Impact of Macroeconomic Shocks on the Individual Banks: Case of Madagascar

The successive financial crises have highlighted the interdependence between the financial system and the real economy. Prudential measures to limit the negative repercussions of the financial crisis, through the Basel I and Basel II agreements, have shown their limits and the Basel III agreements have consequently integrated the macro-prudential component which aims to ensure the stability of the financial sector as a whole within the economy. The financial stability assessment tool known as the “stress test” has also been developed in various forms and its application to the financial sector, particularly the banking sector, is strongly recommended by the Bank for International Settlements. Indeed, the purpose of this study is to assess the resilience of the Malagasy banking sector to macroeconomic shocks and to evaluate its impact on the capitalization of the banking system through the macroeconomic stress test tool. To do so, we used a dynamic panel model. Non-performing loan forecasts are used to obtain capital projections at both the banking system and bank levels under adverse scenarios. The results show that most banks were able to hold capital above the minimum regulatory threshold of 8% under Basel III standards. However, only one bank fails to meet the minimum capital adequacy threshold. Non-performing loan forecasts are used to obtain capital projections at both the banking system and bank levels under adverse scenarios.

The approach used is based on the methodology of Chaibi and Ftiti (2015). represents specific fixed effects, assumed to be independent of the innovation process and is intended to control any unobserved cross-sectional heterogeneity (assumed to be invariant over time) Therefore, the dynamic panel model to be estimated is written as follows: = + ∅ , , −1 + ′ , + ′ + + , With fixed effects N denotes the number of banks and T denotes the dimension of the time series. With i the cross section going from bank 1 to bank 6; and t notes the chronological dimension going from 2005q2 to 2015q2.
the non-performing loan ratio (NPL) of bank i for period t.

Figure 1. Breakdown of Banking Sector Assets
Note. The shares of assets are presented as a percentage. Source: BCM 2015, author The literature on bank-specific credit risk analysis has noted that the determinants of NPLs should not be sought exclusively among macroeconomic variables, which are exogenous to the banking industry.
The peculiarities of the banking sector and the political choices of each bank, in particular with regard to their efforts to improve efficiency and risk management, are likely to influence the evolution of NPLs (Berger & De Young, 1997;Salas & Saurina, 2002;Podpiera & Weill, 2008;Chaibi & Ftiti, 2015;Zheng et al., 2018;Moussa, 2018;Kiemo et al., 2019).
The following table presents the description of the banking variables.
- Louzis et al. (2012) conducted studies on the assumptions associated with each banking variable. The "mismanagement" assumption implies that low profitability is positively related to the increase in future non-performing loans. The "moral hazard" hypothesis implies that the low capitalization of banks leads to an increase in non-performing loans. Bank managers can also increase their portfolio risk by increasing the loan / deposit ratio (more loans not with deposits) and this leads to more non-performing loans. Similarly, the logic holds for the LLP ratio regarding this assumption. Finally, the "size" hypothesis suggests that the size of banks has a negative correlation with non-performing loans.
Variables used throughout the model with detailed data description and correlation assumptions are as follows:  Leverage ratio (LR) for the capital structure is likely to affect credit risk. Highly leveraged capital leads to a tendency to take more risk due to the need to produce higher returns with lower capital. As financial risk increases with leverage, a positive relationship between a banking firm's risk and leverage is expected.


The loan-to-deposit ratio (LtD) examines bank liquidity by measuring funds used in loans from deposits collected. According to previous studies (Louzis et al., 2012;Vuković. 2014;Makri et al., 2014), this indicator should have a positive effect on the NPL.


As a reminder, the ratios of non-performing loans (NPL) which are obtained by the provisions for loan losses, are considered as a means of controlling anticipated loan losses and make it possible to detect and cover large losses on receivables bank loans. Therefore, banks that anticipate high capital losses should build larger provisions to reduce earnings volatility (Hasan & Wall, 2004). Thus, high loan loss provisions indicate high NPL ratios.


