations explain up to 40 percent of inflation variability over the long-run.
- Policy rate coefficient estimates (policy rate lag in inflation equation):
  - 2010Q1-2019Q3: Policy Rate Coefficient -0.35; P-value 0.0180
  - 2010Q1–2018Q4: Policy Rate Coefficient -0.39; P-value 0.0291
  - 2010Q1–2017Q4: Policy Rate Coefficient -0.33; P-value 0.1091
  - 2010Q1–2016Q4: Policy Rate Coefficient -0.18; P-value 0.4932
- Robustness: similar results using EURTND instead of USDTND and using TMM (money market rate) rather than BCT_RATE.
- Exchange rate passthrough regression (dynamic lag regression following Burstein and Gopinath (2014)):
  - β0 denotes instantaneous passthrough; B = Σβτ denotes long-term passthrough (long-term defined as passthrough after 12 months in moving-window analysis).
  - Main passthrough findings:
    - Short- and long-term passthroughs stronger during 2010–15.
    - Passthrough increased when CBT abandoned stabilized exchange rate arrangement.
    - Controlling for broad money growth makes passthrough stronger, especially in 2016–19.
    - Asymmetric responses: short-run appreciations reflected more in prices than depreciations; longer-run depreciations marginally more impactful.
- Selected regression table highlights (Appendix II. Table 1):
  - B (long-term passthrough) estimates:
    - Column (1) 2003-2019: B = 0.2357 *** (0.0569)
    - Column (2) 2010-2019: B = 0.2628 *** (0.0292)
    - Column (3) 2010-2015: B = 0.3590 *** (0.1241)
    - Column (4) 2016-2019: B = 0.2652 *** (0.0640)
    - Column (10) 2010-2019 (asymmetric): B = 0.4175 *** (0.1487)
    - Column (11) 2010-2015 (asymmetric): B = 1.0591 *** (0.2404)
  - beta0 estimates (instantaneous passthrough) include:
    - Column (2) 2010-2019: 0.0380 * (0.0210)
    - Column (3) 2010-2015: 0.0792 ** (0.0334)
    - Column (10) 2010-2019 (asymmetric): 0.1334 *** (0.0369)
    - Column (11) 2010-2015 (asymmetric): 0.3199 *** (0.0426)
  - Δ M3 coefficients reported: -0.0032 * (0.0017); -0.0030 ** (0.0014); -0.0031 * (0.0017); -0.0029 (0.0046)
  - Observations: 211 116 72 44 211 116 72 44 211 116 72
  - Significance notation: *, **, and *** represent significance at the 10, 5, and 1 percent levels.
  - Note: "Standard errors in parentheses. Lags are not reported but available on demand."

### Policy implications and suggested priorities
- Clarify the monetary policy framework to achieve further disinflation.
- Suggested priorities:
  - Strengthen central bank governance and interest rate transmission.
  - Improve communication and transparency; focus communications on inflation projections and outcomes.
  - Announce a medium-term inflation objective; an explicit numerical inflation target to operationalize the CBT’s price stability mandate.
  - Commit to a forward-looking rule-based monetary policy to reduce discretionary time-inconsistency problems.
  - Calibrate MRO volume with reference to autonomous factors’ forecasts to prevent full-allotment pressure on the exchange rate.
  - Maintain balanced collateral policy (since April 2020 collateralized at 50 percent with credits and 50 percent with government securities).
  - Upgrade analytical and macro-forecasting capacity (QPM exists; develop DSGE and an inflation expectations survey).
  - Use prudential policies as first line of defense against financial risks.
- De-cashing and payments modernization: continue efforts started March 2018 to limit cash payments by public administration and adopt regulatory framework for electronic payments.

### Conclusions — key messages
- Evidence consistent with long-run money neutrality and time-inconsistency problems when no clear nominal anchor exists.
- The 2016–18 central bank balance sheet expansion and increases in broad money and credit growth did not generate additional economic growth:
  - Real GDP growth averaged 1.8 percent during 2016–19.
  - Real GDP growth averaged 4.4 percent over 2007–10.
- Modernizing the monetary policy framework while maintaining exchange rate flexibility is expected to reduce real costs of disinflation by anchoring expectations.
- Move to Inflation Targeting is justified by a functioning interest rate transmission that can be strengthened with supportive communication and a clear nominal anchor.
- Tunisia’s financial sector development does not appear to constrain IT implementation; legal independence and prohibition on direct lending to government (Article 25 of Law No. 2016-35) are critical prerequisites.

