• No results found

The influence of consolidation and internationalization on systemic risk in the financial sector

N/A
N/A
Protected

Academic year: 2021

Share "The influence of consolidation and internationalization on systemic risk in the financial sector"

Copied!
39
0
0

Loading.... (view fulltext now)

Full text

(1)

The influence of consolidation and

internationalization on systemic risk in the

financial sector

Abstract

This paper analyses the impact of banking mergers on systemic risk, with in particular if

internationalization prior to acquisition increases systemic risk. By using the marginal expected shortfall methodology for an international sample of mergers, a significant increase in systemic risk is found as a result of mergers in the financial sector. Moreover, if a bank is operating internationally prior to acquisition, this increases systemic risk. Additionally, there is evidence of a too-big-to-fail motive for relatively smaller banks to use mergers to become systemically important. The results confirm that consolidation in the financial sector increases fragility of the financial system.

Keywords: Marginal Expected Shortfall, Internationalization, Mergers and Acquisitions, Banking, Consolidation

Author: Rinke Bakker

University: Rijksuniversiteit Groningen Faculty: Faculty of Economics and Business

Degree program: MSc. International Financial Management Thesis Supervisor: Martien Lamers

Second Assessor: Egle Karmaziene Date submitted: 12/01/18

(2)

2

1. Introduction

The financial crisis that began in August 2007 reached systemic levels with the collapse of Lehman Brothers in September 2008. Systemic risk is now one of the key items for regulators on a

macroprudential level, as the aftermath of the financial crisis demonstrated there were severe financial instabilities that had a significant impact on the real economy (Berger, Molyneux and Wilson, 2014). As each bank has different strategies with respect to risk-taking and how they react to regulations (Laeven and Levine, 2009), it is important to study the many factors that influence systemic risk. One of the main concerns is that consolidation in the financial sector has resulted in a few systemically important banks (2014). The main method in which banks consolidate is through mergers and acquisitions (M&A) (hereafter used interchangeably). However, consolidation has not only been domestic, but also cross-border. In this case the regulatory framework in which banks operate changes, which could allow banks to shift risks to countries with less stringent regulations (Ongena, Popov and Udell, 2013).

Research on systemic risk has gained much traction, with various new methods for assessing systemic risk (Acharya, Pedersen, Philippon and Richardson, 2010; Brownlees and Engle, 2012). However, with these methods, there is no consensus in the literature about whether mergers result in higher levels of systemic risk (Karolyi and Taboada, 2015; Vallascas and Hagendorff, 2011; Weiss, Neumann and Bostandzic, 2014). Moreover, current research does not go in-depth on the influence of internationalization on systemic risk, as international banks could potentially make use of

regulatory arbitrage shifting risk to countries with less stringent regulations (Ongena, Popov and Udell, 2013).

To research how internationalization plays a role in systemic risk change as a result of mergers, this paper extends the work of Weiss, Neumann and Bostandzic (2014). They researched whether bank mergers increase the systemic risk of financial firms, and were concerned with mainly American and European banks, but did not explore how internationalization plays a role, nor why the international regulatory operating network is important. The main research question is how consolidation and internationalization of banks influences systemic risk.

The work of Weiss, Neumann and Bostandzic (2014) provides a strong framework for comparing results with an updated, and a more international sample, and then extending those results by going further in-depth on the internationalization factor. The strictness of regulations in their home country could influence a bank’s decision to shift risk to countries with less stringent regulations, or it might use a merger to become ‘too-big-to-fail’ to gain access to public safety nets that can cover downside risks (Karolyi and Taboada, 2015; Ongena, Popov and Udell, 2013). Internationalization in

(3)

3 this respect is of key importance, because cross-border acquisitions allow banks to make use of regulatory arbitrage, while an already international bank can make use of its extensive network to shift risks away from the regulator. This prediction is in line with research that demonstrates that banks use capital flows to shift funds to countries with less stringent regulations (Houston, Lin and Ma, 2012). This leads to the following questions: How does internationalization influence systemic risk of an acquiring bank as a result of acquisitions? Is there any interaction between

internationalization prior to acquisition and whether the new acquisition is domestic or cross-border? These questions should help regulators develop policies to govern mergers within the financial sector. Moreover, this paper adds to the literature that assesses how mergers influence systemic risk, as there is no clear consensus (Karolyi and Taboada, 2015).

A sample of 528 mergers worldwide announced between 1998 and 2016 was used. Following the popular methodology of Acharya, Pedersen, Philippon and Richardson (2010) on how to assess systemic risk, a significant impact of mergers on systemic risk is found. Moreover, interaction between internationalization and cross-border acquisitions, when controlling for factors such as size and leverage, significantly increases the impact of a bank merger on systemic risk. In other words, an international bank that acquires another bank cross-border will on average become more

systemically relevant, whereas a domestic bank that acquires cross-border does not experience an increase in systemic risk. Thus, the already established network of a bank is key to how systemic risk develops.

The implications of these results are that regulators should consider that mergers in the banking sector generally increase systemic risk, and thus increase the fragility of the financial sector. Cross-border mergers only increase systemic risk when the bank is international prior to acquisition, but when a domestic firm acquires cross-border, there is a reduction in systemic risk. There is evidence of a too-big-to-fail incentive for banks to become systemically important and make use of implicit government subsidies that could result in excessive risk-taking. For the management of banks, this is also useful to consider, as systemic risk is measured by the shortfall of equity, thus their decisions influence firm value during systemic shocks.

This paper is structured as follows. First, the literature on systemic risk and internationalization is assessed. A theoretical model on diversification versus diversity of banks is used to explore how diversification influences systemic risk (Wagner, 2010). The model is applied in the case of

internationalization, which is sometimes called geographic diversification. This is supplemented by theories on regulatory arbitrage, international capital flows and contagion through internal markets. Next, we explain the method of Acharya, Pedersen, Philippon and Richardson (2010) regarding how

(4)

4 to calculate systemic risk, and how this is implemented by Weiss, Neumann and Bostandzic (2014). By following Weiss, Neumann and Bostandzic (2014)in structure and method, the results can be compared, and this provides a strong foundation for exploring internationalization. The section is concluded with hypotheses based on the literature. Then, the sample is explored and the criteria for the final selection used is examined. This is followed by the results based on sub-samples and multivariate regressions. The findings are explored and linked back to the hypotheses and literature to investigate what these results imply. In the final section, the conclusion is followed by the limitations of the research, and directions for further research are explored.

2. Literature review

This section first explores how systemic risk and mergers are related, which is done by using the theoretical work of Wagner (2010) on how the diversification strategies of banks reduce diversity and increase systemic risk. This is caused by an increase in the probability of joint failure of both banks. Based on these insights, Wagner’s (2010) model is applied to mergers of banks and how these influence systemic risk. Next, the various ways in which regulations impact systemic risk are

explored, and similar research on systemic risk and mergers is summarized. The paper of Weiss, Neumann and Bostandzic (2014) is assessed in detail, as their structure forms the basis of this paper. Following this, the link between systemic risk changes due to mergers and internationalization is made. Theories on internal capital flows and cross-border regulations are used to identify in which ways the internationalization of a bank prior to acquisition influences systemic risk. The distinction is made between cross-border acquisitions and domestic acquisitions. Finally, the section concludes with three hypotheses, based on the work of Weiss, Neumann and Bostandzic (2014),

internationalization and the interaction between cross-border acquisitions and internationalization.

