DETERMINANTS OF BANK PERFORMANCE IN A DEVELOPING COUNTRY: EVIDENCE FROM MOROCCO

!is paper aims to de"ne long-term determinants of Moroccan commercial banks performance, for the period 2005-2015, using the Johansen cointegration test. For this purpose, we use bank performance ratios (ROA, ROE and NIM) as dependent variables, and deposits, liquidity ratios, bank-speci"c and macroeconomic variables as explicative variables. Results obtained show that long-term performance of Moroccan commercial banks depends on deposits, short-term, long-term and funding liquidity, the size of the bank and its square, internal and external funding, deposits interest rates and foreign direct investments. !ese results show the signi"cance of bank-speci"c variables as long-term determinants of the performance of Moroccan commercial banks.


Introduction
Bank performance plays an important role in the global nancial system. Indeed, bank performance re ects the banks' nancial health and is a key element in the prosperity of the banks. Regarding performance, it incorporates, according to Dayan et al. (1999), the concepts of e ciency, productivity and growth and is de ned as the ability to create wealth (Ensley, 2014). However, during the last decade, bank performance was strongly impacted mainly as a result of the nancial crisis that a ected banks and nancial systems. Indeed, this nancial crisis was not only limited to bank bankruptcies but caused quasi-bankruptcies, nationalizations and a decline in nancial performance of large nancial institutions.
In Africa, we remark that during the last decade, Moroccan banks have become an important actor in African nancial scene that plays an important role in Africa. According to the African Bank of Development, Moroccan banks are one of the banks' best performing portfolios in Africa and the rst African banks investing in Africa (Ghozali, 2016). Likewise, the eight biggest Moroccan banks (used in the present study) gure in the top banks in Africa (according to relbanks ranking) and are all in the top 15 of best performing banks in francophone Africa. Moroccan banks have also the largest networks in Africa, ahead of Nigerian, South African, Kenyan and Gabonese banks (Wilson, 2015).
is paper aims to de ne long-term determinants of Moroccan banks performance, for the period 2005-2015, using the Johansen cointegration test. For this purpose, we use bank performance ratios (ROA, ROE and NIM) as dependent variables and deposits, liquidity ratios, bank-speci c and macroeconomic variables as explicative variables. e de nition of bank performance determinants allows both researchers and bank managers to identify the most important elements for the survival and prosperity of banks. Banks used in the present paper are all coted in the Moroccan Stock Exchange, also known as Casablanca Stock Exchange (CSE), which is considered as one of the most dynamic stock markets in Middle East and North Africa region (MENA) and part of the MSCI Emerging Markets indices (Ferrouhi & Ezzahid, 2013).
For this purpose, we use the Johansen cointegration test. We use the three main performance ratios (return on assets, return on equity and net interest margins) as independent variables and liquidity ratios (short term liquidity, long-term liquidity and funding liquidity), bank deposits, bank speci c (bank size and its square, internal funding, external funding and deposit interest rate) and macroeconomic variables (foreign direct investment, unemployment rate, growth rate of gross domestic product) as explicative variables. e paper is organized as follows. Section 2 is reserved to the literature review. Variables and data used are presented in Section 3, while Section 4 is reserved to the presentation of research methodology. Results obtained are presented in Section 5. Finally, Section 6 o ers conclusions.