*Source: wpiea2020167-print-pdf - 1. Exchange Rates and Foreign Exchange Interventions*

### 1. Exchange Rates and Foreign Exchange Interventions ........................................................ 10

### 1. Exchange Rates and Foreign Exchange Interventions

### Introduction and recent evolution
- Before 2011, Tunisia operated an exchange rate anchor under which reserve money growth was mostly driven by the accumulation of Net Foreign Assets (NFA).
- Supported by fiscal discipline, NFA increased by an annual average of 20 percent over 2001–10, which assured base money and broad money growth of an average of 13 and 11 percent over the same period, respectively, and contained inflation at an average of 3.4 percent.
- In the wake of the Revolution, the trend accumulation of NFA reverted; the central bank’s asset composition shifted toward growing Net Domestic Assets (NDA) that offset the losses of NFA.
- The monetary base remained stable until 2015. After the 2015 terrorist attacks, the CBT’s balance sheet started to expand steadily due to accelerating demand for cash and foreign currency accommodated by a substantial scaling-up of central bank refinancing operations.
- The balance sheet expansion fueled broad money growth and ultimately inflation that peaked in June 2018 at 7.7 percent.

### Phases of post-2010 monetary history
- The post-2010 decade is broken into three distinct phases:
  - (1) Move to further exchange rate flexibility in the aftermath of the Arab Spring and the emergence of banks’ structural liquidity deficit.
  - (2) Growing volume of central bank refinancing after the 2015 terrorist attacks that led to a loss of control over monetary aggregates and rendered ineffective initial attempts to contain inflation through policy rate hikes.
  - (3) More forceful policy tightening since 2018 to counter accelerating inflation.

### Monetary policy actions and outcomes
- The CBT first increased its policy rate in April 2017, but growing liquidity injections continued to pursue the objective of easing credit supply.
- Negative real interest rates, an increasing volume of refinancing, and the collateral policy accommodated money demand, especially demand for cash and foreign currency fueled by depreciation expectations.
- A clear change in the CBT’s monetary and macroprudential policies occurred in 2018:
  - The policy rate was increased in gradual steps from 5 percent in early 2018 to 7.75 percent in February 2019, slowing the demand for credit.
  - Macroprudential policy was tightened at the end of 2018 and constrained the supply of credit.
- Monetary tightening that began in 2017 contained inflation from July 2018.

### Assessment of monetary policy frameworks
- Monetary targeting has proven to be ineffective because reserve money is mainly composed of currency in circulation (CIC) and bank accounts in foreign currency.
- Money multiplier volatility increased after 2010 due to unstable money demand, reducing the effectiveness of monetary targeting.
- Exchange rate targeting is no longer feasible given the level of foreign exchange (FX) reserves, current account deficit, and inflation differentials with main trading partners.
- The CBT has made important progress toward inflation targeting (IT) by:
  - Implementing an interest rate mid-corridor system.
  - Developing macro-forecasting models.
  - Introducing FX auctions.

### Policy implications and suggested priorities
- Clarifying the monetary policy framework is critical to achieve further disinflation.
- Additional efforts could focus on:
  - Strengthening central bank governance and interest rate transmission.
  - Improving communication.
  - Upgrading the central bank’s analytical and forecasting capacity.

### Contributions of the paper
- Documents changing interest rate transmission and exchange rate passthrough with the move to further exchange rate flexibility.
- Adds the composition of reserve money and the structural liquidity position of the banking system to the list of criteria important for the choice of nominal anchor.

