1 Introduction

Insurance market activity, both as a financial intermediary and as a provider of risk transfer and indemnification, was widely studied in the literature. Thus, according to the literature (See, Skipper 1997; Skipper and Kwon 2007; Haiss and Sumegi 2008; Njegomir and Stojić 2012, etc.), insurance activity contributes to economic growth, because that it: (1) promotes financial stability (Skipper 1997; Skipper and Kwon 2007); (2) facilitates the development of trade and commerce by increasing creditworthiness, lowering the total necessary amount and cost of capital, and decreasing total risk (Skipper 1997; Sawadogo 2019); (3) mobilizes domestic savings (Haiss and Sumegi 2008); (4) allows different risks to be managed more efficiently by encouraging the accumulation of new capital (Dickinson 1998; Skipper and Kwon 2007); (5) fosters a more efficient allocation of domestic capital (Grace and Rebello 1993), (6) helps reduce or mitigate losses (Skipper 1997). In addition, there are likely to be different effects on economic growth from life and nonlife insuranceFootnote 1 given that these two types of insurance protect households and corporations against different kinds of risks (Arena 2008). Moreover, life insurance companies facilitate long-term investments rather than short-term investments as the case for non-life insurance industry does (Liu et al. 2014). Consequently, life and non-life insurance activities can affect the economic growth in diverse ways.

Most empirical studies have analysed the relationship between the development of life insurance and growth in three ways. First, according to the “supply-leading” hypothesis, previous studies have shown that the development of insurance services is a determinant of economic growth (Arena 2008; Haiss and Sumegi 2008; Chang and Lee 2012; Chen et al. 2012; Lee et al. 2013a, b; Outreville 2013). Second, according to the “demand-following” hypothesis, the empirical studies have found that insurance development is influenced by economic growth (Outreville 1990; Beck and Webb 2003; Feyen et al. 2011; Chang and Lee 2012 etc.). Third, there are some studies which have investigated on applying both the supply-leading and demand-following theories, i.e. the causal nexus between insurance activity and economic growth (Ward and Zurbruegg 2000; Kugler and Ofoghi 2005; Lee 2013; Lee et al. 2013a, b; Liu 2016, etc.). Indeed, these authors have shown that there is a two-way causality between insurance development and economic growth.

The relationship between the development of insurance sector and other financial sectors (particularly banking activities) has not received much attention in the empirical literature. However, according to the theoretical literature (Skipper 1997; Rule 2001; Zou and Adams 2006), the relationship between banking and insurance activities can be tenable because of risk transfer between the two sectors. Furthermore, consumer credit may take the form of personal loans (either for general purpose or specified one), the purchase of durable goods (e.g., cars, and furniture), or revolving credit such as credit cards. The more credit consumption there is (ceteris paribus), the more insurance coverage is bought by hypothesis. Thus, the relationship between bank credit and insurance development is probably the most obvious one, as creditors often require insurance coverage for providing credit. Furthermore, given that banks and insurers have mutual disclosures in many areas, banks have unbundled their credit risks to insurance providers mainly through the securing of both credit portfolios and derivatives (Lee 2013). The development of the insurance activity could encourage bank borrowing by reducing companies’ market cost of capital, which influences economic growth by increasing the demand for financial services (Webb et al. 2005). Also, property insurance may facilitate the bank intermediation activity by, for example, partially collateralizing credit, which would reduce the bank’s credit risk exposures by promoting higher levels of lending (Zou and Adams 2006). At the same time, the development of the banking sector facilitates the development of insurance activity through a much more effective payment system, which allows an improved financial intermediation of services (Webb et al. 2005). However, one can have a “saving substitution effect” between insurance activities, particularly life insurance and banks (Haiss and Sümegi 2008) because in market for intermediated saving, insurance companies compete and could reduce banks’ market share in developing countries (Allen and Santomero 2001). Thus, given also that insurance coverage is a secondary product in the market for consumer credit, it seems logical to view development in the credit market in relation to insurance markets.

