2.1 Legal tradition, colonial heritage and institutional development

According to law and finance theory, legal systems have their origins in either the common law or the civil law legal tradition. Most nations that exist today have either adopted independently, or acquired through conquest or colonization, one or the other of these two legal traditions (La Porta et al., Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1997, Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1998; Parent, Reference Parent2018). These legal traditions endure and persist over time, producing ancillary institutions that influence economic outcomes (Ang, Reference Ang2019; Ang and Fredriksson, Reference Ang and Fredriksson2017; La Porta et al., Reference La Porta, Lopez-de-Silanes and Shleifer2008).Footnote ¹ Common law originates from the British legal tradition that provides higher discretion to the courts to develop laws from already decided cases. Through this use of case law or jurisprudence, judges in common law countries establish legal precedents that are the foundations for the development of the legal system. In contrast, the civil law tradition has its historical roots in the codified laws of the Roman Empire. This canon of Roman laws is the basis for much of the legal development of the civil law countries of Europe, and places emphasis on the use of legislated codes or statutes whereby courts or judges are law enforcers but not law developers. The role of the courts is limited to interpretation and application of statutes and not to the continuing development of laws.

The law and finance literature highlights the channels or mechanisms through which legal traditions affect legal and institutional development. Beck and Levine (Reference Beck, Levine, Ménard and Shirley2005) suggest that legal origins influence the development of financial systems through political and adaptability channels that are shaped by the specific legal and institutional environment. According to the authors, the political channel reflects the extent to which a country's judiciary/courts make decisions without interference from political authorities. The political independence gives the courts in common law countries the power and freedom to enforce laws that protect small investors or property. Framing new rules by using previous legal cases also causes a higher degree of adaptability in the legal framework. In contrast, in civil law countries judges are restricted from modifying laws through procedural formalism.

Based on an international database capturing judicial independence and law development in 71 countries, Porta et al. (Reference Porta, Lopez-de-Silane, Pop-Eleches and Shleifer2002) propose that a common law tradition is associated with stronger judicial independence vìs-à-vìs a civil law tradition, and the degree of independence predicts higher economic and political freedom. Beck and Levine (Reference Beck, Levine, Ménard and Shirley2005) use Porta et al. (Reference Porta, Lopez-de-Silane, Pop-Eleches and Shleifer2002)'s database to test whether the described political and adaptability channels through which legal traditions influence the development of financial systems facilitate firms' access to finance.

Alternative views explaining institutional development and financial outcomes in colonies come from the endowment school (Acemoglu et al., Reference Acemoglu, Johnson and Robinson2001). This strand of literature suggests that initial endowments (geography, topography and disease environment) in colonies explain the type of European colonization and subsequent institutional development including legal institutions and financial development (Beck et al., Reference Beck, Levine and Demirgüç-Kunt2003). Specifically, those colonies with hospitable conditions (that had lower mortality rates, higher indigenous population density and sufficient resources) influenced the type of colonization. The type of settlement further influenced the development of institutions (property rights and contract enforcement), which later led to variations in economic outcomes across colonies, since European settlers tended to establish institutions that were similar to those of their home countries. In consideration of these studies, we propose for African countries that informal institutions such as the colonial heritage matters:

2.2 Financial legal institutions and banking systems development

La Porta et al. (Reference La Porta, Lopez-de-Silanes and Shleifer2006) combine the “Antidirector Rights Index” (ADRI) with laws requiring firms to improve their reporting standards and test whether the constructed scores affect the market capitalization of stock markets. They find a strong link and show the superiority of the common law legal tradition in amplifying this causality link. They infer from this evidence that the combination of the ADRI index and scores representing the quality of reporting standards explains financial development better than the ADRI index alone.

Creditor protection is the complement to shareholder protection. La Porta et al. (Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1997) constructed a creditor protection indexFootnote ² to capture the extent to which creditors are protected in solvency and bankruptcy procedures. They find that higher levels of creditor protection affect financial development more in common law than in civil law countries. Levine et al. (Reference Levine, Loayza and Beck2000) assess the effects of creditor rights, contract enforcement and accounting standards on financial intermediation. They find that financial intermediaries only flourish in common law countries in which competent authorities are able to ensure accurate and effective financial reporting and to enforce contracts, and in which the legal system successfully protects creditors when borrowers file for bankruptcy. Djankov et al. (Reference Djankov, McLiesh and Shleifer2007) construct and test an international sample of 129 countries to study how financial development (measured by private credit to GDP) responds to differences in creditor laws. Their findings reveal that financial systems flourish more when the laws protecting creditors are strong, enforcement is guaranteed, and when enough credit information is available. Once again, this finding is more significant in common law than in civil law countries.