*Source: wpiea2020167-print-pdf - 1. Exchange Rates and Foreign Exchange Interventions*

### 1.1 Nominal and Effective Exchange Rates 1.2 Foreign Exchange Interventions

### 1.1 Nominal and Effective Exchange Rates 1.2 Foreign Exchange Interventions

### A. The Fall of the Nominal Anchor and Emerging Structural Liquidity Deficit
- Pre-Revolution: CBT de jure floating, de facto stabilized arrangement; AREAER classification changed over time (floating in 2016; crawl-like in May 2017).
- Structural liquidity:
  - Positive until 2010; turned negative in 2011 due to FX outflows and FX interventions (Figure 2.2).
  - Reserve requirement ratio reduced from 12.5 to 2 percent in May 2011, and then to 1 percent in late 2013.
- Policy rate and liquidity injections:
  - CBT increased policy rate several times to a maximum of 4.75 percent in June 2014.
  - Earlier, CBT decreased its policy rate from 4.5 percent in end-2010 to 3.5 percent in September 2011.
  - Total volume of liquidity injections (mainly 7-day MROs) grew from an average of TND 1.6 billion in January 2011 to TND 4.3 billion in mid-February 2012, de facto sterilizing FX sales in defense of the dinar.

### B. The Gradual Loss of Control Over Monetary Aggregates (2016–mid-2018)
- Balance sheet expansion:
  - CBT balance sheet expanded rapidly after the 2015 terrorist attacks driven by rising NDA and liquidity injections to support credit and FX/cash demand.
  - Total volume of CBT refinancing increased by 60 percent between July 2017 and July 2018 before stabilizing in August 2018.
- Exchange rate and pass-through:
  - Exchange rate passthrough estimations indicate a significantly positive long-term effect of about 25-30 percent.
  - FX accounts at the central bank: share of FX accounts in reserve money rose from 18 percent in 2011 to 35 percent in 2018.
- Inflation and policy actions:
  - Price stability became CBT’s main objective in 2016.
  - CBT increased policy rate by 50 basis points in April 2017 from a low 4.25 percent maintained since October 2015, and again by 25 basis points in May 2017.
  - Monetary tightening was partly counteracted by increases in refinancing operations and multiple objectives (price stability vs. credit/financial stability).
- Distortions to transmission:
  - Cap on outstanding volume of the 7-day MRO introduced in July 2017 weakened passthrough from the policy rate to the money market rate (MMR), widening the spread between MMR and policy rate.
  - Spread increased from an average 10 basis points over January 2015 to July 2017 to an average 41 basis points after July 2017.
  - Cap moved refinancing into other instruments (overnight lending facility) and increased TND liquidity injections through FX swaps.

### C. Monetary Policy Orthodoxy to the Rescue over 2018–19
- Stronger tightening and results:
  - CBT increased policy rate by a cumulative 275 basis points since early 2018.
  - Policy tightening made the policy rate positive in real terms and was combined with containing refinancing volumes (partly via tightening of the Loan-to-Deposit (LTD) ratio).
  - Central bank started buying FX against TND as expectations shifted from depreciation to appreciation, contributing to lower refinancing volumes.
- Early outcomes:
  - Broad money growth halved from its 2017 levels by mid-2018.
  - Credit growth to the economy decreased from 12.7 percent in 2017 to 8 percent in 2018 and 4.9 percent in August 2019.
  - Inflation declined from a peak of 7.7 percent in June 2018 to 5.8 percent in February 2020.
  - Following a policy rate decrease of 100 basis points in March 2020 (Covid-19 shock), inflation increased to 6.2 percent in March 2020. Resurgence likely due to supply shocks (food prices) and extension of loan maturities increasing household purchasing power by about 40 percent.
- Reserve money targeting challenges (August 2019 snapshot):
  - Total reserve money TND 20.9 billion.
  - CIC (currency in circulation) TND 14.1 billion (68 percent).
  - Bank reserves in FX TND 6.5 billion (31 percent).
  - Bank reserves in dinar TND 0.3 billion (1 percent).
  - Bank reserves in TND remain very low since 2011; DIA and MMD (FX accounts) rose over time, complicating reduction of reserve money which would require shrinking CIC and/or FX accounts.
- De-cashing policy:
  - Authorities launched a de-cashing action plan in March 2018 to limit cash payments by the public administration and adopt a regulatory framework for electronic payments.