Literature review
Bank performance has always a racted the interest of researchers and bankers and plays an important role in the global nancial system. is role was greatly displayed by the fall of large banks in 2007-2008, which resulted in a global nancial crisis. eoretical analysis of bank performance determinants identi es two main theories: the Market Power eory that relates bank performance to external factors and the E ciency Structure eory that explains banking performance using internal factors. Each of these theories can be split into two models. us, according to the Market Power eory, bank performance is determined by the behavior of agents on the market and by its structure (Structure-Conduct-Performance model) or by the market shares (Relative Market Power model). As for the E ciency Structure eory, the X-E ciency model postulates that the best performing banks are those with lower costs while the Scale E ciency Hypothesis states that banks achieving high scale economies are the best performers.
Empirically, bank performance is usually expressed in terms of internal and external variables. Internal or speci c variables characterize each bank independently of its environment while external or macroeconomic variables are related to the macroeconomic environment of banks. Various authors studied the determinants of bank performance. us, Hester & Zoellner (1966) were the rst to de ne bank performance determinants from the balance sheet. e authors found that assets elements were positively correlated with bank performance and conversely to liabilities. Other authors studied the relationship between bank performance and internal determinants (Haslem, 1968;Short, 1979;Smirlock, 1985;Akhavein et al., 1997;Bikker &Hu, 2002, andGoddard et al., 2004) and found that there was a positive correlation between a bank's performance and its size.
Regarding external determinants of bank performance, Revell (1979) was the rst to investigate the relationship between bank performance and in ation. Perry (1992) argues that the in uence of in ation on bank performance depends on in ation expectations. Athanasoglou et al. (2006) analyzed a panel data of credit institutions in seven Southeast European countries (Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Macedonia, Romania, and Serbia and Montenegro) during the period 1998-2002. Results show that both bank speci c and macroeconomic variables signi cantly a ect bank performance. us, the concentration in the market and in ation are positively correlated with performance while GDP uctuations have no impact on performance. Kakilli and Çalim (2013) analyzed bank speci c and macroeconomic determinants that a ect Turkish commercial banks' performance during the period 2008-2011. Results show that bank speci c variables have more e ect on bank performance than macroeconomic variables. us, liquidity, gross domestic product, capital adequacy and asset quality have an e ect -positive or negative -on bank performance. Deposits and real exchange rates a ect positively or have no e ect on bank performance, unlike fees and commissions revenue e ect, which is negative or insigni cant. Molyneux and ronton (1992) try to de ne performance determinants of 671 European banks (from 18 countries: Austria, Belgium, Denmark, Finland, France, Greece, Germany, Ireland, Italy, Liechtenstein, Luxembourg, Netherlands, Norway, Portugal, Spain, Sweden, Turkey and United Kingdom) between 1986 and 1989. Results show that liquidity is weakly and negatively correlated to bank performance measured by ROA. Kosmidou (2005) analyzed the performance of the UK banks during the period 1995-2002. e author found that liquidity is positively correlated to return on assets average ROAA and negatively correlated to net interest margins NIM. Pasiouras and Kosmidou (2007) examined European Union bank performance from 1995 to 2001. Results show that liquidity is positively correlated to national banks' performance and negatively correlated to foreign banks' performance. Chen (2009) analyzed bank performance determinants in 12 countries (Australia, Canada, France, Germany, Italy, Japan, Luxembourg, Netherlands, Switzerland, Taiwan, United Kingdom and United States of America) for the period 1994-2006 and concluded that liquidity risk is related to bank performance.

Variables and data
As the aim of this paper is to de ne long-term determinants of Moroccan bank performance for the period 2005-2015 using the Johansen cointegration test, we rst de ne variables and data used in this paper, then we present the model used.
us, in the present paper, bank performance ratios are used as dependent variables, and liquidity ratios, deposits, speci c variables and macroeconomic variables as explicative variables. As theEuropean Central Bank ECB de ned three traditional performance measures, i.e. Return on Assets, Return on Equity and Net Interest Margin (European Central Bank, 2010), we use these three ratios as measures of bank performance. ese ratios are de ned as follows: measures a bank's pro tability relative to its assets and thus the bank's overall performance; measures a corporation's pro tability by revealing how much pro t a company generates with the money shareholders have invested; measures the gap between what the bank pays savers and what the bank receives from borrowers.
Other explicative ratios used are banks' speci c variables (the size of banks calculated using logarithm of the total assets of the bank SIZE; logarithm of the total assets squared SIZE that captures the non-linear relationship; internal funding IF; external funding EF and deposit interest rates DIR), and macroeconomic variables (foreign direct investment FDI, unemployment rate UNE and growth rate of gross domestic product GGDP).
Data used in the present paper were obtained from Moroccan banks' annual reports and from databases of the World Bank, the International Monetary Fund and the Moroccan High Commission for Planning for the period 2005-2015.
As our study concerns Moroccan banks, listed existing banks are presented in Table 1.