Regarding the empirical literature, the previous studies have focused more on the bank credit impact on the development of insurance than the simultaneous effect between insurance and banking development on economic growth. Thus, Outreville (1996), Ward and Zurbruegg (2002), and Beck and Webb (2003) have showed that the development of banking sector is significantly and positively correlated with the development of insurance market. Beck and Webb (2003) have argued that countries whose banking sector is more developed have larger insurance sectors. However, for the simultaneous effect between banking and insurance development, Webb et al. (2005) and Arena (2008) have shown that there is a complementary relationship between baking credit and insurance (life and non-life insurance) markets, while Tennant and Abdulkadri (2010) and Chen et al. (2012) have indicated a competitive relationship between banking credit and insurance market. Moreover, the previous studies, which have investigated the causal relationship between insurance activity and banking credit from a macro-perspective, are byLiu and Lee (2014) and Liu et al. (2014). These authors have found that there is a causal nexus between insurance activity and banking credit in China and G-7 countries. In a recent study, using data from 30 OECD countries during the period 2004–2011, Lee and Lin (2016) have fund that greater globalization leads insurance companies to exhibit a better performance of insurance firms.

However, most previous studies have only analysed the causal relationship between insurance activity and banking credit in developed and emerging countries (Liu and Lee 2014; Liu et al. 2014). Furthermore, none of these studies have examined how globalization affects the impact of banking credit on insurance activity. Indeed, the analysis of the role of globalization in the relation between banking credit and the development of insurance activity is important for policymakers in Africa countries. If globalization impacts insurance markets, then policymakers’ attitude towards globalization may change because it implies that the globalization is the opportunity for the development of insurance sector in Sub-Saharan African. Furthermore, this study aims at filling the gap in the literature and contributing to the existing literature by investigating the long-run linkages between insurance activity (life, non-life insurance, and total insurance) and banking credit to private sector for 20 countries of Sub-Saharan African (SSA) in the period 1990–2017. Thus, we want to know whether the private credit consumption is complementary or substitutionary to insurance (life, non-life, and total insurance) development in the context of the globalization.

This study contributes to the previous works through two main points. First, in contrast to prior studies that have separately examined the direct impact of banking credit to private sector on insurance development (Lorent 2010; Lee 2013; Liu and Lee 2014; Liu et al. 2014) and the impact of globalization on the performance of insurance firms (Lee and Lin 2016), we evaluate how the globalization can influence the impact of banking credit on insurance development by a the Pooled Mean Group (PMG) estimator for panel. The PMG estimator allows short-run heterogeneity, while imposing long-run homogeneity on insurance activity determination across countries.

We investigate the effect of the globalization by using a measure that takes into account the multifaceted globalization factors. Thus, the multidimensional globalization could provide a more comprehensive evaluation than the single indicator of globalization factors. We utilize KOF Globalization Index of Gygli et al. (2018), which consists of the economic, social, and political dimensions of globalization. Indeed, the economic globalization is measured through indicators of actual flows and restrictions, the social globalization is measured by indicators of personal contacts, information flows, and culture proximity, and political globalization depends on the index of a country’s embassies, membership in international organizations, and participation in U.N. Security Council missions (Dreher et al. 2008). These measures enable us to capture the effect of globalization on the bank credit–insurance activity relationship.

The rest of the paper is organized as follows. Section 2 describes the data used in the paper and some stylized facts. Section 3 focuses on the empirical econometric model, while Sect. 4 analyses the results. Section 6 concludes and provides some concluding remarks.

2 Data

The data used in this paper are the annual data from 1990 to 2017 for 20 countries in Sub-Saharan African.Footnote 2 The measure of the real insurance density (life, non-life and total insurance), defined as the average annual premiums per capita and the real banking credit density, indicates the average annual domestic credit to private sector by banks per capita. Indeed, insurance density (life, non-life, and total insurance) shows the average annual premium per capita that one inhabitant in one country spends on insurance products. The banking credit density indicates the average annual domestic credit provided by banking sectors for one inhabitant in the private sector. The annual data for real insurance density (life, non-life, and total insurance) and the real banking credit to the private sector are taken from Global Financial Development Database of Čihák et al. (2012). All variables are measured in constant 2005 $US to be comparable over time.

Furthermore, we have used the indicator of globalization developed by Gygli et al. (2018) which is a composite index measuring globalization for every country in the world along the economic globalization (36%), social globalization (38%), and political globalization (26%) dimension. This index is taken from the KOF Index of Globalization and can be downloaded from http://www.kof.ethz.ch/globalization/.

Finally, we have controlled several other variables used as determinants of the development of insurance. Thus, we have used log (GDP per capita), trade openness, government consumption, inflation, life expectancy, and dependency of young ratio as control variables. These variables are obtained from the World Bank World Development Indicators (WDI). Appendix Tables 5 and 6 present basic descriptive statistics and correlations for the data used in the regressions.