### IV. Monetary Policy Transmission — Empirical Evidence and Passthrough
A. Monetary Transmission to Inflation and Output (VAR evidence)
- VAR setup: 2010Q1–2019Q3 with variables GR (real GDP growth), PI (inflation), USDTND (bilateral exchange rate to US dollar), BCT_RATE (CBT policy rate), M3 growth; oil price included as exogenous variable.
- Impulse responses:
  - A one standard deviation policy rate shock (+0.2 percentage points) leads to a decrease of inflation of about 0.2 percentage points and a decrease of real GDP growth of about 0.4 percentage points after six quarters.
- Strengthening transmission:
  - Coefficient on policy rate lag in the inflation equation is statistically significant at 95 percent over 2010Q1–2019Q4 and its statistical significance has increased over time (Table 1).
- Variance decomposition of inflation (long-run):
  - Policy rate explains about 13 percent of inflation variance (after 9 quarters).
  - Broad money growth explains about 13 percent of inflation variance.
  - Exchange rate fluctuations explain up to 40 percent of inflation variability over the long-run.
- Variance decomposition of real GDP growth:
  - Broad money plays a very limited role in explaining economic growth developments (consistent with long-run money neutrality).
- Table 1 (policy rate coefficient and p-values by estimation period):
  - 2010Q1-2019Q3: Policy Rate Coefficient -0.35; P-value 0.0180
  - 2010Q1–2018Q4: Policy Rate Coefficient -0.39; P-value 0.0291
  - 2010Q1–2017Q4: Policy Rate Coefficient -0.33; P-value 0.1091
  - 2010Q1–2016Q4: Policy Rate Coefficient -0.18; P-value 0.4932
- Robustness: Similar results when using EURTND instead of USDTND and using TMM (money market rate) rather than BCT_RATE.

B. Passthrough from the Policy Rate to the Money Market Rate
- Estimation method: following Gigineishvili (2011) with daily data; long-run pass-through coefficient computed from short-run elasticities and persistence terms.
- Long-run pass-through estimates:
  - Over January 10, 2015–July 11, 2017: long-run policy rate pass-through coefficient = 1.
  - Over July 12, 2017–April 30, 2019 (post-cap on outstanding 7-day MRO): long-run pass-through coefficient = 0.88 (weakened).
- Transmission to bank retail rates:
  - Despite weakening to MMR, transmission from policy rate to bank deposit and lending rates is effective — bank deposit and lending rates co-move with the CBT’s policy rate (Figure 19).

*Source: Tunisian authorities and authors’ calculations (content unit: 1.1 Nominal and Effective Exchange Rates; 1.2 Foreign Exchange Interventions).*

### 17.1 Deposit Rates

### 17.1 Deposit Rates

### Exchange Rate Passthrough to Inflation
- Framework estimated following Burstein and Gopinath (2014) using a dynamic lag regression:
  - Δp_t = α + Σ_{τ=0}^{N} β_τ Δs_{t−τ} + C.X_t + ε_t
  - p_t is the CPI, s_t is the dinar per US$ exchange rate, X_t a vector of control variables (monetary developments: growth in broad money; production costs: changes in oil prices in US$).
  - Coefficients of interest:
    - β_0: instantaneous passthrough.
    - Β = Σ β_τ: long-term passthrough (long-term defined as passthrough after 12 months in the moving-window analysis).
- Estimation frequency and sample:
  - Monthly frequency to isolate subperiods around the Jasmine Revolution and to cover 2011–19 with enough observations.
  - Moving-window estimation: 4-year moving regression window; the value for year y is the β̂_0 from January (y−4) – December y.

### Main Findings on Passthrough
- Three primary results:
  - Passthrough strength by period:
    - Both short- and long-term passthroughs were stronger during 2010–15 (the first years after the change in the exchange rate arrangement).
    - The exchange rate passthrough increased when the CBT abandoned the stabilized exchange rate arrangement.
  - Role of monetary expansion:
    - Controlling for growth of broad money makes the passthrough somewhat stronger, especially in the long run.
    - This effect is particularly visible in the 2016–19 period that saw more pronounced monetary expansion, suggesting monetary expansion resulted in a stronger effect of exchange rate depreciation on inflation.
  - Asymmetric price responses:
    - In the short run, appreciations tend to be reflected more in prices than depreciations (likely due to the large share of regulated prices).
    - In the longer run, depreciations are marginally more impactful than appreciations.
- Robustness:
  - Regression results are in Appendix II.
  - As a robustness check, estimations without the oil price variable produced similar results.

