The interaction between politics and business has prompted many researchers to examine the effect of political connections on firm value (Fisman 2001; Faccio 2006; Yano and Shiraishi 2020). Despite attempts to reduce political influence, policymakers cannot completely ignore the political consequences of their policies. These consequences are particularly important in the democratic context, with the presence of political competition and constitutional protections (Burrier 2018).

The political connections of firms are well documented in the literature. However, the specific impact of political patronage on banks is less studied. In fact, this phenomenon can affect non-financial firms as well as banks. But unlike other firms, banks play an essential role in the overall economy. In emerging markets, banks play a crucial role as a major source of financing, and even more so in the economy as a whole, by establishing the stability of the financial system as a whole (Omran et al. 2008). In addition, banks earn higher profits and are implicitly supported by the state in case of failure. Thus, banks invest in political connections because the benefits that these connections would bring them are higher than the cost they would bear. It is important to note, then, that a strong political connection can be seen as an important intangible asset for extracting more direct rents from government. These rents would not otherwise be available to the bank. On the other hand, the incursion of politics can be detrimental to firm performance according to Dong et al. (2014). Thus, politically backed banks could be viewed as vehicles for financing high-risk, low-return social projects. A handful of recent studies recognize the value of a political presence on bank boards and its impact on bank performance (Braun and Raddatz 2010; Micco et al. 2007), lending and business (Nys et al. 2015). But little attention has been paid to the relationship between political patronage and bank financing decisions (Braham et al. 2019).

Conflicting evidence may be related to different contexts and the extent of political influence. Moreover, the reported results are valid depending on the definition of political connection and the nature of the relationship. Indeed, political influence is defined, in its general meanings, as the ability to directly or indirectly convert political power into economic benefits (Shen 2018). Recent studies have used Faccio’s (2006)Footnote 1 definition to measure political connections, which may not capture the extent to which it might affect firms (Braham et al. 2019, 2020; Dang et al. 2018; Maaloul et al. 2016; Hung et al. 2017; Deng et al. 2018). Nevertheless, using a binary variable to identify political connections lacks important information and may not reflect the strength of political connections. That said, measuring political connections is an important challenge. Therefore, we argue that the effect of political connections on banks is more complicated than reported in previous studies in the literature. It is clear that a stronger political connection can have a larger impact on firms, whether positive or negative. In this paper, we propose an alternative measure that reflects the different political connections instead of using a simple binary variable. We then examine its effect on bank behavior.

Here, we attempt to create a new political connection measure to examine the extent to which political patronage affects banks in the MENA region. We then assess the explanatory power of this political connection measure. For this purpose, a data set of banks is used for the period 2003–2017. Principal component analysis is performed to create an appropriate measure of political connections. Panel regressions are performed to test the appropriateness of this political connections index by examining its impact on leverage and bank risk. The newly constructed political connections index improves the estimation results compared to those obtained with the simple binary variable. Moreover, our results for the MENA region are different from previous ones. This may be due to the inability of the binary variables to capture the true impact of political connections.

Overall, a particular feature of research in the literature is the choice of context. Studies on individual countries, such as Malaysia and China, are more popular. In fact, few recent papers compare the effect of political connectedness across countries because of the heterogeneity of countries and different political scenarios. Moreover, the MENA region may be an interesting platform to examine the effect of political connections for several reasons. First, many countries in this region are known for the significant influence of controlling families, regimes and powerful politicians on the business community. The political economy of the MENA region allows corporate owners and managers to dominate the business sector in exchange for support from certain regimes. Second, discussion of the institutional environment leads to the expectation that politically connected firms in MENA countries will behave differently from those that are not. Third, like emerging economies, financial markets in MENA are dominated by banks (Hamadi and Bassil 2015). Clearly, the impact of political connections would be more prevalent in banks. Its consequences would then be more pronounced at the bank level.

In addition, data on the types of political connections are collected manually, which has not been explored in this region before. We manually identify politically supported executives and shareholders by collecting information on their profiles from company and bank websites. Finally, the existing literature shows that few studies focus on measuring political connections. The design of these studies has primarily used binary variables. Therefore, we extend the work of Braham et al. (2019) by using a different measure of political connections instead of the general binary measures. Due to the lack of information provided by binary variables, we attempt to provide a more informative measure of political connections by considering other forms and aspects. To our knowledge, this is the first study to examine this issue in the MENA region.

The remainder of this article is structured as follows: “Literature review” section presents the literature review. In “Methodology” section describes the data and methodology. In “Results and discussion” section presents and discusses the results of the statistical analysis and panel estimates. Finally, “Conclusions” section concludes the article and “Implications and future research” section provides suggestions for future policy and research.

