What Can We Learn About the Monetary Policy Transmission Mechanism? Evidence from a Peripheral Country After a Political Revolution and COVID-19

Abstract

Interest in empirical studies of monetary policy has grown over the past decade, and particularly since the post COVID-19 pandemic period characterized by a surge in inflation rates in every corner of the globe. Against this backdrop, central banks’ traditional inflation forecast framework has been challenged, leading to renewed analysis of the monetary policy transmission mechanism. Focusing on Tunisia, an emerging small open economy subjected to external shocks, this study focuses on the role played by the monetary authority in the conduct of Tunisia’s monetary policy over the period from 2000 to 2024. This period is characterized by a deceleration of growth and an increase in inflation and unemployment. This work shows also how a VAR model with long-run restrictions justified by economic theory can be usefully applied in the analysis of monetary policy; the effects of the money market rate and other shocks; the relationship between prices and the nominal effective exchange rate; and the relationship between inflation and the output gap.

1. Introduction

The COVID-19 pandemic has profoundly reshaped the global economic and financial landscape. Beyond its direct health implications, it has created a persistent environment of uncertainty that continues to weigh on growth prospects and macroeconomic stability. One of the most pressing challenges for central banks in this context has been the resurgence of inflation. After decades of relative price stability, inflation has returned to levels unseen since the 1980s in many parts of the world. This sudden acceleration has revived old debates about the effectiveness of monetary policy, the credibility of central banks, and the channels through which policy actions influence the real economy. In particular, the capacity of monetary authorities to act as providers of a public good, namely macroeconomic and financial stability, has been put to a severe test (Best, 2024).

Emerging and small open economies have been especially exposed to these developments. They are more vulnerable to fluctuations in global commodity markets and exchange rates, and they often face additional difficulties in maintaining policy credibility under external shocks (Anand & Gulati, 2021; Petersen, 2024). At the same time, the well-known but still unresolved “inflation–openness puzzle” in international macroeconomics (Temple, 2002; Furuoka et al., 2023; Yildirim & Eryigit, 2025; Dardouri et al., 2025) underscores the fact that countries do not experience the same inflationary dynamics when exposed to global integration. These differences highlight the importance of country-specific analyses in order to understand how monetary policy is transmitted under conditions of heightened external dependence and uncertainty.

Tunisia offers a particularly relevant case in this regard. Its economy is structurally dependent on imports of energy, food, and raw materials, making it highly sensitive to international price shocks. The recent trajectory of inflation reflects these vulnerabilities. After peaking at an average of 9.3% in 2023, inflation moderated to 7% in 2024 and fell further to 6% at the beginning of 2025 (TCB, 2025). While this decline is encouraging, it remains fragile, as it was largely driven by the temporary easing of food prices. Future markets for key imported commodities still point to sustained pressures, leaving policymakers in a delicate position. The Tunisian Central Bank (TCB) has therefore been under significant pressure to balance its dual mandate of preserving price stability and safeguarding financial stability, while also supporting the government’s broader economic policy objectives.

In this context, understanding the mechanisms through which monetary policy decisions affect inflation and output becomes crucial. For a central bank such as the TCB, credibility depends not only on its ability to control inflation but also on its capacity to anticipate and explain the channels through which its interventions work. As highlighted in earlier studies (Aguir, 2016, 2018; Mimoun et al., 2024; Masmoudi et al., 2025), the main levers in Tunisia remain the interest rate channel and the exchange rate channel. Both have shaped the trajectory of monetary policy and continue to determine how domestic conditions respond to global shocks. Recent research highlights that the effects of monetary policy are not constant but evolve over time, reflecting changes in the economic environment, policy regimes, and market expectations. For instance, the response of treasury yields to monetary policy shocks has been shown to vary significantly across periods, providing evidence of a time-varying transmission mechanism (Baumeister & Benati, 2012; Gürkaynak et al., 2005). This literature suggests that monetary policy should be interpreted in a dynamic framework, even though our baseline model adopts a reduced-form VAR with fixed parameters.

Yet, despite their importance, the empirical evidence on the effectiveness of these channels in the post-pandemic inflationary environment remains scarce.

While earlier studies have examined aspects of monetary policy transmission in Tunisia (Aguir, 2018; Mimoun et al., 2024; Masmoudi et al., 2025), this paper distinguishes itself in several ways. First, it covers a more recent period marked by the COVID-19 crisis and the subsequent resurgence of inflation, offering new insights into how global shocks reshape the transmission process. Second, it relies on monthly data, which allows us to capture short-term dynamics that may be overlooked in analyses based on lower-frequency observations. Third, the empirical strategy builds on a structural VAR framework that explicitly identifies monetary and exchange rate shocks, rather than relying solely on correlations or reduced form relationships. Finally, the focus on the two main transmission channels—the interest rate channel and the exchange rate channel—enables us to evaluate their relative strength in a small open economy where external dependence remains a central feature. Taken together, these elements provide a sharper and more up-to-date understanding of the Tunisian monetary policy experience.

The remainder of the paper is organized as follows. Section 2 reviews the theoretical and empirical literature on monetary policy transmission channels, with emphasis on emerging and small open economies. Section 3 describes the empirical methodology, focusing on the VAR model and identification strategy. Section 4 presents the results of the empirical analysis, including correlation patterns, impulse responses, and the interpretation of shocks. Section 5 concludes with a summary of key findings and discusses policy implications for the Tunisian Central Bank.

2. The BCT’s Monetary Framework Under Pressure

It is worth starting with a reminder that the main mission of the monetary policy operated by the TCB has consisted in preserving price stability since 2006 and in ensuring financial stability since 2016. The contribution to job creation via non-inflationary growth on the one hand and to the improvement of social welfare through the preservation of purchasing power on the other hand are behind the pursuit of this ultimate goal. This is consistent with the TCB’s “special” mandate: although its main objective is to preserve price stability, there is no precise target assigned to the bank. Therefore, it is of utmost importance for the institution to be credible in its commitment to its main objective in creating the conditions of forward guidance for its monetary policy.

Therefore, in order to influence the money market rate and inflation-related indicators to properly simulate the influence of changes in the key rate on the main economic indicators, this monetary policy framework is based on instruments. Since 2013, the TCB has been using the “GPM” global projection model to forecast short-term inflation.

