Cross-border Bank Acquisitions: Is there a Performance Effect?

Ricardo Correa^*

NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. Recent IFDPs are available on the Web at http://www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at http://www.ssrn.com/.

Abstract:

This paper uses a unique database that includes deal and bank balance sheet information for 220 cross-border acquisitions between 1994 and 2003 to analyze the characteristics and performance effects of international takeovers on target banks. A discrete choice estimation shows that banks are more likely to get acquired in a cross-border deal if they are large, bad performers, in a small country, and when the banking sector is concentrated. Post-acquisition performance for target banks does not improve in the first two years relative to domestically-owned financial institutions. This result is explained by a decrease in the banks' net interest margin in developed countries and an increase in overhead costs in emerging economies.

Keywords: Mergers and acquisitions, performance, international banking

JEL classification: F21, F23, G21, G34

For the last 15 years the international financial system has experienced significant changes that have reshaped its structure and exposure to global shocks. An important issue in this trend has been the increasing presence of foreign banks in developed and emerging countries. The existing literature has associated financial liberalization with an increase in growth (Levine 2005), stability (Crystal et al. 2001), and better credit allocation (Giannetti and Ongena 2005) in emerging economies. It has also become one of the main policy recommendations from multilateral organizations.¹

This paper uses a unique cross-border Mergers and Acquisitions (M&As) database to answer four questions: Which factors influence cross-border acquisitions? Does this type of acquisitions improve the target's performance? Is there any post-acquisition difference in performance for targets in developed and emerging economies? Is it influenced by host-country or home-country characteristics?

The determinants of cross-border acquisitions are evaluated using 220 deals that took place between 1994 and 2003. I estimate a discrete choice model to test the factors that increase the probability of an international takeover. This study finds that the target banks' size, pre-acquisition profitability, and the level of concentration in the host country's banking sector are significant determinants of cross-border deals. For emerging economies, the level of financial intermediation also contributes to the likelihood of acquisitions of domestic banks by Multinational Banks (MNBs).

The effects of bank acquisitions have been studied by using information from deals between local institutions in developed economies and cross-border deals in Europe. The evidence shows limited performance improvements in the post-acquisition period. In contrast, foreign banks in emerging economies are found to be better performers than their domestic counterparts.² This paper focuses on the first two years after a cross-border acquisition to test if foreign acquirers are able to increase the target's efficiency in the short run. Then, I compare if there is a significant difference, in terms of post-acquisition performance, between targets located in emerging and developed economies after a cross-border acquisitions.

Post-acquisition changes in performance are tested using a sub-sample of 102 deals with information for at least two years before and after the cross-border deal. A difference-in-difference analysis is used to control for economy-wide and country-specific effects. As the counterfactual to the targets' profitability measures, I construct a country-specific index that reflects the aggregate performance of local non-acquired banks. I find that acquired banks perform at the same level--and sometimes worse--relative to the country-specific indices after a takeover. This negative change in profitability is mostly explained by a decline in Net Interest Margins. In the post-acquisition period, MNBs have significantly lower margins than domestically owned banks, consistent with a strategy to gain market share in the traditional intermediation business. Loan Loss Provisions decrease after acquisitions, partially compensating the negative effect of the cross-border deal on income.

The next step is to compare deals involving targets located in emerging economies to those associated with targets in developed countries. The targets overall performance is not significantly different for the two groups of banks after cross-border deals. A detailed look at the change in individual components of the banks' income statements shows little differences between banks in emerging and developed countries after an international deal. Nevertheless, there are some contrasts that have to be noted. In particular, median Net Interest Margins and expenditures in non-interest and personnel costs decline in developed countries while the opposite is the case in emerging economies. This result demonstrates the difficulties in improving efficiency in different institutional, economic, and cultural environments.³

Finally, I test for diseconomies in managing foreign subsidiaries due to differences in language, legal origin, and geographical distance. Targets perform better if the home country of the acquirer and the host country share the same language. This factor is particularly relevant in determining post-acquisition Overhead costs in developed and emerging economies. In contrast, differences in neither legal origin nor distance appear to affect performance negatively in the post-acquisition period.

The rest of the paper is organized as follows. Section 1 reviews the literature on cross-border acquisitions and their impact on bank performance. Section 2 describes the empirical methodology used to answer the questions posed in this study. Section 3 describes the data and sample selection criteria. Section 4 presents the main results. Finally, section 5 concludes.

1 Motivation and Related Literature

The literature on cross-border acquisitions has studied the motivation and consequences of this type of deals from different perspectives. A first set of studies analyzes the determinants of cross-border bank acquisitions. The motivation for cross-border consolidation ranges from the "follow-your-customer" hypothesis (Miller and Parkhe 1998; Esperanca and Gulamhussen 2001) to differences in efficiency between acquirers and target banks (Berger et al. 2000). Some studies have explained these deals using arguments from the Foreign Direct Investment (FDI) literature (Goldberg 2004) and New Trade Theory (Berger et al. 2004) literature. Using a sample of OECD countries, Focarelli and Pozzolo (2005) find that it is more likely for MNBs to enter countries "where the expected economic growth is higher", banking sector concentration is lower, and the regulatory environment is less stringent.⁴ In a related study, Claessens and Van Horen (2007) argue that institutional competitive advantages are an important determinant of locational decisions in international banking. MNBs expand to countries with institutions that are similar to the environment that they face in their home country--relative to the institutional environment of competing MNBs in other countries. Lastly, cross-border acquisitions have been relatively scarce compared to their domestic counterpart. Buch and DeLong (2004) attribute this phenomenon to information costs and regulatory restrictions.

