Regional monetary policy transmission in Bangladesh using Bayesian VECM and Taylor rule framework

1 Introduction

Monetary policy is one of the primary instruments available to governments for managing macroeconomic stability, with direct implications for inflation control, credit conditions, and aggregate demand (e.g., Friedman, 1968; Mishkin, 1996). In developing economies like Bangladesh, however, where structural imbalances and regional disparities persist, a one-size-fits-all monetary approach may result in uneven outcomes across subnational regions. These asymmetries can be particularly pronounced in household consumption, a key channel through which monetary policy is transmitted to the real economy, as noted in Taylor (1993). Understanding how monetary policy influences household consumption at the regional level is critical for designing more inclusive and balanced economic policies (Eklou, 2023; Ha et al., 2024).

Extensive literature has examined the aggregate effects of monetary policy on consumption and output in advanced and emerging economies (Bernanke and Gertler, 1995; Christiano et al., 1999; Mishkin, 2007b). However, few studies considered the spatial or regional dimension of monetary transmission, particularly in developing countries. The assumption of homogeneity in policy responses across regions dominates much of the empirical literature, as in Peersman and Smets (2001), but it ignores the heterogeneous economic structures, financial access, and consumption patterns that characterize countries like Bangladesh. In Bangladesh, it is known that the uniform impact of monetary policy across regions is often weaken due to structural rigidities, underdeveloped financial systems, and informal credit markets (Afrin, 2017; Ahmed and Islam, 2004; Brandao-Marques et al., 2020; Deb et al., 2023).

However, there are few existing studies in Bangladesh that have empirically examined the regional impact of monetary policy on household consumption using a dynamic panel framework, and this serves as the motivation for this study. This paper investigates the short- and long-run effects of monetary policy instruments, namely broad money supply, exchange rate, inflation rate, private sector credit, and real interest rate, on regional household consumption in Bangladesh. Using annual data from eight regions over the period 2000–2024, the study adopts a Bayesian Vector Error Correction Model (VECM) framework estimated separately for each region. The model captures both long-run equilibrium relationships and short-run dynamic adjustments while accounting for the joint endogeneity of the variables and small-sample constraints through Bayesian techniques.¹ The Bayesian framework is particularly appropriate in this context because shrinkage priors reduce parameter uncertainty and mitigate overparameterization problems commonly associated with high-dimensional time-series models with limited observations (Bańbura et al., 2010; Karlsson, 2013).

The empirical results indicate that inflation exerts a robust and consistently negative long-run effect on regional household consumption, whereas real interest rates display a positive and statistically meaningful long-run relationship. By contrast, the long-run effects of broad money, exchange rate, and private sector credit remain statistically uncertain because their 95% posterior credible intervals include zero. In the short run, all monetary policy variables, including the real interest rate, exhibit imprecisely estimated effects, suggesting that monetary policy transmission is gradual rather than immediate across regions.² These findings suggest that, despite regional differences in financial development and infrastructure, the estimated monetary transmission mechanism is broadly similar across Bangladesh, reflecting the dominance of nationally implemented monetary policy and common macroeconomic conditions.³ The observed similarity in both short-run dynamics and long-run relationships implies that the core channels of monetary transmission operate consistently across regions, despite underlying structural differences. While regional characteristics such as income variability, credit access, and price sensitivity may differ, these differences do not lead to fundamentally different responses to monetary policy shocks. The positive influence of real interest rates may reflect institutional or behavioral factors, such as weak interest rate pass-through or the role of informal financial markets.⁴

The Bayesian VECM framework offers a robust approach for capturing dynamic interactions among macroeconomic variables. Unlike conventional single-equation models, this approach allows all variables to be treated as jointly endogenous, thereby incorporating feedback effects and improving estimation reliability. The Bayesian structure further mitigates overparameterization by incorporating shrinkage priors, making it particularly suitable for short time-series data. The study also simulates a Taylor rule-based nominal interest rate path to estimate the policy rate necessary for achieving a 5% increase in household consumption. This simulation provides practical insights into the monetary policy gap relative to historical performance.

This study makes several contributions to the literature on monetary policy transmission in developing economies. First, this study presents an empirical analysis using a dynamic panel framework on how regional household consumption responses to monetary policy instruments in Bangladesh, an area rarely analyzed to date. Second, the analysis uncovers significant short-run heterogeneity in consumption responses, challenging the validity of uniform, one-size-fits-all monetary strategies. Third, it integrates a Taylor rule–based policy simulation to retrospectively evaluate the adequacy of historical interest rate settings. Finally, the study offers actionable recommendations for adopting a more flexible, forward-looking, and regionally sensitive monetary framework capable of promoting consumption-driven and inclusive economic growth.⁵

2 Theoretical background and literature review

Traditional macroeconomic theory has suggested that monetary policy, utilizing policy instruments such as money supply, interest rate, and private sector credit, has a direct impact on consumption behavior (e.g., Mishkin, 1996; Taylor, 1993). However, the effectiveness of these transmission channels varies significantly across countries and regions, especially in developing economies like Bangladesh, where structural and financial asymmetries are pronounced. This literature review categorized the existing contributions in alignment with the study’s core research questions and identified important theoretical and empirical gaps.

A large body of research has examined the long-run relationship between monetary variables and aggregate consumption, employing various macroeconomic and time series models. Early contributions grounded in monetary theory emphasized the role of inflation rate, real interest rate, and monetary aggregates in influencing consumption patterns (Friedman, 1968; Taylor, 1993). VECM, based on Johansen’s (1988) cointegration framework, has been commonly used to model short-run deviations from long-run equilibrium. Building on this tradition, this paper adopts a Bayesian VECM framework, which is particularly suitable for capturing dynamic interactions among macroeconomic variables while addressing small-sample limitations through Bayesian shrinkage techniques.⁶

Although national-level studies have extensively explored the impact of monetary policy on consumption, few have addressed how these variables affect regional household consumption, particularly in the long-run. That is, the existing empirical studies largely focused on national-level dynamics and often overlooked within country heterogeneity in regional responses, as noted in Carlino and DeFina (1998, 1999). Moreover, theoretical transmission channels are often hindered by informal financial systems, uneven access to credit, and institutional weaknesses in low-income countries (Edeme et al., 2022; Mishkin, 1996; Mwabutwa et al., 2016; Nguyen, 2020; Golder et al., 2023). This gap is particularly relevant to “Research Question 1,” which investigates the long-run effects of monetary variables, including broad money, inflation rate, exchange rate, private sector credit, and interest rate, on household consumption by region in Bangladesh.