### Evidence and Supporting Figures
- Figure 20 (dynamic estimation of passthrough):
  - Shows estimated short-term (ST) and long-term (LT) passthrough coefficients with 95% confidence intervals for 2010–2019 using a 4-year moving regression window.
  - Note: long-term passthrough in this figure is defined as passthrough after 12 months to accommodate smaller sample sizes.

### The Case for Inflation Targeting (IT)
- Rationale:
  - The CBT is not a traditional monetary targeter; prior to August 2018 liquidity injections targeted credit growth rather than reserve/broad money consistent with low and stable inflation.
  - Liquidity injections collateralization:
    - Until very recently: collateralized at 60 percent with credits against 40 percent with government securities.
    - Since April 2020: collateralized at 50 percent with credits against 50 percent with government securities.
  - Tunisia’s credit-to-GDP ratio is among the highest in the region and compared to other emerging market economies (Figure 21).
  - Since 2018, success in reverting accelerating inflation through policy rate hikes and moving key interest rates into positive territory supports a shift toward an interest rate-based framework consistent with IT.
- Preconditions already in place:
  - Legal and operational features consistent with IT:
    - The CBT cannot directly lend to the government (Article 25 of Law No. 2016-35 of April 25, 2016).
    - Central bank legal mandate is to ensure price stability; implicit inflation target is four percent.
    - O/N interbank rate (MMR) is the operational target per regulation (Circular No. 2017-02).
    - CBT operates a 200 basis points interest rate mid-corridor system and has short-term forecasting models and a medium-term Quarterly Projection Model (QPM).
  - Communication improvements:
    - Monetary policy report “Evolutions Economiques et Monétaires et Perspectives à Moyen Terme” published more regularly since October 2018 and includes CBT inflation projections.
    - May 2019 report upgraded with a new section on international developments.
  - FX auction practices:
    - More competitive FX auctions introduced in August 2018; since December 2018 auctions have become more frequent with smaller volumes.
    - CBT respected monthly net intervention limits under the EFF arrangement since December 2018 and started to buy FX in March 2019.
    - Competitive auctions and lower intervention volumes contributed to accelerated depreciation in Q4 2018; depreciation trend reverted in March 2019 due to tighter monetary policy and stronger tourism receipts and privatization-related FX flows until February 2020.

### Next Steps to Support the Transition to Inflation Targeting
- Recommended areas of focus:
  - Commit to a forward-looking rule-based monetary policy:
    - Reduce discretionary policy to lower time-inconsistency problems; establish a clear nominal anchor.
  - Announce a medium-term inflation objective:
    - An explicit numerical inflation target would operationalize the CBT’s price stability mandate; the objective must be achievable and, over time, achieved to be credible.
  - Strengthen interest rate transmission:
    - Calibrate MRO volume with reference to autonomous factors’ forecasts and to prevent full-allotment pressure on the exchange rate.
    - Collateral policy already more balanced (50 percent government securities) to protect the central bank balance sheet.
  - Upgrade analytical capacity and macro-forecasting models:
    - CBT has a QPM and ARIMA short-term models; under the 2019–21 strategic plan CBT is developing a DSGE model and could develop an inflation expectations survey.
  - Strengthen communications, transparency, and accountability:
    - Focus communications more on inflation projections and outcomes and explain the integrated monetary policy framework to reduce uncertainty and improve policy transmission and accountability.

### Conclusion — Key Messages
- Main conclusions:
  - Findings are consistent with long-run money neutrality and time-inconsistency problems of discretionary monetary policy without a clear nominal anchor.
  - The 2016–18 central bank balance sheet expansion and simultaneous increases in broad money and credit growth did not generate additional economic growth:
    - Real GDP growth averaged 1.8 percent during 2016–19, a level similar to 2011–15.
    - Real GDP growth averaged 4.4 percent over 2007–10 (a period when the CBT was not conducting liquidity injection operations at its initiative).
  - Modernizing the monetary policy framework while maintaining exchange rate flexibility is expected to reduce the real costs of disinflation by anchoring expectations.
  - Move to IT is justified by an already functioning interest rate transmission that can be strengthened with supportive communication and a clear nominal anchor.
  - Financial development level does not appear to constrain IT implementation; Tunisia’s financial sector is more developed than some countries that are already inflation targeters.
  - The CBT’s legal independence and prohibition on direct lending to the government (Article 25 of Law No. 2016-35) are critical prerequisites for a successful transition to IT and for avoiding fiscal dominance.

