Causal effect of mergers and acquisitions on EU bank productivity

Abstract

This paper examines the causal effect of mergers and acquisitions (M&A) on bank productivity (Q) in 23 European Union countries and the short- and long-term relationship among fixed assets (k1), liquid assets (k2), and labour (L) over the period 1990–2013 for a sample of 156 commercial banks, of which 60 entities have acquired at least one other entity. Granger causality tests on our results reveal unidirectional causality from liquid assets to fixed assets. However, the causality between K2 and L is unobservable, and the linkage between fixed assets and labour is bidirectional. The error correction term (ECT) is negative and statistically significant for all models, which denotes the presence of bidirectional relationship among all selected variables and long-term unidirectional causality from mergers and acquisitions to bank productivity. Our long-term dynamic panel estimates indicate that the strategic fit of mergers and acquisitions has the potential to create long-term productivity improvement.

1 Introduction

Over the last two decades, bank consolidation has been a frequent event in the financial sector in developing and developed countries, particularly in the European Union (Pozzolo 2009). Considering the subprime crisis, which negatively affected many advanced and emerging economies, particularly in the banking sector of the European Union, various solutions, such as mergers and acquisitions (M&A), have appealed to these countries. M&A represent external growth and approximately 80% of global foreign direct investment (FDI) flows (Klimek 2014). M&A as a form of banking integration in the EU are one of many strategies of external growth; other strategies include trade agreements, conventions, and cooperation. Furthermore, it is recognized that productivity gains are primarily influenced by external growth, such as mergers and acquisitions (M&A). However, none of these choices is considered as an ideal substitute in an emergency. Thus, many concerns have been raised, including the maximum level of cooperation. But the only consensus is that the larger the bank, the greater the need for cooperation. Nevertheless, the cooperation among commercial banks in the European Union is justified by efficiency gains in terms of profit or profitability. Yet, bank consolidation is not compatible in terms of cost for the financial sector of these consolidated banks. This paper examines the causal effect of mergers and acquisitions (M&A) on bank productivity in 23 European Union nations and the short- and long-term relationships between fixed assets, liquid assets, and labour for a sample of 156 commercial banks, of which 60 entities have undergone at least one acquisition or merger.

This survey is the first interpretive analysis of causality between mergers and acquisitions and bank productivity in the EU. In addition, the analysis examines the causal links between productivity, liquid assets, fixed assets, and labour. The empirical results reveal that the strategic fit of mergers and acquisitions has the potential to create long-term productivity improvement over the period of study.

This study has the following structure. Section 2 tests the short- and long-term effects of mergers and acquisitions on EU bank productivity. Section 3 describes the econometric methods used. Section 4 discusses the results of empirical testing. Finally, Sect. 5 presents the principal conclusions and offers recommendations.

2 Literature review

Among the desired effects of mergers and acquisitions (M&A), we can confidently cite the research on economies of scale. Such economies involve reducing the average cost and surveying the market share of different entities. Empirical studies have focused on this topic, e.g. Cavallo and Rossi (2001) and Vannet (1994), who found economies of scale in the banking sector for the post-merger and acquisition (M&A) period.

However, the research of Berger and Mester (1997), Allen and Rai (1996), and Altunbas and Molyneux (1996) includes a heterogeneous sample of banks from a variety of countries, such as the US and European Union members. Using panel data, Barth et al. (2004) also support the presence of economies of scale for acquired US banks and take into account the strong regulation in the banking sector.

In addition, it should be noted that several studies that have concentrated exclusively on the effects of mergers and acquisitions (M&A) on bank productivity find controversial results compared to M&A studies that focus on the effects on or effectiveness, assets, or economies of scale of the newly merged entity for a variety of countries. Generally, most research has demonstrated that mergers and acquisitions (M&A) generate productivity earnings. In fact, on the one hand, increases in bank size can occur during a merger and acquisition (M&A), and on the other hand, the technological gains that are obtained can increase production factors. In addition, the new strategies that will be maintained by the new managers can result in better allocation of economic resources (efficiency X) and optimize the costs of banks in times of crisis. Finally, Nurboja and Košak (2017) discuss cost efficiency in EU and non-European countries. Their findings show that institutional adjustments outside EU countries should continue to abide by the same EU standards because EU banking systems tend to dominate in terms of measured cost efficiency. In emerging countries, empirical results indicate that merger and acquisition (M&A) efficiency gains are generally weak except when implemented separately (Du and Sim 2016). Because of strong international competition, the challenge for transition countries is to pursue bank mergers and acquisitions (M&A) and public sector privatization as the only solution. Therefore, the banking markets in emerging economies are heavily dominated by foreign capital (Bonin and Wachtel 2003).