3 Empirical strategy

3.1 Specification of PMG test

To examine the long-term or co-integration relationship between banking credit on insurance (life, non-life insurance, and total insurance) development, we estimate the equilibrium relationship between Insurance \( {\text{INS}} \) and \( {\text{BANK}} \) in a panel data context with this model:

$$ {\text{INS}}_{it} = \alpha_{0} + \alpha_{1} *{\text{BANK}}_{it} + \alpha_{2} *X_{it} + \vartheta_{it} . $$
(1)

where \( {\text{INS}}_{it} \) is insurance density (life, non-life insurance, or total insurance) of country i in period t; \( {\text{BANK}}_{it} \) denotes real banking credit density, and \( X_{it} \) is vector of the control variables which include determinants of insurance development documented in the literature such as log (GDP per capita), trade openness, government consumption, inflation, life expectancy, and dependency of young ratio. The error term \( \varepsilon_{it} \) is stationary when the variables are co-integrated. Thus, the dynamic heterogeneous panel model of Pesaran et al. (1999) is an unrestricted error correction autoregressive distributed lag (ARDL) representation. Thus, autoregressive-distributed lag (ARDL) model is an autoregressive model of order p in the dependent variable and an autoregressive model of order q in the explanatory variables. In an ARDL model, the dependent and independent variables enter the right-hand side with lags (p, q):

$$ y_{it} = \mathop \sum \limits_{j = 1}^{p} \lambda_{ij} y_{i,t - j} + \mathop \sum \limits_{j = 0}^{q} \emptyset_{ij}^{'} x_{i,t - j} + \mu_{i} + \varepsilon_{it} $$
(2)

where \( i = 1, 2, \ldots , N \) is country index, \( t = 1, 2, \ldots , T \) is a time index, \( j \) is the number of time lags, and \( \mu_{i} \) denotes country specific fixed effects.

By re-parameterization, with respect to the long-run coefficients \( \alpha \) and the adjustment coefficients \( \varphi_{i} \), the error correction form is given by:

$$ \Delta y_{it} = \varphi_{i} \left[ {y_{i, t - 1} - \alpha_{i}^{'} x_{\text{it}} } \right] + \mathop \sum \limits_{j = 1}^{p - 1} \lambda_{ij}^{*} \Delta y_{i,t - j} + \mathop \sum \limits_{j = 0}^{q - 1} \emptyset_{ij}^{*'} \Delta x_{i,t - j} + \mu_{i} + \varepsilon_{\text{it}} $$
(3)

with \( \varphi_{i} = - \left( {1 - \mathop \sum \nolimits_{j = 1}^{p} \lambda_{ij} } \right) \); \( \lambda_{ij}^{*} = - \mathop \sum \nolimits_{m = j + 1}^{p} \lambda_{im} ,\;j = 1,2? \ldots ,\;p - 1;\;\emptyset_{ij}^{*'} = - \mathop \sum \nolimits_{m = j + 1}^{q} \emptyset_{im} ,\;j = 1,2, \ldots ,q - 2; \;\alpha_{i} = {\raise0.7ex\hbox{${\beta_{i} }$} \!\mathord{\left/ {\vphantom {{\beta_{i} } {\varphi_{i} }}}\right.\kern-0pt} \!\lower0.7ex\hbox{${\varphi_{i} }$}} \).

\( \alpha_{i } \) defines the long-run equilibrium relationship between \( y_{it} \)(insurance density) and \( x_{it} \) (vector of explanatory (banking credit) and control variables). In contrast,\( \lambda_{ij}^{*} \) and \( \emptyset_{ij}^{*'} \) are the short-run coefficients relating insurance (life, non-life insurance, and total insurance) development to its past values and other determinants \( x_{it} \). Finally, \( \varphi_{i} \), coefficient of lagged dependent variable and disturbance, \( \varepsilon_{it} \), are supposed to be normally and independently distributed across i and t with zero mean and variances \( \sigma_{i}^{2} > 0. \) Coefficient \( \varphi_{i} \) is also called adjustment to long-run equilibrium following a change in bank credit and control variables. \( \varphi_{i} < 0 \) ensures that such a long-run relationship exists. Thus, a significantly negative \( \varphi_{i} \) constitutes evidence of a long-run equilibrium relationship. As a result, a significant and negative value of \( \varphi_{i} \) is treated as evidence of co-integration between \( y_{it } \) and \( x_{it} \).

