Finding evidence of Fire-sale
FDI during the global financial
crisis and Euro-crisis: A
firm-level analysis of acquisitions
targeted at the Netherlands,
Sweden, and Greece.
R.A. Prikken (1775839)
Pelsterstraat 10-3, 9711KL, Groningen
roelandprikken@gmail.com, +31(0)651542680
This paper is written by R.A. Prikken as a dissertation for the master programme International Financial Management. The master programme is affiliated to the faculty of economics and business at the Rijksuniversiteit Groningen. This master thesis is written under supervision of Dr. H. Vrolijk, who I thank for sharing his insight and expertise.
Supervisor: Dr. H. Vrolijk Co-assessor: Dr. C.L.M. Hermes Responsibility for all errors in this paper lies solely with the author. Date of submission: 19/06/2015
Abstract
This study uses firm-level variables to examine effects of fire-sale FDI in acquisitions targeted at the Netherlands, Sweden and Greece. First, univariate analysis reveals firm-level variables that make firms more prone to be targets during the financial crises. Second, two Probit models find evidence of fire-sales during the Euro-crisis and not during the global financial crisis. Interestingly, firm-level liquidity has a positive effect on the probability of being acquired by an outsider. Third and foremost, results indicate that firms with lower asset tangibility and higher leverage are more prone to be targets in fire-sale acquisitions.
Keywords: Fire-sale, acquisitions, tangibility, leverage, liquidity, financial crisis, Euro-crisis
Table of content
1. Introduction ...1
2. Literature review ...2
2.1. Fire-sale theory ...2
2.2. Fire-sale foreign direct investment ...3
2.3. Hypotheses ...5
3. Data and Methodology ...9
3.1. Firm-level variables and operationalisation of the hypotheses ...9
3.2. Sample ...10
3.3. Variables ...12
3.4. Analysis ...13
3.5. Constructing the Probit models ...15
4. Results ...18
4.1. Results of the univariate analysis ...18
4.2. Results of the multivariate analysis ...23
5. Discussion ...28
5.1. Comparing the findings with existing literature ...28
5.2. Implications ...32
5.3. Limitations ...33
6. Conclusion ...34
6.1 Recommendations for future research ...35
7. References ...37
Appendix A – Liquidity spirals ...41
Appendix B – Time periods and dates related to the financial crises ...42
Appendix C – Selected countries ...43
Appendix D – Collecting and constructing leverage and tangibility ...44
Appendix E – Descriptive statistics ...45
Appendix F – Graphs...47
Appendix G – Correlation matrix ...53
Appendix H – Welch’s t-tests ...54
Page | 1
1. Introduction
In the history of economics and the history of our planet, crises have always been crucial moments for both organisms and organisations. The strong overtake the weak and those who possess the critical elements for survival prevail, while others perish. Newspapers covering the recent economic crises indicate weak European firms were overtaken by foreigners. The Independent (2012) published: “Everything Must Go! The Great European Fire Sale” (The Independent, 2012). The article covers a story about discounted sales of companies and company assets in Europe, where the weak are cheaply bought by the strong in a regional fire-sale. Research on fire-sales in Europe however, fails to support this claim. Shleifer and Vishny (1992) introduce fire-sales on industry-level, prompted by an industry-wide financial fire. The fire (i.e. financial shock) forces industry-insiders to liquidate assets while all other industry-insiders are financially encumbered. This leads to heavily discounted asset sales to outsiders and creates high private and social (i.e. public) costs, hence the term fire-sale. Krugman (2000) was the first to hypothesise the international applicability of the concept, in his construction of the theory of fire-sale Foreign Direct Investment (FDI). The distinction between the insiders and outsiders herein is regional. Aguiar and Gopinath find the first empirical evidence of fire-sale FDI in 2005. Later, Acharya, Shun and Yorulmazer (2011) and Alquist, Mukherjee, and Tesar (2013) focus on the Asian and South-American crises and find both supporting and contradicting evidence of fire-sale FDI. These mixed results and the recent financial turmoil make it a very interesting topic for me from an academic perspective. Additionally, the possible implications of the results could provide insights to business professionals.
Page | 2 In this research, I will be the first to incorporate firm-level variables related to fire-sale FDI, and one of the first to apply fire-sale theory to the recent global financial crisis and Euro-crisis. This research aims to answer the following question:
“Is there evidence fire-sale FDI happened in Europe during the global financial crisis and Euro-crisis, and which (fire-sale related) firm-level variables determine who is targeted?”
To answer the main question, I will first identify firm-level variables related to fire-sales and financial crises. Second, I will test whether certain firm-level variables change during financial crises, whether the changes differ between targets and acquirers, and determine what makes firms more probable targets. Third, I will test whether the firm-level variables related to fire-sales are important determinants of becoming a target of an outsider during the global financial crisis and Euro-crisis.
The quest for answers commences in the remaining chapters. Chapter 2 features a review of existing literature; it identifies firm-level variables with potential and presents the hypotheses. Chapter 3 introduces the data sample, explains the methodology, and justifies the selected time periods. This is followed by chapter 4, which presents the results of the analyses. The results and their implications are discussed in chapter 5. Chapter 6 concludes the findings of this paper and presents recommendations for future research.
2. Literature review
This chapter provides the theoretical foundation of this paper and reviews literature. Furthermore, it explains the concepts, and builds the hypotheses. The first section covers the basic concepts related to fire-sale theory. These fire-sale principles are extended to a geographical scale in the second section. The third section of this chapter will identify the variables relevant for this research and present the hypotheses.
2.1. Fire-sale theory
Page | 3 that operate in the same industry. Shleifer and Vishny (1992) label these competitors “insiders”. Insiders have the best use for the assets and should be able to offer the highest price.
During an industry-wide shock however, the closest competitors are often facing liquidity problems themselves. This makes insiders unable to raise enough capital to buy the assets and forces the seller to look for a buyer outside the industry (Shleifer and Vishny, 1992). These so-called “outsiders” will only purchase the assets at a discount because they cannot use the assets at their best use. Outsiders lack the knowledge, complementary assets, and other essentials to get the highest value from the assets. The reduction in asset value is a financial loss known as the private cost of a fire-sale. The social costs of fire-sales result from not managing assets at their best use and the effects of one fire-sale on all similar assets in the industry (Shleifer and Vishny, 1992).