Several other studies examine efficiency gains in productivity in the EU after deregulation, particularly for the period 1980 to 1990, in anticipation of the level market (Brissimis et al. 2008). The empirical results reveal that EU banks realized average productivity gains after this deregulation, which occurs towards the end of the reform process for countries that become European Union members. Lichtenberg (1992) concluded that mergers and acquisitions (M&A) could improve a company’s effectiveness following a takeover. In fact, the process used is to argue for the improvement of the total productivity factor for the period of 14 years (7 years before and 7 years after) following a takeover in the bank productivity sector. The findings show that in the pre-merger period, the target framework has a total productivity factor considerably less than other firms. In addition, in the post-merger period, the gap decreases progressively over time. After 7 years of mergers and acquisitions (M&A), the difference between the productivity of acquired entities and non-acquired entities is more important. This gain in productivity is partly due to new management strategies to reorganize the newly merged entities (e.g. a decrease in total occupation, the new organism of economic resources). Conyon et al. (2002) tested the impact of mergers and acquisitions by foreign banks on the bank productivity and wages of consolidating banks in the UK for the period 1989–1994. These authors conclude that these mergers generated a positive and significant effect on wages (i.e. a 3.4% increase) and increased productivity 13%.

Haynes and Thompson (1999) reported the results of an empirical survey of the impact of mergers and acquisitions (M&A) on productivity by using an augmented production function approach covering the period 1981 to 1993. These authors argue for a positive impact of mergers on productivity using financial intermediation activities. They note that the merged gains tend to increase gradually in the post-merger and acquisition period, during which significant cost minimization is observed.

Rezitis (2008) discusses the effect of acquisition activity on output and Greek bank productivity. The empirical results are comparatively contradictory with respect to the theoretical hypothesis. In fact, the author reveals that the effects of M&A on Greek banks are relatively negative with respect to technical output and productivity. It is argued that the reduction in total productivity for the merged bank is due to two main factors: first, the technical short comings of the merged bank increase in the post-merger period; second, economies of scale are lost.

Oberhofer and Pfaffermayr (2013) confirmed evidence of a significant positive impact of acquisitions on employment at acquired businesses. This study examines the post-acquisition growth of acquired businesses and employment and concludes that the acquired targets increase their employment growth rate after the transaction, which for the author is evidence of efficiency gains.

Vennet (1996) examines the effect of mergers and acquisitions (M&A) on the performance of banking entities. His empirical research examined 422 national institutions and 70 multinationals spanning the period 1988 to 1993. The study produced two fundamental results. First, domestic mergers between entities of identical size significantly increased the performance of the merged banks. Second, these mergers and acquisitions (M&A) generated savings for national and multinational firms in the post-merger period. Furthermore, Shams and Gunasekarage (2019) examined the acquisition deals in Australia between public and private acquires firms. The empirical dealings show that public target improved performance in the long run when acquiring a significant holding stake in target firms.

Toumi et al. (2016) examine the dynamic effects of mergers and acquisitions (M&A) on the performance of credit institutions in the EU for the period 2005 to 2013. Their empirical findings reveal that time has negative effects on efficiency gains. However, the composite effects of dummy variables of mergers and acquisitions (M&A) over time generated a positive effect on bank performance. In the case of the EU, Ayadi et al. (2013) analyse the effects of mergers and acquisitions (M&A) on productivity for the period 1996 to 2003. Their empirical results reveal a positive and significant effect of M&A on consolidating banks. Amewu and Alagidede (2018) examine the relation between the stockholder dividends and the announcement of mergers and acquisitions of African banks. Their empirical findings demonstrate a positive relationship between bank productivity and merger and acquisition (M&A) notification. In addition, Alarco (2018) examines the effect of merger and acquisition on production in Latin America for the period 1990–2014. Using an economic model with a production function, the study finds that mergers and acquisitions have the potential to create economic development in selected countries. In addition, bank mergers create added value with respect to the profitability of clients firms (Montgomery and Takahashi 2018). Montgomery and Takahashi’s findings demonstrate that client entities of Japanese banks involved in mega-mergers do not enjoy welfare growth.