Monetary policy in Tunisia faces persistent challenges such as inflation and exchange rate fluctuations. To meet these challenges, the BCT adopts strategies such as adjusting interest rates and controlling monetary aggregates. These strategies aim to balance price stability with the promotion of economic growth and international competitiveness Aguir (2013), Aguir (2016), Aguir et al. (2017), Aguir and Smida (2015), Dardouri et al. (2023a), but these measures have had undesirable effects, such as a significant slowdown in private investment and consumption. Moreover, the negative consequences were accentuated by the continued depreciation of the exchange rate Dardouri et al. (2023b). Analysis suggests that the interest rate is more strongly influenced by the exchange rate than by inflation, highlighting a divergence between official objectives and the BCT’s operational reality. Finally, it is noted that exchange rate depreciation alone is not enough to resolve the structural trade deficit. The nominal effective exchange rate of the dinar depreciated by 3.4% in April 2024 compared with the same month the previous year. Similarly, the real exchange rate depreciated by 2.9%, in line with the average inflation differential with partner countries (TCB, 2025). On an annual average, the dinar depreciated by 6.5% against the dollar in 2024 compared with the previous year, while it appreciated by 0.8% against the euro (TCB, 2025).

The choice of using the exchange rate as a target variable is motivated by its direct correlation with prices. Indeed, both a nominal and real appreciation of the exchange rate can influence relative prices, as it tends to reduce demand for domestic goods whose prices become higher than those of imported goods, thus affecting aggregate demand (Yamani, 2012). Inflation management achieved through a monetary policy focused on targeting the exchange rate has been a short-term measure. Indeed, in an environment of capital mobility, countries that target the exchange rate find themselves unable to respond to domestic economic shocks that differ from those affecting the reference country, whether endogenous or exogenous. This underlines the loss of independence of monetary policy in such a context of capital mobility (Ftiti et al., 2017; Neifar & Smaoui, 2025).

Starting with the first instrument “Required reserves”, the latter helps to regulate bank liquidity and to continue to increase the money supply in line with economic growth. Banks are therefore obliged in this context to constitute a reserve at the central bank as a percentage of the amount of sight deposits, certificates of deposit, time deposits, and special savings accounts, with the exception of home savings deposits, investments, and projects.

We also find the operations at the initiative of the central bank, which can be classified into four categories:

-

The main refinancing operations where the interest rate is close to the policy rate, which can signal the monetary policy stance and help to steer the interest rate.

-

Longer-term refinancing operations to provide additional liquidity.

-

Fine-tuning operations in order to mitigate large fluctuations in liquidity on the interbank market and thus on the overnight rate.

-

Structural operations to resolve a structural liquidity deficit or surplus in the interbank market.

And finally, the operations at the initiative of banks where the TCB offers standing facilities at the initiative of banks, including the marginal lending facility to obtain liquidity at twenty-four hours and the deposit facility to make deposits at twenty-four hours also, allows withdrawal and provides liquidity overnight. Two interest rates applied to the standing facilities are decided by the TCB, where it can modify the conditions. The ceiling of these rates is the rate on the marginal lending facility, and the floor rate is on the deposit facility; the overnight interbank rates fluctuate within this corridor.

Following the financial crisis of 2008, the growth rate in Tunisia has marked a drop from 6.3% in 2007 to 3% in 2010, despite the reduction in the policy rate of the BCT, which has been unable to boost private growth and investment. The slowdown was more dramatic in 2011, with a negative growth of −1.9% after the events of the revolution, although the year 2012 was marked by a slight economic recovery with 3.9%. The increases in lower demand, private consumption, and investment was the reason for the recovery in 2012. However, the continuation of this performance was not possible, and the growth rate of GDP at constant prices in 2010 did not stop its decline; it eventually reached 1% in 2016 under the impact of a change in investment, exports, and the fall of private consumption. Besides that, a negative growth rate of the indices of the volume of exports and imports of −14.3% and −15.3%, respectively, against 9% and 13.4% in 2008, was marked during the year 2009.

An improvement in these two indices following the recovery of economic growth in 2012 was recorded; however, the export index has always shown a negative growth rate except during 2015. Since 2011, the evolution of GDP has strongly correlated with that of industry and services.

Since 2008, the unemployment rate has also increased internationally; it has spread through the financial channel from the United States to Europe, and then to Tunisia via the trade channel, given that our main trading partners are European countries.

Since 2011, the Central Bank of Tunisia (TCB) has gradually increased its policy rate in response to mounting inflationary pressures. Although the interest rate corridor was widened in early 2013, this move did not lead to a reduction in the policy rate, as the Bank prioritized price stability and liquidity support for the banking sector. The successive hikes observed between March 2013 and June 2014 encouraged savings among households and entrepreneurs, even as they coincided with a slowdown in economic growth, particularly in 2013. In the years that followed, the TCB continued to raise its benchmark rate whenever inflation picked up, demonstrating a consistent tightening approach. After a period of relative pause, the Bank once again intervened decisively, lifting the policy rate to 7.5% in March 2025. As illustrated in

Table 1. Decisions of the Board of Directors of the Central Bank of Tunisia (TCB).

Date	Deposit Facility Rate (%)	TCB Policy Rate (%)	Lending Facility Rate (%)
29/06/2011	3.5	4.0	4.5
05/09/2011	3.0	3.5	4.0
29/08/2012	3.25	3.75	4.25
27/02/2013	3.5	3.75	4.5
27/03/2013	3.75	4.0	4.75
25/12/2013	4.25	4.5	4.75
25/06/2014	4.5	4.75	5.0
28/10/2015	4.0	4.25	4.5
25/04/2017	4.5	4.75	5.0
23/05/2017	4.75	5.0	5.25
27/12/2017	4.0	5.0	6.0
05/03/2018	5.75	4.75	6.75
20/06/2018	6.75	5.75	7.75
27/03/2025	6.5	7.5	8.5

Sources: TCB Financial Statistics Report 2025 (TCB, 2025).

, the monetary policy corridor widened considerably between 2013 and 2025, reflecting the Bank’s ongoing effort to anchor inflation expectations and preserve macroeconomic stability.

However, the instability increased in 2015 as a result of terrorist attacks, hence the reduction in tourism revenues and value added services. In this situation, the increase in the TCB’s policy rate only affected private consumption and investment, and the reduction in the latter caused the decline in GDP growth. The fall in FDI flows, and the increase in the cost of imported inputs following the depreciation of the dinar and the increase in interest rates influenced by the increase in the policy rate that encouraged domestic demand and investment, caused an increase in unemployment that was marked from the first quarter of 2015.

An increase in the cost of production was recorded as a result of the increase in the price of imported producer goods in domestic currency, as well as wage inflation. Attempts by the BCT to stabilize prices via the increase in the policy rate, noted as “TD” in Figure 1 failed and the corridor widened again from December 2017 in order to facilitate the refinancing of banks, taking into account the level of the money market rate, noted “TMM”. Where TFD is the Deposit Facility Rate, TFP is the Lending Facility Rate, and INF is the inflation rate.