Consistent with the model of Shleifer and Vishny (1992), Oh (2013) identifies fire-sale effects in acquisitions of financially distressed targets in a distressed industry. His empirical research shows an intra-industry contagion effect in case of fire-sale acquisitions1. The first forced liquidation sends a shock through the distressed industry, signalling a decreased demand and low reference prices (Oh, 2013). The significance of this effect becomes more apparent when the industry-wide shock is so severe that not one, but many distressed organisations offer their assets simultaneously. The contagion effect combined with the simple economics of supply and demand form the first endogenous, or self-enforcing effect of fire-sales.
The fire-sale effects are exacerbated when financial institutions decide to limit or stop extending credit to a distressed industry or market. Financial institutions, for instance banks, often require a certain amount of collateral (e.g. assets) when giving out a loan, specifically, an asset-backed loan. The value of the assets influences the amount an organisation can borrow. Hence, the devaluation of assets negatively impacts the insiders’ ability to get credit. This side-lines insiders and decreases the financial position of industry insiders (Shleifer and Vishny, 2011).
2.2. Fire-sale foreign direct investment
Fire-sale FDI is foreign direct investment targeted at a region where assets are sold for heavily discounted prices due to the financial conditions in the region. The target region is hit by a financial crisis
Page | 4 leading to decreasing asset values, diminishing liquidity, and decreased credit conditions. Financially distressed organisations in the region have to liquidate assets, while other insiders are simultaneously side-lined. The remaining buyers are outsiders that are only willing to purchase the assets at a discount. These outsiders will come from a region less severely struck by the crisis, where credit and liquidity conditions are better, allowing them to engage in fire-sale FDI.
2.2.1. Research on fire-sale FDI
Krugman (2000) was the first to approach fire-sales through this international perspective. He got the impression that fire-sales happened during the Asian crisis of 1997, because FDI into the region increased while domestic and other foreign capital left the area. Other authors adopted his idea of fire-sale FDI and conducted empirical research on the Asian crisis of 1997, the South-American crisis of 1995, and more recently, the global financial crisis and Euro-crisis. These studies focus on the importance of country-level differences in the research of fire-sale FDI.
Aguiar and Gopinath (2005) were the first to empirically research fire-sale effects during the Asian crisis. They find an increase in FDI-inflow despite a decrease in domestic investment. More specifically, they find an increase in foreign merger- and acquisition-activity while the number of domestic mergers and acquisitions decreased. Acharya et al. (2011) research the same crisis and find an increase of foreign investments. They also find that FDI and Foreign Portfolio Investment (FPI) patterns diverge during the crisis. Investors use small FPI to hedge their investment portfolios against financial risks. FPI is based on different motivations, hence the importance of distinguishing it from FDI (Acharya et al., 2011; Weitzel et al., 2014). Acharya et al. (2011) find a drop in FPI and an increase in the frequency and value of Foreign Direct Investments. The authors find that these developments are the result of decreasing asset prices in the affected region and relatively better access to capital for foreign investors. Alquist et al. (2013) examine the Asian and Latin-American crises. They find that the proportion of foreign acquirers increased during these crises. Essentially, Alquist et al. (2013) find an increased likelihood that the acquirer is an outsider.
2.2.2. Fire-sale FDI during the global financial crisis and the Euro-crisis.
Page | 5 In terms of Shleifer and Vishny (1992), there are no outsiders and therefore fire-sales do not occur. Additionally, Desbordes and Wei (2012) see an underdeveloped financial system as a prerequisite of fire-sale FDI. Contrary to Desbordes and Wei (2012), Munnichs (2014) finds evidence of fire-fire-sales in the PIIGS2 countries, particularly during the European sovereign debt crisis (henceforth Euro-crisis). The Euro-crisis perpetuated the recession and crisis in Europe, especially in the Eurozone counties. This makes the Euro-crisis more feasible for fire-sale FDI because it creates an affected region with insiders and a less- or un-affected region with outsiders.
Weitzel et al. (2014) research the Euro-crisis, argue in favour of regional differences, and focus on European countries to shine light on the matter. The authors find no extraordinary patterns in cross border M&A’s and fire-sale FDI in Europe. Weitzel et al. (2014) acknowledge that the selection of only European countries is the paper’s biggest limitation. The interconnectedness of the European financial market limits both differences between countries and potential support of fire-sale theory. The concentrated nature of the Euro-crisis, however, offers potential for outsiders. Therefore, global acquirers are included in the research of this paper.
2.3. Hypotheses
Current literature does not research firm-level variables that particularly benefit or harm firms during a financial crisis and related fire-sales. These variables must exist, for it is clear that not all firms in an affected region are equally subjected to crisis or fire-sale effects. Contrary to previous research on acquirers and country-level variables, the focus here is on firm-level variables that make firms in the target region particularly vulnerable to fire-sales and financial crises.
Brunnermeier and Pedersen (2009) explain how sudden decreases in market liquidity (i.e. ease of liquidating assets) and funding liquidity (i.e. ease of securing funding) influence firms and asset values. Appendix A elaborates on Brunnermeier and Pedersen (2009); this can be reviewed for the workings of the margin-spiral and loss-spiral. The authors find that decreases in market liquidity and funding liquidity
Page | 6 could lead to self-enforcing liquidity spirals. Spirals that diminish country-level liquidity and severely affect asset values. Interestingly, firms with certain characteristics are more heavily affected by diminishing macro-liquidity (i.e. financial crisis). Research on fire-sales and the liquidity spirals provides the basis to identify relevant firm-level variables.