2.1 Data

Our survey involves annual data for the period 2005–2013, whereby bank productivity (Q) is assessed by the sum of loans, headlines, and shares, labour (L) is represented by the number of equivalent full-time employees, fixed assets (K1) represent the value of the (non-financial) fixed assets of the commercial banks, and liquid assets (K2) are represented by deposits and shares. The data are drawn from the balance sheets of commercial banks in the European Union (Bankscope database) (Appendix: Table 11). We perform econometric analyses based on a panel of 23 European countries: Portugal, the Czech Republic, Cyprus, Denmark, Ireland, Poland, Malta, Latvia, Belgium, Hungary, Germany, Finland, Estonia, Romania, Slovakia, Sweden, Spain, Greece, Bulgaria, France, the UK, Luxembourg, and Austria.

At this stage of the analysis, it is important to emphasize that our sample offers diverse reasons for adopting the type of analysis introduced in this study. Our study considers that the sample is sufficiently homogeneous to reveal the long-term effect of mergers and acquisitions on bank productivity (and is as homogeneous as those typically considered in other mergers and acquisitions studies). In contrast, the banking entities of the European Union should increase their likelihood of success to ensure a positive long-term productivity gain. Our sample is selected from the balance sheet of each bank using the intermediation (Table 1) approach of Sealey and Lindley (1977).

Table 1 Descriptive statistics (data in logarithms)

Table 1 summarizes the common sample descriptive statistics. We can detect that the Q distribution is approximately symmetric, whereas the K1 and L distributions are highly skewed. The common means of Q is (21.81032), K1 is (16.67719), K2 is (21.11658), and L is (6.204740). In addition, the coefficient of variation (measured by the ratio: Std. Dev/Mean) for Q is (0.12005), for K1 (0.1711), for K2 (0.1168), and for L (0.3407) in the 23 European Union nations. The normality distributions for these variables of different models of the 156 selected banks in the EU are rejected under the null hypothesis, as confirmed by the Jarque–Bera test.

3 Conclusion and policy implications

This paper examined the causal effect of mergers and acquisitions (M&A) on bank productivity in 23 European Union nations (Q) and the short- and long-term relationship between fixed assets (k1), liquid assets (k2), and labour (L) over the period 1990–2013 for a sample of 156 commercial banks, of which 60 entities have acquired at least one other entity.

Our short-run Granger causality tests reveal bidirectional short-term causality between lnk1, lnk2, lnL, and lnQ, significant at 1% and 5%, and similar causality between lnk1 and lnL. Unidirectional short-term causality from liquid assets to fixed assets was significant at 1%. Bidirectional short-term causality was found between lnk1 and lnL. The short-term causality between lnk2 and lnL was unobservable. The error correction term (ECT) was negative and statistically significant for all selected models at 1%, which indicates a bidirectional relationship among all selected variables and long-term unidirectional causality from mergers and acquisitions (M&A) to bank productivity.

The FMOLS and DOLS long-run estimates support a long-term relationship between all selected variables. The total productivity of commercial banks in the European Union reveals an adjustment process for this productivity of 22%.

In the short- and long-term models, capital and labour well explain the productivity of commercial banks in the European Union, while the gain effects of mergers and acquisitions are unobservable in the short term but apparent in the long term. As we move away from the time of merger, the banks make more productivity gains. This outcome is explained by the positive and significant coefficients associated with the dummy variables (A₀, A₃, and A for FMOLS) and (A₃ and A for DOLS).

As a policy implication of our results, EU countries should encourage their foreign investment banks to increase their merger and acquisition activity. Increasing the strategic fit of the merged banks will help them reduce their dependence and promote capital stock security. In addition, mergers and acquisitions have been a frequent response in European Union countries, and good management has contributed to the success of the integration process.

4 Method

The aims of our study are to determinate the short- and long-term effects of mergers and acquisitions on productivity and to analyse the causal links among production function, liquid assets, fixed assets, and capital–labour.