Replacing \( y_{it} \) and \( x_{it} \) by the different explanatory and control variables, Eq. (3) becomes:

$$ \begin{aligned} \Delta {\text{INS}}_{\text{it}} & = \gamma_{i0} + \varphi_{i} {\text{INS}}_{i, t - 1} + \gamma_{i1} {\text{Bank}}_{it - 1} + \gamma_{i1} {\text{control}}_{it - 1} + \mathop \sum \limits_{j = 1}^{p - 1} \lambda_{ij}^{*} \Delta {\text{INS}}_{i,t - j} + \mathop \sum \limits_{j = 0}^{p - 1} \emptyset_{ij}^{1} \Delta {\text{Bank}}_{i,t - j} \\ & \quad + {\kern 1pt} \mathop \sum \limits_{j = 0}^{q - 1} \emptyset_{ij}^{2} \Delta {\text{control}}_{i,t - j} + \mu_{i} + \varepsilon_{it} \\ \end{aligned} $$
(4)

Does globalization matter in the relationship between banking credit and insurance (life, non-life insurance, and total insurance) development? In order to answer this question, we specify an augmented version of Eq. (4) as follows:

$$ \begin{aligned} \Delta {\text{INS}}_{\text{it}} & = \beta_{i0} + \varphi_{i} {\text{INS}}_{i, t - 1} + \beta_{i1} {\text{Bank}}_{it - 1} + \beta_{i2} {\text{Globalization}}_{it - 1} + \beta_{i3} {\text{Bank}}_{it - 1} *{\text{Globalization}}_{it - 1} \\ & \quad + {\kern 1pt} \beta_{i3} {\text{control}}_{it - 1} + \mathop \sum \limits_{j = 1}^{p - 1} \lambda_{ij}^{*} \Delta {\text{INS}}_{i,t - j} + \mathop \sum \limits_{j = 0}^{p - 1} \gamma_{ij}^{1} \Delta {\text{Bank}}_{i,t - j} \\ & \quad + {\kern 1pt} \mathop \sum \limits_{j = 0}^{p - 1} \gamma_{ij}^{2} \Delta {\text{Globalization}}_{i,t - j} \mathop \sum \limits_{j = 0}^{p - 1} \gamma_{ij}^{3} \Delta ({\text{Bank}}*{\text{Globalization}})_{i,t - j} + \mathop \sum \limits_{j = 0}^{q - 1} \gamma_{ij}^{4} \Delta {\text{control}}_{i,t - j} \\ & \quad + {\kern 1pt} \mu_{i} + \varepsilon_{it} \\ \end{aligned} $$
(5)

Following the approach adopted by Pesaran et al. (1999), it is noted that one of the advantages of ARDL models is that the multipliers of short and long-term are estimated jointly. Furthermore, these models allow the presence of variables that can be integrated in different orders, either I(0) or I(1) or co-integrated (Pesaran et al. 1999). Thus, the underlying ARDL specification dispenses with the unit root pre-testing of the variables. Provided that there is a unique vector defining the long-run relationship amongst variables involved, PMG estimates of an ARDL regression yield consistent estimates of that vector, no matter whether the variables involved are stationary or non-stationary. Finally, we have tested the validity of a homogeneity restriction by using a standard Hausman-type statistic. Thus, if the long-run homogeneity restrictions are valid, MG estimates will be inefficient and the maximum likelihood-based PMG approach will yield a more efficient estimator. This estimator allows the short-run coefficients and error variances to differ freely across groups, and the long-run coefficients are constrained to be the same.

3.2 Cross-sectional dependence tests

One important issue in a panel causality analysis is to take into account possible cross-sectional dependency amongst countries. Indeed, a high degree of economic and financial integration makes a country sensitive to economic shocks in another country as African countries. Thus, cross-sectional dependency may play an important role in detecting causal linkages between insurance activity development and the development of banking sector. Thus, we use the Lagrange multiplier (LM) test for Breusch and Pagan (1980) and cross-sectional dependence (CD) proposition of Pesaran (2004).