Liquidity is the first firm-level variable that could make a difference during a crisis. Liquidity
refers to the amount of firm assets that can be easily transferred to cash or cash equivalents. During a financial crisis, liquidity is expected to differ between targets and acquirers, and expected to be a driver of FDI. Alquist, Mukherjee and Tesar (2014) research liquidity-driven FDI in emerging countries and find that foreign firms’ liquidity positions are often a key determinant of foreign mergers and acquisitions. Normally, this is not feasible in a developed market due to developed financial systems and available liquidity. However, if a financial crisis encumbers the functioning of the financial system and decreases macro-liquidity, I expect developed countries would also be subjected to liquidity-driven FDI. Nevertheless, if a firm’s liquidity position is sufficient to cushion an external shock, the firm could avoid the negative effects of a crisis and liquidity spirals (Brunnermeier and Pedersen, 2009). The above leads to hypothesis one:
H1: During a financial crisis, firm-level liquidity is inversely related to the probability of forced liquidation of assets and the probability of becoming a target of fire-sale FDI.
Leverage is the second firm-level variable of interest and relates to the capital structure of firms.
Organisations that use too much debt in their capital structure prior to a crisis experience difficulties when the crisis hits (Brunnermeier and Pedersen, 2009). One problem arises from the models firms use to determine optimal capital structures. These models account for financial risks but have proven to be flawed and unfit to account for systemic-risk (e.g. the contagion effect of fire-sales). It is exactly the systemic risk that has made highly leveraged firms and the financial system in the Eurozone so vulnerable (Brunnermeier, M., Crockett, A., Goodhart, A., Persaud, A.D., Shin, H. 2009). The risk assessment models assume that changes in one firm do not influence the environment as a whole, while fire-sale theory is based on the opposite (Shleifer and Vishny, 1992; Oh, 2013). The contagion effect is amplified by mandatory marked-to-market accounting, which values assets based on market conditions.
Page | 7 economic crisis hits, the underestimated risk materialises and marked-to-market valuation now works oppositely. The decrease in market conditions decreases asset values while the debt remains the same. The deteriorating balance sheets decrease the ability to secure new credit and increase the need to de-leverage and sell assets. This suggests that financial crises and the contagion effect both affect highly leveraged firms more. This leads to hypothesis two:
H2: During a financial crisis, leverage is positively related to the probability of forced liquidation of assets and the probability of becoming a target of fire-sale FDI.
Short-term debt is the third variable of interest and stands for the amount of debt that materialises
within a year. When the economic climate takes a turn for the worse, short-term financing opportunities become more expensive or disappear. Firms with more short-term debt are more susceptible to the margin- and loss-spirals (Brunnermeier and Pederden, 2009), which further complicates avoiding fire-sales. Ferrando and Mulier (2013) discover that perceived financial constraint increases with short-term debt levels during the crisis in Europe. It is important to note that this is based on the perception of financial constraints on Small and Medium Enterprises (SME’s). Love, Preve and Sarria-Allende (2007) research the Asian crisis and find an increased vulnerability to economic imperfections for firms holding more short-term debt. This leads to the third hypothesis:
H3: During a financial crisis, short-term debt is positively related to the probability of forced liquidation of assets and the probability of becoming a target of fire-sale FDI.
Asset tangibility is the fourth variable that influences firms differently during a financial crisis.
Asset tangibility refers to the extent to which assets have physical form, influencing the degree to which an asset’s value is easy to determine. The value of tangible assets is often easier to determine and more constant than the value of intangible assets (e.g. intellectual property and brand reputation) (Alquist et al., 2014). Overall, the tangibility of a firm’s assets is expected to influence the firm’s financial position three-fold.
Page | 8 ambiguous values and the specificity of the assets. Brunnermeier and Pedersen (2009) describe a second effect of asset tangibility related to the liquidity spirals: their model finds that margins increase more in illiquidity when the asset is more intangible in nature. Third, tangible assets often have a higher fundamental value and are easier to pledge as collateral. This makes it easier to secure a loan (Alquist et al., 2014) and avoid asset sales. Alquist et al. (2014) find that liquidity-driven FDI in emerging countries is focussed on industries with lower asset tangibility. These industries tend to be more financially constrained and thus more likely to be undervalued. This leads to hypothesis four:
H4: During a financial crisis, tangibility of assets is inversely related to the probability of forced liquidation of assets and the probability of becoming a target of fire-sale FDI.
Size can influence a firm’s ability to access external financing and a firm’s ability to hedge against
certain economic shocks. Large firms in general are less dependent on the banking system to receive external financing, while SME’s face higher funding costs and are more dependent on bank-lending (ECB bulletin, 2014). In July of 2014, the ECB published an article that states that SME’s are more often rejected for loans (3% of SME’s report being rejected, against 1% of large firms). Additionally, the ECB finds that the possibility of being rejected for a loan deterred 6% of SME’s from applying, against 2% of large firms.3 In line with the ECB’s bulletin, Kuntchev, Ramalho, Rodriguez-Meza, and Yang (2014) find that the probability of being credit-constrained decreases with firm size. Although size is not a variable directly linked to fire-sale, it makes a suitable control variable. This leads to hypothesis five:
H5: During a financial crisis, firm size is inversely related to the probability of forced liquidation of assets.
Profitability is the sixth firm-level variable of interest. The relevance of variables related to
fire-sales is based on the expectation that the drivers of FDI change. Not profitability but liquidity becomes the driver of the transactions. Fire-sale theory is focussed on liquidity-driven FDI instead of profit-driven FDI. Following the fire-sale theory rationale, I would expect profitability to play a less significant role in FDI during a financial crisis. In essence, if the profitability of two firms is similar during a crisis, the other variables related to fire-sales are expected to determine which firm attracts FDI. Profitability can therefore, analogous to size, be regarded as a control variable. The above leads to hypothesis six:
Page | 9
H6: During a financial crisis, profitability becomes a less important driver of FDI.
3. Data and Methodology
This chapter covers the data collection and methodology of this paper. Since no available framework or guide is available in current literature, it is paramount that the analysis starts with the basics before researching fire-sale effects. The chapter starts with the first section that operationalises the hypotheses to better suit analysis. The second section covers the sample construction and data collection, complemented by a justification of the selected time-periods and regions. Furthermore, the third section operationalises the variables for the analyses. This methodology behind the analyses is presented in the fourth section of this chapter. The concluding section discusses the selection and construction of the statistical model.