2.1 Systemic risk and mergers

Wagner (2010) provides a theoretical model of how diversity and the diversification of banks influence systemic risk. He argues that it is taken for granted that the diversification of individual banks is good for the financial system. This view is based on the idea that an individual bank can diversify its portfolio to reduce individual risk, and in turn lower the fragility of the financial system. However, if banks use similar diversification strategies, this reduces the diversity of banks. The result is that the probability of multiple banks failing at the same time increases; the diversification of these banks will not lower the probability of them failing simultaneously. If the banks’ strategies were not similar, one might barely survive while others fail. The conclusion is that diversification lowers the idiosyncratic risk of individual banks, but can only increase the probability of a systemic crisis. The financial system becomes fragile when these individual risks are simply reallocated across

(5)

5 the system, rather than eliminated (Weiss, Neumann and Bostandzic, 2014). Moreover, if the banks diversify their services, but offer these to the same clients, this is not true diversification, and increases the impact of shocks in these businesses (Berger, Molyneux and Wilson, 2014).

When analyzing a merger of banks using the model of Wagner (2010), this analysis involves two equally diversified (similar) banks. This is based on the theoretical insight of Markowitz (1991) that all investors will hold the same minimum variance portfolio with a riskless asset, with the

components based on individual risk appetite. This insight can be applied to banks also; if they all diversify in a similar manner, they will eventually hold the same portfolio of assets, which leads to zero diversity (Markowitz, 1991; Wagner, 2010). The outcome is that bank mergers can only lead to an increase in systemic risk.

The process of banks becoming more similar began with deregulation in the financial sector, which has allowed the banking sector to integrate financial services, resulting in financial conglomerates. This process in Europe began with the Second Banking Directive of 1989, and with the Gramm-Leach-Bliley Act of 1999 in the United States (US) (De Jonghe, 2010). Large conglomerates were formed through M&A, which led to more diversified banks on an individual level, but also to more similarity among them (Berger, Molyneux and Wilson, 2014.; De Nicolo and Kwast, 2002). Thus, based on the insights of Wagner (2010), these mergers increased systemic risk.

A nuance to the theory of Wagner (2010) is brought by Slijkerman, Schoenmaker and de Vries (2013), who demonstrate that, during systemic shocks, the diversification of banks into insurance activities can be useful, as insurance has different business models, and merging these activities can result in fewer liquidity problems during a shock, without influencing systemic risk.

Another way mergers can influence systemic risk is when the goal is to become too-big-to-fail and make use of implicit government guarantees to take on additional risk, with the public safety net covering the downside risks (Karolyi and Taboada, 2015). This moral hazard problem gained a great deal of attention in the aftermath of the financial crisis, when governments used public funds to bail-out banks. Carbo‐Valverde, Kane and Rodriguez‐Fernandez (2012) suggest that most mergers in the European Union (EU) between 1993 and 2004 were driven by acquirers’ incentive to make use of the safety nets.

When regulation in the home country of the bank is considered too stringent, banks might engage in regulatory arbitrage by shifting operations to countries with less stringent financial regulations (Ongena, Popov and Udell, 2013). When operating in an international network, it becomes more difficult and more expensive for both the regulators and the banks to monitor such activities

(6)

6 (Berger, Molyneux and Wilson, 2014; Carbo‐Valverde, Kane and Rodriguez‐Fernandez, 2012).

Moreover, banks might be able to engage in activities that not permitted in their home country, which in turn increase systemic risk due to higher levels of risk involved.

There are some arguments that indicate that mergers decrease systemic risk, meaning consolidation in the financial sector has led to a more stable system. First, in a concentrated banking system, high market power can lead to higher profits. These profits can shield banks from shocks. Moreover, this can lead to a reduction in the incentives to take excessive risks (Beck, Demirguc-Kunt and Levine, 2006; Hellmann, Murdock and Stiglitz, 2000), which could lower systemic risk. Furthermore, a merger can create an internal money market that increases reserves against shocks (Carletti, Hartmann and Spagnolo, 2007). Second, with fewer banks, it is easier to monitor those in a concentrated banking system as opposed to many diverse banks. Carletti, Hartmann and Spagnolo (2007) argue that supervision in this situation might be more effective, which could result in fewer or less contagious systemic crises. Under the assumption of ceteris paribus, there will be larger banks in a more concentrated system, which in turn are better diversified and lead to a less fragile financial system. The problem is that, according to Wagner’s (2010) model, this can lead to lower

idiosyncratic risk, but does not lower the probability of multiple banks failing at the same time, thus resulting in higher systemic risk.

Most papers seem to indicate that systemic risk increases following an acquisition. Weiss, Neumann and Bostandzic (2014), in their international sample of 420 acquisitions, find that an acquisition results in a significant increase in systemic risk. However, Karolyi and Taboada (2015) state that banks do make use of regulatory arbitrage, but find no significant increase in systemic risk as a result of acquisitions. Both studies are based on marginal expected shortfall (MES), which indicates the expected drop in equity of the firm based on a shock in the market. As this is based on market data, MES assesses systemic risk based on the expectations of the market. Vallascas and Hagendorff (2011) analyzed the impact of 134 European bank mergers on default risk, determining that, on average, bank mergers do not result in an increase in default risk. However, large deals and a sub-sample of the least-risky banks do demonstrate a higher default risk after acquisition. Vallascas and Hagendorff conclude that their sample indicates there is no stabilizing effect when considering European bank mergers. A study by De Nicolo and Kwast (2002) focussed on consolidation in the financial sector between 1988 and 1999, by analysing stock return correlations. They found a positive trend in stock correlation during a period of significant consolidation. This seems to indicate systemic risk during this period increases, which could be caused by mergers, as acquistions mainly drove the consolidation.

(7)

7 The structure and research design of Weiss, Neumann and Bostandzic (2014) as the basis for this paper. They use both MES and a lower tail-dependency measure to assess changes in systemic risk as the result of a merger. Their tail-dependency method was developed by themselves; however, this method only confirms that mergers lead to an increase in systemic risk. Management hubris, the presence of a permanent deposit insurance fund, and government-owned banks in the home

country increase the destabilizing effect of bank mergers. They do consider cross-border mergers, which they find are only significant for US banks, but do not explore internationalization prior to the merger. As regulatory arbitrage could play a significant role, this paper extends their original work to include internationalization.

2.2 Internationalization

This section initially focuses on how the internationalization of a bank, prior to acquisition, influences systemic risk. Next, it is assessed whether there is any difference between an

international bank acquiring domestically and cross-border. Mergers are not the only way for a bank to internationalize, other methods are greenfield projects, which means starting operations from scratch, and joint-ventures. For this paper, the focus is on mergers, as they result in a relatively big change in a small timeframe, that makes it possible to measure the change in systemic risk. One key aspect in this paper is that while cross-border mergers result in an international bank,

internationalization is defined as a bank already operating internationally prior to acquisition.

The link between internationalization (geographic diversification) and systemic risk has not yet been researched extensively; however, the impact of internationalization of banks on their total risk profile has been researched. Berger, El Ghoul, Guedhami and Roman (2016) find that

internationalization increases total risk, due to risks caused operating in foreign markets outweigh diversification benefits of these foreign operation. Banks internationalize to be able to diversify, enter new profitable countries with low competition and less stringent regulations, and make use of economies of scale (Nicolo, Bartholomew, Zaman and Zephirin, 2004). Moreover, banks might be able to gain access to a public safety net if they become too-big-to-fail, as liquidation due to bank failure can become too costly for society.

A more internationalized firm has more options to shift capital to key locations, which can be used in periods of stress to prevent liquidity problems, thus potentially lowering systemic risk. Banks will use deleveraging strategies abroad when a crisis hits; they reduce cross-border credit and entrench in their home country (Forbes and Warnock, 2012). Thus, if an international bank merges domestically or cross-border, it can lower systemic risk, as the bank becomes part of the new operation and gains access to the network of (foreign) funds, which can be used to entrench. An example of this is when

(8)

8 firms reduced cross-border lending following the collapse of Lehman Brothers (De Haas and Van Horen, 2012), indicating that they shifted capital to key locations.