The Lagrange multiplier (LM) test for Breusch and Pagan (1980) statistic is defined by:

$$ {\text{LM}} = T\mathop \sum \limits_{i = 1}^{N - 1} \mathop \sum \limits_{j = i + 1}^{N} \hat{\rho }_{ij}^{2} $$
(6)

where \( \hat{\rho }_{ij} \) is the sample estimate of the pair-wise correlation of the residuals from the pooled OLS estimation. Under the null hypothesis, the LM statistic has an asymptotic Chi square with \( N\left( {N - 1} \right)/2 \) degrees of freedom. Moreover, the LM test is valid for relatively small N and sufficiently large T. Thus, our sample respects this condition because the temporal dimension (T = 27) is superior to number of countries (N = 20). Finally, we have used the new test for cross-sectional dependence (CD) propose of Pesaran (2004). This test is based on the pair-wise correlation coefficients rather than their squares used in the LM test. The CD statistic is given by:

$$ {\text{CD}} = \sqrt {\frac{2T}{{N\left( {N - 1} \right)}}} \mathop \sum \limits_{i = 1}^{N - 1} \mathop \sum \limits_{J = i + 1}^{N} \hat{\rho }_{ij} $$
(7)

Under the null hypothesis of no cross-sectional dependence with the \( T \to \infty \) and then \( N \to \infty \) in any order, CD asymptotically follows a normal distribution. Pesaran (2004) has shown that the CD test is likely to have good small sample properties (for both N and T small). It also turns out to be remarkably robust to major departures from normal errors, particularly for T ≥ 10, and, as predicted by the theory, it is not affected by multiple breaks, so long as the unconditional means of the individual processes remain constant over time (Pesaran 2004).

4 Empirical results

4.1 Unit root and cross-sectional dependence tests

The properties of the variables need to be investigated in order to avoid the possibility of spurious regressions. Thus, we use the tests of first generation of Pesaran et al. (1999) and Maddala and Wu (1999) to test the non-stationarity in our data series. Taking into consideration the low power of the first generation tests, we have also used the second generation of Pesaran (2007) panel unit root test, as an alternative test, which takes account of the cross-sectional dependence due to common factors. The results are displayed in Table 1. The results indicate that Private credit density*Globalization Index, log (GDP per capita), and trade are statistically insignificant. When we apply the unit root tests to the first difference of the two variables, both tests reject the joint null hypothesis for each variable. Thus, the unit root test validates the use of ARDL approach, which demonstrates that long-run relationships can exist between both stationary and non-stationary variables.

Table 1 Results of the unit roots tests
Full size table

To investigate the existence of cross-sectional dependence, we have carried out two different tests (Breusch–Pagan LM and Pesaran CD) and we have reported the final testing results in Table 2. The null hypothesis (H0) is that there is no cross-sectional dependency. The Breusch–Pagan LM and Pesaran CD statistics for all regressions indicate clearly that null hypothesis of non-cross-sectional dependence is rejected at the conventional levels of significance. This finding implies that a shock that occurs in one of these countries spills over onto other countries. Hence, we should take into account this information when examining the causal links between insurance consumption and bank credit to private sector. Thus, PMG method is more appropriate than the country-by-country pooled OLS method.

Table 2 Results of cross-sectional dependence tests
Full size table

4.2 Results of PMG estimations

Before the analysis of the results of PMG, we have tested the possible existence of a co-integrated relationship between the series that allows us to assume a potential long-term relationship between them. Thus, following Kao (1999), the co-integration test (tableau 3 bellow) confirms the existence of a co-integrating vector in all cases. The analysis focuses, first, on the effect of banking credit on insurance (life, non-life) activity and second on the role of globalization in the banking credit impact on development of insurance activity.

PMG estimator was used to model the long-term conditional effect of banking credit on the development of insurance activity. This conditional effect relates to the globalization degree. The results of the Hausman discrimination test show that the hypothesis of homogeneity of the long-term coefficients cannot be rejected. As a result, PMG estimator is robust in explaining the long-run relationship between different model variables. Indeed, being given that PMG supposes homogeneity in long-term coefficients, the interpretations of the results will, therefore, focus on long-term effects obtained by using the PMG estimator. The short-term coefficients can be found in Appendix Tables 7 and 8.

Table 3 presents the long-run coefficients that are of interest to us for the basic model. The adjustment term is always negative and significant, indicating that there is no omitted variable bias. Indeed, the PMG requires a negative coefficient on the error correction term (EC) of between 0 and − 2 (Loayza and Rancière 2006; Huang et al. 2015). The results support this condition. This means that there is an error correction mechanism and that movement of life insurance density (column 1), non-life insurance density (column 2), or insurance density (column 3) of countries in our sample is corrected to 27.04%, 67.70%, or 55.67%, respectively, by the feedback effect.