Recent literature has increasingly focused on regional heterogeneity in monetary policy transmission. Differences in financial development, industrial structure, and institutional quality imply that identical monetary policies can produce divergent effects across space. Carlino and DeFina (1999) demonstrated that United States. regions responded differently to federal monetary shocks depending on banking penetration and economic composition. Obademi et al. (2025) found that regional disparities in credit access drove unequal consumption responses in Nigeria. Similarly, Ahmed and Islam (2004) showed that in Bangladesh, regions with better financial infrastructure exhibited stronger responses to monetary policy. These studies underscore the relevance of “Research Question 2,” which examines short-run regional responses to monetary variables in Bangladesh.

Theoretical frameworks such as the life cycle hypothesis and the permanent income hypothesis suggested that household consumption can gradually adjust following a monetary shock, depending on policy credibility and structural rigidity. Understanding the speed of adjustment is crucial for evaluating the efficiency of the transmission mechanism. This directly informs “Research Question 3,” which addresses how quickly regional consumption reverts to equilibrium following shocks. The Bayesian VECM framework employed in this study allows for the analysis of both short-run dynamics and long-run equilibrium relationships within a unified system, capturing feedback effects among variables.

An intriguing and somewhat contradictory finding in the literature is that there is a positive short-run relationship between real interest rate and consumption, particularly in developing economies. One explanation is that in financially inclusive regions, the income effect of interest earnings may dominate the substitution effect of higher borrowing costs. In the case of India, Reddy (2002) highlighted how financial inclusion and institutional quality shaped these dynamics. This relationship has not been sufficiently studied at the regional level in Bangladesh. “Research Question 4” directly examines whether real interest rate positively or negatively affects household consumption across diverse regions, accounting for local financial conditions and savings behavior.

Existing studies on Bangladesh have largely focused on aggregate outcomes, such as the impact of monetary aggregates on inflation (Younus, 2009; Rana, 2024; Bhattacharya and Patnaik, 2016) or the relationship between monetary policy and consumption/GDP growth (Ahmed and Islam, 2004) but has overlooked regional variation. Yet economic activity, infrastructure, and financial access differ significantly between eastern and western regions in Bangladesh.⁷ Regional disparities result in differences in financial access and infrastructure conditions, which lead to uneven delivery of monetary policy. This study fills this gap by conducting a regionally disaggregated analysis using a Bayesian system-based framework, capturing both long-run and short-run dynamics across Bangladesh’s diverse regions.

3 Data and model

3.1 Data

The study employs the household consumption dataset from the Bangladesh Household Survey (2000–2024) by the Bangladesh Bureau of Statistics (2024).⁸ Data related to key macroeconomic variables such as broad money, exchange rate, inflation rate, private sector credit, and real interest rate are collected from the World Bank’s World Development Indicators (WDI). The region consists of eight administrative divisions, including Barisal, Chittagong, Dhaka, Khulna, Mymensingh, Rajshahi, Rangpur, and Sylhet, which may reveal regional disparities in terms of household consumption and financial access. This study uses household consumption data together with macroeconomic indicators to examine the impact of monetary policy on consumption.⁹

The study focuses on six key variables that shape the transmission of monetary policy, as described in Table 1. Household consumption measures total spending on goods and services, a key driver of economic growth (Deaton and Zaidi, 2002). Broad money to GDP reflects financial liquidity and the effectiveness of monetary policy, and a higher ratio indicates greater liquidity and financial development (Beck et al., 2004). Exchange rate fluctuations influence trade competitiveness and inflation through import costs (Gopinath and Itskhoki, 2010). The inflation rate tracks price stability, a primary concern of central banks (Mishkin, 2007a). Credit to the private sector reflects financial intermediation and corporate investment trends (Beck et al., 2004). Finally, real interest rates determine borrowing costs and savings incentives, which have a direct impact on economic activity (Mishkin, 1996; Taylor, 1993)¹⁰.

Table 1. List of variables and definitions.

(i) Broad money to GDP may not capture the role of informal financial markets, which remains significant in Bangladesh. (ii) Exchange rate values rely on official data and may not fully reflect parallel market fluctuations.

Figure 1 presents the historical trends of these key macroeconomic variables from 2000 to 2024. Household consumption has grown steadily across all regions, while broad money, private sector credit, and inflation rate show upward trends with fluctuations after 2015. The exchange rate and real interest rate display persistent increases, indicating external pressures and a contractionary monetary stance. A visual inspection suggests that variables exhibit non-stationary behavior, as they display persistent trends. These trends underscore the importance of understanding the regional effects of monetary policy. The empirical analysis will assess both long-run equilibrium relationships and short-run adjustments, providing a nuanced view of monetary policy transmission in Bangladesh.

4 Model

To assess the dynamic effects of monetary policy on regional household consumption in Bangladesh, this study employs a Bayesian VECM framework. This approach allows for the estimation of both long-run equilibrium relationships and short-run dynamics while capturing the joint endogeneity among macroeconomic variables within each region. That is, we adopt region-specific time-series models, which allow for flexible estimation of monetary transmission dynamics while accommodating common national shocks.