Some nuance to this hypothesis is brought by Ivashina and Scharfstein (2010), who demonstrate that, depending on the stability of the deposits of the bank and the reliance on short-term debt, banks will cut their lending accordingly. Moreover, the withdrawal of cross-border lending depends on the closeness to the foreign country in terms of geography and lending relationships, as the nearer they are, the better the banks are able to judge risks (De Haas and Van Horen, 2012). The conclusion is that banks can use their geographic diversification to absorb shocks and retreat to their home country in times of stress.

However, if a large portion of banks uses similar internationalization strategies, the systemic risk can increase, according to Wagner’s (2010) model. A logical argument is that banks can transmit shocks in foreign countries across borders to their home country. First, if multiple banks use similar internationalization strategies, they increase systemic risk. Second, contagion through their lending portfolio channels can transmit the shock to the home country (Forbes and Warnock, 2012). Thus, international firms merging can lead to an increase in systemic risk as potential contagion increases.

Another way internationalization can impact systemic risk is the degree to which the acquirer can make use of regulatory arbitrage within their network. While international regulations have been converging over the years, and regulators understand the need to work together, the crisis of 2008 illustrated that problems remain. Acharya (2003) explains that the convergence of standards, such as capital requirements, can lead to a ‘race to the bottom’ if there is no consistency across borders concerning resolutions. The logic is that international banks will take more risks in countries with less stringent regulatory regimes, which forces their competitors to respond in similar fashion and the more country with more stringent regulations losing (large parts of) important banks. This can result in regulators converging on the lowest common denominator of regulations. Houston, Lin and Ma (2012) make the same conclusion. Moreover, they find strong evidence that banks have transferred capital to countries with fewer regulations. This can even be the case within the US and the EU, which have super-national regulator coordinating. In the US, regulators can implement identical rules in different ways, due to differences in their institutional design and incentives (Agarwal, Lucca, Seru and Trebbi, 2014). State regulators are less likely to enforce stringent regulations due to

concerns with the local economy, compared with federal regulators.

One potential way that internationalization can lower systemic risk is that the diversification strategies of banks are sufficiently diverse so that individual diversification can lead to lower probability to default, without increasing systemic risk. Deng and Elyasiani (2008) investigate the

(9)

9 internationalization of bank holding companies and conclude that, based on location, level of activity and geographical distance, internationalization results in lower risk and higher value, but that more remote areas result in smaller risk reductions. Fang and van Lelyveld (2014) find similar results, using a correlation matrix approach for 49 large international banking groups, of lower credit risk as a result of internationalization. Lower individual risk of banks could lower the joint-probability of failure if these diversification strategies do not result in greater similarity among banks.

Specifically, cross-border acquisitions by international banks allow them to shift more risks to less strict environments. Research has shown that cross-border acquisitions and capital flows of banks are more likely to originate from countries that have stronger regulations (Houston, Lin and Ma, 2012; Karolyi and Taboada, 2015). Thus, this effect seems to be stronger for countries

headquartered in developed countries.

2.3 Hypotheses

Based on the literature review, three hypotheses are constructed in relation to changes in systemic risk due to acquisitions. First, the consensus in the literature following the financial crisis seems to suggest that M&A in the financial sector increase systemic risk of the acquirer. Weiss, Neumann and Bostandzic (2014) find confirmation of their concentration-fragility hypotheses (i.e. that mergers significantly contribute to systemic risk). This is in line with the theory of Wagner (2010), who argues that mergers might increase diversification, but that this can lead to greater similarity among banks, resulting in a higher probability of joint failure, thus a higher systemic risk. Therefore, the first hypothesis is that mergers significantly increases systemic risk of the acquirer. This hypothesis is tested by comparing the results of this paper with Weiss, Neumann and Bostandzic (2014), before extending their work.

Second, a link between internationalization (prior to acquisition) and systemic risk as a result of mergers is explored. The argument is that regulatory arbitrage allows international banks to shift risk to public safety nets or, on the other side of the spectrum, shift risks away from stringent regulators to countries with less stringent regulations. This is easier for banks who already have an established international network (Houston, Lin and Ma, 2012; Karolyi and Taboada, 2015). Differences in execution of regulations can even occur when different regulators are enforcing exactly the same regulations, as supervisors have different institutional designs and incentives (Agarwal, Lucca, Seru and Trebbi, 2014). Thus, the second hypothesis is that internationalization prior to acquisitions leads to a significant increase in systemic risk of banks following acquisitions.

Third, the interaction between cross-border mergers and internationalization prior to

(10)

cross-10 border, they can integrate the new target into their international structure. Houston, Lin and Ma (2012) find that banks have shifted capital flows to countries with less stringent regulations, giving merit to the regulatory arbitrage argument. Moreover, banks might transfer shocks in the foreign market to the home market through their internal capital markets (Forbes and Warnock, 2012). Therefore, the third hypothesis is that cross-border acquisitions of internationalized banks positively impact systemic risk after acquisition.

3. Methodology

This section discusses how the systemic risk contribution of an individual bank to the overall system is calculated. A similar method is used to calculate the expected shortfall of a market index, which is used to calculate the adjusted MES. Finally, the multivariate model is presented, which is used for further analysis.

By assessing the differences in MES, the differences in systemic risk that a bank contributes to the system as a result of a merger is identified. The method is identical to that of Weiss, Neumann and Bostandzic (2014), who calculate the delta of the MES based on the difference of post- and pre-MES. When controlling for various externalities, such as other mergers, changes in regulations, accounting variables and deal characteristics, the contribution of the merger that the individual bank adds to the financial system is isolated. The contribution to systemic risk is based on returns on equity, which results in how the market assesses the contribution to systemic risk.

This paper follows the theoretical method developed by Acharya, Pedersen, Philippon and Richardson (2010) on systemic expected shortfall and MES. This a well-used measure in the literature (Berger, Molyneux and Wilson, 2014; Karolyi and Taboada, 2015), and is also used by Weiss, Neumann and Bostandzic (2014). Systemic risk is used to measure the systemic fragility of an individual institution. The contribution of an individual bank is measured as the worst 5% daily equity returns. By using the market to define these days of extreme tail returns (one-sided), a period of general distress is captured. The systemic risk measure is used to identify the capital needs of a financial institution during distress in the market. These are the values greater than the variance at risk (VaR), which we set at 5%. This leads to the expected shortfall conditional to a return smaller than I, with the VaR level, as in Equation 1.

(11)

11 Based on these dates, we calculate the shortfall of company i, on date t. These dates are conditional on the return in the market being lower than I. MES is defined as the mean net return on equity during a market crash. This is shown in Equation 2, with return of firm I conditional on the worst 5% market returns. The sign flipped, as is standard notation for the shortfall.

𝑀𝐸𝑆𝑖,𝑡 = −𝐸𝑡(𝑟𝑖,𝑡+1|𝑟𝑚,𝑡+1 < 𝐼5%) (2)

This paper does not compute MES with a more advanced dynamical model to estimate the beta, which uses threshold autoregressive conditional heteroskedasticity (TARCH) and dynamic

conditional correlation models (DCC) based on the work of Brownlees and Engle (2012). This keeps the results comparable to Weiss, Neumann and Bostandzic (2014), who use a period of 180 days until 10 days prior to the merger announcement to calculate the pre-merger MES, and 10 days to 180 days after the merger completion to assess the post-merger MES. The change in MES is based on the difference between these two variables. The results are in Equation 3.

∆𝑀𝐸𝑆𝑖,𝑡5%= 𝑀𝐸𝑆𝑖,𝑡;[+11,+180]5% − 𝑀𝐸𝑆𝑖,𝑡;[−11,−180]5% (3)

To calculate the change in expected shortfall of the market for a specific region, used to compute the market-adjusted MES, we calculate the expected shortfall of the market during distress, with VaR at 5%, as in Equation 4.

𝐸𝑆𝑚 = 𝐸𝑚(𝑟𝑚+1|𝑟𝑚+1 < 𝐼5%) (4)

This equation is used to calculate the change in expected shortfall of the market when comparing the post-merger period returns. Equation 6 displays, for the post-merger period, the adjusted MES, which is the shortfall of the individual bank minus the shortfall in the market.

𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑀𝐸𝑆𝑖,𝑡5%= 𝑀𝐸𝑆𝑖,𝑡;[+11,+180]5% − 𝐸𝑆𝑚,𝑡;[+11,+180]5% (5)

The main proxy used for internationalization is an international income dummy. Based on research by Berger, El Ghoul, Guedhami and Roman (2016), various factors should be able to capture internationalization. Therefore, foreign assets over total assets, foreign sales over total sales and dummy variables of both these are also explored regarding the degree internationalization and if they are international.

To test the hypotheses from Section 2.3, a multivariate regression is used based on the previously explained dependant variable – delta MES. Various models are used with different control variables to obtain robust findings.

(12)

12 ∆𝑀𝐸𝑆𝑖,𝑡5%= 𝛼 + 𝛽1𝐼𝑛𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖,𝑡−1+ 𝛽2𝐶𝑟𝑜𝑠𝑠𝑏𝑜𝑟𝑑𝑒𝑟𝑖,𝑡+ 𝛽3𝐼𝑛𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛𝑖,𝑡−1

∗ 𝐶𝑟𝑜𝑠𝑠𝑏𝑜𝑟𝑑𝑒𝑟𝑖,𝑡+ 𝛽4𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡−1+ 𝜖𝑖 (6)

4. Data and descriptive statistics

This section provides a detailed description of how the data sample and various sub-samples were acquired. Additional tables and statistics that are referred to are available in the appendix.

The initial dataset was acquired from Bureau van Dijk’s Zephyr M&A database. From this, a global sample was acquired for the following Standard Industrial Classification SIC codes: 6000 to 6162, 6712 and 6719 for both the acquirer and target. This resulted in a sample of credit and depository institutions, as we aim to investigate whether internationalization (and not diversification) and consolidation in the banking sector impact the stability of the financial system. This is of interest as shocks in the financial system can lead to damage in the real economy (Haldane and May, 2011). The selection of SIC codes was made based on this paper being concerned with the detrimental effects of bank mergers on the stability of the financial system, and how internationalization influences this.

Deals with a value lower than 10 million USD, and an acquired stake below 50%, were excluded. The reasoning for this was to only include deals that result in full active control, rather than a potential financial investment, according to IFRS9 and U.S. GAAP. Moreover, this makes the results

comparable with both Vallascas and Hagendorf (2011) and Weiss, Neumann and Bostandzic (2014).

From the initial sample of 1199 firms, only 865 had stock data available from Thomson Reuters DataStream financial data for our required pre-and post-merger intervals. Next, we excluded variables that did not have a deal value from Zephyr available, or accounting variables from Worldscope available through DataStream. This reduced our sample to 675 deals. From these, we excluded deals involving the same company acquiring another company within 360 days. This further reduced our sample to 528, which is a similar number to Weiss, Neumann and Bostandzic (2014).

Table 1 displays the acquiring and target firms per region. Most acquirers and targets are from the US, indicating a high level of domestic mergers. This is explored further in this section. The second largest region of both acquirers and targets is Western Europe; however, there is more cross-border and cross-region activity, caused by both smaller countries and by dividing Europe into a Western and Eastern part. Neither the Middle East nor Oceania have cross-region acquisitions.

(13)

13

Table 1. Distribution of mergers per region

Acquirer Target

Africa

Eastern Europe

Far East and Central Asia Middle East North America Oceania South and Central America Western Europe Sum Africa 9 1 10 Eastern Europe 14 14

Far East and Central Asia 1 28 1 2 32

Middle East 1 2 9 2 14

North America 327 2 1 330

Oceania 1 7 8

South and Central America 1 19 20

Western Europe 3 21 3 4 5 64 100

Sum 12 37 33 9 334 7 29 67 528

This table displays the distribution of acquisitions per region, based on the target region. The acquirer region is listed in the first column, while the target region is listed in the second row. This results in a total sample of 528 mergers, between 1998 and 2017. The sample was retrieved from Bureau van Dijk's M&A database, Zephyr. We kept companies with an SIC code between 6000 to 6162, 6712 and 6719. From these deals, those with a value below 10 million USD and with an acquired stake below 50% were excluded. Additionally, for these companies, stock prices and accounting variables from DataStream were required to be available. These DataStream variables came from Financial data and Worldscope from Thompson Reuters respectively

Figure 2 displays the distribution of announced and completed mergers from 1998 to 2017. There is a sharp increase in merger announcements and completions in 2000. This is likely caused by the deregulation in the 1990s (Berger, Molyneux and Wilson, 2014). This level is maintained until the financial crisis of 2008, which resulted in a 50% drop in announcements that year, and a similar drop in completed mergers in 2009. The M&A market had recovered by 2013 in our sample. From 2016 onward, there is again a sharp decrease in announcements, which is explained by the lack of return data in the post-merger period. In the full sample of Zephyr, the difference in announcements is 21 (70 in 2015 and 49 in 2016). There are no announced mergers in 2017 for the sample due to there being a minimum number of return dates present to calculate the MES.

(14)

14 Table 2 presents the descriptive statistics for all the variables, including the various

internationalization proxies and control variables. All the accounting variables are stated in USD, based on the year prior to the merger announcement. Loan loss provision, total assets and deal value are transformed into the natural logarithm. To normalize the accounting variables, they are winzorized at a 1% and 99% level.

The market index on which both the MES and, similarly, the expected shortfall of the market is calculated as a value-weighted index based on the top 500 available public financial firms ranked according to total revenue in the latest available year. Due to data limitations regarding market values, this results in indices based on 486 US firms, 489 Western European firms, 273 Eastern European firms, 234 Middle Eastern firms, 484 firms from the Far East and Central Asia, 243 from Central and South America, 181 from Africa and 113 from Oceania.

0 10 20 30 40 50 60 Announced Completed

Figure 1. Distribution of mergers per year

The number of acquisitions announced and completed are categorized per year. There is a fall in the number of announcements in 2008, followed by a drop in completed mergers. This change is reversed from 2013 onward, though there is decline in 2016.

(15)

15

Table 2. Descriptive statistics

Observations Mean Median Maximum Minimum Std. Dev.

ΔMES 528 0.21% 0.12% 10.12% -17.86% 1.97%

Acquirer and deal characteristics:

Equity ratio 528 57.166 57.860 93.781 8.223 21.590

Return on assets 528 1.439 1.310 5.482 -0.934 0.883

Non-interest income to total

income 528 0.350 0.275 1.689 0.028 0.281

Loan loss provisions (ln) 528 9.552 9.451 15.501 0.000 3.356

Market-to-book value 528 1.685 1.535 4.479 0.280 0.830

Too-big-to-fail motive 528 0.250 0.000 1.000 0.000 0.433

Pre-merger performance 528 0.075 0.078 1.871 -2.651 0.274

Relative deal value 528 29.828 14.027 918.077 0.009 55.635

Total assets (ln) 528 16.186 15.808 21.343 12.488 2.134

Deal value (ln) 528 18.521 18.257 22.884 16.165 1.554

Cross-border dummy 528 0.165 0.000 1.000 0.000 0.371

Internationalization variables

Foreign income dummy 305 0.167 0.000 1.000 0.000 0.374

Foreign asset dummy 329 0.246 0.000 1.000 0.000 0.431

Foreign assets over total assets 328 7.103 0.000 98.010 0.000 18.552

Foreign income over total income 305 5.646 0.000 155.250 -212.290 24.642

Foreign sales to total sales 350 5.321 0.000 79.600 0.000 12.734

Foreign sales dummy 350 0.289 0.000 1.000 0.000 0.454

Country control: GDP growth 528 2.547 2.596 19.592 -10.895 2.368 Political stability 528 0.367 0.509 1.610 -2.676 0.640 Rule of law 528 1.286 1.591 2.096 -1.114 0.659 Lerner index 524 0.256 0.280 0.590 -2.560 0.154 Regulatory environment: Anti-director rights index (ADRI) 497 3.574 3.000 5.000 0.000 1.053 Deposit insurance 515 349.303 262.000 8798.990 0.000 643.629