Thus, the BCT has reacted to the economic situation by widening the corridor of the key rate and then increasing it because of the persistence of inflation. The BCT seems to be ill-equipped to preserve price stability because of the failure to take into account the real rigidities that are at the origin of inflation persistence, and the constraint of its mandate. Based on these reasons, we will study the transmission mechanisms of monetary policy in Tunisia.

In the conduct of monetary policy, proponents of discretion argue that rules are rigid and incompatible with the complexity of economies, while advocates of commitment argue that rules do not sacrifice long-term stability for short-term gains. The debate about commitment and discretion in monetary policy goes back to the work of Simons (2024), who proposed a rule to ensure price stability. Later, Friedman (1959), in turn, developed a rule for the growth of the money supply. Opponents of discretion identify it as policy actions that are neither systematic nor constrained and whose consequences can be disastrous if the central bank lacks independence. Rieder (2022) argues that the advantage of following a rule is transparency and predictability of monetary policy.

In fact, the better the private sector can anticipate the actions of the monetary authorities, the better it can plan its consumption and investment decisions. Moreover, it will act in accordance with the central bank’s expectations. Economic theory cannot predict with certainty the reactions of economic agents to the decisions of the monetary authority, but it is possible to anchor their expectations. But well before Taylor and his comments on transparency, Lucas and Sargent (1981) insisted on the need for economic agents to understand policy actions.

In emerging economies, the reduction in current account transactions is often misunderstood as rebalancing. In reality, it is the result of adjustments influenced by the actions of central banks to maintain the pre-crisis growth model and avoid a recession, without necessarily aiming for fundamental change Pradhan (2014).

Like discretion, commitment to a rule is associated with systematic actions. But according to McCallum (1993), discretion is not arbitrary and does not exclude systematicity. Rather, it is a process in which the policy instrument is defined to maximize the objectives of policymakers, taking into account the expectations of the private sector. Bernanke (2017) favors a kind of limited discretion in the conduct of monetary policy. He argues that the descriptive nature of the rules cannot be used as a guide, Levieuge et al. (2021). In this spirit, monetary policy doing whatever is necessary through the instruments to achieve the objectives, as long as the level of policy instruments is in the service of the central bank’s objective. For Blinder et al. (2024), the quadratic loss function containing the central bank’s objectives serves discretion or commitment. In the first case, the policymaker ensures that the linear combination is kept equal to zero, and in the second case a commitment to a target is announced. Both behaviors are systematic, but the economic results differ in terms of cost and welfare. According to Ilbas et al. (2013), other factors besides the costs associated with trade-offs come into play. For these authors, the limitations of the “DSGE” modeling affect the robustness of the optimal monetary policy results.

The literature on optimal monetary policy began with the seminal work of Rotemberg and Woodford (1997). In almost all of this literature, the central bank sets the interest rate to maximize welfare. Fiscal policy intervenes only to correct for the inefficiency of the steady state, through a flat tax and a subsidy. The conduct of optimal monetary policy is through the constrained minimization of a quadratic loss function that combines inflation and the output gap Giannoni and Woodford (2003) augmented by the exchange rate Corsetti et al. (2010), mortgage prices Adam and Woodford (2018, 2021), asset prices (Galí, 2014; Caines & Winkler, 2021), or wage inflation (Rhee & Song, 2014; Ida, 2020; Leeper & Zhou, 2021). The optimal approach treats monetary policy through the “DSGE” modality, as an intertemporal optimization problem in terms of first-order conditions and Lagrange multipliers. This assumes a parieto-optimal equilibrium where uncertainty is considered as white noise. The theoretical advantage of this approach is that, unlike simple rules, all information relevant to the conduct of monetary policy is taken into account and the economy is assumed to converge to a stationary state on a predictable path. Yet, Keynes designed his general equilibrium theory without any predictable natural path and without any anchor for expectations, and where decisions that are made under uncertainty may turn out to be systematically wrong.

In analyzing monetary policy, central bank economists face a number of empirical questions. Does the nominal exchange rate help forecast inflation? Does the nominal exchange rate adjust in response to the difference between domestic and foreign inflation, to restore some level of equilibrium to the real exchange rate? How quickly do monetary policy changes affect output and inflation? These questions concern complex relationships between variables of the economic system. It is hardly surprising that a common procedure is to develop empirical studies intended to address these questions, which means to analyze these problems under a set of hypotheses. The risk here is that several models have properties that make them inconsistent. If so, they constitute fragile ground for policy analysis, Jacobson et al. (1999).

Certainly, there is no single model that can provide the best answers to all the questions posed in monetary policy analysis. Nevertheless, we believe that it is necessary to make monetary policy transparent, to estimate a model which offers a coherent framework for studying the above questions. Such a model can be used as a benchmark to analyze the day-to-day functioning of a central bank. In addition, it can provide useful information on aggregate relationships used in theoretical analyzes of monetary targeting, such as the case of Tunisia. We intend to show that a VAR model can serve these purposes (Jacobson et al., 1999).

Although many of the previously used inflation targeting models depend on exogenous variables, the VAR approach endogenously determines all the variables that make up the system. Thus, we can estimate a VAR model in several steps for all the variables.

Previous studies of the VAR model have often focused on the measurement of monetary policy and its macroeconomic and socioeconomic effects. See, for example, Gordon and Leeper (1994), Christiano et al. (1994), Leeper et al. (1996) and Bernanke and Mihov (1998) for studies on the United States, and Cushman and Zha (1997) for a study of Canada and other references. Unlike these earlier VAR studies, identifying a reaction function for monetary policy is not a problem in our article. Part of the reason is that we are interested in a wide range of issues, relevant to monetary policy, but not all directly related to the effects of changes in monetary policy. Another reason is that we believe that it will be difficult to find a sufficiently long series of observations on a single policy instrument Jacobson et al. (1999).

3. Methodology

This section presents the methodological problems and difficulties that VAR (Vector Autoregressive) models can encounter when evaluating monetary policy. Sims (1980a, 1980b) introduced these models during the debates between Clower (1970) and Friedman (2007) on the method of expression between theory and econometrics to justify the causal relationship between money and income. Beginning with his 1958 paper, Friedman relied on the observation of correlation and history between these variables to explain and justify the causal effect of money on money income. Clower rejected the justification of causality by time lags and launched a Keynesian theoretical model that generates the lags noted by Friedman, but where money has no causal role in economic activity.

Clower’s position follows the modeling tradition of the Cowles Commission, in which causality is a theoretical rather than an empirical concept (Leamer, 1985; Zellner, 1979; Cooley et al., 1984; Cooley & LeRoy, 1985). Keynesians insist on estimating model parameters under structural simultaneous equations and prefer to give importance to theoretical assumptions to explain observed correlations between data. They see heterogeneity as a form of constraint on the parameters needed to determine the structural form of the model.