Table 3 Banking credit on insurance (life, non-life insurance, and total insurance) development
Full size table

Table 3 presenting the direct effect shows that banking credit density is positively associated with life and total insurance density in African countries (column 2 and 3), but that it does not have any significant effect on non-life insurance density (column 2). Indeed, in terms of impact, an increase in banking credit of 1% would imply an increase in life assurance activity of 0.017% and 0.0039% for insurance activity. Comparing the effects of banking credit on life and insurance activity, it is easy to find that real banking credit density has a more positive impact on real-life insurance density than real insurance density. The results indicate that life insurance activity and banking credit are complementary for the 20 countries of SSA. Thus, contrary to the works of Allen and Santomero (2001) and Haiss and Sumegi (2008) that have found that there is a competitive relationship between life insurance and bank credit markets because of the storage substitution, our results show a cooperated relationship between bank credit and insurance and confirms the studies of Webb et al. (2005), Lee (2013), Liu and Lee (2014), and Liu et al. (2014). Thereby, this complementarity mainly passes through the bank credit channel and can be explained on two levels. First, the development of the insurance activity covers banks and their customers against a range of risks. Indeed, to support bank loans by protecting customers against risks that might otherwise leave them unable to repay their debts (Rule 2001; Arena 2008). Second, the risk protection offered by insurance companies encourages bank borrowing by reducing companies’ market cost of capital (Grace and Rebello 1993)

Regarding other control variables, the results show that the income level is significantly and positively correlated with non-life insurance and insurance density (column 2 and 3), which is consistent with the findings in the extant literature (Browne and Kim 1993; Enz 2000; Ward and Zurbruegg 2002; Beck and Webb 2003). The trade openness has a positive effect on the development of life, non-life, and insurance activity in SSA, unlike inflation which evidently has a negative effect on life insurance density. The young dependency ratio and life expectancy have a negative effect on life assurance density, but the effect is insignificant. However, their effects are positive for total insurance density.

This subsection examines whether country-specific characteristics such as the level of globalization influence the relationship between banking credit and life and non-life insurance activity in SSA. The interaction between real banking credit density and globalization index are included in the model to portray the accentuation role played by globalization. The regression results from the estimation of Eq. (5) are reported in Table 4. Once again, the diagnostic statistics are favourable. As shown in Table 4, the co-integration test of Kao is valid and the Hausman tests do not reject the long-term homogeneity of coefficients, confirming the choice of PMG estimator. Our results support the prediction that the responsiveness of life, non-life, and total insurance activity to banking credit depends on the level of globalization. Indeed, the results show the direct effect of banking credit and indirect effect between banking credit and globalization are significantly positive at the significance level of 5%. Thus, the interaction coefficients between banking credit density and globalization are 0.00062, 0.00038, and 0.00032, respectively, for life insurance density, non-life insurance density, and total insurance density (see Table 4). This suggests that the positive effect of real banking credit density on development of life, non-life, and insurance activity is more pronounced for countries with high level of globalization. Our empirical evidence suggests that globalization contributes materially to banking credit density impact on insurance activities for SSA. These results are consistent with the findings of Chang and Lee (2012) and Lee et al. (2013a, b), which have found that globalization has a positive effect on the development of insurance activity.

Table 4 Conditional effects of banking activity on insurance development
Full size table

Regarding the control variables, it is clear that the magnitude and the signs of the coefficient estimates are consistent with those of our benchmark regressions in Table 3. Overall, logarithm of GDP per capita and trade openness are positively associated with life, non-life, and total insurance density. Inflation has a negative effect on life insurance density, while young dependency ratio and life expectancy have a positive effect on development of non-life and total insurance activity.

5 Conclusions and implications

The purpose of this study has been to investigate the role of globalization on the bank–insurance relationship for 20 countries of SSA over the period from 1990 to 2017. Thereby, we have used Pooled Mean Group estimator of Pesaran et al. (1999), which considers long-term homogeneity in the behaviour of the insurers across countries while allowing for short-term heterogeneous shocks. On the one hand, the results have shown that the real banking credit density has a positive effect on life insurance and total insurance activity and, on the other hand, this effect of banking credit is accentuated in the countries with more globalization.

As policy implications, economic policies of banking sector development should be used to boost the development of insurance activity services for SSA countries. This is particularly the case in countries which are more integrated in globalization.