Foreign deposit coverage dummy 526 0.844 1.000 1.000 0.000 0.363

Capital regulatory index 515 5.900 6.000 9.000 0.000 1.916

Government-owned banks

dummy 486 0.214 0.000 1.000 0.000 0.411

Independence supervisory power 484 1.694 2.000 3.000 0.000 0.685

Official supervisory power 523 11.669 13.000 14.500 0.000 3.512

Moral hazard index post-crisis 505 3.037 4.618 4.618 -11.862 2.401

The table presents the descriptive statistics of the sample. An in-depth list describing all variables and sources is available in the appendix. Loan loss prevention, total assets and deal value are based on the natural logarithm of their original value. All monetary values are presented in USD. The accounting and internationalization data is based on the last financial year before the announcement of the merger, whereas country controls are from the year of acquisition. Data from the regulatory environment is based on the last available year survey data was present.

(16)

16

4.1 Control variables

To control for various factors that might influence systemic risk (as assessed by capital markets) of banks other than mergers, we control for a variety of firm-specific factors and country factors. Based on the initial sub-sample analysis, we conclude that there are many factors influencing systemic risk, such as concentration and profitability. Moreover, there is an indication that internationalization can influence systemic risk in a similar manner to a cross-border merger, as a domestic firm will also become international after the acquisition. Therefore, this is further explored in multivariate regressions with different control variables to test the hypotheses.

The first control variable is the equity ratio. This factor is commonly expected to have a positive relation when assessing systemic risk (Huang, Zhou and Zhu, 2012), as highly leveraged firms with minimal reserves can face liquidity problems during a systemic shock, but one could argue that low leverage forces banks to increase profitability to survive competition. Another argument is banks with low leverage are preparing a takeover, which results in assuming or renegotiating the debt of the target, resulting in a more normal leverage level (Weiss, Neumann and Bostandzic, 2014). We control for profitability by using return on assets, as seen in Table 4, which can impact systemic risk. For example, low profitability could push banks to become more aggressive in risk-taking. Non-interest income to total income is to consider the impact of more volatile activities, such as investment banking and trading. Brunnermeier, Dong and Palia (2012) state that these non-core items contribute more to systemic risk compared with traditional items, such as liquidity creation. The market-to-book ratio is included to proxy for management hubris, which could lead to excessive risk-taking (Vallascas and Hagendorff, 2011). Pre-merger performance is also used to control for management hubris (Weiss, Neumann and Bostandzic, 2014), calculated as the buy-and-hold return during the pre-merger period. Too-big-to-fail is included as a variable to account for mergers that are motivated by an incentive to become systemically important, and which make use of

government safety nets for downside risk (Karolyi and Taboada, 2015). Too-big-to-fail is defined as a dummy being equal to 1 if banks are part of the lowest quartile of delta MES. Weiss, Neumann and Bostandzic (2014) also include firms that are in the lowest quartile when calculating Merton’s probability to default; however, this is excluded in this paper due to time limitations. Relative deal value and deal value are included to consider the impact of a merger on the newly consolidated bank. The higher the relative deal value and deal value, the more likely the bank will become systemically important. Total assets are included to proxy for a bank already having a too-big-to-fail status. Hovakimian, Kane and Laeven (2012) believe that size and leverage are key drivers of systemic risk.

(17)

17 To control for differences between countries, macro-level factors and regulatory factors are

included. Gross domestic product (GDP) growth, political stability and rule of law are included to control for the operating environment (Weiss, Neumann and Bostandzic, 2014). Differences in the Lerner index are used to assess concentration and competition, as these can influence risk exposure through factors such as market power (Berger, Klapper and Turk-Ariss, 2009). An indication of market power influencing systemic risk is also observed in the sub-sample analysis; however, there are no control variables used.

The regulatory environment in which a bank operates is also important. The anti-director rights index (ADRI) by Porta, Lopez‐de‐Silanes, Shleifer and Vishny (1997), which was revised by Spamann (2009), has been used in various publications to assess shareholder rights. A lack of shareholder rights could allow management to avoid disclosure, which lowers the effectiveness of market discipline and allows management to take excessive risks (Hu and Black, 2008). Data on deposit insurance coverage (i.e. whether foreign deposits are covered, and whether a permanent fund exists) were gathered from the database of Beck, Demirguc-Kunt and Levine (2009). These factors can control for (implicit) government guarantees. If banks have these safety nets available, they can be inclined to take more risks (Karolyi and Taboada, 2015). Finally, to consider the stringency of the regulatory environment in the acquirer’s home market, data are used from Barth, Caprio and Levine (2008). The capital regulatory index, a dummy of government-owned banks, the independence of the supervisory authority, and the official supervisory power are all used to control for differences in the regulatory environment. In this, we follow the approach of (Weiss, Neumann and Bostandzic, 2014) for selection of variables. In general, the expectation is the more stringent the regulations, the lower systemic risk if banks are unable to shift risks to other countries. However, more stringent regulations could also be effective in lowering systemic risk. The existence of government owned banks can be used to control for willingness to bail-out banks. The independence and power of the supervisory authority can indicate the capability of the regulator to pursue a healthy financial system, thus this is expected to lower systemic risk.

5. Results

This section first discusses the MES per region, and how it compares to the market index. This is then used to compare results with Weiss, Neumann and Bostandzic (2014), and confirm the first

hypothesis. Next, a sub-sample analysis is used to explore which factors are relevant when assessing systemic risk. These insights are used for the multivariate regressions from which we explore the second and third hypothesis. This section concludes with a robustness analysis, in which sub-samples

(18)

18 for multivariate regressions and the influence of various internationalization proxies are explored to assess whether methodological choices are driving the results.

Table 3. Changes in marginal expected shortfall

Acquirer systemic risk Expected shortfall market Market adj. expected shortfall

Acquirer region MES pre-merger MES post-merger ΔMES ES mrkt pre-merger MES mrkt post-merger ΔMES mrkt MES adj. pre-merger MES adj. post-merger ΔMES adj. North America 1.6% 1.8% 0.2%*** 2.6% 2.7% 0.1% -1.0%*** -0.9*%** 0.1% (0.002) (0.111) (0.000) (0.000) (0.159) Western Europe 2.8% 3.4% 0.6%*** 3.3% 3.6% 0.3%* -0.5%*** -0.2% 0.3%** (0.011) (0.085) (0.002) (0.157) (0.039)

Far East and

Central Asia 2.1% 1.7% -0.3% 2.6% 2.6% 0.0% -0.6%** -0.9%*** -0.3% (0.177) (0.443) (0.021) (0.000) (0.196) South and Central America 2.2% 2.2% 0.0% 2.8% 2.7% -0.1% -0.6% -0.5%* 0.1% (0.499) (0.440) (0.110) (0.099) (0.437) Eastern Europe 2.6% 2.9% 0.2% 3.3% 3.7% 0.4% -0.7%** -0.8%*** -0.1% (0.357) (0.234) (0.032) (0.006) (0.328) Middle East 0.8% 1.2% 0.5% 3.2% 2.3% -0.9% -2.4%*** -1.1%** 1.3%** (0.125) (0.103) (0.004) (0.018) (0.030) Africa 2.7% 0.5% -2.1% 3.2% 3.5% 0.2% -0.6% -2.9%*** -2.4% (0.136) (0.349) (0.379) (0.003) (0.148) Oceania 3.0% 3.6% 0.6% 3.7% 4.0% 0.3% -0.7%*** -0.3% 0.3% (0.305) (0.383) (0.008) (0.239) (0.225) Total 1.9% 2.1% 0.2%*** 2.8% 2.9% 0.1%* -0.9%*** -0.8%*** 0.1% (0.007) (0.063) (0.000) (0.000) (0.136)