*Source: Central Bank of Tunisia and authors’ calculations.*

### Appendix II. Figure 2. VAR with the Euro Exchange Rate and the Money Market Rate—Responses of Inflation and

### Appendix II. Figure 2. VAR with the Euro Exchange Rate and the Money Market Rate—Responses of Inflation and Output to a Money Market Rate Shock

### Impulse responses (VAR with Euro exchange rate and money market rate)
- Responses shown for horizons: 2, 4, 6, 8, 10, 12, 14, 16, 18, 20 (horizontal axis labels appear as "2   4   6   8   10  12  14  16  18  20").
- Variables with impulse-response panels:
  - GR (growth) — Responses of GR to shocks in: GR, PI, EURTND, TMM_RATE, M3_YOY.
    - Axis ranges indicated in panels: from -0.5 to 1.0 for some GR panels (labels: "-0.5 0.0 0.5 1.0").
  - PI (inflation) — Responses of PI to shocks in: GR, PI, EURTND, TMM_RATE, M3_YOY.
    - Axis ranges indicated in panels: from -0.4 to 0.4 for some PI panels (labels: "-.4 -.2 .0 .2 .4").
  - EURTND (Euro/TND exchange rate) — Responses of EURTND to shocks in: GR, PI, EURTND, TMM_RATE, M3_YOY.
    - Axis ranges indicated in panels: from -0.1 to 0.1 for some EURTND panels (labels: "-.1 .0 .1").
  - TMM_RATE (money market rate) — Responses of TMM_RATE to shocks in: GR, PI, EURTND, TMM_RATE, M3_YOY.
    - Axis ranges indicated in panels: from -0.4 to 0.4 for some TMM_RATE panels (labels: "-.4 -.2 .0 .2 .4").
  - M3_YOY (M3 year-on-year) — Responses of M3_YOY to shocks in: GR, PI, EURTND, TMM_RATE, M3_YOY.
    - Axis ranges indicated in panels: from -1.0 to 1.0 for some M3_YOY panels (labels: "-1.0 -0.5 0.0 0.5 1.0").
- Responses computed using Cholesky identification:
  - "Response to Cholesky One S.D. (d.f. adjusted) Innovations ± 2 S.E."
- Sources: Tunisian authorities and authors’ calculations.

### Variance decomposition (Appendix II. Figure 3)
- Variance decomposition using Cholesky factors (d.f. adjusted).
- Panels report percent of variance explained at horizons 2, 4, 6, 8, 10, 12, 14, 16, 18, 20 for each variable by shocks to each variable.
- Variables and decomposition targets listed:
  - GR variance due to: GR, PI, USDTND, BCT_RATE, M3_YOY.
  - PI variance due to: GR, PI, USDTND, BCT_RATE, M3_YOY.
  - USDTND variance due to: GR, PI, USDTND, BCT_RATE, M3_YOY.
  - BCT_RATE variance due to: GR, PI, USDTND, BCT_RATE, M3_YOY.
  - M3_YOY variance due to: GR, PI, USDTND, BCT_RATE, M3_YOY.
- Percent axis labeled: "0 20 40 60 80 100".
- Sources: Tunisian authorities and authors’ calculations.