Other research, exclusively using firm-level data, focuses on the relationship between creditor institutions and banking development. These studies obtain results that are consistent with the literature that examines the institution/financial development nexus. For instance, Haselmann et al. (Reference Haselmann, Pistor and Vig2009) and Safavian and Sharma (Reference Safavian and Sharma2007) find that creditor protection laws improve lending, reduce interest rates and lengthen loan maturities (Qian and Strahan, Reference Qian and Strahan2007). Creditor laws, registries and information sharing also improve firms' access to finance in developing countries (Peria and Singh, Reference Peria and Singh2014), as banks are likely to offer lower lending rates in an environment where lenders are well protected. In sum, the law and finance literature suggests that countries with strong institutions that protect investors and creditors are associated with better and more efficient financial systems. Strong formal institutions are found to be more likely in common law countries than in civil law countries (La Porta et al., Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1998). Accordingly, we propose:

An alternative explanation is that banking system development is mainly related to institutional development and governance quality. Therefore, the third proposition that is tested in this paper is:

Putterman and Weil (Reference Putterman and Weil2010) propose an alternative way of explaining economic growth by introducing state history and transition to agriculture into the initial endowment conditions. They construct a matrix showing state history population share in each country in year 2000 that is descended from people in different countries in the year 1500. Their analysis shows that human migratory patterns and transition from agriculture do influence current levels of GDP per capita and income inequality in the world.

Another school of thought dealing with the development of institutions focuses on ethnic fractionalization. Proponents of this school argue that countries with higher levels of ethnic diversity are associated with social polarization that adversely affects economic development/growth. The basic concept is that a nation with diverse cultural, linguistic and religious groups will always struggle with implementing policies that are pro-growth and developing political consensus may lead to higher levels of patronage by one dominant ethnic group against the weak one(s) (Easterly and Levine, Reference Easterly and Levine1997). Typically, a dominant ethnic group emerges and diverts resources, or designs rent-seeking policies for their personal benefit, rather than for the general public good (Alesina et al., Reference Alesina, Baqir and Easterly1999: 436; Karnane and Quinn, Reference Karnane and Quinn2019). More recently, Emenalo et al. (Reference Emenalo, Gagliardi and Hodgson2018) combine disease endowment conditions, legal origins and ethnic fractionalization to explain financial development in African countries and find that indeed, these variables explain financial development (in terms of access to finance). Karnane and Quinn (Reference Karnane and Quinn2019) analyze an international sample of 157 countries over the period 1996–2014, to explore the effects of ethnic fractionalization and corruption on economic growth. Their results show that the two measures adversely impact economic growth as they increase political instability, creating an environment that reduces economic growth. There are also strands of literature linking cultural variables to economic outcomes.

One such strand was pioneered by Hofstede (Reference Hofstede1980). His initial international survey led to the development of four cultural dimensions: individualism, masculinity, uncertainty avoidance and power distance. These four measures were further used to predict economic growth, particularly among rich countries. Kwok and Tadesse (Reference Kwok and Tadesse2006) use Hofstede's measures to investigate the determinants of financial systems development using a sample of 41 countries. They find that cultural measures (particularly uncertainty avoidance) led to the development of more bank-based systems than market-based systems. Tabellini (Reference Tabellini2010) relates cultural variables (trust, confidence, respect for others) to economic development. His findings show that, conditional on literacy rates and historical political institutions since the year 1850, cultural variables strongly predict regional out-put. Given this summary of previous literature, we conjecture:

2.3 Banking systems in Africa

Since colonial times, African banking systems were dominated by European banks that operated primarily to serve individuals working in the colonial administration. Cull et al. (Reference Cull, Peria and Verrier2018) show that soon after independence, most African countries nationalized their banking systems, and the majority of banks became state-owned with the exception of a few European banks that continued operating alongside the nationalized banks. In the 1980s and 1990s, many African countries had to initiate structural adjustment programs which required to privatize their inefficient state parastatals, including banks. This brought about a marked rapid change from the dominance of state-owned banks to the emergence of more private and foreign owned banks.