Return on equity (ROE). performance is negatively associated with increases in future bad debts.
In addition, past performance may reflect the high quality of management (Louzis et al., 2012), which has resulted in a decrease in the number of non-performing loans. We expect bank profitability to have a negative impact on non-performing loans. ROE is defined as the amount of net profit expressed as a percentage of equity.


The macroeconomic control variables are the Consumer Price Index (CPI) and the real effective exchange rate (Bitar et al., 2018). For the CPI, high inflation is generally associated with lower interest rates high, which increases the profitability of banks. In this study, we predict a positive relationship with inflation and bank profits. For the real effective exchange rate (REER) variable, the rise in this variable may have mixed repercussions. Empirical studies (Castro, 2013;Nkusu, 2011) include this variable to control external competitiveness. According to Fofack (2005), the appreciation of this variable can weaken the competitiveness of export-oriented firms and make them unable to service their debt. In addition, a real appreciation of the local currency results in higher costs for local goods and services. Nonetheless, an increase in the exchange rate can improve the ability of those who borrow in foreign currency to service their debt (Nkusu, 2011). In this context, the sign of the relationship between the exchange rate and NPLs can be positive or negative.

Projection of the NPLs under the Different Stress Test Scenarios
The results of the panel data estimates will then be used to submit each bank's NPL ratios under adverse macroeconomic scenarios.
In equation (2), the vector of macroeconomic variables, common to all banks, describes the macroeconomic scenarios resulting from an auxiliary model, in our case the GVAR model.
In a stress test exercise, the trajectories of macroeconomic factors are considered "exogenous", in the sense that banks' profits / losses are not passed on to the economy in general. However, in practice, the macroeconomic variables that go into a model are determined using a combination of expert judgment and the output of an auxiliary model and therefore try to implicitly take into account the feedback loop between financial conditions and macroeconomics (Covas et al., 2014).
In the present case, the adverse stress test scenarios come from the results of the GVAR model. As a reminder, the macroeconomic scenarios used for forecasting the ratios of non-performing loans for each bank are as follows:


The base scenario or reference scenario consists of the expected evolution of the economic situation, that is to say the projection of the results of the GVAR model without causing any shock for the system. The method of projecting equity with the scenarios is identical to the approach adopted to measure the capitalization of the aggregate banking sector. As a result, in addition to the measures of four macroeconomic shock scenarios, banks' risk-weighted assets are then increased by 16%. These two categories of scenarios are applied to banks's equity and then if the latter are still within the regulatory standard in terms of equity, despite the CAR ratios level of most banks are still above 8%. The figure   below shows the evolution of the CAR ratio of each bank compared to the CAR of the aggregate sector.

Figure 2. Evolution of Banks' Own Funds between 2005 and 2015
Source: BCM 2015, author

Econometric Estimates and Results
The credit risks under adverse macroeconomic scenarios will be measured and will subsequently be reflected in the level of equity of each bank.
The data is processed using the Stata software.

Descriptive Statistics
The table below presents the descriptive statistics (mean, minimum, maximum and standard deviation) of the banking variables and macroeconomic variables used in this study for the period from 2005 to 2015. The 6 banks which constitute the sample account for more than 95% of the total banking sector assets. With regard to the loan-to-deposit ratio (LtD), the average ratio is 0.51, which means that overall half of the deposits are mobilized as a loan.
For profitability (ROE), the table shows that the banking system is profitable with an average ratio of 0.34. This ratio also varies from one bank to another with standard deviation's value of 0.25.
Looking at the size of the banks, t the average size is 0.16. The verage prices is 5.5%.
The econometric results as well as the results of the stress test implementation provide more elements to analyze and relate the behavior and the situation of banks under the conditions of adverse macroeconomic scenarios.

Econometric Results: Regression of the Panel Model
Before undertaking any econometric analysis, preliminary panel data specification tests were performed.

Stationarity Test
To determine the stationarity of the panel data, a panel unit root test was carried to the study variables.
The panel unit root test involves solving in an autoregressive process AR (1) for the estimated equation (3). If, this means that the dependent variable was dependent on its own delay and that contains a unit root (not stationary) which can lead to erroneous results when testing the significance hypothesis of the explanatory variables | | = 1 (Gujarati, 2009).
The result of the test is a Fisher F statistic shown in the following table:  The Hausman test on the choice between the two models shows a p-value less than 5%. The null hypothesis of preference for the random effect model is rejected, so the fixed effect model is best suited for our analysis.

Heteroskedasticity Test
The Wald test is used to detect heteroskedasticity between individuals. The null hypothesis being homoscedasticity (Leblond & Belley-Ferris, 2004) and the statistic follows a law of degree of freedom N. 2 The table of results below shows the Chi-square statistic: The p-value the test is less than 5%, therefore the null hypothesis is rejected, which shows the presence of heteroskedasticity between individuals.

Autocorrelation Test
To test for the presence of correlation between the errors of individuals, we used the Wooldridge test.
The null hypothesis of the test is the independence of the residuals between individuals.  The p-value is greater than 5%, 0 is accepted. This implies that the instruments are valid.
The results of the various panel model specification tests above thus justify the choice of the dynamic panel model with fixed effects, by using the estimator of GMM.

Regression Results of the Dynamic Panel Model
The following table presents the coefficients of the GMM estimator of the dynamic specification with fixed effects according to the approach of Arellano-Bond (1991). The incorporation of variables specific to each bank in the model does not change the impact of the macroeconomic variable of interest on the credit risk indicator. Indeed, the addition one by one of the banking variables does not modify the sign of the macroeconomic variable of interest and even that of the other variables as shown successively in columns (2), (3), (4) and (1). It also demonstrates the stability of the model.
The results in the above table highlight too that a part from sensitivity to macroeconomic conditions, the NPL ratio is also sensitive to specific bank factors. According to the results of the regression of the panel model, the explanatory variables, including the lagged dependent variable are all statistically significant, except for the CPI.
The estimated coefficients have the expected signs, compatible with the theoretical arguments of the literature. The period-lagged non-performing loan ratio coefficient (is positive and significant. An increase in NPL ratios during the last quarter will be followed by an increase in the current quarter. −1 ) The Loan-to-Deposit ratio (LtD) correlates positively with the credit risk indicator, which supports the moral hazard hypothesis. The capital performance ratio (ROE) is used to test bank mismanagement.
The results show a significant negative relationship between ROE and NPL ratio, suggesting that ROE can significantly influence the level of NPL. The leverage ratio (LR) is a negative determinant of credit risk. Bank size (SIZE) negatively and significantly affects the credit risk indicator. The discussion of the relationship between macroeconomic and banking variables and the indicator of banks' credit risk will be presented in paragraph 4.

Reaction of Credit Risk Indicators to Macroeconomic Shock Scenarios
The results of the panel regression are then used to determine the change in the ratio of non-performing loans under the conditions of the macroeconomic stress test scenarios. The scenarios consist of the The negative shock to GDP results in a sharp increase in the ratio of non-performing loans of all banks in the future horizons. Banks A, B and F are the most affected when looking at changes in the NPL ratio ( Figure 3).

Source: Author's calculations
The following table (Table 10) shows that:


The positive exchange rate shock, i.e., a depreciation of the local currency, has a low positive impact on credit risk indicators for all banks. The variation in the NPL ratio is below 1% for all banks.


The positive shock of the oil price generates an increase of between 0.3% to 0.8% of the NPL ratios of the banks (Table below).
 Finally, a negative variation of one standard deviation in the price of agricultural commodities leads to a modest increase in the ratios of non-performing loans, with a variation of up to 0.68%. Banks A and F are the most affected by the commodity price shock.
In view of the reactions of NPL ratios, banks are more affected by GDP and oil price shocks compared to commodity price shocks and they are weakly affected by exchange rate variation. The following  Source: Author's calculations

Individual Bank Equity Projections
Since the last step of the stress test is to measure the ability of banks to absorb shocks through their equity, changes in the level of banks's credit risk are then reflected in equity through a projection. In fact, deteriorations in credit risk following shocks have an impact on and limit the capital adequacy ratio of banks, and can therefore compromise the system if their capital adequacy ratio falls below standards (Basel III, international standards).
The methods for calculating equity and the assumptions for adverse scenarios are the same as those adopted in the aggregate banking sector. Two adverse scenario assumptions are considered, the first encompasses the different macroeconomic shocks with banks's risk-weighted assets assumed constant, while the second assumes a 16% increase in risk-weighted assets in addition to adverse macroeconomic shocks.
The results on the equity projection under the different scenarios show the banks's credit risk absorption potential. Overall, most banks were able to meet the minimum capital threshold, i.e., the minimum capital ratio of 8%. For the first hypothesis of unfavorable scenarios, it is the negative shock www.scholink.org/ojs/index.php/jepf Journal of Economics and Public Finance Vol. 7, No. 3, 2021 Published by SCHOLINK INC. to GDP that has the greatest impact on the decrease in equity of all the banks, followed by the shock of the oil price. Indeed, these impacts stem from the magnitude of the consequences of these shocks on the deterioration of banks's credit.
For the negative GDP shock, it was banks A, C and F that suffered a sharp decrease in equity. Bank B is the only one that has a CAR ratio below 8% even though the magnitude of the impact of the GDP shock on this bank is less compared to the three previous banks. This is due to the fact that at the end of the second quarter of 2015, Bank B's capital ratio is just above the minimum with a CAR ratio of 8.59%.
The oil price shock also causes a fall in the capital ratio of banks but with a smaller variation compared to the GDP shock. The effects of shocks on the exchange rate and commodity prices on banks' capital ratios remain modest.
Looking at the impact of adverse scenarios with the assumption of a 16% increase in risk-weighted assets, banks's capital ratios experience a considerable drop for all types of shocks while the variation in this drop, compared to the decrease generated by the scenarios of the first hypothesis, differs from one bank to another. Indeed, with the increase in risk-weighted assets, the capital ratios of banks A, D and E are the most affected, followed by those of banks C and F. The capital ratio of bank B experiences less negative variation compared to other banks.
The following paragraph provides a discussion with respect to these adverse scenario results on banks.

Negative relationship between NPL ratio and GDP
The credit risk tends to increase when economic conditions deteriorate. This result aligns with various theories on the business cycle and credit risk that macroeconomic conditions affect the quality of banks's loans. This case demonstrates the vulnerability of some borrowers to macroeconomic shocks and therefore the deterioration of their ability to repay their debts in an unfavorable economic situation.

Relationship between credit risk and control variables: CPI and REER
The inflation rate (CPI) is negatively but not significantly correlated with the NPL ratio, which implies that the credit risk is insensitive to changes in the inflation rate. This negative relationship is explained by the fact that higher inflation weakens the ability of borrowers to service their debt by reducing their real income. This result is consistent with that of Castro (2013), who explains that inflation affects not only the real value of outstanding loans, but also the real income of borrowers. Thus, one effect is practically canceled by the other and the final impact of inflation on credit risk is canceled.
The negative relationship between the Real Effective Exchange Rate (REER) and the NPL ratio indicates that a depreciation (appreciation) of the local currency contributes to a fall (rise) in the ratio of non-performing loans. Indeed, the real depreciation of the local currency (increase in the REER) promotes the competitiveness of exporting companies and thus improves their repayment capacity, thus reducing non-performing loan operations. This result is consistent with the conclusions found by manyBeck, et al. (2013), Chaibi and Ftiti (2015) and Dua and Kapur (2017).

Relationship between NPL and NPL-1 Ratio
The positive sign of the lagged NPL ratio means that NPLs are likely to increase when they increased in the previous year, this is due to amortization. The persistence of growth in the NPL ratio is also confirmed by previous work such as Sorge and Virolainen (2006), Beck et al. (2013).

Positive Relationship between NPL and Loan-to-deposit Ratio LtD
The credit-to-deposit ratio measures liquidity and reflects the attitude of banks towards risk; this ratio is also associated with the "moral hazard" hypothesis. The results which show a significant positive effect on the ratio of non-performing loans confirm the moral hazard hypothesis and align with the conclusions of Dimitrios et al. (2016) and Koju et al. (2018). The increase in the loan-to-deposit ratio indicates a preference for risk and is expected to lead to an increase in non-income producing loans.
Indeed, in order to obtain more profitability, banks grant loans without maintaining the credit quality, which can reduce the quality of the loans and, consequently, increase the ratio of non-performing loans.

Relationship between NPL, Leverage Ratio (LR) and Size (SIZE)
The insignificant negative relationship between the NPL ratio and the leverage ratio does not support the "too big to fail" assumption on risk taking. This assumption predicts that the higher the liabilities relative to the total assets, the higher the probability of impaired loans should be. An explanation could be provided by the work of Louzis et al. (2012) which show that the relationship between the leverage ratio and credit risk is conditioned by the size of the banks. These results suggest that leverage tends to increase NPLs, but this effect only occurs up to a certain size threshold. After this threshold, leverage does not have a positive effect on NPLs, which means that among large banks there is no "too big to fail" effect on NPLs. That is, small banks tend to increase their leverage ratio by expanding loans to unreliable quality borrowers and hence have more non-performing loans. However, the banking sector is mainly made up of large banks, the four out of six banks in our sample hold more than 80% of total assets, For the direct relationship between bank size and credit risk, the results show a significant negative Published by SCHOLINK INC.
other two banks (B and D) are characterized by their size which can be qualified as small compared to the other banks in the sample with a percentage of assets respectively around 5% and 3%. However, their behavior during bad times diverges completely. Figure 5, which shows the evolution of the average ratios of the banking variables, will be used as support for the explanation of the situations of the banks under the conditions of stress test.  bank's credit risk is the lowest of all banks ( Figure 5, NPL). This is due to the fact that this bank's loans are among the most exposed to macroeconomic shocks (see Figure 3 and the results on credit risks). In other words, bank B borrowers are more affected by unfavorable macroeconomic conditions. In addition, this weakness of resistance is also due to the size of the loans granted by this bank in relation to the deposit with a percentage of 75% ( Figure 5, loan on deposit). In terms of risk-taking, the situation of this bank confirms the hypothesis of "moral risk". Indeed, the bank focuses above all on increasing profitability and to do this, it grants loans without maintaining the quality of the borrower, which reduces the quality of loans especially in times of distress and subsequently generates the increase the ratio of non-performing loans. Figure 5 shows that Bank B is among the most profitable compared to other banks by referring to Return on Assets (ROA) which measures the ratio of profit to total assets.

Large Banks (A, C, F and E).
The more capitalized banks are those classified as large banks and hold equity between 8% and 12% after the shock scenarios ( Figure 10). More precisely, three of these banks hold more than 10% of equity even after the extreme scenarios of hypothesis (A, C and F); for one of them (bank E), the capital ratio falls below 10% under the two shock hypotheses.

Banks A, C and F
For banks A, C and F, half of their deposit is allocated to loans (see Figure 5, loan on deposit). As a result, during periods of recession, even if these banks are also exposed to credit risks during unfavorable macroeconomic situations, they manage to hold the necessary capital. These banks therefore diversify their activities by using the other half of the deposits to other income-generating activities such as transaction portfolios, investment portfolios and foreign currency operations; and consequently obtain various sources of non-interest income. These different sources of income allow these banks to be more resilient even during times of distress as their profits do not depend solely on loans.
Indeed, on the basis of the results on the reaction of credit risks, even if one of these banks (A) is the most exposed to risks during recessions while the other two are slightly affected, it manages to maintain a high capital adequacy ratio during periods of shocks. This more pronounced risk aversion is explained by the fact that the borrowers of this bank are sensitive to macroeconomic conditions on the one hand, and on the other hand, the quality of borrowers is lower compared to those of the other two (C and F). In other words, on the basis of 50% of their deposits turned into loans, on average 15% of Bank A's loans are non-performing while for the other two, this ratio is around 8% only. However, As for the other two banks (C and F), they can be considered as selective in terms of granting credit.
Their borrowers are weakly sensitive to the deterioration of the macroeconomic situation. As a result, they manage to hold equity below the minimum threshold during adverse scenarios.

Bank E
For bank E, which is one of the large banks, it differs from the three previous banks (A, C and F) by the fact that it has a capital ratio of less than 10% after the assumptions of macroeconomic shocks ( Figure   4). Indeed, this bank is the least affected by credit risk in a period of macroeconomic shocks and which www.scholink.org/ojs/index.php/jepf Journal of Economics and Public Finance Vol. 7, No. 3, 2021 Published by SCHOLINK INC.
also selects its borrowers such as banks C and F. Unlike the other two (C and F), it strongly favors other sources of income over financing the economy: only 38% of these deposits are converted into loans despite its size. This is how it is the most profitable in terms of return on assets (ROA) (see Figure 5). However, unlike the three big banks (A, C and F), it does not manage to hold as much equity in times of shocks, i.e., above 10%. This is explained by the fact that the equity ratio in normal periods of this bank is lower compared to that of other banks, with a percentage of 10.3% against 13.6% ( Figure 10). Specifically, Bank E's capital and reserves are lower compared to those of the other three banks, or even half, and this consequently weakens the capital adequacy ratio in times of macroeconomic stress. 6% (Figure 4). Specifically, Bank E's capital and reserves are lower compared to those of the other three banks, or even half, and this consequently weakens the capital adequacy ratio in times of macroeconomic stress. 6% (Figure 4). Specifically, Bank E's capital and reserves are lower compared to those of the other three banks, or even half, and this consequently weakens the capital adequacy ratio in times of macroeconomic stress.

Conclusion
The application of the macroeconomic stress test on the banks of Madagascar was carried out in three stages. First, the relationship between the credit risk indicator of all banks with macroeconomic variables and banking variables are estimated using a dynamic panel model from the Allerano-Bond (1991) GMM estimator. Second, NPL ratios are subject to different macroeconomic shock scenarios. The scenarios used for the individual banks are identical to the shock scenarios applied to the aggregate banking sector, generated from the GVAR model. Third, the reactions of the NPL ratios are translated into the shareholders' equity of each bank in order to measure their resistance and resilience to the adverse scenarios.
The results of the macroeconomic stress test with individual banks showed that most banks remain capitalized after macroeconomic shocks. It should be noted, however, that banks' exposure to credit risk does not depend on their size. The different scenarios of macroeconomic shocks affect the credit risk of all banks, including the negative shock to GDP, but to a different degree. The results showed that among the banks most affected by credit risk during the occurrence of macroeconomic shocks are a large bank and a small bank. For the capitalization of banks under adverse scenarios, in addition to macroeconomic shock scenarios, another hypothesis of an increase in banks' risk-weighted assets was also considered. The results at the level of each bank showed that only one bank could not have a capital adequacy ratio higher than the regulatory minimum of 8% after the scenarios. This bank is among the most affected by credit risk. Among the other banks which manage to hold a level of capital higher than the minimum required, there is one which has a sufficiently large capital that although it is also sensitive to credit risk, it can manage to absorb shocks. Some of them, mainly composed by large banks, hold a capital ratio above 10% even after the severe scenarios of the second hypothesis. Indeed, Banks's exposure to credit risk depends in particular on the sensitivity of borrowers in an unfavorable economic situation. In other words, some