In this context, Sims (1972) proposed to present the hypotheses of heterogeneity of money and a causal relationship between money and income for direct and precise econometric tests based on the Granger causality test. According to Sims, the interest of the Granger causality concept is that it offers the possibility of easily and directly testing the existence of one-way causality, which he identifies through a strict exogeneity hypothesis. The theoretical limitations of Granger’s concept of causality have been widely emphasized (Zellner, 1979; Pierce, 1977; Leamer, 1985; Cooley & LeRoy, 1985), and several criticisms have focused on the fragility and lack of robustness of the results of this test (Feige & Pearce, 1979).

The introduction of VAR modeling allowed Sims (1980a, 1980b), after his famous critique, to extend Granger’s causal analysis to a vector of more than two variables. In this review, Sims criticizes the Cowles Commission’s structural models for having too many untested theoretical assumptions and presents VAR models as an alternative to these structural models. However, the sensitivity of the VAR results used by Sims to the choice of specifications has been highlighted by several authors such as (Eichenbaum & Singleton, 1986; Runkle, 1987; Spencer, 1989).

Moreover, the change in the VAR system was necessary to interpret canonical innovations as external shocks to monetary policy in the so-called structural VAR (VARS) framework. Thus, VAR modeling leads to the paradoxical case of where justifying the identification limitations imposed on contemporary innovations, Sims refers to the idea of a causal chain introduced by Wold (1954). Then, the VAR method’s desire to free itself from preconceived theory proved illusory (Pagan, 1987). Moreover, according to Hoover et al. (2008), analyzing the impulse responses of VAR models provides a good example of what Cartwright (2007) considers a “fraud” in the real world. Central bank intervention in structural VAR models is considered a shock to a stable system, and as we know well, monetary policy is not conducted via shocks.

3.1. Strengths and Weaknesses of the Standard VAR

Before presenting the formal specification of the model, it is worth clarifying why a VAR framework with long-run restrictions is particularly suited to the Tunisian case. Other approaches, such as DSGE models or structural equations systems, have been used in the literature to study monetary policy transmission. However, these models rely heavily on strong assumptions regarding rational expectations, parameter stability, and the existence of well-defined equilibrium conditions (Yan et al., 2021). In practice, these assumptions are difficult to sustain in Tunisia, where the economy has been repeatedly exposed to external commodity shocks, exchange rate pressures, and frequent institutional adjustments (Mimoun et al., 2024)

By contrast, a VAR model makes it possible to let the data speak more directly, without imposing such rigid prior constraints. The use of long-run restrictions further allows us to distinguish between permanent and transitory shocks—a distinction that is particularly important in a small open economy like Tunisia, where exchange rate and interest rate disturbances may have persistent consequences (Salem & Bouaziz, 2024). Compared to reduced-form correlations, this approach offers more meaningful economic interpretation while maintaining empirical tractability (Prüser, 2024). For these reasons, a VAR with long-run restrictions provides a balanced and realistic framework to analyze how Tunisian monetary policy is transmitted to the real economy (Mastour, 2020).

Compared to simultaneous equations, the VAR model has the advantage of capturing the variance of the model’s parameters over time, and thus better captures the dynamics of the system, which gives credibility to an economic policy that adapts to changes or shocks in the social and economic environment. The criticisms of simultaneous equations, which are the strength of VAR modeling, can be summarized in three points and are generally due to Sims (1980a), such as a priori constraints, arbitrary causal structure, and treatment mismatch. Note that, unlike a system of simultaneous equations which poses identification problems, vector autoregressive modeling removes the limitations associated with determining structural equations and is therefore less constraining than simultaneous equations, thanks to ignoring the simultaneity assumption, the influence between variables, and the lag of all dependent variables that are considered exogenous.

However, the failure to take into account the hypothesis of simultaneous effects between variables makes the VAR appear to be a theoretical model, which does not reflect economic reality, and which may be subject to economic policy biases. The VAR model is based on assumptions to determine equations that are estimated to have no theoretical basis. This is the great weakness of VAR models, a weakness that has been criticized to the point of having led to the development of so-called “structural” VAR models, i.e., policies known in the social and economic environment. Shocks or innovations are no longer random or anonymous, and their origin is known or determined.

3.2. Vector Autoregressive Model

Although many studies on monetary policy transmission rely on structural VARs (SVARs), our choice of a VAR with long-run restrictions is motivated by both methodological and practical considerations. First, this approach is fully consistent with the SVAR tradition, since the long-run restrictions we impose allow us to identify structural shocks rather than relying on reduced-form correlations (Sargent, 1978; Blanchard & Quah, 1989; Kilian & Lütkepohl, 2017). Second, in the Tunisian case, a full SVAR with several contemporaneous restrictions would be less reliable. The available monthly dataset is relatively short, and the economy has been subject to frequent external shocks and institutional adjustments, which complicates the credibility of short-run restrictions. By contrast, long-run restrictions provide a transparent and economically meaningful way to separate permanent from transitory shocks.

This choice is also consistent with the literature on small open economies, where long-run restrictions have been used effectively to capture the dynamics of exchange rates and inflation (Cushman & Zha, 1997; Salem & Bouaziz, 2024). We therefore consider that a VAR with long-run restrictions strikes the right balance between structural interpretation and empirical tractability for Tunisia.

Before turning to the technical specification of the VAR, it is important to clarify how the structural shocks are identified. In this paper, we rely on long-run restrictions that were widely used in the literature since Blanchard and Quah (1989) and later applied to monetary policy VARs by Christiano et al. (1994), as well as in small open economy settings by Cushman and Zha (1997).

The basic idea is that monetary policy shocks should not generate permanent effects on real activity. Put differently, GDP is assumed to return to its long-run path after a temporary disturbance in interest rates or liquidity conditions. This restriction is consistent with standard macroeconomic theory and avoids attributing long-term growth patterns to purely nominal shocks.

In practice, the Tunisian case requires some additional considerations. Because the economy is highly open and heavily dependent on imports, the exchange rate reacts almost immediately to monetary innovations (Aguir, 2018; Aguir & Smida, 2015). This is not just a theoretical assumption: during discussions with central bank staff and from evidence in previous studies, it is clear that financial markets in Tunisia adjust quickly to policy announcements. Inflation, however, tends to respond more slowly, reflecting price rigidities and the presence of administered prices. The money market rate, as the key policy instrument, is allowed to move contemporaneously with both inflationary pressures and exchange rate fluctuations.