However, despite reforms and the entry of more efficient foreign and domestic privately-owned banks in Africa, African banks are still small. Beck et al. (Reference Beck, Maimbo, Faye and Triki2011) show that the average total assets for an African bank is USD 220 million, whereas the average total assets of a non-African bank is USD 1 billion. Obviously, there are considerable variations among African countries. For example, the total assets of the Standard Bank in South Africa are estimated at USD 100 billion, while the total assets for a median bank in Madagascar is below USD 200 million. Sissy et al. (Reference Sissy, Amidu and Abor2017) note that most African countries have small and shallow economies that do not benefit banks and other providers of financial services to gain from providing financing despite the important role of banks in their financial systems. The World Bank study ‘Making Finance Work for Africa’ highlights that African banking systems are relatively small when compared to the rest of the world, and are the least developed as providers of financing, reaching only about 23% of African households (Beck et al., Reference Beck, Maimbo, Faye and Triki2011).

The African Development Bank report (Nyantakyi et al., Reference Nyantakyi, Sy and Kayizzi-mugerwa2015) shows that African countries, particularly those south of the Sahara, have shallow financial depth, with only 24% of domestic credit lending to the private sector, which is half of the ratio for North Africa and other parts of the world (OECD, Latin America and the Caribbean). Only 21% of firms operating in African obtain credit from African banks, compared to 43% of firms in non-African countries (Beck and Cull, Reference Beck, Cull, Berger, Molyneux and Wilson2014). Mlachila et al. (Reference Mlachila, Cui, Jidoud, Newiak, Radzewicz-Bak, Takebe, Ye and Zhang2016) show that the average amount of loans made by banks to the private sector in African countries account for less than 30% of their GDP. This is lower than the average of more than 45% found in other developing counties. Again, there is a high degree of variation among African countries. Banks in South Africa, Mauritius and Morocco loan more than 50% of their GDP, while the Democratic Republic of Congo, Equatorial Guinea, Guinea, Guinea Bissau and South Sudan loan less than 5% of their GDP. The ratio of liquid liabilities to GDP in African countries is 48%, lower than that of other developing countries and developed countries (80–105%). This figure also varies among African countries. Lower levels of bank deposits to GDP, deposits to loans, higher intermediation costs, higher net interest margins and spreads are among other key features of African banking systems (Beck and Cull, Reference Beck, Cull, Berger, Molyneux and Wilson2014; Honohan and Beck, Reference Honohan and Beck2007),Footnote ³ figures for previous years can be found in Mutarindwa et al. (Reference Mutarindwa, Schäfer and Stephan2020b). For an international comparison, refer to Table A1 in Mutarindwa et al. (Reference Mutarindwa, Schäfer and Stephan2021).

The stark heterogeneity across banking systems calls for exploring the channels through which those different banking systems have been shaped. Appropriate policy recommendations depend on the knowledge about those channels.

3.1 Data

Table 1 provides the names and descriptions of our variables. The data for our study come from a variety of sources. Institutional development data, macro-economic variables, banking development, governance indicators and population figures are all obtained from the World Bank. The data on the sub-classification of legal origins are from Klerman et al. (Reference Klerman, Mahoney, Spamann and Weinstein2011) and Oto-Peralías and Romero-Ávila (Reference Oto-Peralías and Romero-Ávila2014). In addition, we use the countries' profile of legal systems in Africa provided by the Lex Mundi Law Firm Network. Data pertaining to legal systems development (use of case law) are obtained from Guerriero (Reference Guerriero2016) who draws heavily on the International Encyclopedia of Comparative Law database. Data on ethnic fractionalization are sourced from the Harvard University Database developed by Drazanova (Reference Drazanova2019). Data on initial endowment conditions are obtained from Putterman and Weil (Reference Putterman and Weil2010)'s data on the history of nations and state transition to agriculture and from McCord (Reference McCord2012)'s data on Malaria disease prevalence.