*** is significant at 1%; ** is significant at 5%; * is significant at 10%. P-values are displayed between brackets. Significance calculated as changes in MES based on a paired Student’s t-test. MES is calculated as proposed by Acharya, Pedersen, Philippon and Richardson (2010). The pre-merger period is composed of -180 till -10 days till announcement. The post-merger period is designed as +10 till +180 after the completion period. The difference between post- and pre-merger period results in: ∆MESi,t5%=MESi;[+11,+180]5% -MESi;[-11,-180]5% The expected shortfall, calculated as the 5% worst market

outcomes during the event window, is composed of the top 500 financial institutions on revenue. These variables are all calculated based on a total international sample of 528 mergers of banks. Adj. is for the market adjusted expected shortfall. Table 3 presents the MES of the acquirer, the expected shortfall of the market index and the

adjusted MES. Per region, the average of pre-, post- and delta MES is specified. Next, the expected shortfall is displayed, and finally, in the three righthand columns, the adjusted MES. The delta of the MES is significant at the 1% level for North America, Western Europe and the total sample. We find a small economic impact for these regions at 0.2%, 0.6% and 0.2% respectively. When comparing this with Weiss, Neumann and Bostandzic (2014), they find similar values for both Europe compared with North America, and in total. In total, the expected shortfall of the market is significant, at 10%, with a 0.1% difference, which could indicate that shocks have become more severe over time. However, a different explanation could be that there are more post-merger returns during a more volatile period, while there are hardly any announcements during this period. Various adjusted, pre- and

(19)

19 post-merger MES are significant, with the total significance at 1%. This illustrates that the shocks in the market are similar, but slightly more severe than the merging banks’ shortfall. The final column displays the delta MES minus delta expected shortfall, with only the Middle East significant as a region. This could indicate that there is no increase in MES as a result of mergers, but that it is driven by the lack of announcements in more volatile periods.

We confirm the results of Weiss, Neumann and Bostandzic (2014), and our first hypothesis: that mergers and the consolidation of banks generally result in an increase in systemic risk. This is expected, as for example Vallascas and Hagendorf (2011) find a small impact when using distance of default as a systemic risk measurement for their sample of 128 European mergers before the financial crisis of 2008. To further explore this result, sub-samples are assessed in the following section.

(20)

20

Table 4. Sub-sample analysis

MES pre MES post ΔMES N MES pre MES post ΔMES N

Foreign assets

present 2.9% 3.4% 0.5%* 65 Foreign income 3.0% 4.0% 1.0%** 387

(0.079) (0.039) No foreign assets reported 2.1% 1.6% 0.1% 220 No foreign income reported 1.6% 1.7% 0.1% 240 (0.125) (0.190)

No data present 1.9% 2.2% 0.2%** 144 No data present 2.0% 2.2% 0.2%* 196

(0.042) (0.060)

MES pre MES post ΔMES N MES pre MES post ΔMES N

Cross-border 2.7% 3.2% 0.4%* 87 High relevance 2.5% 2.9% 0.4%** 176

(0.072) (0.021)

Domestic 1.7% 1.9% 0.2%** 441 Medium relevance 1.8% 1.9% 0.1% 175

(0.023) (0.132)

Low relevance 1.4% 1.5% 0.1% 177

(0.183)

MES pre MES post ΔMES N MES pre MES post ΔMES N

High

concentration 2.0% 1.8% -0.3% 117 High profitability 2.0% 2.2% 0.2% 175

(0.123) (0.124) Medium concentration 1.8% 2.3% 0.5%*** 188 Medium profitability 1.8% 2.1% 0.3%*** 173 (0.000) (0.006) Low

concentration 2.0% 1.8% -0.3%** 223 Low profitability 1.9% 2.1% 0.1% 180

(0.022) (0.204)

MES pre MES post ΔMES N MES pre MES post ΔMES N

High supervisory power 1.7% 1.8% 0.1% 137 Pre-crisis 1.6% 1.7% 0.2%* 284 (0.186) (0.056) Medium supervisory power 1.8% 1.9% 0.1% 208 During-crisis 4.0% 6.6% 2.6%*** 28 (0.190) (0.001) Low supervisory power 2.2% 2.6% 0.4%** 178 Post-crisis 2.1% 2.0% 0.1% 216 (0.017) (0.320)

*** is significant at 1%; ** is significant at 5%; * is significant at 10%. P-values are displayed within brackets.

The sub-samples are based on dummy variables for foreign assets, foreign income and cross-border acquisitions. The other factors are divided along a percentile distribution. Relevance is calculated as the cross-product of total assets and relative deal value. The crisis period is defined as the beginning of the sub-prime mortgage crisis, end of 2007; whereas post-crisis is defined as early 2009.

Table 4 presents various sub-samples. The first two are based on foreign assets and foreign income, to check whether internationalization has an influence on systemic risk. The difference between domestic and cross-border acquisitions is included as exploratory research for the third hypothesis: that an international firm acquiring cross-border results in an increase in systemic risk. Relevance,

(21)

21 concentration and profitability are included based on the expectations of influence of control

variables. Regulation is included by splitting different levels of supervisory power. Finally, a sub-sample is included based on pre-, during and post-crisis MES, to investigate whether the timing of the merger is relevant for systemic risk.

We find a significant impact of reported foreign income, as it results in 1% higher MES, significant at 10%. Moreover, medium profitability and concentration significantly and positively indicate a higher delta MES. When comparing these result with Weiss, Neumann and Bostandzic (2014), we do not find a confirmation of the theory that high-profitability banks are able to offset the adverse effects of systemic risk. We did find, however, that cross-border mergers seem to result in higher MES compared with domestic mergers, at 10% and 5% significance respectively. Interestingly, low concentration seems to offset the impact of mergers on MES, as it is 0.3% lower at a significance level of 5%. Low supervisory power also increases MES, with 0.4% at 5% significance, indicating that weaker regulators are allowing mergers with detrimental effects to systemic risk.

The crisis period resulted in a significantly higher change in MES, with 2.6%, at 1% significance. A potential explanation is that these are distressed mergers, for example, under pressure from

regulators (Koetter, Bos, Heid, Kolari, Kool and Porath, 2007). The result is that these banks could be deemed to have a high systemic risk after completion (judged by the market) due to the acquisition of risky parts of a failing bank. To follow-up on these findings, control variables are included in the regression models.

5.1 Results multivariate regression

Table 5 presents the results of the first multivariate least squares regressions, focussing on the original model of Weiss, Neumann and Bostandzic (2014) and internationalization. All these

regressions use White’s adjusted standard errors to deal with heteroskedasticity. Regressions 1 to 4 all use region-fixed effects, which are unreported. Table 6 presents the multivariate least squares regressions that explore the influence of the interaction between internationalization and the cross-border dummy. In these regressions White’s adjusted standard errors are also used. Models 6 to 8 use region fixed effects, while these are not present in model 9. The inclusion of region-fixed effects rather than country-fixed effects is a result of the large number of countries with few mergers. In total, nine models are presented, with a different selection of the independent variables of foreign income, internationalization, cross-border acquisitions and the interaction between

internationalization and cross-border acquisitions. A major problem in all the regressions is

multicollinearity, which is caused by high correlation among independent variables. This is measured by variance inflation factors (VIF), which are presented in Appendix 2 to 4. It becomes difficult to

(22)

22 distinguish the individual effect of independent variables when multicollinearity is present. The problem generally becomes worse when including interaction effects, or a moderator. Moreover, there is in some cases a relatively high correlation between internationalization variables and control variables. These are presented in Appendix 5. In the most extreme cases, the problem results in abnormal standard errors (Farrar and Glauber, 1967). As it not only influences that control variables, but also the independent variables we wish to examine, some variables must be dropped to identify their impact. However, a potential problem with this approach is that biased results may be

presented due to an omitted variable bias.