3.1. Firm-level variables and operationalisation of the hypotheses
Throughout fire-sale research, authors focus on regional differences and country-level variables that determine which regions become more active suppliers of FDI (see, e.g. Aquiar and Gopinath, 2005; Alquist et al., 2013; Weitzel et al., 2014). Alquist et al. (2014) show for emerging countries that industries with certain characteristics attract more FDI during a crisis. I combine the findings of these authors to construct operationalised hypotheses and research FDI based on firm-level variables in targets.
The majority of empirical research on the topic requires a clear distinction between outsider and insider, or between target and investor (Aquiar and Gopinath, 2005; Acharya et al., 2011; Alquist et al., 2013; Munnichs, 2014). This immediately points at acquisitions and cancels out other forms of FDI including joint-ventures or Greenfield investments4. I recognise the potential confusion in the term fire-sale FDI, even though this is the terminology used by Krugman (2000) and Weitzel et al. (2014). To eliminate further confusion, the remainder of this paper will focus on fire-sale acquisitions to provide insight in fire-sale FDI5. Adopting the same distinction and terminology both limits confusion, and it benefits the comparability and reproducibility of this research. Other terminology incorporated from
4 Contrary to statements of other authors (see, e.g. Weitzel et al., 2014; Alquist et al. 2013), I expect that analysis of
Greenfield investments could provide new insights in fire-sales. The limited availability of data excludes it for this paper.
5 This will also provide insight in fire-sale FDI in general, because mergers and acquisitions do usually account for up to 80%
Page | 10 sale papers is the distinction between insiders and outsiders (Shleifer and Vishny, 1992; Shleifer and Vishny, 2011).
This paper aims to find evidence of fire-sales by regarding firms in target regions insiders and firms outside the affected region outsiders. The analysis tests whether firm-level variables make firms more prone to be targets during a crisis. Furthermore, it will test whether these variables increase the probability of being acquired by an outsider. To research this, I have operationalised the hypotheses as presented in Table 3.1 below.
Table 3.1. Operationalised hypotheses
3.2. Sample
The sample includes a time-period before the crisis, the financial crisis and the Euro-crisis. I start collecting data after the introduction of the Euro on the 1st of January 1999 and stop at the latest available
data on the 31st of December 2014. For further analysis, the following four periods are defined:
Pre-crisis (1999-2007) Post-crisis6 (2008-2014)
Financial-crisis (2008-2009) Euro-crisis (2010-2013)
6 I do not want to imply that the Post-crisis period are non-crisis years, the Post-crisis period includes all available data after
the financial crisis hit (2008-2014). Hypotheses
H1 Firms with lower liquidity are more prone to be a target (of an outsider) during a financial crisis. H2 Firms that are highly leveraged are more prone to be a target (of an outsider) during a financial
crisis.
H3 Firms that hold relatively more short term debt are more prone to be a target (of an outsider)
during a financial crisis.
H4 Firms with lower asset tangibility are more prone to be a target (of an outsider) during a financial
crisis.
H5 Smaller firms will be more prone to be a target during a financial crisis.
Page | 11 The start of 2008 marks the beginning of the Financial-crisis and the Post-crisis period. Weitzel et al. (2014) select the first quarter of 2008 in their research, yet find no difference when testing with the 3rd quarter of 2007, or the 3rd quarter of 2008 in their robustness test. Based on the events in Europe, and in accordance with Munnichs (2014), I take the 1st of January 2010 as the start of the Euro-crisis. The end of 2009 marks the end of the Financial-crisis period, and the end of 2013 marks the end of the Euro-crisis period. A more elaborate overview of relevant data and events related to the timeframes can be reviewed in Appendix B.
The selected target countries are Greece, the Netherlands and Sweden. These countries are relevant for their location in Europe, the differences between their financial systems, and the degree the crises hit the countries. Appendix C features an elaborate explanation of why these countries are selected, and it explains the potential academic value of the differences. It is beyond the scope of this paper to thoroughly research the differences but it could provide an interesting opening for future research. Greece is selected for the severe impact the crises had on them. The Netherlands is chosen because it was affected by the Euro-crisis, although it is a country with prudent financial policies. The Euro-crisis affected the whole region mainly due to the interconnectedness of the financial system (e.g. systemic risk) and the incompleteness of the Economic and Monetary Union (EMU) (Brunnermeier, Crockett, Goodhart, Persaud, and Shin, 2009). In the EMU, countries lose control over the currency in which they issue debt and cannot guarantee unlimited liquidity. This creates sovereign default risks (Lane, 2012; Benetrix and Lane, 2012). Since Sweden still has their own currency, it is expected to be less affected by these issues. Additionally Sweden faced a banking crisis in 1990s similar to the subprime-crisis (Jonung and Hagberg, 2005). The restructuring of their banking system should have made it less susceptible to such crises. Appendix C provides more information about the countries, EMU, and systemic risk.
The data of the acquisitions are extracted from the Thomson Reuters SDC Platinum database7.
Most papers based on fire-sale theory and mergers and acquisitions are based on data from this database (see, e.g. Aquiar and Gopinath, 2005; Alquist et al., 2013; Weitzel et al., 2014). A transaction is included in the sample if one firm acquires a stake or increases their stake in an organisation in the Netherlands, Greece, or Sweden. This transaction must be completed between the 1st of January 1999 and the 31st of
7 All variables and related data are extracted from the Thomson Reuters SDC Platinum database. Henceforth, all
Page | 12 December 2014, and have a minimum acquired stake of 5%. The 5% threshold is chosen to exclude diversification investments (i.e. Foreign Portfolio Investments) and focus on actual acquisitions (Acharya et al., 2011; Weitzel et al., 2014). Additionally, I exclude transactions that are Leverage Buyouts (LBOs), exchange offers, spinoffs, self-tenders, recapitalisations, and repurchases of own shares. This limits disturbances of actions taken by firms to alter their own capital structure and allows for better analysis of acquisitions (Weitzel et al. 2014). After adding the search criteria above, the database finds 2833 acquisitions of which 1755 domestic transactions and 1078 cross-border acquisitions. This number is divided in 1579 observations for the Pre-crisis periods and 1254 for the Post-crisis period. The latter is divided in 448 observations for the Financial-crisis and 690 for the Euro-crisis. In total there are 859 acquisitions targeted at the Netherlands, 1381 acquisitions targeted at Sweden, and 593 acquisitions targeted at Greece.