Literature Review

Political Connections Literature

Making political connections has many implications for firms. Yano and Shiraishi (2020) examine the economic and political motivations for bank lending in China. The authors find that political connections are not attractive to banks when they are looking for firms that are good customers. However, when banks attach great importance to firms' political connections and recognize the uses of political connections other than financial support, politically connected private firms become interesting to them.

On the one hand, political connections have several advantages: (1) easier access to financial resources such as bank loans on more favorable terms (Fraser et al. 2006; Khwaja and Mian 2005), (2) access to the equity market (Bao et al. 2016), (3) better performance (Johnson and Mitton 2003; Liu et al. 2018), (4) higher probability of bailout (Faccio 2006) and lower cost equity (Boubakri et al. 2012), (5) increased political legitimacy and access to government-controlled resources (Wu et al. 2018). On the other hand, some studies find that political favoritism negatively affects firm value by decreasing the quality of accounting information (Chaney et al. 2011), decreasing efficiency (Boubakri et al. 2008; Leuz and Oberholzer-Gee 2006) and decreasing long-term performance (Claessens et al. 2008; Fan et al. 2007). In addition, there is evidence that political connections can also affect firms' financing decisions. Fraser et al. (2006) focus on developing economies and find a positive and significant link between leverage and political connections. They also suggest that larger and more profitable firms with political patronage tend to take on more debt than firms with only political patronage. Extending the work of Fraser et al. (2006), Bliss and Gul (2012) find that politically connected Malaysian firms have negative equity.Footnote 2 They also find that politically connected firms that borrow more have significantly lower profitability than unconnected firms. Dong et al. (2010) also hypothesize that Chinese firms with strong political connections should take on more debt. In addition, Lim et al. (2012) examine the effect of political patronage on the capital structure of firms listed on the Shanghai Stock Exchange. They find evidence that Chinese firms with stronger political connections are more leveraged.

Political Connections in the MENA Region

Despite the fact that political patronage in the MENA region has been practiced in different ways, the literature on these practices is scarce. For example, in Tunisia, former President Ben Ali and his family were involved in the economy and monopolized most profitable sectors (Rijkers et al. 2017). Tunisian state-owned enterprises hold monopoly positions in many sectors such as industry, transport, banking, infrastructure, imports and exports. These firms are subject to access restrictions but enjoy many advantages such as tax benefits. According to Rijkers et al. (2017), Tunisian firms were willing to forge political ties with Ben Ali's family to invest in these sectors. Maaloul et al. (2016) examine the effect of political connections on the performance and value of Tunisian firms after the revolution. The interaction between business and politics in Tunisia was observed during the Ben Ali regime and intensified after the 2011 uprising, as many businessmen entered politics and became members of political parties and/or parliament. In fact, the Tunisian government controls most of the scarce resources and banks. Also, as in other emerging economies, the Tunisian institutional framework is characterized by the importance of formal and informal relationships, a weak legal system and a high level of corruption. Bencheikh and Boulila (2017) study the impact of the political connection on the performance of Tunisian firms in the period of 2012–2015. This period is characterized by the emergence of a democratic atmosphere and the fight against corruption in its different forms. Regardless of this atmosphere, political connections remain an important determinant of the performance of Tunisian firms.

In Egypt, Dang et al. (2018) argue that throughout the reign of President Hosni Mubarak, state-business relations were dominated by cronyism. Key industries were dominated by a few family business groups. The environment was characterized by the shadowy privatization of public enterprises and privileged access to public bank credit by politically connected businessmen. The Egyptian government's decisions were decisive in all these areas, such as tourist resorts and housing projects. In addition, investments in oil and gas, as well as in some manufacturing sectors, required government approval. Industrial enterprises needed ministerial approval to benefit from energy subsidies. Similarly, Diwan and Schiffbauer (2018) study the determinants of borrowing and find that politically connected firms were more attractive to banks, both because they earned higher profits and because they were implicitly backed by the state in case of default. In fact, businesses under Mubarak were owned by businessmen who were either ministers or members of the ruling party, as well as by businessmen closely connected to Mubarak, such as friends and family members. Therefore, some private banks may also have been directly controlled by politically connected families. In turn, these banks may have lent primarily to related businesses.

Overall, MENA countries share many of the characteristics of developing economies, such as high levels of corruption and a lack of legal protection and strict regulations. This makes political favoritism more pronounced. Braham et al. (2019, 2020) study the effect of political connection in a sample of MENA countries. They provide evidence of the impact of political connection on bank behavior. However, the net effect is still not clear. It is therefore necessary to identify these linkages and quantify their impact. In fact, the scarcity of resources forces a dependence on government control. Firms in these countries must try to create linkages to benefit from resources that would not otherwise be available. This makes them vulnerable to political patronage.