This paper expands the literature reviewed in the last paragraph by analyzing both the determinants of financial FDI at the country level, and also focusing on the target specific characteristics that motivate cross-border acquisitions. The framework used in this study is similar to the approaches followed in Focarelli, Panetta, and Salleo (2002) for Italian banks and Hannan and Rhoades (1987) for U.S. institutions.

A second strand of the literature focuses on the effect of M&As on stock prices and accounting measures of performance. Piloff and Santomero (1998) and Calomiris and Karceski (2000) review the main findings in this literature for U.S. institutions.⁵ The typical analysis of M&As using stock price data, compares the change in returns after a deal is announced. These studies find a negligible effect of M&As deals on stock market value. There is a transfer of wealth from the acquirer to the target shareholders mostly explained by the high premiums paid on these transactions. The lack of stock price information comparable across countries--outside of Europe--has limited the amount of studies using the event methodology to analyze performance effects after cross-border M&As.⁶ In one of the few studies that uses the link between cross-border deal information and stock prices, Amihud, DeLong, and Saunders (2003) find that there is no reduction in risk for those banks that diversify geographically by acquiring financial institutions abroad. Moreover, the cumulative abnormal returns for the acquirers in these transactions are negative and significant.

Another group of studies uses accounting data to asses the effect of M&As on operating performance. Chamberlain (1998) analyzes a sample of deals that took place in the U.S. in the 1980s and finds that these transactions did not yield any operating efficiencies. This result is consistent with similar evidence that shows no improvements in Return on Assets (ROA) or growth in operating income in the same time period (Linder and Crane 1992). A limited number of studies show positive changes in performance after M&A deals in 1980s, for instance, Cornett and Tehranian (1992) find an increase in the post-acquisition Return on Equity (ROE) and operating cash flow, but the authors focus only on 30 mergers between 1982 and 1987. In the 1990s, the observed post-acquisition performance of institutions involved in M&A deals improved on average. Technological changes and the deregulation of national branching by financial institutions are suggested as possible explanations for this difference in the post-acquisition performance of merged institutions (Cornett, McNutt, and Tehranian 2006; Berger, Demsetz, and Strahan 1999).

On the international side, Vander Vennet (2002) studies a sample of European cross-border deals and finds an increase in profit efficiency for target banks on the first year after an acquisition. Nevertheless, the author does not find similar improvements in the cost efficiency and ROA measures. Using a larger sample of cross-border deals, Beccalli and Frantz (2007) find the opposite result: a decrease in profit efficiency and an increase in cost efficiency after cross-border deals. The difference in these findings could be explained by the laxer sample selection criteria used in the latter study. The authors do not restrict the sample of deals to those acquisitions were the target bank's control is transferred to the acquiring institution. Therefore, the results might be driven by the effect of minority share acquisitions. As summarized in these two studies, the effect of cross-border M&As on the targets' post-acquisition performance is inconclusive, and might depend on the location of the target and the level of control of the acquirer over its new subsidiary.

The literature reviewed in this section finds mixed effects in terms of the impact of M&As on banks in developed economies. Alternatively, some empirical studies suggest that foreign bank presence benefits emerging economies in different dimensions. In countries with a larger presence of MNBs, the domestic banking sector is more efficient (Claessens, Demirgüç-Kunt, and Huizinga 2001; Bayraktar and Wang 2004), stable (Crystal, Dages, and Goldberg 2001), capital allocation improves (Giannetti and Ongena 2005), and economic growth is enhanced (Levine 2001).

The current paper expands these last two strands of the literature by using accounting data to assess the effect of cross-border acquisitions on the targets' operating performance. To analyze this effect, I construct a large sample of deals that includes targets in developed and emerging economies and focus on acquisitions where control of the target institution is passed to the foreign acquirer.

2 Empirical Methodology

2.1 Determinants of cross-border acquisitions

This section describes the methodology used to test the first question addressed by this study. Following Vander-Vennet (2002) and Focarelli, Panetta, and Salleo (2002), I use a probit-model to estimate the characteristics of banks that are involved in cross-border acquisitions in comparison to those that are not part of any deal during the sample period. The dependent variable is a binary choice variable equal to one, the year a bank is the target in a takeover where the acquirer is a foreign financial institution. The model to estimate is given by:

$\displaystyle \Pr(Y_{ijt} =1)=\Phi\left( {X_{it-1} ,Z_{jt-1} ,M_{jt-1} } \right)$

(1)

where $Y_{ijt}$ equals one when bank in country gets acquired in year by a foreign bank and zero otherwise; $\Phi$ is the standard cumulative normal probability distribution; $X_{it-1}$ is a vector of bank-specific variables; $Z_{jt-1}$ represents a vector of country characteristics, including macroeconomic aggregates and financial indicators; $M_{jt-1}$ is a vector of variables that describe the regulatory environment and concentration level in the banking sector by country. Estimations include year fixed effects and standard errors are clustered by country.

All explanatory variables enter in the regression with one lag. This specification assumes that buyers take the decision to acquire a target using information available to them at the end of the year before the acquisition takes place. The coefficients on the regressors in this model indicate the change in the probit score in terms of standard deviations, following a one-unit increase in the predictors. To establish the relevant characteristics determining cross-border deals, I test the significance and magnitude of these coefficients.