3.3. Variables
Page | 13
Table 3.2 – All variables and their definitions
Variable Definition Target-specific variables Leverage8 LTD.SH ND.SH SH.TA Tangibility ENTVAL.TANBOOKV
Variables of both parties
Leverage8 N.DEBT.COM.P ST.DEBT.TOTASS TOTLIA.TOTASS Liquidity9 CURRENT Tangibility9 INT.TOTASS TANG.ASS Profitability9 EBITDA.TOTASS Size9 LN.TOTASS Industry and Country dummies10
FINANCIAL
Acquirer-specific variables
FOREIGN FOR.FIN ROW
Long-term debt scaled by shareholders’ equity. Net debt scaled by shareholders’ equity. Shareholders’ equity scaled by total assets.
Enterprise value scaled by the book value of tangible assets.
Net debt scaled by the total of common and preferred equity. Short term debt scaled by total assets.
Total liabilities scales by total assets.
Total current assets scaled by total current liabilities. Intangible assets scaled by total assets.
Total assets minus intangible- and current- assets divided by total assets. Earnings before interest tax depreciation and amortisation scaled by total assets. Natural logarithm of total assets.
Dummy based on SIC-code, (1) for financial firms with a code in range 60-67. Dummy based nationality of acquirer, (1) for foreign acquirer.
Dummy based on SIC-code and nationality, (1) for a foreign financial acquirer. Dummy based on nationality of acquirer, (1) for acquirer outside the Eurozone
Note: In the remainder of this paper and in the output of the statistical analysis, the variables are often preceded by “A.” or “T.”, indicating that the variable belongs to either an Acquiring firm or Target firm respectively.
3.4. Analysis
Analysis of the variables is completed through two general types of analysis. This section starts with the methodology of the univariate analysis and concludes with the methodology behind the multivariate analysis.
3.4.1. Univariate analysis
The first stage of the analysis consists of two parts of univariate analysis: a graphical and statistical analysis of the characteristics and developments of the individual variables over time. The graphs are
8 De Jong, Kabir, and Nguyen (2008)
Page | 14 based on calculated yearly averages. “R”11 is used to compute the means over time of all variables, and
used to extract the summary statistics. I use “R” for all analyses and tests in this paper12. The univariate analysis shows patterns in the data, and indicates changes over time in both acquirers and targets. To get further insight in the developments, the second part of the univariate analysis will test the means of the pre-crisis and post-crisis periods. T-tests are used to test whether the differences between means are statistically significant.
The data extracted from the SDC Platinum database are not complete for all firms, however a standard student’s t-test requires equal sample sizes and equal variances. (Welch, 1947; Ruxton, 2006). The Welch’s t-test is more suitable, since it is insensitive to unequal sample sizes and unequal variances (Welch, 1947; Ruxton 2006). The time periods tested are pre-crisis (1999-2007) and post-crisis (2008-2014). The Welch’s t-tests analyse whether the means of the variables during the years prior to the crisis are equal to the means of the variables after the financial crisis hit. The null hypothesis is thus “Pre-crisis mean” = “Post-crisis mean”. If the difference between the means is significantly bigger than zero, the null hypothesis is rejected: “Pre-crisis mean” ≠ “Post-crisis mean”. The Welch’s t-tests also indicate whether the value during the post-crisis period has increased or decreased relative to the pre-crisis years.
3.4.2. Multivariate analysis
The second phase is the multivariate analysis based on two Probit models. The prior researchers Aguiar and Gopinath (2005), Acharya et al. (2011), and Alquist et al. (2013) find a relative or absolute increase in outsiders. I will analyse which firm-level variables increase the chance of being purchased by an outsider. To assess this, first the response variable becomes the dummy variable: Foreign Acquirer (1), Domestic Acquirer (0). This focusses on cross-border acquisitions, while later the response variable is replaced by acquirers from outside the Eurozone13 (1) and acquirers inside the Eurozone (0). A regression that has a dummy variable as response variable is known as a regression with a dichotomous response variable, or a binary responds model (Abay, 2015). In this analysis, the explanatory variables are one natural logarithm and multiple ratios of continuous variables, thus interval data.
11 The programme is called “R: A Language and Environment for Statistical Computing” and is developed by “R Foundation
for Statistical Computing”, Vienna, Austria (http://www.R-project.org).
12 The programme requires programming skills and relies only on coded commands. Although difficult to learn at first, the
programme is a more suitable than the more user friendly alternatives, if the research requires computing many tests and re-running regression analyses. The scripts and codes used for all analyses in this paper can be requested from the author.
Page | 15 Models most commonly used for these kind of regressions are Logit and Probit models, also known as generalised linear models. Traditionally, Logit was more popular because its calculations were easier to do without a computer. Now, the divide seems only based on preferences and previous research. Logit models are more frequently applied in social sciences while Probit models are more frequently used for business research. Therefore, I have chosen to select the Probit model. As aforementioned, I will replace the variable A.FOREIGN with A.ROW to test whether fire-sale effects can be shown for transactions with acquirers from outside the Eurozone. I will adapt the Probit model accordingly. Additionally, Weitzel et al. (2014) exclude financial firms from their sample to remove firms under different regulatory standards. To tests whether this makes a difference for this research, the Probit models will be repeated with a data set that excludes financial targets14.
The second Probit model has A.ROW as dependent variable, where acquirers from outside the Eurozone are grouped together. In addition to the Eurozone countries, I have to decide whether I will include Swedish acquirers as insiders too. Even though Sweden does not have the Euro as their currency, the Swedish economy is highly interconnected with the Eurozone countries. The banking sector of Sweden, and the economy is linked to the Eurozone through trade agreements and membership of the European Union (EU). Membership of EU entails that the primary distinctions between Sweden and the Eurozone are the currency and related connection to the ECB. The results of the univariate analysis in chapter four will determine the role of Swedish transactions in the analysis based on Probit model (3.7).