### Regression results (Appendix II. Table 1)
- Sample periods and columns:
  - Columns labelled by sample windows: 2003-2019, 2010-2019, 2010-2015, 2016-2019 (repeated across model specifications).
  - Three model variants: Baseline; With money variable; With asymmetric effects.
- Selected coefficient estimates and standard errors (parentheses) preserved exactly as in table:
  - Δ Oil:
    - Column (1) 2003-2019: 0.0007 (0.0022)
    - Column (2) 2010-2019: -0.0008 (0.0032)
    - Column (3) 2010-2015: -0.0065 * (0.0033)
    - Column (4) 2016-2019: 0.0030 (0.0078)
    - Column (5) 2003-2019 (with money variable): 0.0006 (0.0023)
    - Column (6) 2010-2019: -0.0012 (0.0030)
    - Column (7) 2010-2015: -0.0071 ** (0.0031)
    - Column (8) 2016-2019: 0.0031 (0.0071)
    - Column (9) 2003-2019 (asymmetric): 0.0011 (0.0024)
    - Column (10) 2010-2019: -0.0065 ** (0.0030)
    - Column (11) 2010-2015: -0.0159 * (0.0077)
  - beta0:
    - Column (1): 0.0197 (0.0172)
    - Column (2): 0.0380 * (0.0210)
    - Column (3): 0.0792 ** (0.0334)
    - Column (4): 0.0108 (0.0274)
    - Column (5): 0.0213 (0.0172)
    - Column (6): 0.0426 ** (0.0175)
    - Column (7): 0.0802 ** (0.0308)
    - Column (8): 0.0179 (0.0353)
    - Column (9): 0.0495 (0.0339)
    - Column (10): 0.1334 *** (0.0369)
    - Column (11): 0.3199 *** (0.0426)
  - betaDepr0 (reported only for some specifications):
    - Values: -0.0409 (0.0446); -0.1219 ** (0.0566); -0.2783 *** (0.0743)
  - B:
    - Column (1): 0.2357 *** (0.0569)
    - Column (2): 0.2628 *** (0.0292)
    - Column (3): 0.3590 *** (0.1241)
    - Column (4): 0.2652 *** (0.0640)
    - Column (5): 0.2392 *** (0.0549)
    - Column (6): 0.2684 *** (0.0308)
    - Column (7): 0.3534 *** (0.1267)
    - Column (8): 0.2721 *** (0.0785)
    - Column (9): 0.1751 (0.1602)
    - Column (10): 0.4175 *** (0.1487)
    - Column (11): 1.0591 *** (0.2404)
  - Bdepr:
    - Values reported: 0.0720 (0.1863); -0.1919 (0.2066); 0.0166 (0.8162)
  - Δ M3:
    - Values reported in last block: -0.0032 * (0.0017); -0.0030 ** (0.0014); -0.0031 * (0.0017); -0.0029 (0.0046)
  - Constant:
    - Multiple entries, examples:
      - 0.0024 *** (0.0003)
      - 0.0028 *** (0.0002)
      - 0.0026 *** (0.0005)
      - 0.0024 *** (0.0007)
      - 0.0024 *** (0.0003)
      - 0.0027 *** (0.0002)
      - 0.0025 *** (0.0005)
      - 0.0024 *** (0.0008)
      - 0.0022 *** (0.0008)
      - 0.0035 *** (0.0010)
      - 0.0010 (0.0038)
- Observations:
  - "Observations 211 116 72 44 211 116 72 44 211 116 72"
- Significance notation:
  - "* ,  **,  and  *** represent significance at the 10, 5, and 1 percent levels."
- Note: "Standard errors in parentheses. Lags are not reported but available on demand."
- Model labels at table foot: "Baseline", "With money variable", "With asymmetric effects".

### Appendix III. Definitions and Frameworks

### Monetary policy frameworks and anchors
- AREAER defines three frameworks with a clearly identified nominal anchor: (1) Exchange Rate Anchor; (2) Monetary Aggregate Target; and (3) IT Framework (IMF, 2018a).
- Countries with no explicitly stated nominal anchor but monitoring various indicators are classified as "other" by AREAER.
- Exchange Rate Anchor:
  - Monetary authority buys/sells FX to maintain exchange rate at predetermined level or within a range.
  - Associated arrangements: no separate legal tender, currency board, pegs (stabilized arrangements) with/without bands, crawling pegs (or crawl-like), other managed arrangements.
- Monetary Aggregate Targeting:
  - Authority uses instruments to achieve target growth rate for a monetary aggregate (reserve money, M1, M2); targeted aggregate becomes nominal anchor or intermediate target.
- IT (Inflation Targeting):
  - Public announcement of numerical inflation targets; institutional commitment to achieve targets typically over a medium-term horizon.
  - Key features: increased communication, increased accountability; policy guided by deviation of inflation forecast from target, with forecast acting as intermediate target.