In the first step, we apply different unit root tests for the series to determine the order of integration. When selected series include a unit root, the second step is to investigate the long-term relationship between all considered variables using panel cointegration tests. Finally, we study the long-term relationship and causality linkages between all variables by the appropriate dynamic approach of panel cointegration using fully modified ordinary least squares (FMOLS) and dynamic ordinary least squares (DOLS).

In our sample, we adopt fictitious variable quads that explain the event of the merger and acquisition as follows: A₀ indicates that the merger has occurred; A₁ indicates that the merger occurred 1 year previously; A₂ indicates that the merger occur 2 years previously; A₃ indicates that the merger occur 3 years or more previously; A indicates that the merger can occur at any time.

4.1 Unit root test

The existence of a unit root for selected variables is evaluated by several tests, such as the test of Levin et al. (2002), LLC, IPS, Fisher-ADF, and Fisher-PP. LLC, Fisher-ADF, and Fisher-PP assess the homogeneity of dynamic autoregressive distribution coefficients for all selected variables. However, IPS supposes a heterogeneous unit root in compliance with an alternative hypothesis.

The panel unit root tests assume all of the variables are not stationary at the level of intercept or trend (Table 2). However, all variables in the table of the panel unit root test with respect to the initial difference (Table 3) are stationary. Thus, we conclude that Q, k1, k2, and L are integrated in order one I (1). Therefore, the FMOLS and DOLS techniques are appropriate for the entire series.

Table 2 Panel unit root results: series in level

Table 3 Panel unit root test results: series in first difference

4.2 Cointegration test

The findings of the panel unit root test for productivity, liquid assets, fixed assets, and labour indicate that these four variables are integrated in the order I (1). One can observe that all selected variables are stationary with respect to first differences. Therefore, we can apply panel cointegration methods to examine the long-term relationships between Q, K1, k2, and L. The alternative of the long-term cointegration relationship is mentioned by Pedroni (1999, 2004) and Kao et al. (1999). The panel PP-statistic and panel ADF-statistic for each dimension and the group PP-statistic and group ADF-statistic are less than 1% (Table 4). In addition, according to Kao, the ADF test is less than 1% (Table 5). Thus, the cointegration procedure reveals long-term relationships between bank productivity, liquid assets, fixed assets, and labour for the European Union countries.

Table 4 Results of Pedroni panel cointegration test

Table 5 Kao et al. (1999) residual cointegration test results

4.3 Model specification

The fundamental objective of our econometric model is to analyse the causal linkage of merger and acquisition with commercial bank productivity in the European Union. Therefore, we estimate a Cobb–Douglas production purpose, where k1, k2, and L represent the input variables and Q represents the output. The principal purpose of this framing is rather simple, and it is performed to demonstrate the theoretical relationship between merger and acquisition and commercial bank productivity. The output (Q) of commercial European banks i at time t can be expressed as follows:

$$Q_{\text{it}} \; = \;{\text{tech}}L_{\text{it}}^{\alpha } K_{\text{it}}^{\beta }$$

(1)

where L and K are the different factors of production, tech is a parameter that describes the developmental level of the technology of the commercial European banks, and α and β are coefficients that denote the effect of various factors on total production. To estimate the model, it is important to linearize it in logarithmic form. Model (1) appears as follows:

$${\text{Ln}}(Q_{\text{it}} ) \; = \; {\text{Ln}}\left( {\text{tech}} \right) + \alpha {\text{Ln}}(L_{\text{it}} ) + \beta {\text{Ln}}(K_{\text{it}} )$$

(2)

A benefit of this method is that the econometric model can include the impact of technological change in the effect of mergers and acquisitions on the productivity of commercial banks in the European Union. Thus, we can observe the effect of mergers and acquisitions on productivity when banks are consolidated by inserting dummy variables (i.e. A₀, A₁, A₂, A₃ and A). Therefore, model (2) can be expressed as follows:

$${\text{Ln}}(Q_{\text{it}} ) \; = \; {\text{Ln}}\left( {\text{tech}} \right) + \alpha {\text{Ln}}(L_{\text{it}} ) + \beta {\text{Ln}}(K_{\text{it}} ) + \mathop \sum \limits_{j} \gamma_{j} {\text{merger}}j_{i,t}$$