Table 1. Variables description

Sources: (a1) Global Financial Development Database; (a2) World Bank: Doing Business projects; (a3) World Development Indicators; (a4) World Population estimates; (a5) World Governance Indicators; (b) Maoz and Henderson (Reference Maoz and Henderson2013), Klerman et al. (Reference Klerman, Mahoney, Spamann and Weinstein2011), Oto-Peralías and Romero-Ávila (Reference Oto-Peralías and Romero-Ávila2014), and La Porta et al. (Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1997); (c) Guerriero (Reference Guerriero2016); (d) Putterman (Reference Putterman2004); (e) McCord (Reference McCord2012), Emenalo et al. (Reference Emenalo, Gagliardi and Hodgson2018: 354); (f) Drazanova (Reference Drazanova2019); (g) Penn World Table 9.1, Groningen University Growth and Development Center.

3.2 Description of legal system and financial legal institutions

Table 2 groups the countries in our sample into the four categories. Similar to La Porta et al. (Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1997) and (Reference La Porta, Lopez-de-Silanes and Shleifer2008), we use categorical variables to indicate legal traditions. We categorize the countries in our sample based on British common law legal origin, mixed common law, French civil law and mixed civil law. To capture the type of colonization we distinguish between settler communities (Settler) and purely extractive colonization regimes (Extraction). We indicate the technique of developing the legal system by a dummy variable Case law following Beck et al. (Reference Beck, Levine and Demirgüç-Kunt2003), Guerriero (Reference Guerriero2016) and La Porta et al. (Reference La Porta, Lopez-de-Silanes, Pop-Eleches and Shleifer2004). This variable measures whether courts apply legal precedents established by case law when adjusting legal systems to respond to new legal and economic circumstances. A value of 1 is assigned to those countries that use case law and 0 otherwise.

Table 2. Legal traditions classifications

Source: Oto-Peralías and Romero-Ávila (Reference Oto-Peralías and Romero-Ávila2014); Klerman et al. (Reference Klerman, Mahoney, Spamann and Weinstein2011) and, Lex Mundi.

^a Swaziland changed its name in 2018 to Eswatini.

Financial legal development institutions are drawn from the World Bank's Doing Business Reports (DBR) pertaining to minority investor and creditor protection rights. Minority investor's protection rights are measured by the Investor protection index capturing how countries' laws protect small shareholders from expropriation by block-holders and management. Creditor rights protection is measured using the index Creditor rights. This variable captures the extent to which regulation and laws protect creditors from losses arising from loan defaults (Djankov et al., Reference Djankov, McLiesh and Shleifer2007). Enforcing contracts measures the time that it takes creditors in a given country to enforce debt contracts. We also use Kaufman and Kraay (Reference Kaufman and Kraay2008)'s measures of countries institutional development namely: Control of corruption, and Regulatory quality.

3.3 Measurement of initial endowments and culture

We follow Acemoglu et al. (Reference Acemoglu, Johnson and Robinson2001) and Putterman and Weil (Reference Putterman and Weil2010) in measuring precolonial conditions by initial endowments. Acemoglu et al. (Reference Acemoglu, Johnson and Robinson2001) argue that European colonizers adopted different colonization strategies depending on the initial conditions found in the colonized region, such as the relative prevalence of tropical diseases and the population density of the indigenous people. Oto-Peralías and Romero-Ávila (Reference Oto-Peralías and Romero-Ávila2014) also explore the population density measure, and suggest that some densely populated precolonial regions limited European occupation, while other densely populated regions were favorable to European settlement and were used for resource extraction. We were unable to access Acemoglu et al. (Reference Acemoglu, Johnson and Robinson2001)'s data on population density, so we opted for Putterman and Weil (Reference Putterman and Weil2010)'s data on population migration and statehood from the year 1500 measured in the year 2000. Two measures are derived from this database, namely: state history and transition to agriculture. State history is a binary variable that measures whether a country had sub-tribal governments, geographical boundaries and/or if the country was governed by its indigenous people or by outside countries, by the year 1500.

In addition, we adopt McCord (Reference McCord2012)'s malaria prevalence variable to capture settler mortality. Recently, Emenalo et al. (Reference Emenalo, Gagliardi and Hodgson2018) employed this measure to explore the effect of disease burden on the development of financial systems in African countries. Finally, we add the dummy variable EFindex which indicates the extent to which people belonging to a certain country differ in terms of ethnic identity, where 0 indicates no fractionalization and 1 high ethnic diversity.