By combining these restrictions, the identification strategy separates permanent and transitory effects and captures some structural features of Tunisia as a small open economy. This approach is admittedly a simplification, but it provides a tractable and credible framework for interpreting the dynamics of monetary transmission without imposing excessively rigid assumptions.

Although the primary objective of the central bank is to ensure price stability, it plays a key role in macroeconomic stabilization because its ability to achieve its inflation target depends on the influence of monetary policy on activity. The central bank thus has a dual mandate and reacts by lowering the policy rate when the economic situation deteriorates and inflation falls. Thus, the fear of recession linked to the shock of the 2007–2008 financial crisis led to expansionary monetary policies by central banks, which initially led to a drop in interest rates. However, the magnitude of the shock necessitated the expansionary nature of monetary policy restricted to the minimum rate (zero floor). Consequently, they resorted to unconventional measures, leading to an increase in the size of their balance sheet and an adjustment in its composition. The objective of this work is to evaluate and analyze monetary policy using VAR models.

To this end, we present a methodology that allows us to estimate the elasticity of the activity of a monetary policy instrument and then use the sequence of inflation rates since 1987 to determine the extent to which the monetary policy stance contributes to the other explanatory variables of Tunisia. This approach also allows us to predict the transmission effect of monetary policy through the expected evolution of the inflation rate. Since monetary policy reacts to activity as well as to a set of other economic variables, the estimation of the impact of monetary policy cannot be obtained by a direct regression on the main indicators of monetary policy. Measuring the multiplier requires first identifying the exogenous monetary policy shocks for Tunisia. The effect of monetary policy is measured by the projection method proposed by Jordà (2005), which consists of estimating the response function of GDP at different horizons to previously identified shocks. Based on the estimated elasticities, the evolution of the inflation rate, whose effect is reflected in the commodity price space, we then calculate the contribution of monetary policy to activity between 1987 and 2020. A forward-looking analysis of the impact of monetary policy is subject to the Lucas (1976) critique for the stability of macroeconomic relationships over time.

Statistically, the VAR model consists of variables which are treated identically without any exceptions or external conditions and with the same lag length for each. The simplest form is the unrestricted VAR model, which is written as follows:

$$\begin{gathered} Y_{1t}={\varnothing }_{11}^{\left(1\right)}{Y}_{1,t-1}+\dots +{\varnothing }_{11}^{\left(p\right)}{Y}_{1,t-1}+\dots +{\varnothing }_{1n}^{\left(1\right)}{Y}_{n,t-1}+\dots +{\varnothing }_{1n}^{\left(p\right)}{Y}_{n,t-p}+{\epsilon }_{1t} \\ ⋮ \\ {Y}_{nt}={\varnothing }_{n1}^{\left(1\right)}{Y}_{1,t-1}+\dots +{\varnothing }_{n1}^{\left(p\right)}{Y}_{1,t-p}+\dots +{\varnothing }_{nn}^{\left(1\right)}{Y}_{n,t-1}+\dots +{\varnothing }_{nn}^{\left(1\right)}{Y}_{n,t-p}+{\epsilon }_{nt} \end{gathered}$$

where

• n: number of variables;
• p: number of delays;
• ∅_ij: coefficients of the model variables in the polynomial matrix ∅(p) in the delay operator;
• ε_t: white noise process.

A more synthetic writing in the reduced form was given by the following (Gourieroux & Monfort, 1990)

$$Y_t=\alpha +{\varnothing }^{*}\left(L\right)Y_t+{\epsilon }_t$$

where ${\varnothing }^{*}\left(L\right)$ est is the (n, n)-sized matrix of ${\varnothing }_{ij}$ coefficients, such that $\varnothing \left(L\right)={\varnothing }^{*}\left(L\right)-I$ in the autoregressive writing, with $\varnothing \left(L\right)Y_t={\epsilon }_t$, where $\varnothing \left(L\right)$ is an (n, n)-sized matrix of lag polynomials in L of degree p.

Writings (2) et (3) express that a particular economic phenomenon ($Y_{1t})$ is studied according to its past values $(Y_{1,t-1},\dots ,Y_{1,t-1})$ and the past of other variables ($Y_{2,t-1},\dots ,Y_{2,t-2};Y_{3,t-3};Y_{n,t-1},\dots ,Y_{n,t-p})$. The originality of this model is that it does not contain purely exogenous variables:

-

It is composed of as many equations as there are selected variables.

-

Each variable is dependent in its generic equation and is then determined by its past and that of the (n − 1) other variables.

The result of such a model is to identify the interactions between the different components of an economic phenomenon. The analysis of the regression coefficients shows us the direction of causality between two variables when it exists, and the extent of the temporal dynamics:

-

When, for example, $Y_{1,t}$ is caused by $Y_{4,t-5}$ (in this case $\left|{\varnothing }_{14}^5\right|$ is significantly different from 0), we can verify the nature of the causal link, here unidirectional, since it was a realization of $Y_4$ five periods earlier that influenced the contemporaneous value of the variable $Y_1$.

-

In the same way, when $Y_{1,t}$ is caused by $Y_{4,t-5}$, then the reaction of the variable $Y_1$ occurs five periods later, we obtain here an expression of the dynamics of a process from its components.

The main objective of a VAR model is to reveal a set of causal relationships in the sense of Granger (1969) as pointed out by Fackler and Krieger (1986).

It is true that the Granger causality tests reported in Section 4 suggest that only a limited number of variables significantly Granger-cause each other. However, this does not invalidate the VAR approach. First, Granger causality tests are sensitive to sample size and lag selection, and weak causality in the statistical sense does not necessarily imply economic irrelevance. Second, the VAR is not only designed to detect predictive relations but to capture the joint dynamics of the system and to allow the identification of structural shocks (Sims, 1980b; Kilian & Lütkepohl, 2017). In this sense, even in the absence of strong pairwise Granger causality, the VAR remains a valid tool to analyze how monetary and exchange rate shocks propagate through the Tunisian economy.

3.3. Data Description

The correct identification of monetary policy shocks requires homogeneity of monetary policy, which assumes that the instrument studied does not change during the period under review. Based on discussions with officials from the Central Bank of Tunisia and using case studies of BCT decisions, we have selected variables that can be included in the monetary policy rule.

In our analysis of monetary policy in Tunisia, we used monthly data from 2000M1 to 2024M4. Our VAR framework includes the following variables:

IPC: consumer price index;

TMM: money market rate;

REER: index of REER;

REAL_GDP: real gross domestic product.

These variables are taken from the database of the Central Bank of Tunisia (BCT), with the exception of REAL_GDP which is extracted from the database of the Tunisian National Institute of Statistics (INS).

