To investigate how the COVID-19 pandemic influenced the CEMAC financial system under conditions of information sharing, this research employs an autoregressive framework. By leveraging both the available data and established prior knowledge about the behavior of the studied excess liquidity variables, the estimation is conducted through a Bayesian Vector Autoregressive (BVAR) model, following the empirical methodology of Ghenimi and Brahim Omri (22). The priors are selected using the Minnesota approach, given the uniformity in monetary policy decisions and the simultaneous adoption of COVID-19-related measures across CEMAC. A Bayesian model is well-suited to capturing the complex, evolving interactions among financial macroeconomic indicators and health-related information—such as the presence of the COVID-19 virus—thus providing a more nuanced understanding of how these relationships shape excess liquidity.

Under the random shock framework, the classical VAR specification model can be expressed as:

Here, A₀ is defined as the residual matrix for the reduced form of the model, and e_t is the residuals vector. In matrix form, the equation takes on the following structure:

The central assumption in estimating the BVAR model is that the variables included are treated as endogenous. Unlike traditional VAR estimation, Bayesian methods determine parameter distributions assumed proportional to the likelihood of observed data to guide parameter estimation. Following the approach of Garchs and Odendahl (23), the BVAR model can be represented as:

Stochastic volatility is taken into account, enabling us to significantly improve the BVAR model’s density forecasts. By integrating the dig (.) operator, we generate the variance-covariance matrix according to the law N(μ,ϕ). In this work, we assume a posteriori distribution of the parameters to be estimated according to Bayes’ theorem:

Formally, the parameters are determined from the following specification:

Z_t the [(np+1)*1] is the vector of endogenous variables, including a constant term. In this study, the key variables are Debt, Credit_Brat, Liquidity_short, Liquidity_long, Deposits, Equity, and Covid¹. Represented in matrix form, the equation is:

With:

εt denoting the vector of residuals. The variance-covariance matrix of these residuals characterizes their distribution. Functions F and P represent the likelihood and the assumed prior probability distribution of the parameters to be estimated. Thus, the posterior probability distribution emerges from combining the prior information with the observed data.

The prior of Minnesota, as suggested by Ghenimi and Brahim Omri (22) and building on the works of Gininone and Odendahl (24), is applied here to incorporate prior knowledge on model variables. These priors typically assume that any variable exhibits at most a unit root tendency, ensuring that the specification accounts for both stationary and potentially non-stationary behaviors of the data over time. The SIMS and ZHA priors may also be employed if required, depending on whether the data manifest non-stationary properties. Ultimately, chosen values for the hyperparameters (e.g., λ₁ = 0,1;λ₂ = 0,99;λ₃ = 1;λ₄ = 100) guide the shrinkage level and influence how each variable’s historical values shape the posterior distributions.

To ensure the reliability of the outcomes obtained with the Minnesota prior, we will apply a cointegration test. The ARDL model will serve as the foundation for this staggered lag evaluation. Depending on the presence and order of stationarity and cointegration among the variables, the chosen model will be either an error-correction model (ECM) or a vector error-correction model (VECM).

As shown in Appendix 2, which summarizes the descriptive statistics, we analyzed 1488 observations. On average, permanent capital accounts for approximately 12.5% of banks’ total assets. The equity risk coverage ratio averaged 13.16 points and ranged down to a minimum of 58.5 points. Non-performing loans, shareholders’ equity, deposits, gross loans, and overdue loans constituted around 0.7%, 12.8%, 74.17%, 57%, and 0.9% of total balance sheet value, respectively. Both long- and short-term transformation capacities remained robust, reaching 109 points (long-term liquidity) and 186 points (short-term liquidity). The COVID severity index exhibited a notably high average of 97.22, with a minimum value of 27.78. Congo and Chad were the most affected countries. Meanwhile, the average information coverage rate—a measure reflecting the percentage of adults covered—stood at only about 8% across the CEMAC region (see Appendix 4).

Figure 4 presents the initial impulse response results for the period before COVID-19 emerged. The goal here is to demonstrate how information sharing influenced liquidity conditions in CEMAC prior to the pandemic. Two primary scenarios can be distinguished.