3.5. Constructing the Probit models
The Probit model is a dichotomous responds model that limits the estimated probabilities
between zero and one. A general dichotomous responds model is shown in (3.1), for which the function G only takes on values that are between zero and one for all real numbers z (i.e. 0 < G (z) < 1). This is the advantage over linear probability models that would give non-sense predictions when the dependent variable is binary. G is known as a non-linear Cumulative Density Function (CDF) that is positively affecting the value of z. Then, (3.2) is the index through which “Pr (y = 1jx)” is a vector of x.
The non-linear function for G then determines whether it is the Logit of Probit model. For the Probit model, G is standard normally distributed, the probability of success is between zero and one, and CDF is non-linear. Therefore, G can be represented by the non-linear CDF in (3.3). This integral has the
Page | 16 standard normal density as shown in (3.4), also guaranteeing a probability of success between one and zero for all parameters and variables. The function of G increases rapidly when xβ diverges from zero, but the effect on G is negligible if xβ approaches an infinite value. This means the effects are not constant and are not as easily interpreted as a linear model.
In a Probit analysis, the marginal effect of one unit increase in an explanatory variable on the response variable differs with the value of the explanatory variable. This can be solved with R by using the function “probitmfx” to find the marginal effect of the explanatory variable on the responds variable. The function calculates the marginal effect when all explanatory variables are at their mean and one explanatory variable increases with one unit (Fernihough, 2014).
(3.1)
(3.2)
(3.3)
(3.4)
Based on the correlation matrix and the Welch’s t-tests I have selected the variables for the Probit analysis. I use the Akaike Information Criterion (AIC) as a measure of relative quality of the Probit model. R-squared or Adjusted R-squared are not viable measure of fit for a Probit model and a pseudo R-squared for the Probit model is difficult to interpret. Based on AIC, the programme automatically fits and remodels to find the best model relative to the alternatives15. The relative model fit based on AIC balances between a better fit and additional variables. The programme removes all variables but size from the model based on the total data set. It shows that for the aggregate data, size is the most relevant predictor. The
15 The data set has to be adapted to fit this method since only the transactions with all variables observed can be used to
Page | 17 programme does not provide enough support to omit a particular variable. The variables are based on different time-periods, different regions, and different numbers of observations. This complicates finding the best model fit. Manually removing and substituting variables and measurements in the model did not show significant improvement either. I therefore choose the variables based on the literature review and univariate analysis. Accordingly I find no obvious reason to assume violation of the assumptions of the Probit model, although it should be considered when interpreting the results.
To assess the effect of the explanatory variables on the response variable, I execute the Probit regression for the aggregate data and each country individually. This is repeated for each of the time-periods to assess the differences between them. The model used for the analysis is presented with general terms in (3.5), and the measurements that are used are presented in model (3.6). In this model, A.FOREIGN is the response variable for the explanatory variables for the i’th transaction with coefficient “
𝛽
” and error term“𝜀”.
R analyses only the transaction for which all variables are available. To test for evidence of fire-sales based on transactions with an acquirer from outside the Eurozone, A.ROW will replace A.FOREIGN and form the model presented in (3.7).Page | 18
4. Results
This chapter presents the statistical results of both the univariate and multivariate analyses. The descriptive statistics are presented first, followed by the Welch’s t-tests as the last part of the univariate analysis. The second section presents the correlation matrices and the results of the multivariate analysis, concluding this chapter with the marginal effects of Probit model (3.7).
4.1. Results of the univariate analysis
Table 4.1 presents the descriptive statistics of the variables most relevant for the second phase of the analysis. Appendix E presents the descriptive statistics of all variables. Both include the number of observations, the mean, the standard deviation, the minimum value, and maximum value. The descriptive statistics are based on observations of the variables in the aggregate data. The descriptive statistics show the most observations for Sweden. The data show that on average targets in the Netherlands have the highest leverage and the most foreign acquirers. In Greece, the targets have the highest short-term debt ratios, are largest, and most profitable. Swedish targets relatively hold the most intangible assets and have the highest liquidity.
Page | 19
Table 4.1. Descriptive statistics.
Variables: Leverage Short-term debt Tangibility Liquidity Size Profitability Region Measurement: T.TOTLIA.TOT T.STDEBT.TOT T.INT.TOT T.CURRENT T.LN.TOT T.EBITDA.TOT A.FOREIGN All countries Observations 1197 1197 1197 948 1197 1197 2833 Mean 0.5883 0.0956 0.1549 1.9516 4.9419 0.0211 0.3805 St.Deviation 0.4665 0.2928 0.2140 2.7765 2.5271 0.2269 0.4856 Minimum -0.1416 0.0000 0.0000 0.0000 -2.3026 -1.7843 0.0000 Median 0.5865 0.0209 0.0392 1.3772 5.0441 0.0306 0.0000 Max 10.5561 8.7659 0.9591 40.0000 14.4754 2.0923 1.0000 Netherlands Observations 287 287 287 208 287 287 859 Mean 0.7344 0.1195 0.1147 1.4296 5.6730 0.0611 0.4715 St.Deviation 0.8098 0.5397 0.1805 0.9806 3.0230 0.1930 0.4995 Minimum 0.0000 0.0000 0.0000 0.0000 -2.3026 -1.4377 0.0000 Median 0.6379 0.0216 0.0041 1.2627 6.1196 0.0458 0.0000 Max 10.5561 8.7659 0.7255 7.3563 14.4754 2.0923 1.0000 Greece Observations 234 234 234 198 234 234 593 Mean 0.5951 0.1593 0.0798 1.5808 5.7096 0.0796 0.2917 St.Deviation 0.2630 0.1889 0.1360 1.7047 1.9571 0.1289 0.4549 Minimum 0.0000 0.0000 0.0000 0.2198 1.5892 -0.4293 0.0000 Median 0.5924 0.0858 0.0095 1.2924 5.6553 0.0665 0.0000 Max 1.4774 0.9314 0.8227 17.6465 11.2856 0.5959 1.0000 Sweden Observations 676 676 676 542 676 676 1381 Mean 0.5239 0.0634 0.1980 2.2873 4.3658 -0.0161 0.3621 St.Deviation 0.2658 0.1164 0.2376 3.4352 2.3145 0.2578 0.4808 Minimum -0.1416 0.0000 0.0000 0.0000 -2.3026 -1.7843 0.0000 Median 0.5597 0.0046 0.0931 1.4939 4.3155 0.0050 0.0000 Max 2.0000 0.8638 0.9591 40.0000 12.3525 1.2655 1.0000
Note: Variables taken from target firms are preceded by “T.” and variables taken from acquiring firms are preceded by “A.”