Political Connections Measurement

Recent studies on political patronage rely on a simple binary measure to identify politically connected firms. The existing literature on political patronage does not examine this aspect in detail. We note that the majority of previous studies have mostly followed Faccio (2006) in defining political connections (Braun and Raddatz 2010; Bliss and Gul 2012; Fan et al. 2007; Wu et al. 2012).

The measurement is the most important challenge, as results may depend on the types and nature of these political ties. For example, Otchere et al. (2020) study the impact of political connection on risk taking. The authors consider three groups of politically connected firms, their industry rivals as well as non-rival controls. They draw a large sample of 48 countries from a data set by Faccio (2006, 2010). Dang et al. (2018) examine the stock return of Egyptian firms with and without political connection. The authors use an event study method and a binary indicator to identify firms politically connected to President Mubarak’s regime, for which at least one of the largest shareholders is associated with the president’s family. Maaloul et al. (2016) examine the effect of political connections on the financial performance of companies listed on the Tunis stock exchange. The authors define a politically connected company as having at least one of its members belonging to a political party or having a key government position or informal ties with politicians. They assign a binary variable for politically connected firms. Hung et al. (2017) examine the effect of political connections on Chinese banks’ performance and risk using a dummy variable to proxy for a politically connected CEO who held government positions before joining the bank. Also, Braham et al. (2019) examine the effect of political connections on bank financing decisions in the MENA region. The authors use a general measure, which is a binary variable, to identify whether the bank is politically connected or not. Some other studies consider state ownership as the ultimate form of political connection. So they classify state-owned firms as politically connected. For example, Deng et al. (2018) investigate the relationship between political connections and the investment activities of Chinese firms. The authors use two binary variables to proxy for ascribed political connections—state-owned enterprises and acquired political connections—which proxy for privately owned enterprises. Also, Fu et al. (2017) use a survey method to study how firms’ political connections affect their access to financing and performance. The authors employ binary variables to identify formal and informal links of the board members to government. In a different context, Panda (2015) examines the political connections hypothesis for the beneficiary identification process. Using an Indian household survey data set, the political connections variable is defined as a dummy that equals one if the household head is connected to a local political executive (somebody close to him/her or a family member), and equals zero otherwise. Additionally, Liu et al. (2018) provide evidence that political connections through state ownership or politically connected managers enable firms to enhance their performance. The latter type of connection is a valuable asset for firms. Furthermore, Kusnadi et al. (2015) identify a vital channel through which political connections are beneficial for non-state-controlled firms in terms of mitigating the threat of political extraction in China. Firms in provinces with more developed institutions (non-state-controlled firms) hold more cash reserves than those in provinces with less developed institutions (state-controlled firms). Also, Kobeissi and Sun (2010) examine the relationship between ownership structure and bank performance in the MENA region. They find that MENA foreign-owned banks performed significantly better than other types of banks. Chen et al. (2017) study the relationship between political connections and firm value. The authors create a process to measure political connection. An index is compiled on the basis of the current and former positions and political rank of the chief executive officer, board chair, directors and other senior officers, with government and parliamentary bodies at the national, regional and local levels. However, due to differences in political and governmental systems, the employment of this construct would not be applicable in the MENA region. Similarly, Su et al. (2019) establish a political connection index by aggregating scores of firms’ chief executive officers, board chairs, directors or other senior officers. These scores are based on their government and deputy positions to examine the contribution of political connections to corporate innovation. They define political connection as having a top manager or board member who is a current or former official of the central government, local government or the military. This threshold approach ensures that the firm’s top managers or board members with political connections have relatively strong power and authority in the government. Obviously, the impact of political connections may depend on the nature and types of these links. According to Khaw et al. (2019), the impact of political connections depends on the incentives to establishing these ties by considering two types of politically connected (PCON) leaders. Having a PCON executive chairperson leads to profit from relationship-based capitalism and is negatively related to the cost of debt, while having a non-PCON executive chairperson leads to being charged with a higher cost of debt. For example, Chen et al. (2018) analyze the relation between firm value and political connection in China. They find that the relation depends on ownership. The value of state-owned enterprises decreases with high political connections. For non-state-owned enterprises, the relation is nonlinear due to the costs and benefits of political connections. That is, the firm value increases initially at a lower level of connections and then decreases at a higher level. Also, Wong and Hooy (2018) examine the effect of different types of political connections on firm performance in the Malaysian context. The authors find that the effect of political connections on performance is not the same for all types of connections. In a more recent study, Sharma et al. (2020) investigate the link between political connections and firms’ export performance. The authors use a political connections measure that takes into account the political relationships of all board members (chairperson, CEO, executive directors and independent directors). Political connections are categorized into three types: (1) the position held currently or positions in the past as an official in the administrative hierarchy of government, (2) being a representative in the People’s Congress and (3) a committee member in the Chinese People’s Political Consultative Conference. A measure that reflects the average political relationship per board director is used. But political relationships are treated equally at all levels (county, city, province and national). The authors find significant differences in these effects depending on the type of firm ownership.