Research methodology
e existence of a cointegration relation between two or more variables means that these variables are long term correlated, even if they are not stationary (variables should be stationary at the same level). us, cointegration demonstrates the stable long-term equilibrium relations between integrated variables in the same order. ere are two main methods for examining the cointegration relationships, namely Engle and Granger test and the Johansen cointegration test. Whereas the rst method allows detecting only one cointegration relation, the Johansen cointegration test allows detecting several types of cointegration. us, we apply the Johansen cointegration test.
To de ne long-term determinants of Moroccan banks' performance, our methodology consists on the application of the Johansen cointegration test that requires, as a sine qua none condition, the integration of the studied variables in the same order.
us, the rst step of our methodology consists o the application of unit root tests (Augmented Dickey Fuller, 1981 andPhillips Perron, 1988) to determine the order of integration. We then apply the Johansen cointegration test ( Johansen, 1988), which is a multivariate cointegration test that allows the detection of multiple cointegrating vectors and tests the following equation: where Z t is the column vector of p-variables, Γ and Π are coe cients matrices to test, μ 0 and μ 1 are column vectors of constant terms and trend coe cients, Δ is the di erence operator, and υ t the Gaussian error of dimension p. e coe cient matrix Π, also called impact matrix, is equal to the number of cointegrated independent vectors and indicates the rank of the matrix and contains information relating to long-term relationships (Awokuse, 2003).