4.1 General VECM specification

VECM for each region and time t is expressed as follows:In Equation 1, Y_t is a vector of endogenous variables defined as Y_t=[C_t,BM_t,EX_t,INF_t,CR_t,R_t]′, where C_t denotes household consumption, and BM_t,EX_t,INF_t,CR_t, and R_t represent broad money, exchange rate, inflation rate, private sector credit, and real interest rate, respectively. The matrix Π captures the long-run relationships among the variables and can be decomposed as Π=𝛼𝛽′, where 𝛽 represents the long-run cointegration vectors, and 𝛼 denotes the speed of adjustment toward equilibrium. The matrices Γ_*j* capture short-run dynamics, and 𝜀_*t* is the vector of error terms.

Δ Y t = Π Y t − 1 + ∑ j = 1 p − 1 Γ j Δ Y t − j + ε t (1)

4.2 Bayesian estimation framework

To improve estimation efficiency in a relatively small sample, the VECM is estimated within a Bayesian framework. The Bayesian approach incorporates prior distributions on model parameters, which helps mitigate overparameterization and improve the stability of estimates. Specifically, Bayesian shrinkage is implemented using independent normal priors on the model coefficients, allowing parameters to be centered around zero while retaining sufficient flexibility to capture the underlying data dynamics. This shrinkage mechanism is particularly suitable for the present study because each regional model is estimated using approximately 25 annual observations, where conventional maximum likelihood estimation may suffer from parameter uncertainty and instability. By shrinking less informative coefficients toward zero, the Bayesian framework improves estimation reliability while preserving important dynamic relationships among the variables. Separate Bayesian VECMs are estimated for each region to allow regional responses to a common national monetary policy to be examined individually, which is appropriate given the centralized nature of monetary policy in Bangladesh.¹¹

Posterior inference is based on posterior medians and corresponding 95% credible intervals. Throughout the empirical analysis, parameters whose credible intervals exclude zero are interpreted as providing stronger statistical evidence, whereas coefficients with credible intervals containing zero are interpreted more cautiously as indicating uncertain associations rather than conclusive effects. The Bayesian estimation employs independent Normal shrinkage priors ( λ = 0.3 ) , 4 , 000 posterior draws, and a burn-in period of 1,000 iterations.

4.2.1 Full matrix representation

For a VAR of order p with n endogenous variables, the system is written as follows:

[ y 1 , t y 2 , t ⋮ y n , t ] = [ c 1 c 2 ⋮ c n ] + A 1 [ y 1 , t − 1 y 2 , t − 1 ⋮ y n , t − 1 ] + ⋯ + A p [ y 1 , t − p y 2 , t − p ⋮ y n , t − p ] + [ ε 1 , t ε 2 , t ⋮ ε n , t ] (2)

In Equation 2, each coefficient matrix A j is defined as follows:

A j = [ a 11 ( j ) a 12 ( j ) ⋯ a 1 n ( j ) a 21 ( j ) a 22 ( j ) ⋯ a 2 n ( j ) ⋮ ⋮ ⋱ ⋮ a n 1 ( j ) a n 2 ( j ) ⋯ a nn ( j ) ] (3)

In Equation 3, a_kl^(j) represents the coefficient corresponding to the effect of the l-th variable on the k-th equation at lag j.

4.2.2 Stacked matrix form

Stacking observations over time, the system becomes:

Y = XB + E (4)

In Equation 4, Y is a T × n matrix of endogenous variables, X is a matrix of lagged variables and intercept terms, B is the coefficient matrix, E is the error term matrix. This can also be represented as a vectorized representation of the system as follows:In Equation 5, vec ( · ) denotes the column-stacking operator, I n is the n × n identity matrix, ⊗ denotes the Kronecker product, and vec ( E ) is the stacked error term.

vec ( Y ) = ( I n ⊗ X ) vec ( B ) + vec ( E ) (5)

4.2.3 Bayesian prior specification

Assuming vec ( B ) ∼ N ( 0 , Ω ), where the prior covariance matrix is diagonal. Specifically,

Ω = [ σ 1 2 0 ⋯ 0 0 σ 2 2 ⋯ 0 ⋮ ⋮ ⋱ ⋮ 0 0 ⋯ σ k 2 ] (6)

In Equation 6, this diagonal structure reflects independent normal prior distributions, which imply there is no prior covariance across parameters and enables shrinkage toward zero.

4.3 Error correction matrix representation

The long-run equilibrium relationship is captured through the error correction matrix Π, defined Π = α β ′, where α = [ α 1 α 2 ⋮ α n ] and β ′ = [ β 1 β 2 ⋯ β n ]. Thus, the matrix Π is explicitly given by:

Π = [ α 1 β 1 α 1 β 2 ⋯ α 1 β n α 2 β 1 α 2 β 2 ⋯ α 2 β n ⋮ ⋮ ⋱ ⋮ α n β 1 α n β 2 ⋯ α n β n ] (7)

In Equation 7, the Error Correction Term (ECT) is denoted asEC T t − 1 = β ′ Y t − 1 . When substituting the ECT into the VECM in Equation 1, it is as follows:

Δ Y t = α ( EC T t − 1 ) + ∑ j = 1 p − 1 Γ j Δ Y t − j + ε t (8)

In Equation 8, the consumption equation becomes:

Δ C t = α 1 ( EC T t − 1 ) + ∑ j = 1 p − 1 λ j Δ C t − j + ∑ j = 0 p − 1 δ j Δ X t − j + ε 1 , t (9)

In Equation 9, α 1 measures the speed of adjustment toward equilibrium, λ j and δ j capture short-run dynamics.

4.4 Region-specific estimation

The model is estimated separately for each region, allowing the dynamics of monetary transmission to be examined at the regional level. This approach avoids imposing restrictive cross-sectional assumptions and enables a direct comparison of coefficient magnitudes and response patterns across regions (Pesaran, 2024). Although the estimation is conducted independently for each region, the similarity of estimated coefficients and credible intervals provides evidence on whether monetary transmission is homogeneous and/or heterogeneous across regions.