Overall, the findings from the existing literature are inconclusive. This may be related to the failure of simple binary measures to capture the variations and strength of political patronage due to lack of information. Hence, an important challenge is to find the appropriate measure of political connection. In this research, we suggest the construction of a new measure that incorporates different forms of political links and political aspects to overcome the inefficiency of binary measures.


Sample and Data Description

The sample contains 67 banks operating in the following MENA countries: Bahrain, Egypt, Jordan, Kuwait, Lebanon, Morocco, Oman, Qatar, Saudi Arabia, Tunisia and United Arab Emirates. Annual financial data and key ratios are obtained from the Thomson Reuters database for the period of 2003–2017. Information on ownership structure and shareholding are obtained from banks’ websites and financial reports. Macroeconomic and institutional indicators are provided by The World Bank database.

Political connections data are hand collected. The names of shareholders and managers of each bank are drawn from banks’ official websites. Then, we refer back to their individual biographies and information about their profiles, their current and former position held, their relationships and political background to identify cases of political connections for each bank. The information is collected from various sources such as banks’ official websites and business websites: Bloomberg (, Zonebourse (, Marketscreener ( and Leaders (

Political Connection Index Construction

In this section, we provide details to identify political links and describe the construction of the index. Measuring political connections is the most important challenge. The specification of the index allows detecting four channels through which political links are built in the MENA region. Unlike Braham et al. (2019, 2020) using binary variable derived from whether or not political connections exist, our measure reflects the importance of each channel assigned to its weight in the index.

Political Connections Identification

Cases of political patronage are identified following Braham et al. (2020): First, the names of shareholders and individuals serving on the board of directors of each bank are drawn from banks’ official websites. Second, we refer back to their individual biographies and information about their profiles and relationships to manually identify politically connected shareholders and managers. Binary variable is then constructed, indicating 1 if the bank is politically backed and 0 otherwise. Our identification takes into account the political relationships of shareholders and board members (chairperson, CEO and directors). The tenure of board members in the bank is also considered in the identification process allowing the observations to vary in time. We note that for banks operating in MENA countries, only banks with information on their managers were retained. Regardless of whether information on the presence or absence of political links was retained, banks whose data on their managers and shareholders was not available on the websites were eliminated from the sample.

Then, we categorize political connections into four types: president or royal family members; individuals having government position in past or present; parliament members, political parties’ members or politicians; persons having ties with previously mentioned persons. Hence, we determine four variables of political connection regarding channels or type of connection: royal and president family members (fam), ex or current government official (gov), parliament or political party members (par), informal ties (inf). The variables are defined in Table 1.

Table 1 Political connections determinants

Political Connection Index

The idea is to construct a political connections measure considering formal and informal political links. In order to create a new index measuring political connection, a linear specification f is assumed to define the following measure of political connection (POLindex):

$${\text{POLindex}} = f\left( {fam,gov,par,\inf } \right).$$

The aim is to create a single variable that summarizes in an appropriate way the different aspects of political patronage. Hence, the principal component analysis (PCA) is firstly used. PCA is the classical multivariate statistical method to extract essential information for index construction. The goals of PCA are to extract the principal information from the data set as a set of new orthogonal variables called principal components. They are extracted according to their importance in terms of explained variance. The explained variance represents a measure of the summary power of the data. By only selecting the most significant components, it thus simplifies the variable description. The coefficients (αi) can be properly given from the eigenvectors of the most significant components defining the principal components (PC).

A preliminary step is to calculate the Kaiser–Meyer–Olkin (KMO) index (Kaiser 1974), a measure of sampling adequacy. KMO index evaluates the extent to which the selected variables can form a coherent set. It measures the proportion of variance among variables that might be common variance. The lower the proportion, the more suitable the data is to warrant PCA analysis. The formula for the overall KMO index is given by this formula:

$$KMO_{j} = \frac{{\mathop \sum \nolimits_{i \ne j} r_{ij}^{2} }}{{\mathop \sum \nolimits_{i \ne j} r_{ij}^{2} + \mathop \sum \nolimits_{i \ne j} u_{ij}^{2} }},$$

where the correlation matrix is R = [rij] and the partial correlation matrix is U = [uij].

In practice, KMO index returns values between 0 and 1 with higher values indicating that there is a statistically acceptable factorial solution.