The figure shows that a positive shock to information sharing enhances customer deposit variability, which subsequently increases gross credit availability around the fourth and sixth months. As banks in the sub-region share information more effectively, credit access improves, buoyed by rising bank deposits. Consequently, as gross credit grows due to improved information exchange, both short- and long-term excess liquidity decline during the same timeframe. In other words, information sharing curtails excess liquidity (25) across the CEMAC zone.

Saxegaard (26) underscores that reliable, shared information can influence credit rationing practices. In a setting with lower levels of information asymmetry, financial institutions are less vulnerable to mounting bad debts and corresponding excess liquidity concerns. This environment enables banks to better calibrate their lending strategies, thereby expanding financing options through credit extension (7, 27, 28).

At this juncture, our main objective is to examine how COVID-19-induced shocks have influenced banking behavior within the CEMAC sub-region. More specifically, we want to understand how banks respond by adjusting their short- and long-term liquidity positions and how they mobilize deposits and credit supply.

To address potential endogeneity issues among variables such as short- and long-term liquidity, we apply suitable shock constraints. These constraints provide an economic interpretation of the identified shocks. In other words, the pandemic is treated as an external impetus influencing financial indicators, not vice versa. Thus, we categorize and interpret pandemic-related shocks according to two criteria: those affecting banking activities and those influencing prudential, stability-oriented behavior. The rationale behind these choices is rooted in banks’ roles in upholding financial stability chiefly through adherence to regulatory frameworks and by financing the region’s economies.

By grouping banks based on whether they are public, small, or large, Demirgüç-Kunt et al. (13) observe that smaller institutions, in particular, experienced a drop in their returns during the COVID-19 period. However, liquidity enhancement measures in the market helped counter these effects, essentially ensuring that banks could protect their rent extraction capabilities.

The analysis of CEMAC countries presented in the subsequent figure sheds light on the shifting balance between short- and long-term credit transformation capacities. The results (see Figure 5) show that a positive shock to the severity of the COVID-19 pandemic simultaneously affects (in the short run) and diminishes excess liquidity conditions.

Examining the behavior of banks over both the short and long term, the first two quarters indicate that a positive COVID-related shock actually increased excess liquidity. These effects began to manifest starting around the fifth month. Under these conditions, where banks hold significant surplus liquidity, the overall transmission of lending shocks took on a positive trajectory. The subsequent reduction in excess liquidity can be attributed to an uptick in gross credit extended to the economy.

Within the same timeframe, the pandemic’s favorable shocks on excess liquidity have implications for the growth of gross credit and deposits, as observed around the sixth and ninth months. The economic downturn triggered by pandemic-related health measures (such as quarantine and the suspension of business activities) prompted banks to concentrate more heavily on short-term domestic credit rather than allocating funds to investment or longer-term loans.

Referencing Pagano and Jappelli (7), as well as Mamatzakis and Kalyvas (29), who stress the link between bank excess liquidity and information asymmetries, the present circumstances bring these insights into focus. The pandemic, it appears, has helped mitigate excess liquidity within CEMAC’s banking sector by facilitating greater access to short-term credit. This can be explained by the BEAC’s decision not to reinstate liquidity drawdowns initiated in early 2020 (30). Despite the pandemic’s severity and a decline in deposits at the start of the first 3 months, the trend remains unchanged. According to Andrianarison and Nguem (31), banks’ involvement under these conditions is vital. It should enable firms to offset both income declines and various difficulties, including fulfilling their tax obligations, considering the drop in demand and other constraints on household income.

In developed economies, interventionist policies aimed at mitigating the implications of COVID-19 have included credit facilities for borrowers, a change in monetary policy strategies and borrower assistance (facilities), monetary policies, and in short, liquidity support based on prudential measures (13). In CEMAC, banks faced significant exposure to prudential measures. Beyond pandemic-related effects, compliance constraints influenced bank behavior and activity. The COVID-19’s severity positively impacted systematical bank capital and particularly information sharing (see Figure 6). Initially (the first three (3) months), the pandemic’s severity affected equity capital and information sharing. Despite deposit declines mentioned at the beginning of the month, shareholders’ equity compensated for credit requests, aided by the BEAC’s loan drawdowns. Contrary to existing literature (32), CEMAC made counter-cyclical adjustments to equity capital during the pandemic, rather than pro-cyclical adjustments observed in recessions. During an economic recession, banks typically ration credit to riskier agents (33, 34). The pandemic increased risk positively and affected past-due receivables from the second month onwards, following the positive shock of the COVID pandemic.