Following Focarelli and Pozzolo (2000), four sets of variables are included in these estimations. The first group of variables consists of ex ante measures of bank profitability, size, capital, and lending activity.⁷ The second set of variables is taken from the literature on the determinants of economic growth, and includes real GDP, inflation, GDP per capita growth, and Private Credit to GDP--a measure of financial intermediation. The third group includes variables that proxy for regulatory restrictions and bank concentration.⁸ These proxies measure the structure of the banking sector in the host country and implicit limitations to bank entry. Finally, the last group of variables measures the level of financial development in the host country, proxied by the value of stock market and private and public bond market capitalization to GDP.

2.2 Performance effect

The second question outlined in this paper analyzes the change in performance for target banks after a cross-border acquisition. In order to measure this change, I have to determine what the bank's performance would have been if the acquisition had not taken place. This study draws on Cornett, McNutt, and Tehranian (2006) and measures the counterfactual of the target's performance with a country-specific bank index. The effect of the deal is calculated by subtracting this benchmark from the acquired-bank's performance indicators, and comparing this measure between the before and after acquisition period. This estimation technique controls for possible differences in accounting methods across countries, regulatory environments, and country specific-economic activity.

The empirical methodology in this section follows Chamberlain (1998). The target's performance is assumed to be given by:

$\displaystyle r_{\tau i} =\mu_{z} +c_{\tau i} +\eta_{\tau i}$

(2)

where $r_{\tau i}$ represents the performance proxy for target at event time $\tau$ ; $\mu$ $_{z}$ is a constant treatment effect; $c_{\tau i}$ is an unobserved target control effect; and $\eta$ $_{\tau i}$ represents a target specific error term.

The control effect ( $c_{\tau i})$ is measured with error using the country ( specific industry index. This measure is defined as:

c $\displaystyle _{\tau\mbox{j}} =c_{\tau i} +\varepsilon_{\tau j}$

(3)

It is assumed that $\eta$ $_{\tau i}$ and $\varepsilon$ $_{\tau j }$ are mutually and cross-sectionally independent, but could be correlated over time. Then, by subtracting (3) from (2) I obtain:

$\displaystyle r_{\tau i} - c_{\tau j} =\mu_{z} +\eta_{\tau i} - \varepsilon_{\tau j} =\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{\tau i}$

(4)

With this expression I can compute the pre-acquisition ( $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{bi} )$ and post-acquisition ( $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{ai} )$ relative performance measures by averaging all $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{\tau i}$ in each period. These measures will proxy for the treatment effect $\mu$ $_{z}$ with an error that is independent across observations. Using the sample distributions of $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{bi}$ and $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{ai}$ , I test for changes in the target's relative performance ( $\rho$ ) after an acquisition. By subtracting $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{bi}$ from $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{ai}$ , $\rho$ plus an error term ( $\nu_{i})$ are obtained:

$\displaystyle \mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{ai} -\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\mu }} _{bi} =\rho+\nu_{i} =\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\rho }} _{i}$

(5)

The Sign Test and $\mathord{\buildrel{\lower3pt\hbox{$\scriptscriptstyle\frown$}}\over {\rho }} _{i}$ are used to examine the null hypothesis that the number of positive and negative relative differences are equal.⁹In other words, this method tests if cross-border acquisitions had an effect on the acquired banks' performance. The only requirement for the Sign Test is that each $\nu_{i}$ has to come from a continuous median zero distribution.

Bank performance is measured using three accounting ratios: Return on Average Assets (ROA), Return on Average Equity (ROE) and the Cost to Income Ratio.¹⁰ In addition, I analyze the post-acquisition change in four revenue and cost components: Net Interest Margin, Non-Interest Income, Overhead, and Loan Loss Provision.¹¹

Finally, to answer the question about the differences in post-acquisition performance by targets depending on the level of the development of the host country, I divide the sample between targets located in emerging countries and those in developed economies. Following Barth, Caprio, and Levine (2001), a bank is defined as being located in a developed country, if GDP per capita in the host-country is above 10,000 dollars (2000 U.S. dollars). Then, performance and other income indicators are compared using the Sign Test, Wilcoxon Test, and the Median Test.

2.3 Performance, economic integration, and information costs

The third set of tests deal with the effect of economic integration and information costs on the target's performance after a cross-border acquisition takes place. Buch and DeLong (2004) find that information costs and regulation decrease the amount of cross-border M&A activity.¹² The following empirical specification includes these factors to measure their effect on post-acquisition bank profitability:

$\displaystyle y_{ijt} =\alpha_{0} +\alpha_{1} Yr0+\alpha_{2} Yr12+\alpha_{1} Yr3^{+}+X_{jh} {\beta}^{\prime}+Z_{jt} {\gamma}^{\prime}+\upsilon_{i} +\eta_{j} +\varepsilon_{ijt}$

(6)

where $y_{ijt}$ is the performance proxy for year , country , and deal . This variable is a transformation of the original balance sheet ratios into percentile ranks in the distribution of all non-acquired banks by country.¹³ This method makes it possible to control for changes in the distribution of the relevant variables over time, as well as comparing the target banks to their relevant peer group. Yr0, Yr12 and Yr3 $^{+}$ are indicator variables equal to 1 for the year the deal takes place, for the first and second year after the deal, and for the third year and after, respectively; $X_{jh}$ is a vector of bilateral variables representing information costs and the level of integration between the host country and the home country ; $Z_{jt}$ is a vector of macroeconomic aggregates and banking structure variables; $\eta$ $_{j}$ and $\upsilon$ $_{i}$ are host-country and target fixed effects, respectively.