### Monetary Policy Frameworks—The Classics (Appendix III. Table 1)
- Inflation Targeting:
  - Policy Objective: Inflation
  - Intermediate Target: Inflation forecast
  - Operating target: Interest rate
  - Instruments: OMOs, standing facilities, Reserve Requirements
- Monetary Targeting:
  - Policy Objective: Inflation/Price Stability
  - Intermediate Target: Broad money / Reserve money
  - Instruments: OMOs, standing facilities, Reserve Requirements
- Exchange Rate Anchor:
  - Policy Objective: Exchange Rate/Price Stability
  - Intermediate Target/Operating target: Exchange Rate
  - Instruments: Interest Rate, FX Interventions, Liquidity management operations

### Appendix IV. Primacy of Price Stability Versus Multiple Objectives
- Multiple-objective targeting may erode effectiveness and credibility of monetary policy.
- Case study: India
  - 1998: adopted "multiple indicator approach" with no explicit nominal anchor.
  - Persistent high inflation led to adoption of Flexible IT in early 2014; officially adopted by Reserve Bank of India in February 2015.
  - CPI inflation target: medium-term objective of 4 percent with a band of +/- 2 percent; two-year "glide path" to reduce inflation toward target.
  - Lesson: a clear inflation objective can strengthen policy communication and anchor expectations.
- Credibility implications:
  - Sustained deviations from inflation objective can cause loss of credibility and weaken central bank’s ability to deliver price stability.
  - Less credible regimes may need to prioritize price stability and respond aggressively to inflation shocks.
- Flexibility with credibility:
  - Credible central banks have more room to manage trade-offs and may treat sharp inflation increases as temporary (example: Czech National Bank cut rates in mid-2008 despite inflation almost twice target).
  - International evidence: oil-price shocks of the 1970s had larger effects than post-1990s; commitment to low inflation improved trade-offs (Blanchard and Gali, 2007).
- Policy implication:
  - Keep price stability as focal point of monetary policy to anchor inflation expectations and allow better management of trade-offs.
  - Use prudential policies as first line of defense against financial risks.
- Source cited: IMF (2015).

### Appendix V. Tunisia Financial Development—Comparison to Peers
- Appendix V includes:
  - Figure 1: Financial Development Index of MCD Countries (2017).
  - Figure 2: Financial Institutions Development in MCD Countries (2017).
- Source: Financial Development Index Database.

### Appendix VI. Selected Countries’ Monetary Policy/Inflation Reports—Main Content
- Table lists central banks, report titles & frequency, publication date, and which core content items appear in reports.
- Core content fields include: International developments; Inflation developments & forecasts; Economic developments & forecasts; Labor Market; Financial Markets/Conditions; Fiscal Policy Assumptions & risks; Inflation Fan Chart/Forecast; GDP Fan Chart/Forecast.
- Examples from the table:
  - Central Bank of Tunisia — "Evolutions Economiques et Monétaires et Perspectives à Moyen Terme" — Quarterly — May 2019
    - Content ticks: International developments, Inflation developments & forecasts, Economic developments & forecasts, Financial Markets/Conditions, Fiscal Policy Assumptions & risks.
  - Bank of Ghana — "Monetary Policy Report" — 4-6 per year — March 2019
    - Content ticks: International developments, Inflation developments & forecasts, Economic developments & forecasts, Financial Markets/Conditions.
  - Bank of Canada — "Monetary Policy Report" — Quarterly — July 2019
    - Content ticks: International developments, Inflation developments & forecasts, Economic developments & forecasts, Financial Markets/Conditions, Fiscal Policy Assumptions & risks.
  - Reserve Bank of India — "Monetary Policy Report" — Biannual — April 2019
    - Content ticks: International developments, Inflation developments & forecasts, Economic developments & forecasts, Labor Market, Financial Markets/Conditions, Fiscal Policy Assumptions & risks, Inflation Fan Chart/Forecast, GDP Fan Chart/Forecast.
- Sources: Central banks’ websites and the authors.

*Sources: Tunisian authorities and authors’ calculations; Financial Development Index Database; central banks’ websites and the authors; IMF and cited references as listed in source document.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020167-print-pdf.pdf_