Then, the principal components are extracted according to their importance in terms of explained variance of the original variables. The first PC explains a percentage of variance greater than the second one and so on. Generally, a percentage around 50–60% of variance explained of the first principal components is considered a good value of summary power of the data. Hence, the index can be estimated with the coefficients given by the eigenvector defining the retained principal components.

Other Variables Definition

In order to examine the effect of political connection, some variables related to bank characteristics, institutional and macroeconomic variables are considered. The definitions, references and expected signs are presented in Table 2.

Table 2 Variables names and definitions

Effect of Political Connection Index on Bank Leverage and Risk

In order to evaluate bank behavior, the effect of political connection on leverage can be specified in a panel econometric model as follows:

$$\begin{aligned} leverage_{it} & = \beta_{0} + \beta_{1} POLindex_{it} + \beta_{2} profit_{it} + \beta_{3} {\text{s}}ize_{it} \\ & \quad + \beta_{4} efficiency_{it} + \beta_{5} loandep_{it} + \beta_{6} gdp_{it} \\ & \quad + \beta_{7} inflation_{it} + \beta_{8} polstab_{it} + \beta_{9} goveff_{it} + \mu_{i} + \varepsilon_{it} \\ \end{aligned}$$

where i denotes the individual (i = 1, 2, …, 67), t denotes the year (t = 2003, …, 2017), β0 is a constant term, β1β9 are the coefficients to be estimated, μi is the unobserved time-invariant individual effect and εit is the error term that should be white noise independent to μi.

In panel data, two panel estimation methods are performed using fixed effects and random effects models. It appears necessary to firstly verify the homogeneous or heterogeneous specification of the model to determine if the parameters are perfectly identical or vary across individuals. Fisher test with the null hypothesis of homogeneity of individual effects is performed when running the fixed effect model estimation. Then, to distinguish between the fixed and random effects model, we run the Hausman (1978) test with the null hypothesis that the coefficients estimated by the efficient random effects estimator are the same as the ones estimated by the consistent fixed effects estimator.

Finally, the objective of our paper is to create a new measure of political connections. Hence, in order to evaluate the reliability of the new constructed index, we compare the estimation results from panel regression based on information criterion, by replacing the new variable of political connections with the simple binary variable as following:

$$\begin{aligned} leverage_{it} & = \beta_{0} + \beta_{1} POL_{it} + \beta_{2} profit_{it} + \beta_{3} size_{it} + \beta_{4} efficiency_{it} \\ & \quad + \beta_{5} loandep_{it} + \beta_{6} gdp_{it} + \beta_{7} inflation_{it} + \beta_{8} polstab_{it} \\ & \quad + \beta_{9} goveff_{it} + \mu_{i} + \varepsilon_{it.} \\ \end{aligned}$$

For robustness check, the same analysis is conducted for risk. Credit risk variable is used as the main source of risk exposure for banks (Tanda 2015). The regression is as follows:

$$\begin{aligned} risk_{it} & = \alpha_{0} + \alpha_{1} POLindex_{it} + \alpha_{2} profit_{it} + \alpha_{3} {\text{s}}ize_{it} + \alpha_{4} efficiency_{it} \\ & \quad + \alpha_{5} loandep_{it} + \alpha_{6} oplev_{it} + \alpha_{7} gdp_{it} + \alpha_{8} inflation_{it} \\ & \quad + \alpha_{9} polstab_{it} + \alpha_{10} goveff_{it} + \mu_{i} + \varepsilon_{it,} \\ \end{aligned}$$

where i denotes the individual (i = 1, 2, …, 67), t denotes the year (t = 2003, …, 2017), α0 is a constant term, α1α9 are the coefficients to be estimated, μi is the unobserved time-invariant individual effect and εit is the error term that should be a white noise independent to μi.

Then, a comparison of the estimation results is done by replacing POLindex with the binary variable POL as following:

$$\begin{aligned} risk_{it} & = \alpha_{0} + \alpha_{1} POL_{it} + \alpha_{2} profit_{it} + \alpha_{3} {\text{s}}ize_{it} + \alpha_{4} efficiency_{it} \\ & \quad + \alpha_{5} loandep_{it} + \alpha_{6} oplev_{it} + \alpha_{7} gdp_{it} + \alpha_{8} inflation_{it} \\ & \quad + \alpha_{9} polstab_{it} + \alpha_{10} goveff_{it} + \mu_{i} + \varepsilon_{it} \\ \end{aligned}$$

Results and Discussion

Principal Components Analysis

As a preliminary analysis, some descriptive statistics of the variables used in PCA are presented in Table 3.

Table 3 Descriptive statistics of PCA variables

Tables 4 and 5 report some statistics of banks political connections each year in the sample, the number of politically connected banks and the means values regarding the four types of connections by year and by country.