As discussed by Berger and DeYoung (2001), there are diseconomies in managing subsidiaries that are located at longer distance relative to their parent bank's location. The same argument applies to other dimensions of distance like the difference in language and legal systems across countries. Vector controls for these factors as it includes a dummy indicating if the country of the acquirer and target share the same principal language (Same Language); another indicator variable equals one if both countries have similar legal systems (Same Legal).¹⁴Log distance measures the geographical distance between the host country and home country of the acquirer; Same Region is a dummy variable equaling one if the target and acquirer are located in the same region. In addition, following Berger et al. (2004) I include an index of comparative size (Similar GDP) and an index measuring comparative economic development (Similar GDP PC) between the home and host countries.¹⁵These indices range from 0 to 1, with a value of 1 indicating that both countries have the same size or the same GDP per capita. These set of variables will measure the effect of economic integration and information cost on the target bank's performance.

3 Data Description

To estimate the models defined in the previous section, I construct a sample of banks involved in cross-border deals between 1994 and 2003. For this purpose, two databases are matched: the first one includes bank financial data and the second has information on cross-border acquisitions. Data on banks' financial statements is collected from the Bankscope database maintained by Bureau van Dijk. This dataset contains annual statements for listed and unlisted banks in 179 countries starting at the beginning of the 1990s. For M&A information, I use the Zephyr database from Bureau van Dijk, the SDC Platinum database from Thompson Financial Securities Data, and individual bank webpages.

In addition to bank information, controls at the country level are also included in the estimations. Macroeconomic and financial aggregates are from the World Development Indicators (WDI) database and the Financial Structure and Development database published by the World Bank. The Banking Freedom index is constructed by the Heritage Foundation.¹⁶ It has values between 0 and 100 and measures the stringency of financial regulation in a country. Higher values for this index imply a more liberalized banking sector. Institutional variables are taken from La Porta, Lopez-De_Silanes, and Shleifer (2002), and bilateral data were compiled by Rose and Spiegel (2004).

The next two sub-sections describe the sample selection process for banks included in the estimations described in sub-sections 2.1 and 2.2. The last sub-section also outlines the construction of the control indices used in the performance estimations.

3.1 Sample selection

Two bank samples were constructed to estimate the regressions described in the previous section. The first one includes all financial institutions classified as Commercial Banks in Bankscope between 1994 and 2003 (3295) that have at least one cross-border deal in the sample period.¹⁷ Table 1 shows the distribution of banks across countries. A large percentage of the sample is represented by financial institutions from the United States (27.3%), Germany (5.5%) and France (5.3%). Amongst emerging economies, Brazil (2.9%), Argentina (2%), and Panama (1.8%) have the largest shares.¹⁸ The second sample is limited to a group of banks acquired in cross-border transactions.

Table 1: Banks and deals by country

Deal data is from Zephyr, SDC and the banks' webpages. Bank data is from Bankscope. The deals' sample period ranges between 1994 and 2003. Bank balance sheet and income statement information covers the period between 1994 and 2004.