4.5 Taylor rule evaluation and policy prescription using Kalman filter-based output gap

This study extends its empirical strategy by integrating a monetary policy evaluation and recommendation framework based on the Taylor rule. The output gap is estimated through a state-space model applying the Kalman filter to household consumption data. Given the absence of GDP data, household consumption serves as a robust proxy for aggregate demand, and we derive cyclical component for policy calibration by utilizing the smooth trend of household consumption. This study adopts a forward-looking Taylor rule framework to evaluate and prescribe interest rates in Bangladesh.

The Taylor rule is denoted by the following equation:In Equation 10, i t denotes the nominal policy interest rate at time t, r^*=6.5% represents the median real interest rate, and π_t is the actual inflation rate. The inflation target is set as the median of historical data at π^*=5.5%. The variable y ˜ t captures the output gap, proxied as the deviation of log consumption from its potential level, while the parameters α = 1.5 and β = 0.96 is calibrated from the observed relationship between the policy interest rate and the consumption-based output gap. To avoid confusion, these coefficients are interpreted here as monetary policy response parameters rather than structural production elasticities.¹²

i t = r ∗ + π t + α ( π t − π ∗ ) + β y ˜ t (10)

To construct the output gap, y ˜ t, we apply a univariate state-space model to log consumption, defined as y t = log ( C t ). The model consists of two equations. Specifically, the observation equation and the state (trend) equation are as follows:

y t = μ t + ε t , ε t ∼ N ( 0 , σ ε 2 ) (11)

μ t = μ t − 1 + η t , η t ∼ N ( 0 , σ η 2 ) (12)

In Equation 11, μ t represents the potential or trend component of log consumption, and the disturbance term ε t captures transitory shocks. The Kalman filter is used to estimate μ t recursively, producing a smooth trend path for consumption. Hence, the output gap is defined as y ˜ t = y t − μ ˆ t .

In Equation 12, the Kalman filtering method is preferred over simple detrending approaches, such as linear trends or Hodrick-Prescott filtering because it provides adaptive learning, flexibly adjusts to structural breaks, and generates more realistic dynamics of the output gap in emerging market environments where consumption patterns are highly volatile.

Using the estimated output gap, we construct the Taylor-implied rate, denoted as i t Taylor, and compare it with the observed nominal policy rate, denoted as i t actual. The policy gap is measured by Δ i t = i t actual − i t Taylor . The stance of monetary policy is then evaluated according to the following classification: if Δ i t > + 0.5, the policy rate is considered too high and therefore contractionary; if Δ i t < − 0.5, the policy rate is deemed too low and expansionary; and if ∣ Δ i t ∣ ≤ 0.5, the policy stance is considered appropriate. The 0.5 percentage point threshold is adopted as a pragmatic tolerance band, consistent with recent IMF work that defines a neutral monetary policy stance as an interest rate gap within the range of −0.5 to 0.5 percentage points (Tiedemann et al., 2024). This provides a rigorous benchmark for assessing whether the Bangladesh Bank has historically aligned its policy rates with macroeconomic fundamentals.

Finally, to establish a forward-looking interest rate policy aimed at an average annual household consumption growth rate of 5% over the next 24 years, the target output gap is set as the deviation between projected consumption-based output and its steady-state path corresponding to an average annual compound growth rate of 5%. This target reflects the medium-term potential growth benchmark for Bangladesh as documented in the literature (e.g., Bangladesh Bank, 2024; World Bank, 2017), where long-run real GDP expansion has averaged about 5–6% per year. Specifically,

y ˜ t target = log ( 1.05 ) ≈ 0.0488 , ∀ t ∈ { T + 1 , , … , , T + 24 } .

Assuming that inflation expectations remain anchored at their recent average, denoted as π ¯, the prescribed interest rate path is calculated as follows.

i t target = r ∗ + π ¯ + α ( π ¯ − π ∗ ) + β · log ( 1.05 ) .

This counterfactual interest rate sequence provides a basis for policy calibration designed to support sustainable, consumption-driven growth. It highlights the importance of rule-based forward guidance in stabilizing expectations, reducing uncertainty, and fostering pro-growth monetary conditions in Bangladesh.

5 Findings and discussions

5.1 Estimated results

Table 2 shows the results of the ADF unit root tests reveal an order I(1) of integration among the variables. As presented in the descriptive statistics in Appendix A. Table A1, the macroeconomic variables and consumption series exhibit different distributional characteristics. However, these do not determine stationarity, which is formally assessed through unit root test. Since all variables are integrated of order one, I(1), first-order differences are required to achieve stationarity.¹³

Table 2. Unit root test results.

(i) ADF tests are estimated to have a constant and linear trend. (ii) p-values are shown in parentheses. (iii) ***: p < 0.01, **: p < 0.05, and * p < 0.10.

Table 3 reports on the Johansen trace and maximum eigenvalue tests for cointegration across regions. The null hypothesis of no cointegration and at most one cointegrating vector is consistently rejected at the 5% level across all regions. However, the null hypothesis of at most two cointegrating relationships cannot be rejected. These results indicate the presence of two cointegrating relationships in all regions. The strong consistency between the trace and maximum eigenvalue tests further reinforces the robustness of this finding. The existence of cointegration confirms that the variables share a stable long-run equilibrium relationship, thereby justifying the use of a VECM. Moreover, the similarity of results across regions suggests that the long-run structure is broadly consistent, even though short-run dynamics may differ.

Table 3. Johansen cointegration test.

(i) p-values are shown in parentheses. (ii) ***: p < 0.01 and **: p < 0.05. (iii) Critical values based on MacKinnon et al. (1999).