(3)

where i denotes the bank (1; 2;…; 157), j denotes the number of years post-merger and acquisition (j = 0, 1…, 3 and more), and t denotes the year (t = 2005…, 2013). $\alpha$, $\beta$ are the parameters to be estimated and measure the influence of the variables of the model (labour and capital), and $\gamma_{j}$ measures the effect of mergers and acquisitions. To analyse the temporal effect of mergers and acquisitions on bank productivity in our sample, we adopt the following formalization:

$${\text{Ln}}(Q_{\text{it}} ) \; = \; {\text{Ln}}\left( {\text{tech}} \right) + \alpha {\text{Ln}}(L_{\text{it}} ) + \beta {\text{Ln}}(K_{\text{it}} ) + \mathop \sum \limits_{j = 0}^{j = 3} \gamma_{j} Aj_{it} + \alpha_{i} + u_{it} ,$$

(4)

where i denotes the bank (1, 2…, 157), j denotes the number of years post-merger and acquisition (j = 0, 1…, 3 and more) and t denotes the year (t = 2005… 2013). $\alpha$, $\beta$ are the parameters to be estimated and measure the influence of the variables of the model (labour and capital), $\gamma_{j}$ measures the effect of mergers and acquisitions, and Aj_it is a dummy variable that reflects the time horizon in which the dynamics of mergers and acquisitions are realized. For example, $A_{{11\left( {t = 2005} \right)}}$ indicates that in 2005 banks (i = 1) were 1 year post-merger. $\alpha_{i }$ is the bank fixed effect. Table 6 provides the number of sample banks by country and the banks that performed mergers and acquisitions.

Table 6 Number of acquisitions and number of sample banks by country

4.4 Granger causality test

Granger causality is used to analyse the causal links among variables. Engle and Granger (1987) observe that if two variables that share a unit root are integrated, a vector autoregression (VAR) on first differences will be poorly specified. In this research, lnQ, lnK1, lnK2, and lnL are integrated in order I (1) and therefore have a long-term association. To analyse this association, we adopt an empirical model with an error correction term (ECT) augmented with a lagged period. The Granger causality test is based on the following representation:

$$\Delta { \ln }Q_{it} = \alpha_{i} + \mathop \sum \limits_{j = 1}^{q} \beta_{1i} \Delta \ln K1_{it - j} + \mathop \sum \limits_{j = 1}^{q} \beta_{2i} \Delta { \ln }K2_{it - j} + \mathop \sum \limits_{j = 1}^{q} \beta_{3i} \Delta { \ln }L_{it - j} + \mathop \sum \limits_{j = 0}^{j = 3} \gamma_{j} Aj_{it} + \delta_{1i} {\text{ECT}}1_{it - 1 } + \varepsilon_{it} ,$$

(5)

with:

$${\text{ECT}}1_{it } = { \ln }Q_{it} - \widehat{{\beta_{1i} }}{ \ln }k1_{it} - \widehat{{\beta_{2i} }}{ \ln }K2_{it} - \widehat{{\beta_{3i} }}{ \ln }L_{it} - \gamma_{j} Aj_{it} ,$$

(6)

where Δ represents the first difference of the variable and q indicates the lag order automatically specified by the Schwarz information criterion (SIC) and the Akaike information criterion (AIC). The outcome of this lagged vector autoregression (VAR) indicates that all the criteria exhibit a maximum lag equal to the unit (VAR (q = 1)). The ECT is obtained from the long-term cointegration relationship (Eq. 4), $Aj_{it}$ represents the dummy variables of mergers and acquisitions, and $\varepsilon$ is the random error term.

Table 7 presents the results of a short-run Granger causality test (pairwise Granger causality tests). The results suggest bidirectional short-term causality among lnk1, lnk2, lnL, and lnQ, significant at 1% and 5%, and similar causality between lnk1 and lnL. Unidirectional short-term causality from liquid assets to fixed assets is significant at 1%. Bidirectional short-term causality is found between lnk1 and lnL. The short-term causality between lnk2 and lnL is unobservable.

Table 7 Panel pairwise granger causality tests

With respect to Eq (5), the error correction term (ECT) is negative and statistically significant for all models at 1% (Table 8), which implies a bidirectional relationship between all variables and a long-term unidirectional relationship between mergers and acquisitions (M&A) and bank productivity.