3.4 Measurement of banking systems development

We use data from the Global Financial Development Database (GFDD) (Demirguc-Kunt et al., Reference Demirguc-Kunt, Klapper, Singer, Ansar and Hess2018) to measure the development of a country's banking system. The World Bank's typology includes three characteristics of a bank: depth, breadth and intermediation. Two ratios, the ratio of private credit provided by domestic banks to the private sector to Gross Domestic Product (GDP) (Private credit),Footnote ⁴ and the ratio of liquid liabilities to GDP (Liquid liabilities to GDP) measure bank depth. The amount of deposit resources mobilized by banks as a percentage of a country's GDP (Deposits to GDP) represents the banking system's breadth. Two other variables, namely Loans to deposits and Net interest margins (NIM), indicate the level of bank intermediation. A higher value of Loans to deposits and a lower value of NIM indicate a more developed banking system.

3.5 Control variables

We control for the macro-economic environment by using a proxy of the country's average income level, log(GDPpercapita). Following Oto-Peralías and Romero-Ávila (Reference Oto-Peralías and Romero-Ávila2014) and Emenalo et al. (Reference Emenalo, Gagliardi and Hodgson2018), we propose that higher incomes increase demand for financial services which ultimately drive differences in countries' financial development. We use GDP per capita, which is converted to international dollar using purchasing power parity rates (Emenalo et al., Reference Emenalo, Gagliardi and Hodgson2018). The size of a country's population is also included as a control variable expressed as a natural logarithm of the country's population size, denoted as log(Population). Allen et al. (Reference Allen, Carletti, Cull, Qian, Senbet and Valenzuela2014) argue that higher population densities create scale economies and improve financial development. The extension is that higher population densities provide a higher number of depositors allowing financial institutions to accumulate more savings, and this in turn increases intermediation and a more efficient provision of financial services. We also use the variable share of government spending relative to GDP as a measure of state activity and provision of public goods.

4.1 Sample description

In our sample, lending to private sector and liquid liabilities to GDP are, on average, 16.4 and 28.7%, respectively. These findings closely corroborate previous results in Beck et al. (Reference Beck, Fuchs, Singer and Witte2014), who find that private credit and liquid liabilities to GDP in Africa are, on average, 18 and 32%, respectively. These averages are considerably less than 34 and 47% (respectively) found in developing countries outside of Africa.

Thirty-five percent of the sample are countries that were once French colonies and use civil law as the basis of their legal system. Countries that were colonized by other European states but still practice civil law account for 28% of the sample. British colonies that use common law as the basis for their legal system account for 28% of the sample. Countries that were not British colonies but use common law as their legal basis make up 10% of the sample. An examination of the initial conditions variables reveals that settler colonies account for 33% of the countries in our sample, while 67% belong to the extraction colonies group (see Tables A2 and A3 in the online supplementary materials of Mutarindwa et al., Reference Mutarindwa, Schäfer and Stephan2021).

Minority shareholders and creditors (creditor rights and enforcement of contracts) are relatively more protected in common law than in civil law countries. Our results reveal that common law countries have higher regulatory quality and reduced levels of corruption as compared to civil law countries, indicating superior governance institutions. Our banking development variables indicate that private credit to GDP in common law countries is higher than in civil law countries (see Table A4 in Mutarindwa et al. (Reference Mutarindwa, Schäfer and Stephan2021) summarizing descriptive statistics for the legal traditions origin and also provides t-tests on differences between the legal traditions).

There are remarkable differences in the sample with respect to private sector lending. South Africa has the highest ratio of private credit to GDP (above 60%), while countries such as the Democratic Republic of Congo, Chad, Guinea and Guinea Bissau have the lowest (below 10%). A longitudinal comparison reveals that Botswana and Cape Verde's private credit to GDP ratio show observable improvements over time, while this ratio has decreased over time in Chad, Congo and Gabon. Some countries have banks with extremely large liquid liabilities to GDP, and a larger proportion of loans to deposits. Our intermediation variable indicates that countries such as Angola, Central African Republic, Democratic Republic of Congo, Gabon, Ghana, Liberia, Malawi, Siera Leone, Tanzania, Uganda and Zimbabwe have relatively higher interest margins (see Table A5 in Mutarindwa et al. (Reference Mutarindwa, Schäfer and Stephan2021) which summarizes the development of our banking systems' indicators by country over time).