The results regarding the lag order selection criteria clearly indicate that lag-1 outperforms lag-0 across all evaluation metrics, including Akaike information criterion, Schwarz information criterion, Hannan-Quinn information criterion, and final prediction error. In addition, the likelihood ratio test strongly rejects the null hypothesis of no additional lag, confirming the statistical relevance of including one lag.¹⁴ The improvement in log-likelihood and the substantial reduction in information criteria values demonstrate that lag-1 provides a significantly better fit while maintaining parsimony. Importantly, the consistency of results across all regions indicates a common dynamic structure. Therefore, a lag length of one is selected for all regional VECM estimations, ensuring both efficiency and comparability.

Table 4 provides the estimation results of the Bayesian VECM using independent normal prior distributions with regional consumption as the dependent variable. Considering that there is no cross-sectional dependency, estimations are performed separately for each region. The reported estimates are based on posterior medians together with their corresponding 95% credible intervals. Accordingly, coefficients whose credible intervals exclude zero are interpreted as providing stronger statistical evidence, whereas coefficients with credible intervals including zero are interpreted more cautiously as indicating uncertain associations rather than conclusive effects. We examine the estimation results in Table 4 in the form of answers to the four research questions presented in Chapters 1 and 2.

Table 4. Bayesian VECM estimation results.

(i) Values represent posterior means, with 95% credible intervals in brackets. (ii) A coefficient is considered to have strong posterior support if the credible interval does not include zero. (iii) If the credible interval includes zero, the effect is interpreted as statistically uncertain.

5.1.1 Research question 1

How do broad money, exchange rate, inflation rate, and private sector credit influence regional household consumption in the long-run?

Table 4 reveals a high degree of consistency in both magnitude and direction, suggesting that monetary policy effects are broadly similar across regions in Bangladesh, as also suggested by the substantial overlap in credible intervals. The coefficients of inflation rate are negative in all regions. Importantly, the credible intervals for all regions exclude zero (e.g., [−2.012, −0.074] for Barisal), providing strong posterior evidence that inflation reduces household consumption.¹⁵ This finding is consistent with standard monetary theory, where inflation erodes real purchasing power and suppresses consumption (Mishkin, 2007b). Additionally, the exchange rate shows positive coefficients in all regions, but the credible interval includes zero in all cases (e.g., [−0.019, 1.650] for Barisal), indicating that the effect is not precisely estimated. While the direction suggests a potential positive relationship through remittance inflows or export income (Krugman and Obstfeld, 2009), the evidence remains statistically weak and does not support a robust long-run effect.

The coefficients of private sector credit are positive, and the credible intervals include zero in all regions. This suggests that credit may support consumption through improved financial access (Beck et al., 2004), but the posterior evidence remains weak and uncertain, indicating that the effects of credit channel are limited in the long-run. The real interest rate shows a strong and consistently positive long-run effect in all regions. The credible intervals exclude zero in all cases (e.g., [0.420, 2.359] for Barisal), indicating a statistically meaningful relationship. This suggests that income and stability effects may dominate substitution effects in the long-run, particularly in a context where financial markets are imperfect and savings behavior plays a significant role (Blanchard, 1985; Campbell and Mankiw, 1989). In contrast, broad money coefficients exhibit negative, and the credible intervals include zero in all regions. This indicates that the long-run effect of liquidity on consumption is statistically uncertain, suggesting that increases in monetary aggregates do not necessarily translate into higher consumption.

In sum, the long-run results provide strong posterior evidence that inflation rate and real interest rate are robust long-run determinants of household consumption across all regions, whereas the effects of exchange rate, private sector credit, and broad money remain statistically uncertain under the Bayesian framework.

Although Bangladesh exhibits substantial regional disparities in financial development, infrastructure, and household income, these structural differences do not necessarily imply heterogeneous monetary policy transmission. One plausible explanation is that monetary policy instruments, including policy interest rates and monetary aggregates, are implemented uniformly by Bangladesh Bank at the national level. Moreover, the widespread integration of financial markets, national banking regulations, common inflation dynamics, and interregional trade may contribute to similar consumption responses across regions. The estimated similarity in posterior coefficients therefore suggests that regional structural differences influence the magnitude of economic development rather than fundamentally altering the direction of monetary policy transmission. Consequently, the findings should be interpreted as evidence of broadly homogeneous transmission mechanisms operating within a nationally integrated monetary system, despite underlying regional differences.

5.1.2 Research question 2

What are the short-run effects of changes in monetary policy variables on regional household consumption in Bangladesh?

The short-run dynamics also display strong uniformity across regions. However, in contrast to the long-run results, the credible intervals suggest that short-run effects are generally not precisely estimated. Although inflation rate and real interest rate are estimated as negative and positive coefficients, respectively, and the credible intervals include zero in all regions. This indicates that the short-run effect of inflation rate and interest rate on consumption is not statistically distinguishable from zero. Similarly, the coefficients of broad money, exchange rate, and private sector credit exhibit credible intervals that include zero in all regions, indicating no strong evidence of short-run effects.

Overall, the short-run results suggest that monetary policy variables do not exert statistically strong immediate effects on consumption when evaluated using 95% credible intervals. This suggests that monetary transmission may be weak or delayed in the short run, although the evidence is not sufficiently strong to draw definitive conclusions regarding the effectiveness of short-run monetary transmission.

5.1.3 Research question 3

How quickly does regional household consumption return to equilibrium after a monetary shock?

The negative signs of ECT are consistent with theoretical expectations of convergence toward equilibrium (Engle and Granger, 1987), but the credible intervals include zero in all regions. This indicates that the speed of adjustment is statistically weak and imprecisely estimated. In other words, there is no strong posterior evidence that deviations from long-run equilibrium are systematically corrected in the short-run. This finding suggests that adjustment mechanisms may be slow or uncertain, potentially reflecting structural rigidities, financial frictions, or weak monetary transmission channels (Blanchard, 1985).

5.1.4 Research question 4

What is the role of real interest rate in shaping regional household consumption?