The first model (1) is based on Weiss, Neumann and Bostandzic (2014). The model suggests that the too-big-to-fail motive (defined as systemically unimportant banks), the Lerner index and the

shareholder rights index ADRI significantly influence the change in MES, with 1%, 10% and 1% significance respectively. With stronger shareholder rights, on average, the MES drops by 0.29%; whereas a too-big-to-fail motive increases the systemic risk by 0.88%. For a more concentrated banking system, measured by the Lerner index, the systemic risk is lower with -1.03%. Next, we continue to investigate whether internationalization influences MES as a result of an acquisition.

The second model (2) includes accounting and deal characteristics and the foreign income dummy, but country and regulatory controls are excluded. The inclusion of foreign income leads to a drop in the number of mergers by 124, to 305. Total assets is included, but relative deal value is excluded to prevent multicollinearity problems. For Model 2, there is only a significant impact of the too-big-to-fail motive at 0.6%, with a significance level of 1%.

(23)

23

Table 5. Multivariate regression models marginal expected shortfall

(1) Model reference paper (2) Internationalization w. accounting (3) Internationalization w. accounting and country (4) full model (5) full model w. cross-border Internationalization variables:

Foreign income dummy 0.78% 0.96% 1.36%** 1.70%**

(0.214) (0.133) (0.036) (0.019)

Cross-border dummy -0.25% -1.05%

(0.585) (0.213

Acquirer and deal characteristics:

Equity ratio 0.01% 0.01% 0.01% 0.01%** 0.01%*

(0.203) (0.171) (0.180) (0.043) (0.060)

Return on assets 0.23% 0.16% 0.40% 0.45%

(0.213) (0.405) (0.114) (0.041)

Non-interest income to total income -0.17% -0.82% -0.91% -0.65% -0.61%

(0.666) (0.132) (0.118) (0.320) (0.294)

loan loss provisions over interest income 0.73% 0.69% 0.54% 0.46% 0.27%

(0.203) (0.554) (0.649) (0.697) (0.792)

Market to book value -0.01% -0.04% 0.02% -0.11% -0.11%

(0.945) (0.868) (0.943) (0.663) (0.663)

Too-big-to fail-motive 0.88%*** 0.60%*** 0.75%** 0.73%*** 0.74%***

(0.000) (0.004) (0.013) (0.007) (0.007)

Pre-merger performance 0.24% -0.10% 0.25% -0.33% -0.29%

(0.724) (0.892) (0.735) (0.691) (0.723)

Relative deal value -0.01% -0.01% -0.01% -0.01%

(0.285) (0.434) (0.404) (0.334) Total Assets (ln) 0.00% (0.540) Country control GDP growth 0.00% -0.16%* -0.22% -0.26%* (0.960) (0.098) (0.151) (0.056) Political stability 0.00% 0.29% -0.19% -0.22% (1.000) (0.253) (0.590) (0.434) Rule of law 0.85% -0.28% -0.13% (0.152) (0.722) (0.903) Lerner index -1.03%* -0.16% -0.66% -0.82% (0.053) (0.781) (0.239) (0.148)

Regulatory environment and deposit schemes:

ADRI -0.29%*** -0.21% -0.19%

(0.006) (0.154) (0.169)

Deposit insurance

Foreign deposit coverage dummy 0.08% 1.57% 1.60%

(0.898) (0.167) (0.147)

Independence supervisory power -0.11% 0.29% 0.33%

(0.534) (0.163) (0.112)

Observations 461 305 305 272 272

Region fixed effects Yes Yes Yes Yes Yes

R-squared -0.102 0.119 0.149 0.240 0.250

Adjusted R-squared -0.059 0.080 0.095 0.176 0.187

*** is significant at 1%; ** is significant at 5%; * is significant at 10%. P-values are displayed within brackets. Dependent variable is acquirer change in MES due to acquisition, ΔMES. All models are estimated with White’s robust standard errors to adjust for heteroskedasticity. Models 1 to 4 use region-fixed effects. In regressions 1 to 4, various control variables are dropped due to correlation and the resulting multicollinearity with internationalization variables. A full description of each variable and its source is available in Appendix 1.

When accounting and country controls are included, as in Model 3, we find that the too-big-to-fail-motive is significant at 5% and GPD growth at 10%. These variables have a coefficient of 0.75% and -0.16% respectively. However, GDP growth is barely significant, with a very small economic impact.

When accounting and country controls are included, as in Model 4, the full model is specified. This also results in a significantly higher R-squared of 0.176, compared to models 1 to 3. The foreign

(24)

24 income dummy is significant at 5%, with a 1.36% increase in systemic risk as a result of a merger if the acquirer is international prior to acquisition. The equity ratio is statistically significant, but economically insignificant at 0.01%. Similar all the other models, the too-big-to-fail motive is significant, at 10%.

The final model (5) of Table 5 also includes the cross-border dummy. This results in slightly higher impact of internationalization at 1.70%. Moreover, GDP growth is significant at 10%, with a coefficient of -0.26. This model results in the best fit, measured a by R-squared of 0.187.

To summarize, table 5 shows that an international bank will on average have a higher increase in systemic risk and thus a destabilizing effect on the financial system. This gives merit to the theory that internal capital markets can cause shocks in foreign countries by influencing the company through internal capital markets (Forbes and Warnock, 2012). Another explanation is that companies make use of regulatory arbitrage to transfer risk across borders to countries with less stringent regulations, which in turn increases systemic risk. However, none of the included regulatory

variables are significant. The too-big-to-fail-motive is significant for all models, with a value of 0.75% on average in models 3 to 5. Thus, relatively less systemically risky banks will on average result in a higher systemic risk after acquisition. By increasing their size, banks can increase implicit subsidies, such as deposit insurance (John, John and Senbet, 1991). This finding is different from both Buch and DeLong (2008) and Karolyi and Taboada (2015), who find a limited impact of this incentive on risk and systemic risk respectively. Recent developments, such as the European Bank Recovery and Resolution Directive, which aims to make use of the bail-in of creditors and shareholders to bear the costs of bank failure, lower the incentive of attempting to make use of implicit government subsidies (Berger, Molyneux and Wilson, 2014). However, Acharya, Anginer and Warburton (2013) state that the Dodd-Frank act did not lower the expectation of investors that the government will shield them from significant losses. Similarly, the market is expected to adjust its expectation, which can result in a higher delta MES. Therefore, harmonization of regulations could be one of the items to improve financial stability and offset the increase in systemic risk.

Table 6 present the multivariate regressions for hypotheses 3, that the interaction of

internationalization prior to acquisition and a cross-border acquisitions results in higher systemic risk. To assess this, three different internationalization proxies are used next to the foreign income dummy. In models 6 to 8 region fixed effects are included, but excluded in model 9. All these models suffer moderate multicollinearity, defined as a VIF between 2.5 and 5, which is reported in Appendix 2.