4.2 Econometric models

To test the four propositions, we use the CRE approach (Mundlak, Reference Mundlak1978; Wooldridge, Reference Wooldridge2010). This econometric model has the advantage that it allows to estimate the effects of time-invariant variables, such as, legal origin or colonial settlement type. Furthermore, it relaxes the requirement of the random effects model that unobserved heterogeneity may not be correlated with the explanatory variables as those correlations are explicitly modeled by including the group-specific averages of time-variant explanatory variables into the model which is known as Mundlak formulation.

The CRE model can be written as (Schunck, Reference Schunck2013; Schunck and Perales, Reference Schunck and Perales2017):

(1)$$y_{it} = \beta _0 + \beta _wx_{it} + \beta _2c_i + \pi \bar{x}_i + \mu _i + \lambda _t + \nu _{it}$$

where y_it is the dependent variable for country i in year t, β_w corresponds to the within estimates, x$\bar{x}_i$ are group specific means of variables and π indicates the difference between within and between estimates, π = β_w − β_b. μ_i denote individual random effects uncorrelated with the error term ν_it, λ_t denotes time effects and x_it the other explanatory variables of the model. It should be noted that if H ₀: π = 0 is not rejected, a pure random effects model would be the appropriate model. Under the alternative H ₁: π ≠ 0, the CRE specification is supported. It is also worth noting that the CRE formulation corresponds to an augmented regression model test where a Hausman test on the random versus fixed effects specification is conducted.

Schunck (Reference Schunck2013) shows that the CRE model is numerically equivalent to a so-called hybrid model specification that encompasses both within and between estimates of time-variant variables:

(2)$$y_{it} = \beta _0 + \beta _w( x_{it}-\bar{x}_i) + \beta _2c_i + \beta _b\bar{x}_i + \mu _i + \lambda _t + \nu _{it}.$$

Because the between group estimates $\hat{\beta }_b$ have a direct interpretation, we prefer to report the results from the hybrid model over the CRE specification results.Footnote ⁵ The within estimate $\hat{\beta }_w$ shows the effect of a time-varying variable on the outcome at the country level, while the between estimate $\hat{\beta }_b$ is interpreted as the long-term impact of that variable. At the same time we can identify the impact of time-invariant variables denoted as c_i with this approach.

In the first set of estimates using the CRE model, time-invariant legal tradition and colonial type as well as other time-varying country controls are used as explanatory variables as shown below:

(3)$$\eqalign{Creditor/Investor\,protection_{it} = \, & f( Legal\,tradition_i, \;Colonial\,type_i, \;Controls_{it}) \cr & \quad + \mu _i + \lambda _t + \omega _{it}} $$

where the dependent variables describe countries i = 1, … N financial legal development in year t.

The variable Legal tradition_i represents the legal origins category consisting of common law (British), common law (mixed), civil law (French) and civil law (mixed), and the variable Colonial type_i describes settler versus extraction colonization. Controls_it denotes the country controls including country's population size (using log) and macro-economic variables including growth rate and GDP per capita in country i at time t. ω_it denotes the error term.

The second set of estimations contains our main results. We explain depth, breadth and the level of intermediation of the countries' banking sectors with legal tradition, financial legal institutions, colonial type, initial endowment, culture and institutional development and quality and additional control variables such as population size and GDP per capita.

(4)$$\eqalign{Banking\,System\,Development_{it} = \, & f( Creditor/Investor\,protection_{it}, \;\,Legal\,tradition_i, \;\cr & Colonial\,type_i, \;\,Initial\,Endowment_i, \;Culture_i, \;\cr & Institutional\,development\,and\,governance\,quality_{it}, \;\cr & Controls_{it}) + \mu _i + \lambda _t + \varepsilon _{it}} $$

where μ_i and λ_t denote country- and years-effects, respectively, and ɛ_it denotes the error term. These models are also estimated by employing the CRE estimator which allows identification of the effect of time-invariant variables as well as provides within and between estimates for the time-varying variables.