Real interest rates exhibit a clear distinction between long-run and short-run effects. In the long-run, the effect is positive and statistically meaningful in all regions, with credible intervals excluding zero. This indicates strong posterior evidence that higher real interest rates are associated with higher consumption. This may reflect income effects, improved macroeconomic stability, and enhanced consumer confidence, particularly in environments with limited access to credit (Campbell and Mankiw, 1989; Mishkin, 1996). On the other hand, the interest rate coefficients remain positive in the short-run, but the credible intervals include zero in all regions. That is, the immediate effect of interest rate on consumption is uncertain and not precisely estimated. The fact that long-run effects are consistently observed across regions while short-run reactions are weak suggests that interest rate channels generally operate in a uniform but gradual manner, with their impact appearing primarily over the long term.

5.1.5 Sum-up

The results in Table 4 show that while Bangladesh’s monetary policy transmission mechanism is structurally weak in the short-run, this vulnerability becomes more pronounced in the long-run. Also, the transmission of monetary policy on consumption shows similar across all regions. Inflation rate emerges as a key determinant of household consumption, exerting a consistently negative effect in all regions, reflecting the erosion of real purchasing power. In contrast, real interest rate exhibits a positive long-run relationship with consumption, suggesting that income and stability effects dominate traditional substitution effects in a financially constrained environment.

Other monetary variables do not display statistically robust effects, indicating limited effectiveness of conventional transmission channels. The absence of significant short-run dynamics, together with the weak error correction mechanism, further suggests that consumption adjustment is slow and that monetary shocks are not rapidly transmitted to households. A more important result is that the consistency of monetary policy regarding consumption across all regions implies that monetary policy operates in a broadly uniform manner. However, this uniformity reflects weak transmission rather than strong effects, highlighting structural constraints of the monetary transmission process.

5.2 Policy evaluation and projection

The historical evaluation of Bangladesh’s interest rate policy from 2000 to 2024 reveals a consistent contractionary bias in Table 5. In all 26 years, the nominal interest rate exceeded the Taylor Rule–implied rate by more than 0.5 percentage points, with an average rate gap exceeding 4% (Clarida et al., 2000; Taylor, 1993). Even during economic downturns (e.g., 2008 and 2020), policy rates remained significantly above the Taylor benchmark, indicating insufficient monetary accommodation (Mishkin, 2007a; Woodford, 2001). This persistent over-tightening suggests a strong anti-inflationary stance but may have come at the cost of suppressed private investment and limited counter-cyclical support (Blanchard et al., 2010). Despite a gradual narrowing of the rate gap, no year reflects an “appropriate” policy stance (Taylor, 1999).¹⁶

Table 5. Historical interest rate policy evaluation (2000–2024).

Policy is “Too High (Contractionary)” if rate gap >0.5, “Too Low (Expansionary)” if rate gap < −0.5, and “Appropriate” if −0.5 ≤ rate gap ≤0.5.

These findings highlight the need for a more balanced, rule-consistent monetary framework that supports both price stability and output growth, as noted in Clarida et al. (2000) and Svensson (1997). Incorporating forward-looking indicators and enhancing responsiveness to real-time macroeconomic conditions can improve the effectiveness and inclusivity of Bangladesh’s monetary policy.

Table 6 illustrates the projected trajectory of nominal interest rates derived from a dynamic Taylor Rule alongside the path of consumption assuming a 5 % annual compound growth rate. The Taylor Rule incorporates a neutral real interest rate of 2% (Taylor, 1993), an inflation target of π ∗ = 5.5 %, and observed inflation of π = 7.5 %, implying a consistent positive inflation gap. As a result, the rule suggests a gradual increase in the nominal policy rate from 12.55% in Period 1 to 13.64% in Period 24. Consumption projections, based on the compound growth formula C t = C 0 ( 1.05 ) t, rise from 1.10 to 9.85 over the same period, reflecting steady expansion. The concurrent increase in both interest rate and consumption highlights a policy environment where moderate tightening is compatible with sustained demand growth. This underscores the potential for calibrated rule-based monetary policy to maintain macroeconomic stability while supporting long-run consumption objectives aligned with a 5% growth target, as in Mishkin (2007b) and Svensson (1997).

Table 6. Dynamic Taylor rule and projected consumption response.

Consumption is projected assuming 5% compound growth per period: C t = C 0 · ( 1.05 ) t, C 0 = 1.0.

The recommended nominal policy rate of approximately 13% should not be interpreted as a definitive policy prescription but rather as a model-based benchmark derived under the maintained assumptions of the Taylor rule framework. Specifically, the estimate is conditional on a long-run household consumption growth target of 5%, historical inflation dynamics, and the consumption-based output gap estimated using the Kalman filter. Consequently, alternative assumptions regarding inflation expectations, the equilibrium real interest rate, or the target consumption growth path would produce different policy rate recommendations. Therefore, the proposed interest rate should be interpreted as an indicative benchmark for policy evaluation rather than an exact optimal policy rate.

From a policy perspective, maintaining a nominal policy rate around this level involves important trade-offs. Although such a rate may support medium-term consumption growth under the model assumptions, persistently high policy rates could increase borrowing costs, dampen private investment, constrain credit expansion, and impose additional financial burdens on households and firms. Conversely, an excessively accommodative monetary stance could generate inflationary pressures, encourage excessive credit growth, and increase macro-financial vulnerabilities. Accordingly, Bangladesh Bank should interpret the Taylor rule recommendation alongside broader macroeconomic indicators, including inflation expectations, financial stability considerations, exchange rate developments, and external economic conditions, rather than relying exclusively on the model-based policy rate.

To evaluate the robustness of the Taylor-rule recommendation, a sensitivity analysis was conducted by varying the assumed household consumption growth target from 4 to 6% while holding all other model parameters constant. The results, reported in Appendix Table A3, indicate that the recommended nominal policy rate changes only marginally, ranging from 13.19 to 13.20%. This limited variation suggests that the estimated policy recommendation is not sensitive to reasonable changes in the assumed consumption growth target. Therefore, the baseline recommendation of approximately 13% appears to be robust under alternative growth scenarios, providing additional confidence in the stability of the policy simulation.