Country	Total Banks: Banks	Total Banks: Percentage	Total Deals: Deals	Total Deals: Percentage	Performance Deals: Deals	Performance Deals: Percentage
Albania	5	0.2%	0	0.0%	0	0.0%
Argentina	66	2.0%	11	5.0%	1	1.0%
Australia	25	0.8%	1	0.5%	1	1.0%
Austria	47	1.4%	3	1.4%	2	2.0%
Belarus	9	0.3%	1	0.5%	0	0.0%
Belgium	35	1.1%	7	3.2%	6	5.9%
Bolivia	11	0.3%	2	0.9%	1	1.0%
Bosnia-Herzegovina	15	0.5%	2	0.9%	1	1.0%
Brazil	94	2.9%	12	5.5%	6	5.9%
Bulgaria	22	0.7%	5	2.3%	3	2.9%
Cameroon	4	0.1%	1	0.5%	0	0.0%
Canada	47	1.4%	2	0.9%	0	0.0%
Chad	3	0.1%	0	0.0%	0	0.0%
Chile	24	0.7%	4	1.8%	2	2.0%
Colombia	23	0.7%	2	0.9%	2	2.0%
Croatia	32	1.0%	4	1.8%	2	2.0%
Czech Republic	17	0.5%	7	3.2%	2	2.0%
Denmark	53	1.6%	3	1.4%	2	2.0%
Dominican Republic	24	0.7%	1	0.5%	0	0.0%
Egypt	28	0.8%	4	1.8%	2	2.0%
El Salvador	7	0.2%	1	0.5%	0	0.0%
Estonia	5	0.2%	3	1.4%	0	0.0%
Finland	5	0.2%	1	0.5%	0	0.0%
France	173	5.3%	12	5.5%	6	5.9%
Germany	182	5.5%	12	5.5%	8	7.8%
Ghana	10	0.3%	1	0.5%	0	0.0%
Hong Kong	14	0.4%	0	0.0%	0	0.0%
Hungary	27	0.8%	4	1.8%	1	1.0%
Indonesia	49	1.5%	4	1.8%	2	2.0%
Ireland	15	0.5%	0	0.0%	0	0.0%
Italy	110	3.3%	1	0.5%	1	1.0%
Jamaica	6	0.2%	1	0.5%	0	0.0%
Japan	133	4.0%	0	0.0%	0	0.0%
Kenya	23	0.7%	0	0.0%	0	0.0%
Republic of Korea	13	0.4%	0	0.0%	0	0.0%
Latvia	19	0.6%	7	3.2%	1	1.0%
Lebanon	43	1.3%	1	0.5%	0	0.0%
Lithuania	10	0.3%	6	2.7%	0	0.0%
Luxembourg	102	3.1%	4	1.8%	2	2.0%
Macau	5	0.2%	1	0.5%	1	1.0%
Macedonia (Fyrom)	10	0.3%	2	0.9%	1	1.0%
Mexico	36	1.1%	6	2.7%	3	2.9%
Mongolia	3	0.1%	0	0.0%	0	0.0%
Morocco	7	0.2%	1	0.5%	1	1.0%
Netherlands	21	0.6%	2	0.9%	2	2.0%
New Zealand	8	0.2%	0	0.0%	0	0.0%
Nicaragua	8	0.2%	1	0.5%	1	1.0%
Norway	12	0.4%	3	1.4%	2	2.0%
Pakistan	19	0.6%	0	0.0%	1	1.0%
Panama	59	1.8%	3	1.4%	0	0.0%
Paraguay	18	0.5%	1	0.5%	0	0.0%
Peru	16	0.5%	3	1.4%	1	1.0%
Philippines	22	0.7%	1	0.5%	1	1.0%
Poland	39	1.2%	11	5.0%	7	6.9%
Portugal	21	0.6%	1	0.5%	0	0.0%
Romania	14	0.4%	4	1.8%	2	2.0%
Russian Federation	80	2.4%	0	0.0%	0	0.0%
Slovakia	12	0.4%	7	3.2%	4	3.9%
Slovenia	17	0.5%	3	1.4%	3	2.9%
Spain	74	2.2%	7	3.2%	3	2.9%
Switzerland	157	4.8%	8	3.6%	3	2.9%
Thailand	7	0.2%	1	0.5%	1	1.0%
Tunisia	15	0.5%	1	0.5%	1	1.0%
Turkey	10	0.3%	0	0.0%	0	0.0%
Uganda	12	0.4%	1	0.5%	0	0.0%
Ukraine	29	0.9%	0	0.0%	0	0.0%
United Kingdom	63	1.9%	2	0.9%	1	1.0%
Uruguay	31	0.9%	2	0.9%	1	1.0%
United States	900	27.3%	12	5.5%	6	5.9%
Venezuela	37	1.1%	5	2.3%	2	2.0%
Western Samoa	3	0.1%	1	0.5%	0	0.0%
Total	3295	100.0%	220	100.0%	102	100.0%

To construct the first sample, the Bankscope dataset is matched to an M&A database, which is comprised of information for all cross-border acquisitions between 1994 and 2003.¹⁹ This paper requires two conditions for a deal to be defined as a cross-border acquisition: first, the transaction has to give the acquiring bank a majority stake (more than 50%) in the target bank, provided that it previously held either no shares or a minority stockholding in the target. Additionally, the headquarters of the target bank has to be located in a country different from the home-country of the ultimate parent of the acquirer. The result is 328 deals matched to Bankscope.

The next step is to exclude all bank-year observations that are defined as outliers in terms of their income and balance sheet components.²⁰This restriction reduces the number of deals to 220 as shown in Table 1. One third of the deals involve targets in the United States, France, Germany, Brazil, Argentina, and Poland. Panel A in Table 2 shows that 174 of these targets were acquired by Western European institutions. The preferred destinations of these acquirers are Western and Eastern European countries (56 and 55, respectively), closely followed by Latin American (40) targets.

Table 2: Deals by region - Panel A: All Deals

Deal data is from Zephyr, SDC and the banks' webpages. The deals' sample period ranges between 1994 and 2003.

Target	Acquirer: Latin America	Acquirer: Eastern Europe	Acquirer: East Asia	Acquirer: Western Europe	Acquirer: US and Canada	Acquirer: Oceania	Acquirer: Middle East	Total
Latin America	7	0	0	40	7	0	1	55
Eastern Europe	0	8	1	55	2	0	0	66
East Asia	0	0	3	3	1	0	0	7
Western Europe	1	3	0	56	5	0	1	66
US and Canada	1	0	1	10	2	0	0	14
Oceania	0	0	0	1	0	1	0	2
Africa	0	0	0	9	0	0	0	9
Middle East	0	0	0	0	0	0	1	1
Total	9	11	5	174	17	1	3	220

Table 2: Deals by region - Panel B: Performance Deals

Target	Acquirer: Latin America	Acquirer: Eastern Europe	Acquirer: East Asia	Acquirer: Western Europe	Acquirer: US and Canada	Acquirer: Oceania	Acquirer: Middle East	Acquirer: Total
Latin America	0	0	0	17	2	0	1	20
Eastern Europe	0	1	0	25	1	0	0	27
East Asia	0	0	2	2	1	1	0	6
Western Europe	1	3	0	33	0	0	1	38
US and Canada	1	0	1	2	2	0	0	6
Oceania	0	0	0	1	0	0	0	1
Africa	0	0	0	4	0	0	0	4
Middle East	0	0	0	0	0	0	0	0
Total	2	4	3	84	6	1	2	102

Table 3 displays summary statistics for this sample. Acquired and non-acquired banks are similar in terms of their level of equity as shown in Panels A and B, but the median size, defined as Real Assets, is larger for the former group. The three performance measures for non-acquired banks, ROA, ROE, and the Cost to Income Ratio, have larger medians in the first two cases and lower in the last case, relative to the target banks. These statistics show that the median acquired bank was less profitable than its non-acquired counterpart during the sample period.