Results
Unit root tests results for the period 2005-2015 show that performance ratios (return on assets ROA, return on equity ROE and net interest margins NIM), liquidity ratios (short-term liquidity SL, long-term LL and funding liquidity FL), banks' deposits (DE-POSITS), bank's speci c ratios (banks' size SIZE, square of banks' size SIZE , internal funding IF, external funding EF and deposit interest rate DIR) and macroeconomic ratios (foreign direct investment FDI, unemployment rate UNE, growth rate of gross domestic product GGDP) are integrated in the rst order. Indeed, results obtained using ADF and PP unit root tests show that all the variables studied are integrated in the rst order. ese results indicate that the Johansen cointegration test can be applied to these variables.
However, since the Johansen cointegration test is sensitive to lag structure of VAR model, the optimal length of the o set must be determined before applying the test.
To determine the lag order of the VAR, Akaike Information Criterion AIC and the Schwarz Criterion SC are used. Based on Schwarz criterion, the optimal lag length proposed is 1 (Kasri & Kassim, 2009).
Tables 3 and 4 present results of the Johansen cointegration test for Model 1: We remark that the null hypothesis (H 0 : r = 0) for this model is rejected at 5% signi cance level while assumptions r = <1; r = <2; r = <3 are not rejected. us, Table 1 shows the existence of a cointegration vector in "Trace Statistic", as the 897.3518 value exceeds the critical value 752.6548 at 5%. Similarly, we note the existence of a cointegration vector in "Max-Eigen Statistic", as the 294.4035 value exceeds the critical value at 5% values 130.5684. ese results thus indicate the existence of a single balance and one cointegrating equation. We remark that the null hypothesis (H 0 : r = 0) for this model is rejected at signicance level 5%, while alternative hypotheses r = <1; r = <2; r = <3 are not rejected on the same threshold. us, the table shows the existence of a cointegration vector in "Trace Statistic", as the 925.7361 value exceeds the critical value 507.8793 at 5%. Similarly, we note the existence of a cointegration vector in "Max-Eigen Statistic", as the 320.2091 value exceeds the critical value 150.5152 at 5%. ese results thus indicate the existence of a single cointegration equation. e existence of cointegration between these variables implies the existence of a longterm equilibrium which governs relations between these variables (Tables 5, 8 and 11). us, as the purpose of our study is to de ne long-term determinants of Moroccan banks performance and relationship between speci c variables, macroeconomic variables and Moroccan banks performance between 2005 and 2015, results obtained show that: mance measured by ROA. Indeed, deposits are generally reinvested by the bank, which allows them to achieve a higher pro t. A similar result is obtained for profitability as measured by ROE. However, the relationship between deposits and bank performance measured using NIM is signi cant and negative; measured for the three models. us, when short term liquidity increases, bank performance increases; mance for Models 1 and 3. However, the impact of long-term liquidity is not signi cant for Model 2; for Model 2 and negative one for Model 3. However, results for Model 1 show that the impact of funding liquidity on bank performance is insigni cant; Bank size (SIZE) measured by the logarithm of the total banks assets has a signi cant and positive impact for the three models. us, the larger the bank is, the higher its performance is; tive impact for Models 1 and 3. However, this relationship is negative for Model 2; e impact of internal funding is positive for Model 3; us, when banks rely on external funding sources, they are less e cient; payment of their deposits has a signi cant and negative impact on commercial banks' performance in Morocco. is result seems reasonable since it means that when the bank pays more interest on deposits, its performance decreases; Regarding the macroeconomic variables, one variable only has a signi cant impact on bank performance, i.e. foreign direct investment FDI. Indeed, the relationship between bank performance measured by performance ratios ROA and ROE and foreign direct investment is signi cant and positive for Models 1 and 2. e impact of foreign direct investment for Model 3 is insigni cant; duct growth is insigni cant for the three models. Results obtained show that deposits are generally reinvested by the bank, which allows them to achieve a higher pro t, short and long term liquidity increases banks' performance and foreign direct investment in Morocco increases Moroccan banks' performance. Bank size is considered as a long-term determinant of Moroccan commercial bank performance in so far as the larger the bank is, the higher its performance is. Our ndings also show that banks that rely on external funding sources are less e cient and that when bank pays more interest on deposits, its performance decreases. However, results relating to deposits, the funding liquidity, and internal funding depend on performance ratio used.
us, long-term performance of Moroccan commercial banks depends on deposits, short-term, long-term and funding liquidity, the size of the bank and its square, internal and external funding, deposits interest rates and foreign direct investments. ese results show the signi cance of bank-speci c variables as long-term determinants of Moroccan commercial banks' performance.

Conclusion
is paper aims to de ne long-term determinants of Moroccan banks' performance for the period 2005-2015 using the Johansen cointegration test. For this purpose, we use three bank performance ratios as dependent variables, and deposits, liquidity ratios, bank-speci c and macroeconomic variables as explicative variables. Results for Models 1 and 2 show the existence of a cointegration relationship between bank performance, deposits, liquidity ratios, speci c bank ratios and foreign direct investment, while results for Model 3 show the existence of a long-term relationship between bank performance, deposits, liquidity ratios and bank speci c variables.
Our ndings show that deposits are generally reinvested by the bank, which allows them to achieve a higher pro t; short and long term liquidity increases bank performance, and foreign direct investment in Morocco increases Moroccan banks' performance. Bank size is considered as a long-term determinant of Moroccan commercial banks' performance in so far as the larger the bank is, the higher its performance is. Our ndings also show that banks that rely on external funding sources are less e cient and that when bank pays more interest on deposits, its performance decreases. However, results relating to deposits, the funding liquidity, and internal funding depend on performance ratio used.
us, we conclude that long-term performance of Moroccan commercial banks depends on deposits, short-term, long-term and funding liquidity, the size of the bank and its square, internal and external funding, deposits interest rates and foreign direct investments. ese results show the signi cance of speci c variables as long-term determinants of Moroccan commercial banks performance.