Table 4 Summary statistics of banks’ political connections by year
Table 5 Distribution of banks’ political connections by country

Connections by family and government officials are the most common forms of political patronage for all countries. Particularly, Tunisian banks establish political links with parliament members.

Following Table 5, we examine the extent to which POLindex in these countries may have changed between the period 2003–2010 and the subsequent period, i.e., before and after the Arab Spring: see Table 6. Although the changes for Egypt and Tunisia between these two periods would be of particular interest, no significant changes are observed in the data. This suggests that either the Tunisian banks were not significantly linked to the Ben Ali family, or that this relationship was replaced by others. But this does not contradict a possible effect of political patronage. There are significant changes only for Lebanon and Qatar.

Table 6 Distribution of banks’ political connections by country and periods

Then the results from the PCA are provided. The sample adequacy measure KMO is firstly calculated which has the value of 0.42 and Cronbach’s alpha is 0.26. The reliability coefficient is not large; this may be related to the nature of the variables used in the analysis. In Table 7, the principal components are calculated.

Table 7 Principal component eigenvalues and contribution rate

In Table 7, eigenvalues of the principal components PC are presented. The highest eigenvalue retains more explained variance among others. The contribution rate of the first principal component is 31.64% with an eigenvalue more than 1. However, the cumulative contribution of the three first components is about 83.23%. As the PCA helps determine which linear combinations of the variables matter most, these combinations are made in such a way that most of the information contained in the initial variables is stored in the first components. The principal components retained in the analysis are reported in Table 8.

Table 8 Principal component’s eigenvectors

Table 8 presents the eigenvectors of the three retained components. The principal components are sometimes seen as variables hidden behind the initial variables, which are the only observable ones, and which the method therefore makes it possible to bring to light. Based on the results of Table 8, it is possible to summarize the information distributed over the variables. The eigenvalues can also be interpreted as correlation coefficients between the variables and the components. Thus, an examination of Table 8 of the eigenvectors allows us to see easily that the first component is defined by the dimensions fam (0.6235) and gov (−0.7364), and that gov plays a negative role on the first component. Hence, the first component could explain the political links based on government and ruling family, which are more relevant and effective. The second component, on the other hand, is defined by the links of par (0.8005), which characterize parliamentary and political party relationship. The third component is defined by the dimension inf (0.9338). Hence, it is associated with the informal ties which are less relevant. These values correspond to the regression coefficients when we try to explain the variables using the components.

The variables weighted by their coefficients given by eigenvectors for components 1, 2 and 3 are averaged to define a linear combination. The index is then calculated as linear combination of the retained components weighted by their explained variance reported in Table 7 The index is given by the following specification:

$${\text{Polindex}} = 0.3164*{\text{comp}}1 + 0.2659*{\text{comp}}2 + 0.2541*{\text{comp}}3.$$

In this way, the constructed index reflects the contribution of each type of links. Finally, the retained components and then it is used as an independent variable in the subsequent analysis.

Effect of Political Connection Index on Bank Leverage

Before conducting our analysis, preliminary statistics of the variables for the whole sample of the analysis are presented in Table 13 of Appendix. We also split our sample regarding state ownership to compare the characteristics of state-owned banks with non-state-owned ones. Table 14 provides summary statistics for the two sub-samples. As shown in Tables 13 and 14, state-owned banks, regardless their political ties, have higher values of leverage, lower efficiency and higher loan loss provisions. For country-level variables, inflation is higher for non-state-owned banks in average, and government effectiveness is higher for non-state-owned banks. Regarding political connection, state-owned banks without political ties have higher values of leverage, loans to deposits, operating leverage. However, for non-state-owned banks establishing political links makes these values higher in average compared to banks without these connections.

Along with the descriptive statistics, we also check for collinearity and correlations. The variance inflation factorFootnote 3 (VIF) values vary between 1.08 and 2.78 with the average equals to 1.63, which implies the absence of multi-collinearity problem.

The sample consists of 67 banks resulting in 1005 observations from around 1455 observations of the total sample of banks randomly selected in the Thomson Reuters database over the period 2003–2017. In this way, we eliminate banks with no available information on political connections in the boards for the empirical analysis. However, a sample selectivity problem may occur when the observations are present according to some selection rule. As a sample selection bias test, we propose to use a t test based on a comparison of mean for the two samples. The t test is used to determine if the means of two sets of data are significantly different from each other. The test statistic follows Student distribution under the null hypothesis H0 that there is no significant difference in the mean.