Table 3: Summary statistics - Panel A: Acquired Banks

Bank Balance Sheet and Income Statement data is from Bankscope. The sample period is 1994 to 2003. The variable Real Assets is defined in terms of millions of 2000 U.S. dollars. The rest of the variables are defined in terms of percentage points.

Statistic	Obs.	Mean	Median	Std. Dev.	Min.	Max.
Real Assets	1576	6357	1075	15618	5	150292
Equity to Avg. Assets	1578	12.22	9.28	10.8	1.0	95.2
ROA	1578	1.02	0.84	2.0	-8.8	11.8
ROE	1577	9.09	9.34	18.5	-96.9	135.4
Cost to Income Ratio	1578	71.80	67.55	27.6	3.4	232.4
Net Loans to Avg. Assets	1577	48.37	49.56	20.7	0.0	98.8
Net Interest Margins	1578	4.82	3.80	3.9	-1.8	27.8
Non-Interest Inc. to Avg. Ass.	1578	2.73	1.86	3.3	0.0	54.6

Table 3: Summary statistics - Panel B: Non-acquired banks

Statistic	Obs.	Mean	Median	Std. Dev.	Min.	Max.
Real Assets	30096	11244	854	54661	0	1352996
Equity to Avg. Assets	30393	12.66	8.79	13.6	0.0	100.0
ROA	30404	1.09	0.92	1.7	-10.0	12.0
ROE	30367	10.55	10.07	19.2	-100.0	928.0
Cost to Income Ratio	30404	65.22	63.45	24.3	0.0	244.0
Net Loans to Avg. Assets	30106	51.68	55.99	23.6	0.0	100.0
Net Interest Margins	30404	4.08	3.53	3.4	-2.3	28.0
Non-Interest Inc. to Avg. Ass.	30404	2.50	1.28	4.3	0.0	92.5

For the performance estimations described in section 2.2, I restrict the sample to banks with at least two years of information before a cross-border acquisition and two years after.²¹ This creates a sample of 102 deals shown in the last two columns of Table 1. A significant share of targets is located in Germany (7.8%), Belgium (5.9%), Brazil (5.9%), Poland (6.9%), and the United States (5.9%). The share of Argentinean (1%) banks in this sample decreases relative to the full set of deals in this country due to missing and outlier observations attributed to the banking crisis in 2001. Panel B in Table 2 shows that most of the acquirers are based in Western European countries (84). Financial institutions in Western Europe are mostly involved in deals within the region (33) or in Eastern European (25) and Latin American (17) countries.

Figure 1 shows the number of all matched deals by year, and the number of deals used in the performance estimations. Most of the deals are clustered around the last years of the 1990s. Data restrictions for the performance estimations reduce the sample of deals considerably.

Figure 1: Number of cross-border deals by year

The total number of yearly cross-border deals in the sample is denoted by bars with diagonal patterns. Bars with horizontal patterns denote the number of deals used in the performance estimations.

Data for Figure 1 immediately follows.

Data for Figure 1

Year	All	Performance
1996	1	0
1997	17	7
1998	35	21
1999	32	19
2000	51	20
2001	37	19
2002	31	16
2003	16	0
Total	220	102

To estimate the regressions in section 2.3, I relax the restriction of having at least two years of information before and two years after the deal to one year before and one year after. This change increases the sample to 132 cross-border deals for the period between 1994 and 2003.

3.2 Control indices

As it was described in section 2.2, to calculate the change in performance before and after a cross-border acquisition, I have to control for overall changes in banking activity at the country level. This study uses the same methodology as Cornett and Tehranian (1992) and Linder and Crane (1992), and calculates banking industry indices for each country in the sample.

The selection of banks included in these indices starts with the sample of non-acquired banks described in the previous sub-section. Countries with less than five banks with non-missing information in any year between 1994 and 2004 are excluded. With this sample of banks, averages for the relevant performance and income statement variables are computed. These indices by country and variable are used as the counterfactual to the target banks' profitability measures.

In section 2.3, $y_{ijt}$ was defined as a percentile rank transformation of the performance ratios. The peer group used to calculate these ranks is the same sample of banks used to compute the industry indices by country.

4 Results

4.1 Determinants of cross-border acquisitions

Table 4 shows the results of the probit estimation described in equation (1). Columns (1) through (3) include bank, country, and banking market characteristics as regressors. These columns differ in the performance proxy used in the estimations. The coefficients for ROA and ROE are negative, and positive and significant for the Cost to Income Ratio. All this coefficients are significant at the 1% level. This finding suggests that there is a higher probability for ex ante poorly performing banks of being acquired in a cross-border deal. In addition, larger banks are more likely to be targets, especially if they are located in smaller countries with low levels of financial intermediation. This is supported by the coefficients on Log Assets, Log GDP, and Private Credit to GDP, respectively. Finally, Concentration has a positive and significant coefficient, with a similar level across the three columns.

Table 4: Determinants of cross-border acquisitions

The empirical model in equation (1) has been estimated using a probit specification. The dependent variable equals one if a bank is acquired by a foreign institution in year and zero otherwise. The model is explained in section 2.1; the sample is defined in section 3.1. The model is estimated for the 1994-2003 period. Columns (1) through (6) differ in the performance proxy included. In columns (1) and (3) profitability is measured by the Return on Average Assets (ROA). Columns (2) and (5) include the Return on Average Equity (ROE). In columns (3) and (6) performance is defined as the Cost to Income Ratio. Columns (4) to (6) include Financial Development proxies in addition to the variables included in the first three columns. All estimations include time fixed effects.

Dependent Variable	ROA (1)	ROE (2)	Cost to Income Ratio (3)	ROA (4)	ROE (5)	Cost to Income Ratio (6)
Performance	-0.0511***	-0.0060***	0.0052***	-0.0430**	-0.0053***	0.0054***
Performance: std error	[0.0160]	[0.0018]	[0.0011]	[0.0180]	[0.0019]	[0.0013]
Log Assets	0.0815***	0.0821***	0.0909***	0.0609**	0.0616**	0.0723***
Log Assets: std error	[0.0212]	[0.0211]	[0.0209]	[0.0248]	[0.0249]	[0.0248]
Equity to Assets	0.0015	-0.0005	0.0009	0.0003	-0.0012	0.0000
Equity to Assets: std error	[0.0021]	[0.0021]	[0.0021]	[0.0026]	[0.0024]	[0.0026]
Net Loans to Assets	-0.0005	-0.0006	0.0002	0.0000	-0.0001	0.0007
Net Loans to Assets: std error	[0.0016]	[0.0016]	[0.0016]	[0.0021]	[0.0021]	[0.0020]
Non-Interest Income to Total Income	0.0733	0.0714	0.0066	0.1522	0.147	0.0475
Non-Interest Income to Total Income: std error	[0.1161]	[0.1176]	[0.1132]	[0.1153]	[0.1182]	[0.1177]
Log GDP	-0.0830***	-0.0838***	-0.0909***	0.0018	0.0073	-0.0073
Log GDP: std error	[0.0202]	[0.0204]	[0.0197]	[0.0510]	[0.0518]	[0.0512]
GDP Per Capita Growth	-0.0073	-0.0071	-0.0068	-0.0090**	-0.0092**	-0.0092***
GDP Per Capita Growth: std error	[0.0056]	[0.0059]	[0.0063]	[0.0037]	[0.0038]	[0.0035]
Inflation	-0.0033	-0.0035	-0.0034	-0.0022	-0.0028	-0.0029
Inflation: std error	[0.0049]	[0.0051]	[0.0048]	[0.0069]	[0.0072]	[0.0068]
Private Credit to GDP	-0.4325***	-0.4388***	-0.3890***	-0.4480***	-0.4748***	-0.4069***
Private Credit to GDP: std error	[0.1062]	[0.1081]	[0.1059]	[0.1209]	[0.1242]	[0.1190]
Banking Freedom Index	-0.0029	-0.0026	-0.0026	0.0006	0.0006	0.0011
Banking Freedom Index: std error	[0.0024]	[0.0024]	[0.0024]	[0.0030]	[0.0030]	[0.0033]
Concentration	0.9003***	0.8970***	0.8453***	1.2550***	1.3021***	1.2055***
Concentration: std error	[0.2034]	[0.2064]	[0.2086]	[0.3203]	[0.3246]	[0.3212]
Market Cap. to GDP	-	-	-	-0.0011	-0.001	-0.0011
Market Cap. to GDP: std error	-	-	-	[0.0007]	[0.0007]	[0.0007]
Priv. Bond Mkt. Cap. to GDP	-	-	-	-0.2482*	-0.2553*	-0.2237
Priv. Bond Mkt. Cap. to GDP: std error	-	-	-	[0.1314]	[0.1316]	[0.1374]
Pub. Bond Mkt. Cap. to GDP	-	-	-	-0.2268	-0.2267	-0.2433
Pub. Bond Mkt. Cap. to GDP: std error	-	-	-	[0.2930]	[0.2922]	[0.2923]
Observations	20575	20554	20575	16776	16762	16776
Countries	66	66	66	33	33	33
LR chi2	228.5	227.4	280.6	758.3	816.9	933.6
Pseudo R²	0.09	0.09	0.10	0.09	0.09	0.10

Robust standard errors clustered by country in brackets.
* significant at 10%; ** significant at 5%; *** significant at 1%.

The results on the performance variables could be explained, as in Vander Vennet (2002), by the expected comparative advantage of international banks in managing large financial institutions. Better technology, geographical diversification, and management skills are factors that may induce MNBs to acquire targets of considerable importance in local market where they could exert some market power and turn around the profitability ratios. The result on the relation between bank concentration and the probability of a cross-border deal differs from the evidence found in Focarelli and Pozzolo (2005). These authors find that this variable has a negative effect on cross-border bank entry using a sample of OECD countries. Nevertheless, their results only apply to the distribution of cross-border holdings of OECD banks in 1998, rather than a dynamic analysis of entry across years.