4.3 The relationship between legal origin, colonial settlement type and financial legal institutions

Table 3 reports the results for the CRE model. Our main interest is to examine whether the strength of the financial legal institutions is strongly linked to the legal tradition. As is common in the research initiated by La Porta et al. (Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1997, Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1998) (hereafter LLSV), the variable Legal tradition describes only the dichotomous categories of common and civil law. The significance of the Legal tradition coefficients strongly confirms Proposition 1a. Common law is linked to stronger investor protection and creditor rights and to more prompt enforcement of debt contracts. Not entirely surprisingly, results for the dichotomous categories of legal origin are in line with previous findings in the LLSV literature (Beck et al., Reference Beck, Levine and Demirgüç-Kunt2003; Djankov et al., Reference Djankov, Glaeser, La Porta, Lopez-de-Silanes and Shleifer2003; La Porta et al., Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1997). The type of colonization (settler versus extraction) also matters. The coefficient of the variable Colonial type:settler is in line with Proposition 1b as it is related to a higher level of investor protection. Interestingly, former settler colonies are associated with weaker contract enforcement mechanisms, in the sense that in those countries it takes, on average, more days to enforce a contract.

Table 3. Relationship between legal traditions, colonial settlement type and investor/creditor protection (CRE model estimates)

Notes: Cluster robust t statistics in parentheses. ^∗p < 0.10, ^∗∗p < 0.05, ^∗∗∗p < 0.01. Reference categories: Legal origin: Civil law (French), Colonial type: extraction, (t) denotes time invariant, (b) denotes between, (w) within estimates. Parameters log(GDPpercapita) (b) and log(Population) (w) not reported.

Enforcing contracts in days, see Table 1.

The other macro indicators are of minor relevance in the CRE regression approach. The level of economic development (measured by GDP per capita) has no significant impact on any of the dependent variables. For the sake of brevity, we only report for this variable the within estimate, since it reflects improvement of income in the respective country. The size of a country's population affects the existence of strong creditor rights positively. For this variable we only report the between estimates, as it reflects differences across countries in terms of different market sizes.

Table 3 also reports in columns 4–6 results using the legal origins sub-groups as explanatory variables in the CRE approach. It can be seen that British common law judicial origins have stronger creditor rights and better investor protection than civil law countries. Overall, the Bayesian information criterion (BIC) values indicate that the models shown in columns 1–3 are superior as increasing the models' complexity in columns 4–6 is not compensated by a much better model fit. Therefore, in our attempt to explain banking sector development, the following analyses focus only on the common law versus civil law distinction, and we disregard the mixed legal origins subgroups.

4.4 Determinants of banking system development in African countries

We turn now to our main question of interest: How do legal institutions shape the evolution of banking sectors in African countries and what role does the legal and colonial heritage play?

Table 4 reports the regression results from the CRE model. Creditor rights (b) and Enforcing contracts (b) influence banking system development, however, creditor rights have, in contrast to Proposition 2, a negative effect on the depth and breadth of the banking system. Only the within estimates of Creditor rights (w) show the expected influence on breadth and intermediation. It should be noted that an increase in Creditor rights (w) lowers the cost of banking. Clearly, as many coefficients of the legal institutions are insignificant or have an unexpected sign, the obtained results from the CRE estimation do not support Proposition 2.

Table 4. Institutional determinants of banking system development (CRE model estimates)

Notes: See Table 3.

Proposition 2 is only partly confirmed with respect to formal institutions. The indicator Legal tradition shows a significant influence on development, in particular banking sectors in countries with a Common Law origin have a significant higher depth and breadth. However, in terms of intermediation, the cost of banking is higher and the loans to deposit ratio is lower for countries with common law origin.

Colonial heritage affects banking system development. Countries that were settled by colonizers have deeper and broader financial systems than countries that experienced purely extractive colonization. Those findings confirm the second part of Proposition 2, and are compatible with the notion that, in contrast to the native population, settlers had the power to influence political and judicial decision making in support of constructing and developing a decent banking system in the colony where they settled. Although legal origin and colonial heritage matter, current legal institutions show only a modest impact on banking system development in African countries.

The set of estimates shown in Table 5 describes the role played by institutional development and governance quality in the development of banking systems in African countries. Specifically, we employ the variables Regulatory quality and Control of corruption as measures for the overall governance quality in the respective country. We also include the share of government spending as an additional control to investigate the degree to which the government contributes to banking sector development, relative to the private sector.