These findings also have important implications for Bangladesh Bank’s current interest-rate-based monetary policy framework. The empirical results indicate that inflation remains the dominant long-run determinant of household consumption, whereas the short-run transmission mechanism appears relatively weak. Consequently, policy decisions should focus not only on adjusting the policy interest rate but also on improving the effectiveness of monetary transmission through deeper financial markets, enhanced banking sector efficiency, and broader financial inclusion. Furthermore, the Taylor rule simulation provides a useful benchmark for evaluating the consistency of policy decisions with macroeconomic fundamentals. Nevertheless, such rule-based recommendations should complement, rather than replace, policymakers’ judgment by considering inflation expectations, exchange rate developments, and financial stability risks.

5.3 Robustness checks

This section evaluates the stability, reliability, and structural validity of the Bayesian VECM framework through a comprehensive set of robustness checks. Specifically, alternative lag specifications are considered to address potential dynamic misspecification (Lütkepohl, 2005), while prior shrinkage variations are employed to assess sensitivity to Bayesian prior assumptions (Koop and Korobilis, 2010). A potential concern in dynamic models is lag misspecification, which may lead to biased estimates of persistence and short-run dynamics. To address this, an extended lag structure is used. That is, the model is re-estimated using two lags instead of one, as shown in Table 7.

Table 7. Bayesian VECM results: Lag-2 specification.

Values represent posterior means, with 95% credible intervals in brackets.

In Table 7, the second lag of the dependent variable remains positive across all regions, with 95% credible intervals excluding zero, indicating continued persistence in regional consumption dynamics. By contrast, the second-lag coefficients of the macroeconomic variables have credible intervals that include zero suggesting limited posterior support for additional short-run macroeconomic effects. Overall, these findings indicate that the main substantive conclusions remain broadly robust to the lag-2 specification.

Additionally, when considering the Bayesian framework, the results may vary depending on the choice of prior hyperparameters, particularly the shrinkage parameter (λ). To examine this, the model is estimated under alternative values of λ (0.1, 0.3, and 1), corresponding to weak, moderate, and strong shrinkage, respectively. The results are summarized in Table 8.

Table 8. Shrinkage robustness.

Values represent posterior means, with 95% credible intervals in brackets.

Under weak shrinkage (λ = 0.1), the coefficients of inflation rate and real interest rate are negative and positive, respectively, and are found to be statistically significant. However, as the degree of shrinkage increases (λ = 0.3 and λ = 1), the credible intervals widen and all coefficients become statistically insignificant. This pattern reflects the well-known trade-off in Bayesian estimation: stronger shrinkage reduces variance at the cost of attenuating signals. Nevertheless, the fact that key variables remain significant under weak shrinkage suggests that the baseline findings are not purely driven by prior assumptions, although inference becomes more conservative under tighter priors.

In addition, a restricted specification excluding the error correction term is estimated to examine the role of long-run equilibrium adjustment within cointegrated systems, as suggested by Engle and Granger (1987). This test examines whether the short-run dynamics are sensitive to the imposition of a cointegrating relationship. The results indicate that the lagged dependent variable remains strongly positive and statistically significant, consistent with the baseline specification. Moreover, all macroeconomic variables remain statistically insignificant. These findings suggest that short-run dynamics are robust to the exclusion of the long-run adjustment mechanism, and that the primary driver of regional variation is persistence rather than equilibrium correction.¹⁷

Subsample analysis is further conducted to detect potential structural changes and parameter instability over time (Hansen, 2001), and the results are shown in Table 9. The subsample results reveal several notable differences. First, the ECT becomes negative and statistically significant across all regions, indicating stronger adjustment toward long-run equilibrium in the subsample period. Second, broad money exhibits a positive and significant effect, while the exchange rate shows a negative and significant impact. Other variables remain statistically insignificant. These findings suggest that the transmission mechanism may be time-varying, with stronger long-run adjustment and clearer monetary channels in the subsample period. This provides evidence of structural heterogeneity over time, rather than instability in the baseline results.

Table 9. Subsample stability.

Values represent posterior means, with 95% credible intervals in brackets.

Finally, an important requirement for the validity of a VECM is that the system satisfies stability conditions, which ensures that the dynamic process is stationary and converges following shocks. This condition is evaluated by examining the roots of the characteristic polynomial, where all roots must lie strictly inside the unit circle. The results indicate that, for all regions, the maximum modulus of the roots is well below unity, ranging between 0.6925 (Dhaka) and 0.7117 (Chittagong). This confirms that the estimated systems are dynamically stable across all specifications. In particular, the similar magnitude of the roots across regions suggests a high degree of homogeneity in dynamic adjustment patterns.

6 Conclusion

This study contributes to the field of monetary economics as one of the systematic studies investigating regional monetary policy transmission in Bangladesh using the Bayesian VECM framework. Unlike previous national-level analyses, this research combines region-specific econometric evidence with a Taylor rule–based policy evaluation to assess the historical effectiveness of interest rate settings over a 25-year period. The findings identify a robust long-run relationship for inflation, alongside a consistently positive long-run association for real interest rates. These findings are consistent with the existing monetary policy literature, which emphasizes that inflation remains one of the principal channels affecting household consumption and macroeconomic stability (Mishkin, 2007b; Deb et al., 2023). On the other hand, other monetary variables, including broad money, exchange rate, and private sector credit, do not exhibit statistically robust effects, suggesting that traditional transmission channels may be limited or weak in Bangladesh (Deb et al., 2023; Rana et al., 2024).