According to the Central African States Bank, the pandemic period of COVID-19 has not really reduced the risky context of the financial system positively and is highly correlated to the level of bad and doubtful debts. Similarly, in relation to the pandemic context, the health policies of containment and border closures in particular can be indexed to explain the decline in equity capital, the result of the ensuing recession.

The presence in the short and the long term of the cointegrated relationships between financial variables confirms once more the results obtained previously from the error-correction model (see appendices). It emerges that in the nonlinear effect or in the long term, the pandemic chock conducts to observe non-negligible implications on the liquidity of CEMAC banks both in the short and long term.

According to Figure 1, loan demand exhibited a notable increase during the initial phases of the COVID-19 pandemic. This rise can be attributed to firms and households seeking financial support to manage liquidity constraints amid economic disruptions caused by lockdowns and decreased revenue streams. The demand for loans was primarily concentrated in short-term credit facilities, reflecting an urgent need for working capital rather than long-term investments. However, banks exercised caution in their lending practices due to elevated credit risks and uncertainties surrounding borrower repayment capacity, exacerbated by widespread economic instability.

The increase in loan demand aligns with global patterns observed during economic crises, as noted by Demirgüç-Kunt et al. (13). However, the unique challenges within the CEMAC region, such as weak credit reporting systems and limited financial inclusion, amplified the effects of information asymmetry, constraining the effective allocation of credit. This underscores the importance of policy measures aimed at improving transparency and enhancing borrower information through credit registries.

The analysis of the classical variance decomposition confirms some essential cases of the financial activity during the pandemic COVID-19 in the CEMAC zone. The variance of information sharing is at least 99% self-influenced in the short term and the long term. The COVID-19 effect on the variance of information sharing remains negligible, expressing a low level of information sharing between banks. Despite the strong variation in short-term liquidity explained by its own variance along the 10 periods studied, information sharing and the shock of the severity of the COVID pandemic contribute at least 0.76% and 0.11%, respectively, in the long term. As for the variance of long-term liquidity, it controls at least 90% of its variances through its own adjustments. At least 6% of the variance is explained by the level of bank deposits. Deposits drive long-term lending. While by considering the long term, the classic variance in equity is influenced by the gross credit from banks to the economy (at least 20%) influenced by overdue loans in the same time dimension.

The findings of this study shed light on critical dynamics in the CEMAC banking sector during the COVID-19 pandemic. The observed reduction in excessive liquidity, approximately 12% over the initial 6 months of the pandemic, underscores the influence of pandemic-induced economic disruptions on banking activity (35, 36). This reduction aligns with increased credit provision and declining deposit levels, reflecting adjustments made by banks to address evolving liquidity conditions.

Key insights from the study emphasize the interplay between liquidity trends and information asymmetry. For instance, the observed downturn in equity investments and the corresponding decline in information sharing highlight structural inefficiencies that hindered optimal resource allocation during the pandemic. These inefficiencies further accentuate the critical role of robust information-sharing mechanisms in mitigating systemic risks.

The BVAR model results reinforce the need for policy interventions to address information asymmetries. By establishing comprehensive credit registries, policymakers can enhance the transparency of borrower information, thus promoting sustainable lending practices. Additionally, the integration of borrower assistance programs could mitigate the negative effects of economic downturns on the region’s banking sector.

From a policy perspective, the findings also call for a reassessment of regulatory frameworks within CEMAC. As observed, the absence of functional credit bureaus exacerbated the challenges of liquidity management, particularly during periods of heightened uncertainty. Strengthening the region’s financial infrastructure by prioritizing the development of information-sharing platforms could foster greater resilience against future economic shocks.

Finally, this study acknowledges certain limitations, such as its reliance on secondary data and the absence of real-time feedback mechanisms. Addressing these limitations in future research could further enrich the understanding of liquidity dynamics and policy impacts within the region.

The findings of this study highlight critical policy directions to address excess liquidity and improve information sharing within the CEMAC banking sector: