liquidity risk in the mena region banking sector: does ...

The Journal of Developing Areas Volume 53

No. 1

Winter 2019

LIQUIDITY RISK IN THE MENA REGION BANKING SECTOR: DOES BANK TYPE MAKE A DIFFERENCE? Suzanna ElMassah* Cairo University, Egypt Zayed University, United Arab Emirates Shereen Mostafa Bacheer Cairo University, Egypt Zayed University, United Arab Emirates Ola Al Sayed Cairo University, Egypt Sultan Qaboos University, Oman ABSTRACT Liquidity risk is a challenge facing banks in their efforts to maintain financial stability. Islamic banks are under added pressure with the constraint of having to adhere to Sharia’h principles. The goal of this paper is to investigate the determinants of liquidity risk in Islamic and conventional banks in the Middle East North Africa (MENA) region. The generalized, least squares model is utilized to estimate the determinants of liquidity risk in 257 banks (90 Islamic and 167 Conventional) over the period 2009-2016, in which the struggles of both types of banks to mitigate the impacts of the global financial crisis were observed. A dummy variable representing the bank type is included to allow for comparison between liquidity risk determinants in both types of banks. The model investigated the impact of four bank specific variables and a macroeconomic one on bank liquidity represented by five alterative ratios. The results show a positive effect of bank size on liquidity risk of all sample banks, thus demonstrating that both bank types follow the “too big to fail” rule. Capital adequacy has a positive impact on the liquidity risk of all sample banks irrespective of bank type. Return on assets has no significant effect while credit risk has a negative impact on liquidity risk of both bank types. That is, higher credit risk encourages a more conservative liquidity management policy in both bank types, despite the theoretical fact that Islamic banks have higher credit risk due to the “risk sharing principle”. Similarly, real per capita GDP has a positive impact on liquidity risk of conventional and Islamic banks, reflecting their procyclical lending behaviour. Evidently, bank type in the MENA region does not affect the determinants of a bank’s liquidity risk; Islamic and conventional banks use different terms for their practices, but in reality mobilize funds the same way. This is due to the fact that both banks operate under the same micro- and macroeconomic conditions and are both influenced by the same domestic and international liquidity regulations. Introducing more efficient financial products and having a unified regulatory and supervisory framework can offer Islamic banks better opportunities. JEL Classifications: G11, G20, G21, G32, G33, Z1 Keywords: Liquidity risk, Liquidity ratio, Islamic banks, MENA region, Islamic finance, Bank size, Capital adequacy, ROAA, Credit risk. Corresponding Author’s Email Address: [email protected]

148

INTRODUCTION Liquidity is considered as a crucial indicator of a bank’s performance, it refers to the ability to meet its financial obligations on time1, while liquidity risk (LR) refers to the possibility of failing to meet these obligations. Accordingly, LR management refers to a bank’s strategies to balance the demand for liquidity on the liability side with the supply of liquidity on the asset side. Liquidity problems occur if a bank fails to balance those two sides, does not have sufficient internal liquidity reserves or fails to obtain funds from external sources (Ismal 2010). During the global financial crisis of 2008, LR was one of the greatest challenges faced by banks. Many banks failed, despite the unprecedented levels of liquidity support offered by central banks to help maintain financial stability (Vodova 2011). This crisis drew attention to Islamic banking, which emphasizes risk sharing. Islamic banks were established mainly to provide a variety of religiously acceptable financial services to Muslim communities (Hassan and Lewis 2007). Academics and policymakers alike highlighted the advantages of Islamic financial products, which were less dependent on debt instruments and more reliant on equity for greater risk-sharing (Moheildin 2012). These products are very attractive to those segments of the population that demand financial services consistent with their faith-based beliefs (ElMassah 2015). LR is as important to Islamic banks as it is to their conventional counterparts. Yet, having to comply with Sharia’h principles, which prohibit most existing interest-based instruments and put very restrictive conditions on debt instruments reselling, is an extra challenge for the liquidity management by the Islamic banks (Ariffen 2012). Since Islamic banking is a relatively new branch of finance, emerging only in the 1970s, there has been a growing body of empirical studies to evaluate their performance. Accordingly, the main goal of this paper is to find out if the bank type makes a difference in the determinants of LR. We investigate the impact of some bank-specific and macroeconomic variables on the LR of both types of banks in the period 2009-2016. The main value added by the study is using five different liquidity ratios as proxies for LR. The study uses annual data of all 257 banks in 14 MENA countries2: 167 conventional and 90 Islamic. We focus on the period after the global financial crisis, which witnessed the struggles of the majority of banks to mitigate the impact of the crisis, given the lack of available data of Islamic banks prior to 2009 is an additional reason for choosing the sample period. In Section 2, the relevant literature is reviewed, while Section 3 introduces the methodology, data and hypotheses development. Empirical results and their discussion are presented in Sections 4 and 5, respectively. Finally, Section 6 provides the conclusions of the study, together with certain policy recommendations. LITERATURE REVIEW Among the various performance criteria, liquidity has usually been at the top of the list. Empirical literature in this area can be categorized into three groups: studies evaluatingthe

149

liquidity level of Islamic banks, studies comparing liquidity in Islamic banks vs. their conventional counterparts, and studies focusing mainly on MENA countries. A number of empirical studies aimed at detecting the main determinants of LR in Islamic banks. (Ahmed, Akhtar & Usman 2011) investigated the main factors affecting liquidity levels in six Pakistani Islamic banks over 2006-2009. Liquidity was found to be negatively affected by the debt-to-equity ratio and the non-performing loans ratio, while it was positively impacted by bank size and the asset utilization ratio. (Ahmed, Ahmed & Naqvi 2011) used evidence from Pakistan to show the main determinants of liquidity, namely leverage, tangibility and bank age. However, profitability and bank size were found to have an insignificant impact on the sample banks’ liquidity levels. Using evidence from Malaysia, (Sulaiman, Mohamad & Samsudin 2013) showed that liquidity was influenced by macroeconomic variables representing the economic cycle. Moreover, they proved that the return on assets (ROA) positively impacted liquidity, while bank size negatively affected it. (Ramzan and Zafar 2014) concluded that only bank size affected the liquidity of Islamic banks in Pakistan. Other variables were proved to have an insignificant relationship with liquidity. (Mobin and Ahmad 2015) analyzed the liquidity levels of 15 Malaysian Islamic banks. Among five tested key independent variables, they proved that liquidity had a significant relationship with bank size and bank specialization. A second group of studies focused on comparing the liquidity levels of Islamic and conventional banks. (Iqbal 2012) used a ratio analysis to compare the liquidity of five Islamic banks to 22 conventional counterparts in Pakistan during 2007-2010. The results highlighted the better liquidity positions of the Islamic banks in comparison to the conventional ones, both banks’ liquidity levels were positively related to the capital adequacy ratio, the ROA and the return on equity (ROE), whereas the non-performing loans ratio was found to have a negative impact on liquidity. (Anam et al. 2012) compared the liquidity of six conventional vs four Islamic banks in Bangladesh in the period 20062010, and concluded that in Islamic banks, liquidity was negatively affected by the net working capital and the ROE, while positively affected by the bank size, the capital adequacy ratio and the ROA. (Jaffar and Manarvi 2011) compared the performance of five Islamic and five conventional banks in Pakistan in the period 2005-2009. Islamic banks have performed better with regard to liquidity position and possessing adequate capital, while conventional banks excelled in management quality and earning ability. Using a comparative analysis, (Akhtar, Ali & Sadaqat 2011) evaluated liquidity levels in Pakistani Islamic and conventional banks, and found out that liquidity was insignificantly affected by the bank size and the networking capital of both types of banks, but the capital adequacy ratio of conventional banks and the ROA of Islamic banks have positive and significant impacts. A third group of studies, and yet a relatively limited one, focused on assessing the liquidity of Islamic banks compared to conventional entities, in the MENA region in particular. (Ben Khediri, Charafeddine & Ben Youssef 2015) investigated the features of Islamic vs. conventional banks in the Gulf Cooperation Council (GCC) countries in the period 2003-2010. The results showed that Islamic banks were, on average, more profitable, more liquid, better capitalized and had lower credit risks than conventional ones. Similarly, (Hares, AbuGhazaleh & Galfy 2013) examined the financial ratios of the banking industry in the GCC (conventional vs. Islamic banks), they highlighted that Islamic banks were more liquid than conventional banks in the period 2003-2011, and have

150

lower loans-to-deposits ratios, lower deposits-to-assets ratios, higher cash reserves and balances at the central bank (or other financial institutions) to deposits ratios, and higher cash and portfolio investments to total deposits ratios. (Parashar and Venkatesh 2010) compared the performance of Islamic and conventional banks in the GCC countries before and during the 2008 financial crisis. They concluded that during the crisis, conventional banks suffered more than the Islamic institutions in terms of liquidity. However, no previous studies have investigated the determinants of LR in MENA conventional and Islamic banks; this study fills this gap. METHODOLOGY, DATA AND HYPOTHESES DEVELOPMENT MODEL This study focuses on investigating LR in the MENA banking sector by testing the factors that affect five liquidity ratios. A panel data model is employed that captures all behavioral differences between individuals, thus taking into account the individual heterogeneity of separate banks, dummy variables were used to differenciate between Islamic and conventional banks. The study follows (Vodova 2011) in estimating each liquidity ratio using the following functional form:

𝐿"# = 𝛼" + 𝛽( 𝐷"# 𝑋"# + 𝛽+ 𝐷"# 𝑍"# + 𝜀"# Where 𝐿"# is the liquidity ratio for bank (i) over time (t), 𝑋"# is a vector of explanatory variables for bank (i) over time (t), 𝑍"# is a vector of country-specific macro variables for bank (i) over time (t), 𝐷"# is a dummy variable (Islamic bank=1, commercial bank=0), 𝛼" is an individual-specific constant term, βs are coefficients that represent the slopes of variables, and ε0 is the error term. So we have different x variables. We created a dummy (Islamic/ conventional) and in the model we included for example dummy*x1, as a result we estimated 2 slopes for x1, one for Islamic, and the second for conventional. By using the interaction terms with a dummy variable, we were able to estimate different slope effects for the two groups. DATA The study uses the annual data of all the banks working in 14 MENA countries during 2009-2016. There were 257 banks, 90 Islamic and 167 conventional. Table 1 presents the sample’s breakdown.

151

TABLE 1. SAMPLE BREAKDOWN Country UAE Bahrain Egypt Jordan Kuwait Lebanon Oman Qatar KSA Syria Iran Iraq Tunisia Yemen Total

Conventional Banks 21 12 23 11 5 39 6 6 7 10 1 10 12 4 167

Islamic Banks 8 19 3 3 8 2 3 5 4 3 20 7 1 4 90

Total 29 31 26 14 13 41 9 11 11 13 21 17 13 8 257

Five different liquidity ratios are used as proxies of LR facing the sample banks. Table 2 presents the dependent variables: the five ratios adopted in the study and their correlation with the bank LR. TABLE 2. FORMS OF LIQUIDITY RATIOS IN RELATION TO LR Liquidity ratio L2 L5 L6

L7

L8

Calculation

Relation to LR

(Loans / Total Assets)*100

Indicates the percentage of a bank’s assets that is tied up in loans. Higher L2 means lower bank liquidity. ↑ L2 → Higher LR Higher L5 means lower bank liquidity. ↑ L5 → Higher LR

(Loans / Customer & Short Run Funding)*100 Loans / (Customer & Short Run Funding + Other Funding – Hybrid Capital – Subordinated Debt)*100 (Liquid Assets / Customer & Short Run Funding)*100

(Liquid Assets / Customer & Short Run Funding + Other Funding – Hybrid Capital – Subordinated Debt)*100 Source: Constructed by the researchers

Similar to L5 with the exception of capital instruments Higher L6 means lower bank liquidity. ↑ L6 → Higher LR This is a deposit run-off ratio, which indicates the percentage of customer and short-term funds that could be met in case of sudden withdrawal. Higher value means higher capacity to absorb a liquidity shock. Higher L7 means higher bank liquidity. ↑ L7 → Lower LR Similar to L7, but looks at the amount of liquid assets available to borrowers as well as depositors. Higher L8 means higher bank liquidity. ↑ L8 → Lower LR

This study is guided by the prior theoretical and empirical literature to choose the independent variables that influence a bank’s liquidity ratios. Four bank-specific factors

152

are considered and one macroeconomic factor to control for economic conditions. Table 3 shows the independent variables used in the quantitative analysis. TABLE 3. IDENTIFICATION OF INDEPENDANT VARIABLES Bank-specific variables Bank Size Return on Average Assets (ROAA) Credit Risk

Measured by Log of total assets: tangible or intangible Net income to total assets: It indicates the efficiency and ability of a bank to generate profit by using its assets Loan Loss Reserves/Gross Loans: Indicates the creditworthiness of a bank Macroeconomic variable (control variable) Growth rate Log of real per capita GDP. Source: Constructed by the researchers

Data points are collected annually from 2009 to 2016 for the bank-specific data from the BankScope database, while the GDP annual data is derived from the International Financial Statistics of the International Monetary Fund (IMF). The panel is unbalanced, as some of the banks did not report over the full study period. HYPOTHESES Based on the literature and the description of variables, the study tests the following hypotheses to investigate the LRs determinitants of MENA banks. BANK SIZE AND LIQUIDITY Large banks, according to the “too big to fail” theory, give little concern to safe levels of liquidity. The bank size will result in an increase in its LR. This relationship is supported by studies; (Ahktar, Ali & Sadaqat 2011), (Anam et al. 2012), (Sulaiman, Mohamad & Samsudin 2013), (Laurine 2013), (Iqbal 2012) and (Naveed, Akhtar & Usman 2011). Accordingly, we develop the following hypothesis: 𝐻2: Bank size in the MENA positively affects LR. 𝐻2 (𝑎): An increase in a bank’s size encourages the bank to invest in large loan portfolios, so (𝐿2 , 𝐿5 and 𝐿6 ) tend to increase for big banks. 𝐻2 (𝑏): The levels of liquidity (𝐿7 and 𝐿8 ) that the bank should keep usually fall with an increase in the bank’s size as a result of higher loan ratios (𝐿2 , 𝐿5 and 𝐿6 ). CAPITAL ADEQUACY AND BANK LIQUIDITY A bank with great financial strength is more confident and maintains lower levels of liquidity by disbursing more loans, hence exposed to a higher LR. This impact is supported by; (Ahktar, Ali & Sadaqat 2011), (Anam et al. 2012), (Iqbal 2012) and (Naveed, Akhtar & Usman 2011). Accordingly, we propose the following hypothesis:

153

𝐻5: Capital adequacy of MENA banks positively affects their LR. 𝐻5 (a): A higher capital adequacy ratio, which reflects more financial strength, will encourage a bank to increase its loanable funds. With higher loans (𝐿2 , 𝐿5 and 𝐿6 ), banks are exposed to higher LR. 𝐻5 (b): As a result of increasing the size of a loan portfolio, the liquidity ratios (𝐿7 and 𝐿8 ) tend to fall, keeping high exposure to LR. ROAA AND BANK LIQUIDITY A bank with higher (ROAA) seeks more risky/less liquid investments to gain more profits, and is therefore exposed to higher LR. Prior studies by (Ahktar, Ali & Sadaqat 2011), (Anam et al. 2012), (Sulaiman, Mohamad & Samsudin 2013) and (Iqbal 2012) support the same conclusion. Thus, we develop the following hypothesis: 𝐻6: Return on average assets of MENA banks positively affects their LR 𝐻6 (a): In cases of generating more profits (higher ROAA), banks will be exposed to higher LR due to the positive effect of ROAA on loans ratios (𝐿2 , 𝐿5 and 𝐿6 ). With an increase in profits, a bank’s financial leverage will increase, creating motivation for a bigger loan portfolio. 𝐻6 (b): When a bank is more efficient in generating earnings from assets (higher ROAA), it will grant more loans, keeping low levels of liquidity (measured by ratios 𝐿7 and 𝐿8 ), thus exposing its portfolio to higher LR. CREDIT RISK AND BANK LIQUIDITY Credit risk refers to a bank’s loan defaults. The relationship between credit and LRs is discussed in the classic financial intermediation theory and in the industrial organization approach to banking (Imbierowicz 2014). Later on, empirical literature based on both branches investigated the direction of this relationship. Recent empirical studies referred to the negative relationship between both risk types, as a higher credit risk encourages banks to hedge against it by keeping more liquid assets (Gatev, Schuermann & Strahan 2009). Moreover, during times of macroeconomic uncertainty or stress, households and corporations will shift towards bank deposits, which are usually perceived as a safer alternative (Acharya & Naqvi 2012). Accordingly, the following hypothesis is developed: 𝐻7: Credit risk of MENA banks negatively affects their LR 𝐻7 (a): Higher credit risk has a negative effect on the loans ratios (𝐿2 , 𝐿5 and 𝐿6 ), since it encourages a bank to follow more conservative lending behavior by decreasing its loan portfolio, and eventually it becomes less exposed to LR.

154

𝐻7 (b): A bank with low-quality loans keeps more liquid assets to hedge its portfolio against liquidity shocks. Accordingly, higher credit risk affects ratios (𝐿7 and 𝐿8 ) positively, thus reducing the bank’s exposure to LR. COUNTRY GROWTH AND BANK LIQUIDITY When a bank works in a growing economy, it will loosen its lending and funding behavior and therefore become exposed to higher LR. Prior empirical literature (Sulaiman, Mohamad & Samsudin 2013) and (Nikomaram, Taqhavi & Diman 2013) found a positive impact of real GDP growth rate on the banking sector’s LR, thus concluding that economic growth boosts banking LR due to the increase in banks’ offered loans. Accordingly, we propose the following hypothesis: 𝐻8: Real per capita GDP positively affects LR of MENA banks. 𝐻8 (a): A bank responds to an increase in real GDP by granting more loans to benefit from economic expansion. Accordingly, ratios (𝐿2 , 𝐿5 and 𝐿6 ) rise, and the bank is exposed to higher LR. 𝐻8 (b): In line with 𝐻8 (a), granting more loans is usually accompanied by keeping less liquidity (lower 𝐿7 and 𝐿8 ) and hence experiencing higher. EMPIRICAL RESULTS DESCRIPTIVE STATISTICS The trend of MENA banks’ liquidity ratios during 2009-2016 is represented in Figure 1, which shows that liquidity ratios on average, except for L2, were higher for Islamic banks compared to their onventional counterparts over the whole study period.

155

FIGURE 1. LIQUIDITY RATIOS IN MENA BANKS – AVERAGE DEVELOPMENTS FOR 2009-2016

Source: Constructed by the researchers

Similarly, Figure 2 illustrates a comparison of MENA banks’ latest average liquidity ratios in 2016. From the figure, it is obvious that all 2016 liquidity ratios, except for L2, were higher for Islamic banks than for conventional ones. This figure tries to show average and actual values, compare how they are spread around the average.

156

FIGURE 2. AVERAGE LIQUIDITY RATIOS IN 2016

Source: Constructed by the researchers

Table 4 illustrates the main descriptive statistics of the independent variables during the period 2009-2016 for all Islamic and conventional banks in the MENA countries. While the table shows that most of the independent variables have moderate dispersion in general, it is notable that there is a greater variability between Islamic banks and their counterparts across the years for the Equity / Total Assets and Loan Loss Reserve / Gross Loans variables . TABLE 4. DESCRIPTIVE STATISTICS OF INDEPENDANT VARIABLES FOR 2009-2016 Variable Bank type Conventional

Log(Total Assets in USD)

Islamic Conventional

Equity / Total Assets


Return on Average Assets (ROAA)


Loan Loss Reserve / Gross Loans


Log(pc real GDP)

Islamic Source: Constructed by the researchers

Obs.

Mean

Std. Dev.

Min.

Max.

1641

11.30

6.94

0.00

18.81

1641

4.03

6.75

0.00

18.39

1641

9.99

9.51

0.00

70.85

1641

6.24

15.58

0.00

99.78

1641

8.32

10.05

-121.06

116.20

1641

1.53

10.12

-167.44

49.00

1589

4.50

6.32

0.00

46.58

1589

2.44

9.54

0.00

100.00

1272

7.06

4.25

0.00

11.51

1272

2.60

4.43

0.00

11.51

TESTS AND REGRESSION RESULTS Several pre-estimation tests were implemented before the empirical estimation. First, to decide between the fixed effects and random effects panel data model specification, we employed Hausman specification test. The fixed effects models assume that all individual

157

differences are captured by differences in the intercept parameter. In contrast, the random effects models again assume that all individual differences are captured by the intercept parameters, but the individuals in the sample are randomly selected and thus are treated as random rather than fixed. Random individual differences are included in the model by specifying the intercept parameters to consist of a fixed part that represents the population average and random individual differences from the population average. The test results are summarized in Table 5 and indicate that in all cases the null hypothesis is rejected, therefore fixed effects models should be employed. TABLE 5. HAUSMAN SPECIFICATION TEST (H_0: DIFFERENCE IN COEFFICIENTS NOT SYSTEMIC AND THE RANDOM EFFECTS MODEL) L2 L5 chi2 78.81 57.80 Prob. >chi2 0. 0000 0. 0000 Decision Fixed effects Fixed effects Source: Constructed by the researchers

L6 57.60 0. 0000 Fixed effects

L7 69.23 0. 0000 Fixed effects

L8 156.93 0.0000 Fixed effects

The Wooldridge test for autocorrelation in panel data and the modified Wald test for group-wise heteroscedasticity in the fixed effect regression model are summarized in Tables 6 and 7, showing the existence of autocorrelation and heteroscedasticity in all the models. TABLE 6. WOOLDRIGE TEST FOR AUTOCORRELATION IN PANEL DATA (𝐇𝟎 : NO FIRST-ORDER AUTOCORRELATION) L2 L5 chi2 189.654 57.80 Prob. >chi2 0. 0000 0. 0000 Autocorrelation Yes yes Source: Constructed by the researchers

L6 34.715 0. 0000 yes

L7 497.852 0. 0000 yes

L8 30.061 0.0000 yes

TABLE 7. MODIFIED WALD TEST FOR GROUP-WISE HETEROSCEDASTICITY IN FIXED EFFECT REGRESSION MODEL (𝐇𝟎 : σ_i^2 =σ^2 for all i) L2 L5 chi2 1.3e+05 4.9e+07 Prob. >chi2 0. 0000 0. 0000 Heteroscedasticity Yes yes Source: Constructed by the researchers

L6 3.1e+05 0. 0000 yes

L7 1.2e+07 0. 0000 yes

L8 3.1e+05 0.0000 Yes

To accommodate the autocorrelation and heteroscedasticity, the generalized least squares model was used as an alternative to the fixed effects model, which allows for unbiased estimation in the presence of autocorrelation within panels and cross-sectional correlation and heteroscedasticity across panels. The results of applying GLS on the model are presented in Table 8.

158

TABLE 8. EMPIRICAL ESTIMATION RESULTS BY GLS MODELS Net Loans / Tot. Assets

Net Loans / Dep. & ST Funding

Net Loans / Tot. Dep. & Bor.

Liquid Assets / Dep. & ST Funding

Liquid Assets / Tot. Dep. & Bor.

L2

L5

L6

L7

L8

Conventional banks Log(Total Assets in USD) Equity / Total Assets Return on Average Equity (ROAA) Loan Loss Reserve / Gross Loans Log(pc GDP)

2.540*** (8.91)

0.968* (2.34)

2.398*** (4.60)

-0.412 (-0.83)

-0.467 (-1.39)

0.339*** (6.13)

1.290*** (13.01)

2.220*** (19.44)

1.052*** (7.73)

0.208* (2.47)

-0.064** (-3.25)

-0.108*** (-3.39)

-0.120*** (-3.71)

-0.064 (-1.39)

-0.076 (-1.64)

-0.472***

-0.719***

-0.496***

-0.059

0.269*

(-7.65)

(-7.90)

(-5.37)

(-0.52)

(2.36)

7.796***

11.074***

7.327***

-4.936***

-3.572***

(19.83)

(20.28)

(10.13)

(-6.64)

(-7.42)

5.850***

4.070***

6.933***

-0.668

-1.905

(9.64)

(3.54)

(8.34)

(-0.34)

(-1.62)

-0.281*** (-5.89)

0.507*** (3.40)

1.144*** (8.38)

1.842*** (5.79)

1.122*** (4.80)

0.011 (0.67)

-0.079 (-2.42)

-0.053 (-1.45)

-0.094 (-1.27)

-0.028 (-0.65)

-0.217***

-0.770***

-0.563***

0.005

0.274

(-5.14)

(-7.06)

(-4.64)

(0.02)

(1.41)

3.366***

7.242***

1.645

-5.616

-2.659

(3.34)

(3.99)

(1.21)

(-1.70)

(-1.36)

-66.581***

-68.698***

-70.471***

68.892***

63.182***

(8.66)

(10.35)

real

Islamic banks Log(Total Assets in USD) Equity / Total Assets Return on Average Equity (ROAA) Loan Loss Reserve / Gross Loans Log(pc GDP) _cons

real

(-13.30) (-9.48) (-9.26) Source: Constructed by the researchers Notes: *p