The study’s strength lies in bridging two underexplored areas: regional dynamics in monetary policy transmission and the historical misalignment of actual policy rates with rule-based prescriptions. The results further indicate that monetary transmission is structurally weak in the short-run, with limited immediate effects and an imprecise adjustment mechanism, while more discernible relationships emerge over longer horizons. This suggests that monetary policy in Bangladesh operates more through gradual, long-horizon adjustments rather than immediate demand-side responses. This finding is also consistent with previous studies reporting sluggish monetary policy transmission in developing economies because of financial frictions, incomplete interest-rate pass-through, and institutional constraints (Jamilov and Égert, 2014; Gregor and Melecký, 2018). Despite this, the similarity of results across regions suggests broadly uniform transmission patterns.

However, several limitations should be acknowledged. The use of annual data and a relatively narrow set of regional indicators may mask intra-regional variation and higher-frequency dynamics. Moreover, important structural factors, such as regional inequality, infrastructure, and financial access, are not explicitly modeled. What remains underexplored in the literature is a deeper understanding of why monetary transmission remains weak and how structural or institutional constraints mediate this process. The roles of expectations, credibility of monetary authorities, and informal financial systems also warrant further investigation. Future research should extend this framework by incorporating higher-frequency detailed data, and structural characteristics that shape regional responses to monetary shocks. Integrating structural or expectation-augmented models can enhance analytical depth. Future studies may also examine nonlinear and state-dependent monetary transmission mechanisms, which have recently received increasing attention in the monetary economics literature (Ha et al., 2024). Ultimately, this study emphasizes the need for forward-looking, rule-based, and structurally informed monetary policy in Bangladesh.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. While preparing this manuscript, the author(s) used Grammarly for language editing and grammatical corrections. After using this AI tool/service, the author(s) reviewed and edited the content as needed. All intellectual content, analysis, and interpretations were produced solely by the authors, taking full responsibility for the publication's content.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpos.2026.1894199/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Angelo Leogrande, Università Lum Jean Monnet, Italy

Reviewed by: Uttam Golder, Jashore University of Science and Technology, Bangladesh

Masud Rana, Pabna University of Science & Technology, Bangladesh

Gülbahar Atasever, Muğla Sıtkı Koçman University, Türkiye

Footnotes

The Bayesian VECM framework is particularly suitable for this study as it accommodates small sample sizes through Bayesian shrinkage techniques.

The short-run positive relationship between real interest rate and consumption is likely to reflect the dynamics of informal credit market and weak inflation rather than traditional monetary channels.

The contractionary policy gap identified through the Taylor rule suggests that Bangladesh Bank’s stance prioritized inflation control at the expense of household consumption growth.

Jamilov and Égert (2014) and Gregor et al. (2021) showed that interest rate pass-through is often sluggish and incomplete in developing economies due to underdeveloped financial systems and market imperfections,

It is worthwhile to note that the absence of prior empirical work on regional monetary policy transmission in Bangladesh underscores the novelty of this study’s contribution.

Bayesian VECM approach is particularly suitable for this study as it captures dynamic interactions among macroeconomic variables while accounting for endogeneity and small-sample limitations through Bayesian estimation techniques. This framework allows for region specific estimation while enabling an assessment of whether monetary transmission patterns are consistent across regions.

The disparity between eastern and western Bangladesh is usually understood as the result of uneven industrialization, differences in market connectivity, disparities in infrastructure, and unequal access to finance. Particularly, Eastern regions, linked to the Dhaka–Chattogram growth corridor, has been better placed to benefit from manufacturing expansion, trade integration, and urban growth. In contrast, several western regions have been highly dependent on agriculture and have shown slower progress in terms of poverty reduction and welfare improvement. Although this gap may have narrowed to some extent, spatial disparities remain visible and continue to reflect differences in infrastructure, local development capacity, and region-specific constraints.

https://bbs.gov.bd

Although Mymensingh Division was officially established in 2015 by separating four districts from Dhaka Division, the Bangladesh Bureau of Statistics reports the survey data according to the current eight-division administrative structure (Bangladesh Bureau of Statistics, 2024). Accordingly, the analysis consistently employs the eight administrative divisions throughout the sample period (2000–2024). Furthermore, household consumption data are available from the Bangladesh Household Survey at five-year intervals. To obtain an annual series consistent with the annual macroeconomic variables, the household consumption data were temporally disaggregated using the Denton proportional benchmarking method, which preserves the original benchmark values while generating annual estimates (Denton, 1971; Dagum and Cholette, 2006).

The descriptive statistics for these variables are summarized in Table A1 in the Appendix A.

This approach is consistent with the Bayesian VAR/VECM literature (e.g., Bańbura et al., 2010; Karlsson, 2013).

The discount factor β is given by β = 1 1 + r , where r is the real interest rate.

The descriptive statistics (Appendix A. Table A1) indicate mild skewness across the macroeconomic variables, such as broad money (−0.869), exchange rate (0.363), inflation rate (−0.137), private sector credit (−0.615), and real interest rate (−0.152), and consumption also shows slight asymmetry (−0.280). These distributional features have no bearing on the unit-root results, which are determined solely by formal stationarity tests rather than normality assumptions.

For conciseness, the result table of the lag order selection criteria is omitted.

The prior and posterior distributions of the Bayesian VECM coefficients by region are shown in Figure B1 of Appendix B.

To verify the robustness of the results, we re-estimated the Taylor Rule simulation using the real GDP output gap instead of the consumption gap (see Appendix A; Table A2). The resulting pattern is nearly identical to the baseline case: in every year from 2000 to 2025, the nominal policy rate exceeded the Taylor-implied rate by more than three percentage points, and the policy stance remained consistently contractionary. The magnitude and direction of the rate gaps are closely aligned with those derived from the consumption-based model, confirming that the conclusion of a persistent tightening bias is insensitive to the choice of output variable. Both specifications indicate that monetary policy in Bangladesh has systematically operated above the rule-based benchmark across the sample period.

To save space, the results are omitted and can be obtained from the authors upon requests.