It is important to note that a significant share of food and energy prices in Tunisia are subject to government controls and subsidies. This implies that the official CPI may not fully capture the market-based response of consumer prices to monetary policy shocks. In practice, this institutional feature tends to dampen the short-run pass-through from monetary innovations to inflation. Nevertheless, we rely on the CPI because it represents the official target of the Central Bank of Tunisia and remains the most relevant indicator for policy credibility. As a robustness check, we also considered the CPI, excluding administered components (when available), and the results remain qualitatively similar. This limitation is acknowledged, and it partly explains why some of the inflation responses reported in Section 4 appear relatively weak compared to other small open economies (IMF, 2024 on Tunisia; Aisen & Veiga, 2008).

4. Empirical Results

Variables included in the monthly VAR model include non-farm GDP, consumer price index (IPC), TMM (money market rate), and nominal effective exchange rate (NEER). These choices are justified as follows: GDP excluding agriculture is a better measure of activity since total GDP is more dependent on climatic conditions. A significant portion of food and energy prices are controlled by the state and therefore do not respond to monetary policy shocks. Thus, we prefer to use the prices determined by market transactions.

The interest rate was used to promote growth and improve the stability of the banking sector as the rate cut was supposed to reduce the share of bad debts in total bank loans. Changes in the interest rate have been infrequent and asymmetric. In addition, the exchange rate has been used to protect the competitiveness of the real sector. The BCT allowed the exchange rate to depreciate in effective terms because it kept in mind rising unit labor costs relative to those of major partner and competitor countries. Domestic inflation and GDP react to a monetary policy shock with a certain delay.

To make this work more concrete, we start with the correlation analysis of a representative sample of four indicators, which are generally used to construct indices of the economic situation of Tunisia, in particular the consumer price index, the money market rate, the nominal effective exchange rate, and GDP. The parameters in the table refer to the correlation coefficients between the different study variables. Of the 10 correlations that fill the off-diagonal elements of the table, no value is greater than 0.7 in absolute value (

Table 2. Correlation matrix.

Correlation
Probability	TMM	NEER	INFL	GDP
TMM	1.000000
	-----
NEER	0.639983	1.000000
	0.0000	-----
INFL	−0.438892	−0.663670	1.000000
	0.0000	0.0000	-----
GDP	0.175729	0.414097	−0.103376	1.000000
	0.0098	0.0000	0.1308	-----

Source: authors’ calculations.

).

As mentioned in the methodology and dataset above, the results were tested by individual null hypotheses such as the lag test. Moreover, null hypotheses were also tested to select the well-fitting and statistically significant models like the Granger causality test, residual serial correlation VAR, impulse response, and test of stability conditions.

The Granger causality test is used in this work to find out whether or not the change in variables occurs by the money market rate. In the VAR model, each set of variables is regressed based on its past value and the value of the other variables presented. To analyze the relationship between the variables, it is necessary to introduce the number of optimal lags of the variables, as well as the matrix of the correlation between the “white noise” of the different equations. One of the main tests in using the VAR model is the causality test between variables (

Table 3. Results of the estimation of the VAR model.

Variables	TMM	NEER	INFL	GDP
NEER(−1)	0.027465	1.169585	0.021760	−0.00511
	(0.02580)	(0.06988)	(0.02789)	(0.05575)
	[1.06448]	[16.7380]	[0.78020]	[−0.10244]
NEER(−2)	−0.024392	−0.171304	−0.025990	0.018387
	(0.02596)	(0.07030)	(0.02806)	(0.05609)
	[−0.93964]	[−2.43663]	[−0.92619]	[0.32780]
GDP(−1)	−0.020994	−0.110273	0.036405	0.930189
	(0.03281)	(0.08886)	(0.03547)	(0.07090)
	[−0.63982]	[−1.24093]	[1.02639]	[13.1201]
GDP(−2)	0.029109	0.118684	−0.030455	−0.029970
	(0.03239)	(0.08771)	(0.03501)	(0.06998)
	[0.89883]	[1.35318]	[−0.86996]	[−0.42829]
C	0.324503	−0.923115	1.028711	0.226287
	(0.27897)	(0.75550)	(0.30155)	(0.60276)
	[1.16324]	[−1.22186]	[3.41144]	[0.37542]

Source: authors’ calculations.

).

The objective of the study is to analyze the reaction of monetary policy in Tunisia following shocks through the VAR model. The results show that the real effective exchange rate had a positive and significant impact on the money market rate and this result was not confirmed by the Granger causality test.

Most countries today do not intervene in the foreign exchange market, where exchange rates vary depending on market conditions. This is called a floating exchange rate policy, which is the case for Tunisia, even if the Central Bank of Tunisia controls exchange rates. This is happening in a lot of emerging countries, which depend a lot on imports. For Tunisia, keeping inflation under control requires controlling the prices of imported goods, and therefore the exchange rate. The change in the exchange rate by the central bank is determined by the components of currency supply and demand. An appreciation of the exchange rate reflects the fact that the national currency becomes more and more expensive on the foreign exchange market, and depreciation reflects a decrease in the scarcity of the national currency. But controlling the scarcity of money is similar to controlling the money supply. Thus, the more money a central bank creates, the more national currency will be obtained in the currency market, where the exchange rate will depreciate.

Tunisia is classified as belonging to a “moving parity regime” by the IMF, which means that the nominal effective exchange rate (NEER) must remain within a narrow range of 2% from a statistically identified trend for six months (while tolerating some outliers). The annualized rate of change in the NEER must be at least 1% with the condition that the currency decreases or increases in a sufficiently continuous manner.

The phenomenon of the depreciation of the dinar took into account the deterioration of the fundamentals of the economy. Thus, in an exceptional context pushed by the IMF, Tunisia has benefited, within the framework of a stand-by credit, from the accumulation of foreign exchange reserves to the detriment of the fixed exchange rates. The result of this devaluation fueled inflation by transmitting the change in the exchange rate to domestic prices: it is the “Pass-through” of the exchange rate to price, which is the case of our result where the exchange rate nominal headcount affects inflation positively.

In terms of GDP, its effect on the money market rate is negative and significant. Theoretically, these signs are reflected in the rise in key rates, which in turn is transmitted to lending interest rates, which could negatively impact investments, but Tunisian investors seem more sensitive to the deterioration of the business climate than to the reduction in the cost of financing projects. Conversely, its decline would be unable to positively impact the investment decision and promote growth. On the contrary, it could discourage saving with a negative real interest rate, thereby penalizing national saving, which is already threatened by the preference for real assets such as real estate. Not to mention that the interest rate was used to promote growth and improve the stability of the banking sector as the rate cut was supposed to reduce the share of bad loans in total bank loans. Changes in the interest rate have been infrequent and asymmetric (mostly downward).