4.1.1 Results of the Welch’s t-tests
Page | 20 distribution (i.e. bell curve). Based on the significance level, the variables will be referred to as slightly significant, significant, and highly significant for the 1%, 5%, and 10% level, respectively.Here, I will focus on the results related to the hypotheses. Appendix H explains the results for acquirers and targets. The tables present the hypothesis (for the targets) and related variable, the variable as used in the t-tests, the expected development based on the hypothesis and the observed development. If the post-crisis mean (Post-Mean) is bigger than the pre-crisis mean (Pre-Mean), the observation is indicated with a “+”, whereas an opposite development is indicated with a “-”. The tests provide evidence that more targets score lower or higher on a firm-level variable, indicating an increased or decreased likelihood of becoming a target.
For Liquidity, the tests show no significant results. For Leverage there are more results. For (TOTLIA.TOTASS), the tests for targets show a slightly significant increase in the Netherlands, a slightly significant decrease for Sweden, and a significant increase in Greece. The targets in Greece and the Netherlands are higher leveraged in the post-crisis period. This is partial support for hypothesis two. The decrease among Swedish targets could be related to the reforms that were implemented after their banking crisis in the 90s. For the Netherlands, I find a highly significant increase for LTD.SH in targets. TOTLIA.TOTASS is selected as the primary measurement of leverage, and thus used in the Probit model. In line with the third hypothesis of Short-term debt, the tests find a highly significant increase in short-term debt held by Greek targets.
The fourth hypothesis is related to Tangibility. In line with the hypothesis, TANG.ASS significantly decreases in both Dutch and Swedish target firms. Further support of the hypothesis follows from testing INT.TOTASS; the tests show highly significant increases for the aggregate data, Sweden, and Greece, while the increase is significant in the Netherlands. Based on the results, INT.TOTASS is selected as the primary measurement of Tangibility.
Page | 21 Additionally, I have tested the composition of acquirers based on nationality and industry. A relative increase in foreign acquirers is regarded evidence of fire-sales by Aguiar and Gopinath (2005), Alcharya et al. (2011), and Alquist et al. (2013). Alquist et al. (2013) argue that financial foreign acquirers are most inclined to benefit from fire-sales during the crisis. Here, the share of foreign acquirers increases in Greece and decreases in Sweden, while the share of foreign financial acquirers increase in all regions. Caution is needed when interpreting these results: the foreign acquirers are not necessarily outsiders because intra-European transactions are included, and therefore it is not proof of fire-sales.
The results for Swedish targets are in line with all but one hypothesis. There seems no clear divide between Swedish firms and the two Eurozone countries, except for the leverage variable. This suggests that the connections with Eurozone countries through trade and the EU had a strong impact on the economy. Based on these results I decide to include Sweden, together with other Eurozone countries, as insider in the analysis with Probit model (3.7).
Table 4.2. Comparison of the means (all countries)
Hypothesis - Variable Variable Expected Observed Pre-Mean Post-Mean t-statistic Targets: H4 – Tangibility T.INT.TOTASS + + 0.0963 0.2009 -9.1161*** H4 – Tangibility T.TANG.ASS - - 0.9274 0.5734 2.5052** H5 – Size T.LN.TOTASS - - 5.4421 4.5497 6.3033*** H6 – Profitability T.EBITDA.TOTASS - - 0.0582 -0.0079 5.2589***
Industry variable T.FINANCIAL 0.1457 0.1746 -2.0811**
Acquirers:
Short-term debt A.STDEBT.TOTASS 0.1012 0.0776 3.1104***
Tangibility A.INT.TOTASS 0.0890 0.1622 -6.1315*** Size A.LN.TOTASS 6.8134 7.2188 -2.1152** Profitability A.EBITDA.TOTASS 0.0801 0.0499 3.1097*** Leverage A.TOTLIA.TOTASS 32.5060 188.0249 -2.2614** Industry A.FINANCIAL 0.4522 0.5917 -7.4600*** Liquidity A.CURRENT 1.7450 2.6886 -1.9485* Industry x Nationality A.FOR.FIN 0.1425 0.1914 -3.4491***
Page | 22
Table 4.3. Comparison of the means (the Netherlands)
Hypothesis - Variable Variable Expected Observed Pre-Mean Post-Mean t-statistic Targets: H2 – Leverage T.TOTLIA.TOTASS + + 0.6508 0.8096 -1.7440* H2 – Leverage T.LTD.SH + + 0.6356 1.1596 -2.9175*** H4 – Tangibility T.INT.TOTASS + + 0.0920 0.1352 -2.0629** H5 – Size T.LN.TOTASS - - 6.1483 5.2448 2.5906** H6 – Profitability T.EBITDA.TOTASS - - 0.0848 0.0398 2.0473** Acquirers: Leverage A.NDEBT.COM.P 1.1400 0.6357 1.7975* Tangibility A.INT.TOTASS 0.1056 0.1975 -4.2000*** Liquidity A.CURRENT 1.6068 2.0436 -1.7156*
Industry x Nationality A.FOR.FIN 0.1595 0.2116 -1.9057*
Note: The *, **, and *** placed on the t-statistic scores represent statistical significance at the 10%, 5%, and 1% level. Variables taken from target firms are preceded by “T.” and variables taken from acquiring firms are preceded by “A.”