Hence, the descriptive analysis is carried out through a comparison between two sub-samples made up of banks with available data (sample 1) on political connections and banks with missing data on political connections (sample 2). The results are presented in Table 15 of Appendix A comparison in means of the variables show that there are indeed differences identified by student test but not very significant for leverage as well as for other variables. Hence, there is not important difference between the samples and this should not skew or contradict our results.

In order to test the effect of political connections on bank leverage, panel data regressions are estimated, where leverage ratio is the dependent variable and Polindex is the independent variable. Moreover, to test for possible nonlinear effects of political connection, we extend the model by index interacting with other bank characteristics and country-level variables (political stability and government effectiveness) that may be related to political patronage. The estimation results of model (2) are presented in Table 9.

Table 9 Polindex interaction effects on leverage: iterative estimations

We note that we proceed to an iterative procedure. In this way, we extend the work of Braham et al. (2019) to find out which variables contribute the most to leverage and risk, especially the interaction terms. Also, the work of Braham et al. (2019, 2020) is extended by controlling for institutional as well as country-level variables. Hence, a maximum number of interactions are explored.

Fixed and random effect estimations are conducted along with Hausman test to retain the adequate model. For each iteration, the less significant variable is eliminated. In this way, we end with the only significant variables. In Table 10, only the initial and last iterations are reported in the results.

Table 10 Pol interaction effects on leverage: iterative estimations

According to the calculated Fisher statistic, individual effects have to be included in the model. The Hausman test rejects the null hypothesis. Thus, the fixed effect model has to be retained in the analysis.

Regarding the last iteration, we find that the political connection index is negatively significant. This result contradicts the hypothesis that politically connected firms tend to have higher leverage. The coefficients of profitability, loans to deposits and government effectiveness are significant and negative. However, the coefficients relative to the interaction terms are significant but positive. Obviously, political connections have an indirect impact on leverage through these variables.

In contrast to Braham et al. (2019) that politically backed banks are more leveraged, political connection is negatively associated with leverage. Notably, the interaction term of political connections with government effectiveness in Table 9 is not significant. However, the last iteration technique has shown the important significance of government effectiveness in the relationship with leverage, which could not be detected at the first glance. A comparison between the findings implies that the effect of political patronage on leverage may also depend on political and institutional environment (proxied by government effectiveness).

Besides, we examine the interaction effects of political connection for the two sub-samples of banks. Results are presented in Table 16 of Appendix. For non-state-owned banks, the findings are similar to previous ones. Comparing state-owned banks with their counterparts, we find that political stability exhibit a direct and indirect impact on leverage through political connection. Overall, country governance variables play an important role for state-owned and non-state-owned banks.

Political Connection Index Reliability

To evaluate the new constructed index, it is replaced by a simple binary variable (POL) in the regression model (3). We also make the binary variable interact with other bank characteristics and country-level variables. The estimations are conducted through an iterative procedure, the first and last estimations are presented in Table 10.

Regarding political connection variable, there is positive and significant effect on leverage. Also, nonlinear effect is found through size and efficiency.

In terms of explained variance, the adjusted R2 is 0.039 for the binary variable model estimation and 0.063 for the model including the index of political connections. The results confirm that the new measure outperforms the use of the simple classical binary variable as proxy for political connections.

We examine the interaction effects of political connection for the two samples of state-owned and non-state-owned banks. Results are presented in Table 17 of Appendix. The binary variable of political connection is not significant for non-state-owned banks and exerts only indirect effect on leverage through the other variables; however it is significant for their state-owned counterparts. Besides, the gain in terms of explained variance is not significantly improved in the two models. When comparing the findings regarding the two models, it is judicious to note that the contrasting effect of political connection may be related to the measurement of political connections. Moreover, after controlling for some institutional factors, it is notable that the net effect of political connection depends significantly on governance factors. Table 11 provides the estimation results of model (3) for the whole sample and the two sub-samples. Comparing the estimation results using two variables of political connections, the classical political connection measure (pol) is also not significant for state-owned and state-owned banks. The results for control variables are found to be the same: profitability, the ratio of loans to deposits and government effectiveness has negative and significant effect on bank leverage.

Table 11 Polindex interaction effects on risk: iterative estimations

Robustness Check

For more robustness check, the same analysis is conducted for the bank risk. It would be interesting to compare the analysis for banks which are largely state-owned with those that are not. Estimation results of the two sub-samples are provided in Appendix (Tables 18 and 19) for models 4 and 5, respectively. Table 11 presents the estimation results of model (4) including interaction terms with political connection index.

The new political connection index POLindex is found to be significant and positively related to risk taking. Profitability, efficiency and government effectiveness are negatively significant, whereas political stability is positively related to risk. These findings are valid for non-state-owned banks as reported in Appendix (Table 18). As political factors could affect bank risk taking (Razgallah et al. 2019), we find that (polstab) is positively related to risk. That is, when political instability increases, bank risk increases. The result is in line with Ng et al. (2020), there is positive association between banks’ loan loss provision and the level of political instability. Profitability is negatively related to risk, indicating that more profitable banks are less likely to fail (Psillaki et al. 2010). However, the negative association between efficiency and risk contrasts with previous literature on the MENA region (Srairi 2013). Hence, the negative association may be related to an implicit effect of political connection (Braham et al. 2020). For state-owned banks, political connection index is negatively related to risk. Thus, the role of political patronage in enhancing bank risk taking is more prevalent for non-state-owned banks.

Next, we estimate model (5) including interaction terms with the binary variable to compare the estimations with previous results. Results are presented in Table 12.

Table 12 Pol interaction effects on risk: iterative estimations

Regarding the classical political connections variable (pol), the calculated R2 is not significantly improved. The coefficient related to this variable is significantly negative which is in contrast to previous results. The signs of coefficients regarding this variable changes as well as the interaction term (pol*polstab). Hence, the unique effect of (pol) depends also on the interaction term. A possible explanation of this finding is that increasing political instability may also decrease banks’ certainty of deriving benefits from their political connections which results in lower risk to response to market pressure. Also, political connection measure, as binary variable, may lack important information captured by the index. Although results regarding bank characteristics are the same, the contrasting effect of political connection on risk taking may be related to political connection measurement as the index reflects the strength of political links.


In this paper, the extent to which political patronage affects banks in the MENA is examined by creating a new measure of political connection. Our work contributes to the literature as follows: On one hand, we provide more informative measure based on different types of political connections. The idea is to construct a measure of political connections by considering formal and informal political ties. Then, we classify political connections into four categories: members of the royal family or the president; individuals who have held or are holding a government position; members of parliament, members of political parties or politicians; and people with connections to the above-mentioned people. Therefore, we determine four political linkage variables regarding the channels or type of linkage: members of the royal family and the president, former or current government officials, members of parliament or political parties, informal links. In this way, the new index enhances the estimation results compared to those obtained with binary variable. The construction of this index reflects the weight of the dimensions considered and allows capturing the strength of the linkages.

On the other hand, our finding that political patronage has negative relationship with bank leverage is in contrast with the hypothesis that being politically backed enable firms to have easier access to debt financing (Faccio 2006) and enjoy preferential treatment by governments to acquire more capital (Khwaja and Mian 2005) than their non-connected peers. The result is also in contrast with studies regarding MENA region (Braham et al. 2019). A possible explanation is that the impact of political connection may be related to institutional and political environment. When there are interaction terms the impact of political depends on the other variables (Braham et al. 2020).

Second, after controlling for some institutional factors, we find that government effectiveness decreases bank leverage. When government effectiveness increases, the benefits driven from the government are eroded, forcing banks to be more responsive to market pressures and resulting in a lower level of leverage. Also, the unique effect of Pol depends on political stability. In fact, political stability may be useful for banks’ to take benefits from their political connections which results in higher risk.

Third, we not only find evidence that political connection is an important determinant of bank leverage (Braham et al. 2019), we also show that the impact of political connection may depend significantly on types of political connection. As most existing studies treat all political connections equally, using binary variable lacks important information that may explain the differences in the results. Finally, a comparison between two political connections variables provides evidence that our measure may outperform the use of binary variables. The concern with such indicators is the lack of important information and their failure to capture variations and strength of political links. Our index, in contrast, is able to take into account different aspects of political connections.

Implications and Future Research

Our findings provide insights on the mechanisms of political rent seeking in the MENA region. Political patronage does not increase leverage for banks in MENA region consistent with rent seeking hypothesis. It is double edged sword and has perilous implications that should be carefully considered by investors and regulators.

We highlight the importance of political connections for emerging countries as a determinant of banks' funding choices. Because of the importance of the banking sector in emerging economies, it is strongly influenced by certain political aspects, and politically connected banks with the privilege of being bailed out in case of distress may engage in riskier and inefficient activities. The monitoring and evaluation of these banks by economic actors must be carefully considered.

Besides, it is important to consider institutional aspects that characterized the region along with corporate governance to provide more evidence on the mechanism of political patronage within banks. It is also important to note that results may differ from developed to emerging economies, as well as from individual or multiple countries as political systems and regimes are not the same for all countries. Hence, it should a greater extension would be to conduct country-level analysis and identify important and revolutionary changes in a given country.

Finally, measuring political connections is still an important challenge as the net effect on companies’ value and financing decisions may depend on the strength of these linkages. Future studies should address this issue in more detailed ways as well as refining the measures and methods used for the investigation.