Table 8 Granger causality test results

Table 8 describes the results of short- and long-run Granger two-step causality tests between all selected variables. The total productivity of commercial banks in the European Union exhibits an adjustment process for this productivity of 22% because the values of the ECT are negative and statistically significant at 1%.

The coefficient of the dummy variable (A₀) is positive and significant at 5%. Thus, mergers and acquisitions (M&A) had a positive effect on total productivity. Thus, the merged banks experienced efficiency gains in productivity at the moment of acquisition.

However, the effects of the dummy variables (A₁, A₂, A₃, A) were all non-significant. This outcome implies that at the moment of acquisition the banks were starting to develop new strategies to restructure their new labour and capital stocks and adopting new organizational strategies and human resources management. Therefore, it would be difficult in the short term to attribute an efficiency gain in productivity only to the fact that the banks experienced mergers and acquisitions (M&A).

Regarding the size effect, which is essentially expressed by the factors of production (lnK1, lnK1, and lnL), we note that these factors are positive and significant. In fact, mergers enable banks to benefit from an increase in size because this increase means they will have more resources in terms of capital and labour. However, despite this improvement, in the short term, all production factors are unaffected.

Therefore, the scale of bank productivity cannot be revised by changing the quantity of all production factors. These findings are similar to those of Carbó and Molyneux (2009), who examined a Spanish sample over the period 1986–2000 and concluded that approximately one-third of Spanish savings banks benefited from significant cost reductions due to mergers. Our results indicate that productivity improvements are not general but heavily dependent on the identity of the merged banks.

4.5 Long term with FMOLS and DOLS

For all selected variables with the same order of integration I (1) for different models, we estimate the long-term coefficients by using the fully modified ordinary least squares (FMOLSs) and dynamic ordinary least squares (DOLSs). For the panel data, FMOLS was developed by Pedroni (2001, 2004), while the DOLS approach was developed by Kao and Chiang (2001) and Mark and Sul (2003). These two techniques are used in the event of a unique long-term cointegration vector. The results of long-run estimates using these two techniques are reported in Tables 9 and 10. The coefficients of the long-term elasticities are approximately similar according to the two techniques.

Table 9 Long-run estimates with FMOLS

Table 10 Long-run estimates with DOLS

The long-term elasticity of productivity with respect to capital stock is on average 0.196 and higher than the short-term elasticity (0.191). However, the case of the labour stock is different. Here, the long-term productivity elasticity is on average 0.28 and lower than the short-term elasticity (0.35). Thus, we can conclude that in the short- and long-term models, capital and labour well explain the productivity of commercial banks in the European Union, while the gain effect of mergers and acquisitions is unobservable in the short term. Therefore, short-term effects (i.e. A₁: 1 year after the merger and acquisition; A₂: 2 years after the merger and acquisition) are negative and statistically insignificant.

As we progress away from the time of acquisition or merger, the banks make more productivity gains. This outcome is explained by the positive and significant coefficients associated with the dummy variables (i.e. A₀, A₃, and A for FMOLS and A₃ and A for the DOLS). Consequently, we can state fairly confidently that these mergers and acquisitions do not generate dynamic efficiency (expressed in terms of productivity gains) until the third year post-merger and acquisition. We can also note that as the merged or integrated banks progress from the time of acquisition they become more productive. This statement is supported by the fact that the coefficients associated with capital and labour 3 years post-merger are relatively higher than those associated with previous years.

Based on our results, we can assume that mergers and acquisitions create banks’ productivity improvements in the EU. As previously mentioned, one reason for this result could be related to the resolution of the problems that the banking entities face during the long-term integration of the culture of the merged entities and organizational issues. These findings are similar to those of Amel et al. (2004).

Finally, the deregulation process in the banking industry that has occurred in most developed countries, particularly in the European Union, with the subsequent increase in the level of competition, forced banking entities to react to a new competitive scenario. Mergers and acquisitions were a frequent response in many European countries, and good management of the integration process and the consolidating banks clearly contributed to the success of mergers and acquisitions. In fact, problems related to the integration process may be more similar in a sample with a high level of homogeneity (as in this paper) than in heterogeneous samples. Nagano and Ushijima (2018) examined the effect of the deregulation process on an interregional bank branch in Japan over the period 2000–2012. Their empirical findings show that geographical distance increases the probability of interregional branch closure.

Appendix

See Table 11.

Table 11 Data from Bankscope databases