(25)

25

Table 6. Multivariate regressions interaction internationalization

(6) Interaction Foreign sales dummy (7) Interaction Foreign sales ratio (8) Interaction Foreign asset dummy (9) Interaction Foreign asset ratio Internationalization variables:

Foreign sales dummy 0.67%

(0.216)

Foreign sales ratio -0.01%

(0.278)

Foreign asset dummy 0.19%

(0.700)

Foreign asset ratio 0.02%

(0.178)

Cross-border dummy -2.33%*** -0.34% -3.46%*** 0.09%

(0.000) (0.654) (0.000) (0.867)

Internationalization*cross border 2.50%*** 0.00% 2.89%*** -0.01%

(0.001) (0.971) (0.000) (0.597)

Acquirer and deal characteristics:

Equity ratio 0.01% 0.01%* 0.01%** 0.02%**

(0.318) (0.056) (0.049) (0.013)

Return on assets 0.05% 0.23% 0.13% -0.01%

(0.749) (0.291) (0.488) (0.976)

Non-interest income to total income -0.49% -0.10% -0.10% -0.59%

(0.263) (0.854) (0.833) (0.298)

loan loss provisions over interest income 0.22% -2.25% -2.11% 0.34%

(0.733) (0.056) (0.134) (0.640)

Market to book value 0.16% 0.11%

(0.516) (0.506)

Too big to fail motive 0.72%*** 0.36%** 0.36%** 0.63%***

(0.001) (0.047) (0.038) (0.001) Pre-merger performance -0.60% -0.35% 0.23% 0.02% (0.404) (0.634) (0.750) (0.973) Country control GDP growth 0.14% (0.136) Political stability -0.18% (0.418) Lerner index -0.81%* -0.82%** (0.063) (0.039)

Regulatory environment and deposit schemes:

ADRI -0.27%** -0.18%

(0.029) (0.136)

Foreign deposit coverage dummy 0.01%

(0.982)

Capital regulatory index -0.02% -0.02%

(0.748) (0.444)

Government owned banks dummy -1.20% -1.63%*

(0.109) (0.088)

Observations 350 309 288 328

Region fixed effects Yes Yes Yes No

R-squared 0.102 0.143 0.164 0.062

Adjusted R-squared 0.062 0.080 0.105 0.036

*** is significant at 1%; ** is significant at 5%; * is significant at 10%. P-values are displayed within brackets. Dependent variable is acquirer change in MES due to acquisition, ΔMES. All models are estimated with White’s robust standard errors to adjust for heteroskedasticity. Internationalization is defined as the previously used foreign ratio posted above. Models 6 to 8 use region-fixed effects. In regressions 6 to 9, various control variables are dropped due to correlation and the resulting multicollinearity with internationalization variables. A full description of each variable and its source is available in Appendix 1.

Model 6 uses a foreign sales dummy to assess the interaction. In this model only accounting controls are taken into account to minimize multicollinearity. Both the cross-border dummy and interaction term are significant at 1%. A cross-border merger will result on average in a reduction of 2.33%. However, this is nullified if the firm engaging in the merger is already international prior to

(26)

26 acquisition, as the coefficient of the foreign sales dummy times the cross-border dummy is equal to 2.50%.

When foreign sales ratio is used as internationalization proxy in model 7, the multicollinearity problem is less severe, allowing the inclusion of additional country controls and regulatory controls. In this case cross-border and interaction term is no longer significant. The equity ratio is significant at 10%, but economically insignificant at 1%. As with all models, the too-big-to-fail motive is

significant. Additionally, the anti-director rights index and Lerner index are significant at 5% and 10% respectively. These results indicate that a more concentrated banking system and better shareholder rights lead to a reduction of systemic risk after acquisition.

Model 8 uses the foreign asset dummy as internationalization. While internationalization itself is not significant in this case, both the cross-border dummy and internationalization multiplied by the cross-border dummy are significant at 1%. In this case effect is similar as in model 6, an international firm will, on average, experience a 2.1% increase in systemic risk; while a firm that diversifies

internationally (i.e. a cross-border acquisition) could lower systemic risk by 2.3% on average. There is a significant negative relation between delta MES and the Lerner index and between delta MES and the government owned bank dummy. In both cases these lower on average the systemic risk result of an acquisition. A potential explanation for the fact that the presence of government owned banks reduces systemic changes of an acquisition, is that the market could expect the government to bail-out banks in the worst-case scenario.

The final model (9) in Table 6 uses the foreign asset ratio as proxy for internationalization. Berger, El Ghoul, Guedhami and Roman (2016) use this as key-variable to proxy for internationalization. The fact that it is not significant is a strong indicator against the hypothesis that internationalization positively influences delta MES. In this regression no country fixed effects are included and only accounting and deal characteristics are used, to reduce the impact of multicollinearity. The only economically and statistically significant factor is the too-big-to-fail motive at 1%.

In summary, all the regressions in which the too-big-to-fail factor is included are significant both statistically and economically. Thus, there is support for the argument that banks have an incentive to make use of implicit subsidies related to systemic importance, which results in an increase in systemic risk. The results of table 6 imply there is no relation between internationalization of merging banks and financial instability, whereas Table 5 does present results that there is a relationship. Due to moderate multicollinearity problems with the models in Table 6 we conclude that internationalization of merging banks increases system fragility. Arguments proposed in the literature review that support this, are that it only decreases diversity among banks, thus increasing

(27)

27 the overall joint-likelihood a systemic shock (Wagner, 2010). Moreover, the increase can also be caused by the additional exposure in a foreign country, which can transfer a shock through internal capital markets (Forbes and Warnock, 2012).When looking at our third hypothesis, the relationship between systemic risk and internationalization is affected by whether it is a cross-border acquisition. If it is an international firm is acquiring cross-border, there is an indication that systemic risk

increases. However, we conclude these findings are inconclusive due to measurement problems when using an interaction term.

5.2 Robustness analysis

To test whether methodological choice affect the results, sub-sample analyses are used to compare pre- and post-crisis results. Moreover, as the sample consist of a large number of US banks, there is a split between US and non-US banks in Table 7. Moreover, this provides additional tests for

whether internationalization is a source of strength or weakness, and whether this has changed over time. In multivariate regressions in Table 7 only region fixed effects are used in the pre-financial period.

For model 10 based on only US banks, deal and accounting characteristics are used, as the country and regulatory controls are similar over time. This results in a sample of 232 mergers, with

internationalization and the cross-border dummy being significant at both 10%. Internationalization increases systemic fragility as a result of a banking merger, whereas a cross-border merger results in a decrease. This confirms the second hypothesis that internationalization significantly influences systemic risk.

For non-US mergers in model 11, internationalization prior to acquisition and cross-border mergers both aren’t significant. The loan loss provision, GDP growth, foreign deposit coverage dummy variable and the government owned bank variable are significant. Interesting here is that loan loss provision ratio is significant and reducing systemic risk. This could be caused by more conservative estimates for the required provision are valued by the market, which lowers systemic risk. However, for this theory additional evidence is required.

Model 12 and 13 split the sample in two, from 1999 to august 2007 and from the crisis period up to 2016. These results indicate that internationalization prior to the crisis did not influence system fragility caused by mergers, whereas during and after (unreported) there is a significant impact of internationalization on systemic risk. From these results the conclusion can be derived that

References

Related documents

Parallellmarknader innebär dock inte en drivkraft för en grön omställning Ökad andel direktförsäljning räddar många lokala producenter och kan tyckas utgöra en drivkraft

Närmare 90 procent av de statliga medlen (intäkter och utgifter) för näringslivets klimatomställning går till generella styrmedel, det vill säga styrmedel som påverkar

• Utbildningsnivåerna i Sveriges FA-regioner varierar kraftigt. I Stockholm har 46 procent av de sysselsatta eftergymnasial utbildning, medan samma andel i Dorotea endast

I dag uppgår denna del av befolkningen till knappt 4 200 personer och år 2030 beräknas det finnas drygt 4 800 personer i Gällivare kommun som är 65 år eller äldre i

Denna förenkling innebär att den nuvarande statistiken över nystartade företag inom ramen för den internationella rapporteringen till Eurostat även kan bilda underlag för

Utvärderingen omfattar fyra huvudsakliga områden som bedöms vara viktiga för att upp- dragen – och strategin – ska ha avsedd effekt: potentialen att bidra till måluppfyllelse,

Den förbättrade tillgängligheten berör framför allt boende i områden med en mycket hög eller hög tillgänglighet till tätorter, men även antalet personer med längre än

Det har inte varit möjligt att skapa en tydlig överblick över hur FoI-verksamheten på Energimyndigheten bidrar till målet, det vill säga hur målen påverkar resursprioriteringar