Columns (3) through (6) include three additional proxies for financial development. Missing observations reduce the number of countries and deals covered from 66 to 33 and from 214 to 125, respectively. The coefficients on the performance measures are still significant, and with the same sign as in previous estimations. The coefficient on Priv. Bond Mkt. Cap. to GDP enters with a negative and significant sign in two out of the three estimations. More developed capital markets compete with the banking sector in the allocation of financial resources. Firms' access to arm's length finance reduces the banks' market power and makes entry less attractive for international banks.²²

In Table 5 I estimate the model described in section 2.1 dividing the sample between potential targets located in emerging and developed economies. Columns (1) through (3) show the results for the former group. As in Table 4, the coefficients for the three performance proxies, bank size, Private Credit to GDP, and concentration are significant. These results suggest that MNBs are attracted to poor performing large banks in concentrated banking markets with low levels of financial intermediation. Columns (4) through (6) display the same estimations, restricting the sample to developed economies. In this case, performance and concentration have significant coefficients. In contrast to the estimations including banks in emerging economies, GDP per capita growth has a negative and significant coefficient. This result implies that there is a higher probability of cross-border acquisitions taking place in in years and countries that are growing at a slower place. Another interesting finding comes from the value of the coefficient on Non-Interest Income to Total Income. It is positive in the three estimations and significant in two, and differs from the values observed in emerging economies. This result implies that acquirers target banks with a significant revenue stream that is not tied to interest income in developed countries. It is consistent with a larger reliance on income from fees tied to capital markets in these countries.

Table 5: Determinants of ross-border acquisitions: Emerging vs. Developed Economies

Dependent Variable	Emerging Economies: ROA (1)	Emerging Economies: ROE (2)	Emerging Economies: Cost to Income Ratio (3)	Developed Economies: ROA (4)	Developed Economies: ROE (5)	Developed Economies: Cost to Income Ratio (6)
Performance	-0.0438**	-0.0050**	0.0050***	-0.0786***	-0.0090***	0.0052***
Performance: std error	[0.0197]	[0.0022]	[0.0015]	[0.0239]	[0.0021]	[0.0015]
Log Assets	0.1675***	0.1662***	0.1748***	0.0169	0.0195	0.0267
Log Assets: std error	[0.0278]	[0.0278]	[0.0279]	[0.0294]	[0.0292]	[0.0275]
Equity to Assets	0.0034	0.0014	0.003	0.0004	-0.002	-0.0005
Equity to Assets: std error	[0.0027]	[0.0027]	[0.0026]	[0.0033]	[0.0032]	[0.0034]
Net Loans to Assets	-0.0005	-0.0007	0.0005	0.0001	0.0001	0.0006
Net Loans to Assets: std error	[0.0019]	[0.0019]	[0.0018]	[0.0027]	[0.0028]	[0.0026]
Non-Interest Income to Total Income	-0.1621	-0.1775	-0.1903	0.2514**	0.2652**	0.1456
Non-Interest Income to Total Income: std error	[0.2120]	[0.2111]	[0.1826]	[0.1150]	[0.1147]	[0.1255]
Log GDP	-0.0784**	-0.0810**	-0.0837**	-0.0521**	-0.0579**	-0.0649**
Log GDP: std error	[0.0382]	[0.0380]	[0.0378]	[0.0238]	[0.0245]	[0.0260]
GDP Per Capita Growth	0.0183	0.0185	0.021	-0.0124***	-0.0126***	-0.0123***
GDP Per Capita Growth: std error	[0.0151]	[0.0151]	[0.0148]	[0.0020]	[0.0020]	[0.0020]
Inflation	-0.0051	-0.0052	-0.0051	0.0204	0.0214	0.0224
Inflation: std error	[0.0057]	[0.0058]	[0.0058]	[0.0387]	[0.0384]	[0.0391]
Private Credit to GDP	-0.6998***	-0.7068***	-0.6214***	-0.3034*	-0.3313**	-0.2559
Private Credit to GDP: std error	[0.2355]	[0.2311]	[0.2211]	[0.1586]	[0.1634]	[0.1634]
Banking Freedom Index	0.0012	0.0011	0.0014	-0.0044	-0.004	-0.0046
Banking Freedom Index: std error	[0.0029]	[0.0030]	[0.0030]	[0.0029]	[0.0031]	[0.0031]
Concentration	0.8089**	0.7842**	0.8174**	0.9287***	0.9190***	0.8406***
Concentration: std error	[0.3545]	[0.3581]	[0.3638]	[0.2211]	[0.2314]	[0.2419]
Observations	6192	6173	6192	14383	14381	14383
Countries	45	45	45	22	22	22
LR chi2	113.5	127.7	150.9	4448	6153	1830
Pseudo R²	0.06	0.07	0.07	0.08	0.08	0.08

Robust standard errors clustered by country in brackets.
* significant at 10%; ** significant at 5%; *** significant at 1%.

4.2 Performance effect

This section displays the results for the difference-in-difference estimations described in section 2.2. Tables 6 through 8 provide distributional characteristics on the acquired banks (Targets), control country-indices (Industry), and on the differences between these two measures (Targ-Ind). The columns headings in Tables 7 and 8 indicate pre-acquisition (Before), acquisition-year (Yr0), post-acquisition (After), and changes (Change) in the performance and income statement items of target banks. The Sign Test statistically evaluates the null hypothesis of a median equal to zero for Targ-Ind in each one of the target bank's acquisition stages.²³

Table 6 shows summary statistics for the sample of 102 deals in the two pre-acquisition years and compares them to the country-industry indic