Table 5. Institutional determinants of banking system development with government quality indicators and share of government spending in GDP (CRE model estimates)

Notes: See Table 3.

The CRE results highlight that governance quality matters for most dimensions of banking system development, and even renders some of the previously significant factors as insignificant. This shows that the positive effect of the common law tradition on financial development is captured by a higher degree of institutional development and governance quality in those countries. Thus, a better governance quality is conducive for banking system development, which supports Proposition 3. We also find that the governmental spending is conducive to banking sector development, in particular in terms of private credit and loans to deposit. Government spending is a signal for the importance of the state in the economy. For one, it is conducive for the supply of liquid and safe assets. Such assets are essential for banks. Accordingly, a positive relationship arises naturally between the spending variable and the development of the banking system. Moreover, the flourishing of the banking sector depends on the quality of regulation and supervision (Mutarindwa et al., Reference Mutarindwa, Schäfer and Stephan2020a), and thus on governmental money for the regulatory and supervisory infrastructure. Also, we find that across countries, those countries with a higher share of government spending as a percentage of GDP have lower banking costs.

A BIC comparison of the models presented in Table 5 with the models presented in Table 4 reveals that the models that include institutional development and governance quality variables have a similar or even somewhat lower BIC score. This indicates that those models are slightly better in explaining banking sector development considering the increased model complexity due to including more variables.

Proposition 4 is a consideration of alternative explanations of banking sector development and concludes our empirical analysis. These estimations are reported in Table A6 (Mutarindwa et al., Reference Mutarindwa, Schäfer and Stephan2021). We include four additional time-invariant variables compared to our baseline estimates of Table 4: initial endowment proxied by Malaria variable, ethnic diversity captured by the EFindex, State history and Transition to agriculture. We add the variable Case law to indicate whether the country uses case law to establish legal precedent.Footnote ⁶

No clear pattern emerges that helps to explain which parts of banking sector development are influenced by these four additional variables. Although higher EFindex values have a positive effect on private credit and loans to deposit ratios, malaria occurrence and late transition to agriculture have a detrimental effect on banking sector development. Surprisingly, the application of case law appears to have a negative influence on all variables of banking sector development, except for the cost of banking. What is even more striking is that the BIC score for all models using the four additional time-invariant variables are lower when compared to baseline models reported in Table 4, which we view as supporting Proposition 4.

4.5 Robustness tests

The first set of robustness checks concern potential endogeneity of financial institutions. Although legal origin, colonization type or initial endowment is less of a concern, the legal institutions (Creditor rights, Investor protection and Enforcing contracts) are potentially endogenous in explaining banking sector development. The Hausman–Taylor estimator (Hausman and Taylor, Reference Hausman and Taylor1981) uses instruments to explain depth, breadth and the level of intermediation in the countries' banking sectors. Specifically, we instrument Creditor rights, Investor protection and Enforcing contracts with all exogenous variables in the specification (Table A7, Mutarindwa et al., Reference Mutarindwa, Schäfer and Stephan2021).

By using instruments, in particular the legal origin, we follow earlier scholarly study, e.g. Levine et al. (Reference Levine, Loayza and Beck2000), La Porta et al. (Reference La Porta, Lopez-de-Silanes and Shleifer2006), Caprio et al. (Reference Caprio, Laeven and Levine2007) and Beck et al. (Reference Beck, Demirgüç-Kunt and Maksimovic2006) who use legal tradition to explain minority investor protection, creditor rights, contract enforcement and property rights. By including the colonial type as an instrument, we borrow from Acemoglu et al. (Reference Acemoglu, Johnson and Robinson2001). Using instruments change the results considerably, see Table A7 in Mutarindwa et al. (Reference Mutarindwa, Schäfer and Stephan2021), but overall the support for Proposition 2 remains still quite weak.Footnote ⁷

Another potential concern is that cluster robust standard errors might be downward biased as the number of clusters is 40 or below and several regressors vary only at the country level. To investigate this issue, we computed bootstrap standard errors with asymptotic refinement, which are expected to have lower finite sample bias, see Cameron et al. (Reference Cameron, Gelbach and Miller2008).Footnote ⁸ The findings of this exercise reveal that the estimated standard errors reported in Table 4 are quite accurate showing little downward bias and most significant results remain using the bootstrap standard errors.