We thus move on to the analysis of the impulse response of the VAR model. The analysis of the impulse response is an important step in econometric analysis, which uses autoregressive vector models. The main objective is to describe the evolution of the variables of the model in reaction to a shock (see Figure 1 and Figure 2). In accordance with the economic literature, we used Cholesky’s recursive decomposition to identify interest rate and exchange rate shocks.

The VAR model equation is estimated at quarterly frequency for Tunisia. For the latter, the effect of monetary policy is captured by the estimated monetary shock on Tunisia. The estimate of this equation also incorporates lags of the endogenous variable and of the monetary policy instrument. The elasticities thus obtained are represented in Figure 1 and Figure 2. It appears that the transmission times for monetary policy are quite short, in accordance with the results usually highlighted in the literature. The effects are generally significant from the start of the period. The time for the impact to be significant is too short in Tunisia, since the coefficient is significant three quarters after the shock.

In particular, the effect of monetary policy would be stable on the real effective exchange rate and on economic growth. There is also a strong heterogeneity in Tunisia, since the impact of the monetary policy of the TCB is very important in Tunisia. Finally, for the other variables, the impact of monetary policy is significant.

These multipliers allow us to analyze the impact of a one percentage point increase in the money market rate on inflation, the effect of which becomes more significant after three quarters and then becomes stable from the fifth trimester.

Assuming that this increase occurs in the first quarter of year (t) and then that the inflation rate returns to zero at the rate of a quarter point decrease in the rate each quarter, inflation increases by 0.2 point in (t + 1) and then 1 point in (t + 2) in Tunisia. The effect becomes zero in the fifth trimester after the shock. The effect of monetary policy would be stable on inflation, the real effective exchange rate and economic growth. There is also a high degree of heterogeneity in Tunisia, since the impact of the BCT’s monetary policy is very significant in Tunisia. Finally, for the other variables, the impact of monetary policy is significant.

The impact is stronger for the effect of monetary policy on economic growth, with an increase of 0.1 point in (t + 1) and (t + 2). Thus, the increase in the money market rate is 0.5 point from the year of the shock and then reaches 2 point (t + 1) and (t + 2).

The graph shows the response functions of inflation when the shocks are determined from the implicit rate estimated by Wu and Xia (2016). While these implicit rates are quite different from those of Krippner (2013), the estimated parameters of the monetary multiplier are close. These results suggest that the responses of monetary policy on the different variables presented in the graph are robust.

Thus, all responses to a money market rate shock tend to stand out as statistically significant but weak. A one-percentage-point increase in the money market rate leads to a depreciation of the NEER of about 0.8% after two quarters, with the effect dissipating after the fifth quarter. GDP decreases by approximately 0.2% after three quarters following a one-percentage-point increase in the money market rate, before gradually returning to baseline. Inflation increases by about 0.3% after two quarters in response to a 1% depreciation of the NEER, although the effect is temporary and vanishes after the fourth quarter.

The chart shows the inflation response functions when shocks are determined using the implied rate estimated by Wu and Xia (2016). While these implied rates are quite different from Krippner (2013), the estimated money multiplier parameters are close. These results suggest that the monetary policy responses on the different variables presented in the chart are robust.

The exchange rate regime in Tunisia is officially akin to a directed float with no predetermined path announcement. However, the IMF considers that the dinar has been very stable vis-à-vis the euro and dollar basket, particularly from May 2006 (with the exception of the year of the financial crisis of 2008). The IMF classifies Tunisia, in its new classification, among the countries with a stabilized rate regime compared to a basket. Thus, the exchange rate is determined on the interbank market, and the Central Bank of Tunisia intervenes to regulate the liquidity of the market according to its own rates. During the 1990s, the nominal effective exchange rate was targeted as an intermediate objective and the stability of the real effective exchange rate as the final objective.

While from the beginning of the 2000s, the TCB widened the scope of the indicators taken into account in order to be closer to an equilibrium exchange rate compatible with the preservation of the competitiveness of Tunisian exports, this has resulted in an almost continuous decline in the nominal effective exchange rate.

The control of the exchange rate by the TCB was facilitated by the restrictions imposed on capital transactions and, in particular, on short-term capital. The same VAR model is intended to identify the effects of an exchange rate shock.

An exchange rate shock shows a negative and almost stable effect on GDP, so we expect a statistically significant and rapid drop in GDP. The maximum effect is observed during the 4th trimester after the shock, and then the variation gradually subsides. Pass-through to consumer prices is stable and close to the threshold of statistical significance. We keep in mind that the usual transmission mechanism of the exchange rate channel takes into account the link between the interest rate and the exchange rate. Changes in the interest rate trigger changes in the exchange rate, which is only effective if capital flows are liberalized. To this extent, the exchange channel observed in Tunisia is not the exchange channel in the literal sense of the term, due to the asymmetric opening of the capital account.

Variance decomposition refers to the breakdown of the variance of the forecast error for a specific time horizon. Variance decomposition can indicate which variables have short- and long-term impacts on CPI. In addition, the variance decomposition involves considering the percentage of fluctuation in a time series attributable to variables at selected time horizons.

The decomposition of the total variance of GDP shows the share of the total variance attributable to each shock. After 3 years, 97% of the GDP variance is attributable to its own shock, about 30% to the exchange rate shock, 11% to the TMM shock, and 8% to the foreign demand shock. Four years after a shock, 37% of the price variance is explained by its own shock, 43% by the domestic demand shock, and 17% by the exchange rate shock. The interest rate has a negligible impact (1.5%). The variance of the nominal effective exchange rate is explained by its own dynamic at 53%, 25% by the GDP shock (proxy of economic fundamentals) and 11% by the interest rate shock. The interest rate is explained almost equally by its own shock and that of foreign demand (around 31–33% of the variance), while price and domestic GDP shocks play only a minor role.

The sense of economic causality is an essential element in formulating economic policy or in making forecasts. Consequently, in order to draw the necessary lessons in the case of Tunisia, the proven relationship of the variables leads us to analyze the Granger causality test by an econometric estimate of this causality, the results of which are shown in

Table 4. Causality tests according to Granger.

Null Hypothesis:	Obs	F-Statistic	Prob.
NEER does not Granger cause TMM	211	1.89916	0.1523
TMM does not Granger cause NEER		0.90930	0.4044
INFL does not Granger cause TMM	235	3.26413	0.0400
TMM does not Granger cause INFL		1.10878	0.3317
GDP does not Granger cause TMM	211	1.36468	0.2578
TMM does not Granger cause GDP		0.48841	0.6143
INFL does not Granger cause NEER	214	0.29350	0.7460
NEER does not Granger cause INFL		4.49458	0.0123
GDP does not Granger cause NEER	214	1.07712	0.3425
NEER does not Granger cause GDP		1.00976	0.3661
GDP does not Granger cause INFL	214	0.89930	0.4084
INFL does not Granger cause GDP		0.19260	0.8250

below.

The meaning of economic causality is an essential element in the development of economic policy or in forecasting. Consequently, in order to draw the appropriate lessons in the case of Tunisia, the proven relationship of the variables leads us to analyze the Granger causality test by an econometric estimation of this causality, the results of which are shown in Table 4 below. The Granger causality test showed that the nominal exchange rate «NEER» had an effect on the consumer price index at a significance level of 5%. This is confirmed in the baseline model.

To assess the stability of the empirical findings, a series of robustness checks were conducted. First, we considered alternative lag lengths in the VAR specification. While the baseline model relies on two lags, as suggested by standard information criteria, additional estimations with one and three lags were performed. The qualitative shape and timing of the impulse responses remained broadly unchanged, which suggests that the conclusions are not overly sensitive to the lag structure (see Kilian & Lütkepohl, 2017).

Second, we extended the model to account for external shocks that are particularly relevant for Tunisia as a small open economy. In particular, international oil prices were included to capture commodity price shocks, and Euro-area GDP was added as a proxy for external demand conditions. These two variables are central for Tunisia, given the country’s strong dependence on imported energy and its close economic ties with European partners. Similar approaches have been adopted in recent studies on small and emerging economies (Salem & Bouaziz, 2024; Trabelsi, 2025; Kilian, 2009). Incorporating these external factors did not alter the direction or persistence of the main results. The responses of inflation and output to monetary policy shocks remained broadly consistent with the baseline specification, although the estimated pass-through from exchange rate shocks to prices was slightly attenuated once oil prices were controlled for.

Overall, these robustness checks reinforce the credibility of the baseline results. They confirm that the identified transmission channels—particularly the roles of the interest rate and the exchange rate—are not artifacts of specific modeling choices, but rather reflect underlying features of the Tunisian economy.

A further robustness exercise was conducted to assess whether the results are sensitive to the ordering of variables in the Cholesky decomposition. We experimented with alternative orderings, including placing inflation first and the exchange rate last, instead of the baseline sequence. The main impulse responses remained qualitatively unchanged: monetary policy shocks continued to affect the exchange rate and contemporaneously inflation with a lag, while output responses were small and temporary. This suggests that our results are not driven by arbitrary ordering assumptions, but rather reflect structural features of the Tunisian economy (Kilian & Lütkepohl, 2017).

5. Conclusions

This analysis showed that monetary policy affected the output and price level and that the effect of monetary policy was strongest after two quarters; however, the importance of each channel was small.

The baseline VAR model suggested that all responses to a money market rate shock tend to stand out as statistically significant but weak. The reaction of the nominal effective exchange rate is close to being statistically significant and demonstrates that changes in the interest rate have an effect on the exchange rate. The aggregate product reacts negatively and the magnitude of the reaction is close to being statistically significant 2–3 quarters after the rate hike. The price index shows no reaction apart from small oscillations around 0. It follows from these observations that the impact of the interest rate is generally weak.

The effects are generally significant from the beginning of the period. The time lag for the impact to be significant is too short in Tunisia, since the coefficient is significant one quarter after the shock. The results also indicate that the effects of monetary policy are related to the heterogeneity of financial structures, labor markets, the degree of openness of economies, nominal or real rigidities, etc.

The view of the transmission mechanisms of monetary policy in Tunisia, as given by VAR models, is comparable to that of other emerging economies characterized by tight financial markets and administrative controls. The exchange rate tends to exert a noticeable effect on the economy, both on the real sector and on prices. We have tried through this study to clarify the framework of monetary policy in Tunisia and to present a more rule-based monetary policy in order to avoid the problem of temporal incoherence.

Some recommendations that can be addressed to the TCB based on this study are that the TCB should now take steps to increase exchange rate flexibility. In addition, it should create an exit strategy to relax the existing interest rate ceilings with a view to removing them and improving the monetary transmission mechanism.

Over the past decade, monetary policy in Tunisia has managed a floating exchange rate regime in which the TCB intervenes in the market with a view to obtaining a slight rate of depreciation of the real exchange rate against a weighted basket of currencies based on the country’s main trading partners and competitors.

However, this strategy has shown its limits due to the lack of sustainability of growth. The Tunisian economy must be able to create enough jobs for its population. Thus, the transition to another exchange rate regime will help preserve the TCB’s foreign exchange reserves, facilitate external adjustment, and support the demand for money by reducing the absorption of liquidity due to interventions in the foreign exchange market.

The findings of this paper confirm that monetary policy in Tunisia faces limits in terms of transmission and credibility. Still, there are practical steps that the Central Bank of Tunisia (BCT) could take to make progress. These steps need to be sequenced because the institutional and economic constraints do not allow for a sudden shift. Short term: The BCT could start by gradually widening the exchange rate band. This is important because past episodes, such as the exchange rate pressures of 2018, showed how rigid corridors can amplify market tensions and trigger abrupt depreciations once the band becomes unsustainable. A gradual approach would help avoid panic and allow agents to adjust their expectations step by step. At the same time, liquidity management operations—reserve requirements, open market interventions—should be fine-tuned to prevent excessive swings in the money market rate. Medium term: Beyond technical adjustments, the Bank needs to improve its ability to forecast inflation. This is not just a matter of more complex models but also of collecting and using timely data. The inflation surge of 2022 highlighted how difficult it was to anticipate the combined effect of imported food and energy prices. Developing stronger forecasting tools, and communicating them in a clear and regular way, would make policy more predictable and help to anchor the expectations of households and firms. Long term: The natural horizon is to move towards a formal inflation-targeting framework. This would require defining a transparent numerical target, ensuring the operational independence of the BCT, and building a record of accountability. Such a regime would not happen overnight, but the gradual steps taken in the short and medium run could pave the way for this long-term goal. By moving progressively along these stages, the BCT could strengthen its credibility and make its interventions more effective, while avoiding the disruptive costs of abrupt policy changes.

One limitation of our approach is that it assumes constant parameters over the sample period. However, monetary policy effects are known to be time-varying, especially when measured through financial market variables such as treasury yields (Gürkaynak et al., 2005; Primiceri, 2005). Incorporating such dynamics—possibly through a TVP-VAR or models allowing for structural breaks—would be a valuable extension for future research on Tunisia.

Despite this limitation, our findings provide new evidence on the channels of monetary transmission in Tunisia and offer operational insights for strengthening the effectiveness of monetary policy in a small open economy.