Table 4.4. Comparison of the means (Sweden)
Hypothesis - Variable Variable Expected Observed Pre-Mean Post-Mean t-statistic Targets:
H2 – Leverage T.TOTLIA.TOTASS + - 0.5469 0.5089 1.9262*
H3 – Short-term debt T.STDEBT.TOTASS + + 0.0518 0.0711 -2.2363**
H4 – Tangibility T.INT.TOTASS + + 0.1169 0.2509 -8.1408*** H4 – Tangibility T.TANG.ASS - - 0.9249 0.4423 2.2324** H5 – Size T.LN.TOTASS - - 5.0106 3.9448 6.1881*** H6 – Profitability T.EBITDA.TOTASS - - 0.0239 -0.0422 3.4578*** Industry T.FINANCIAL 0.1293 0.1822 -2.7216*** Acquirers: Tangibility A.INT.TOTASS 0.0906 0.1513 -3.5305*** Leverage A.TOTLIA.TOTASS 15.2831 162.0799 -1.8809* Industry A.FINANCIAL 0.4814 0.6780 -7.5352*** Nationality A.FOREIGN 0.4116 0.3150 3.7451*** Industry x Nationality A.FOR.FIN 0.1486 0.1822 -1.6828*
Note: The *, **, and *** placed on the t-statistic scores represent statistical significance at the 10%, 5%, and 1% level. Variables taken from target firms are preceded by “T.” and variables taken from acquiring firms are preceded by “A.”
Page | 23
Table 4.5. Comparison of the means (Greece)
Hypothesis – Variable Variable Expected Observed Pre-Mean Post-Mean t-statistic Targets:
H2 – Leverage T.TOTLIA.TOTASS + + 0.5584 0.6358 -2.2552**
H3 – Short-term debt T.STDEBT.TOTASS + + 0.1188 0.2042 -3.5014***
H4 – Tangibility T.INT.TOTASS + + 0.0563 0.1059 -2.7680*** H6 – Profitability T.EBITDA.TOTASS - - 0.1032 0.0534 3.0211*** Acquirers: Tangibility A.INT.TOTASS 0.0696 0.1268 -2.2732** Size A.LN.TOTASS 6.3737 7.3605 -2.4158** Profitability A.EBITDA.TOTASS 0.0867 0.0432 3.9677*** Industry A.FINANCIAL 0.4005 0.4925 -2.1317** Nationality A.FOREIGN 0.2449 0.3831 -3.3972***
Industry x Nationality A.FOR.FIN 0.1097 0.1891 -2.4894**
Note: The *, **, and *** placed on the t-statistic scores represent statistical significance at the 10%, 5%, and 1% level. Variables taken from target firms are preceded by “T.” and variables taken from acquiring firms are preceded by “A.”
4.2. Results of the multivariate analysis
Table 4.6 presents a correlation matrix with the variables eligible for Probit model (3.6). This correlation matrix does not include all firm-level variables from the hypotheses presented earlier. Due to multicollinearity in the complete correlation matrix, I have chosen to omit the short-term debt variable. Appendix G presents further explanation and the complete correlation matrix including the omitted variable. Table 4.6 shows there is no strong correlation between the response and explanatory variables, and there is no strong correlation between the explanatory variables.
Table 4.6. Limited correlation matrix
Variables taken from target firms are preceded by “T.” and variables taken from acquiring firms are preceded by “A.”
The results of the Probit analysis based on model (3.6) do not provide convincing support for the expectations and hypotheses. The results are inconclusive, sometimes even contradictory. Accordingly,
Region Leverage Liquidity Tangibility Size Profitability
A.FOREIGN T.TOTLIA.TOT T.CURRENT T.INT.TOT T.LN.TOT T.EBITDA.TOT
Page | 24 the model does not show which firm-level variables that make a firm more prone to be a target. Appendix I presents the results and findings of the Probit analysis based on model (3.6). The results for the global financial crisis are in line with Calderon and Didier (2009), who state that fire-sale theory is not applicable to the global financial crisis. All firms are insiders in their view, leaving no outsiders to engage in the fire-sales. Results are also in line with Desbordes and Wei (2012), who argue that fire-sale theory is only applicable to developing countries. The lack of fire-sale evidence is also in line with the paper of Weitzel et al. (2014) who focus on intra-European transactions only.
I, however, am not prepared to accept this conclusion without exhausting all available options. Based on the paper of Brunnermeier et al. (2009) and Weitzel et al (2014), the interconnectedness of the European financial market could limit analysis of fire-sale effects. This would mean that the distinction between insiders and outsiders in model (3.6) is incorrect. The cross-border acquisitions from other Eurozone countries also receive the outsider status in (3.6). Weitzel et al (2014) conclude that they find an interconnected European market without significant frictions or fire-sales between European counties. Following this reasoning and the interconnectedness of the European financial system (Brunnermeier et al. 2009), the Eurozone countries are insiders during the Euro-crisis. In model (3.6) the presence of too many insiders (i.e. Eurozone acquirers) could then distort the fire-sale effects of the outsiders (i.e. firms from a region less affected by the Euro-crisis). Therefore, to further test for fire-sale effects, I regard acquirers from outside the Eurozone as outsiders and acquirers inside the Eurozone as insiders. A.FOREIGN is now replaced by dummy variable A.ROW that has a value of (1) if the acquirer comes from outside the Eurozone and a value of (0) if the acquirer comes from inside the Eurozone. The analysis for A.ROW is executed with the data set cleaned of financial targets, and the analysis is conducted with Probit model (3.7)16.
The descriptive statistics of A.ROW are presented in Table 4.7 and the new correlation matrix is shown in Table 4.8. Later on, the Welch’s t-tests are presented in Table 4.9. Reviewing the descriptive statistics shows that the Netherlands has the largest share of acquirers from outside the Eurozone. The adapted correlation matrix in Table 4.8 shows that there is no strong correlation among the explanatory variables, or between the explanatory variables and the response variable. The t-tests show that during the
16 Model (3.6) is executed with the data set including and excluding financial targets. This does not provide significantly
Page | 25 post-crisis period, the share of acquirers from outside the Eurozone increased significantly in the Netherlands. This is regarded evidence of fire-sales by Aguiar and Gopinath (2005), Alcharya et al (2011) and Alquist et al. (2013). The other regions show an insignificant increase in the aggregate data and in Greece. Furthermore, the results show a highly significant decrease in Sweden.
Table 4.7. Acquirers from outside the Eurozone
A.ROW All Countries Netherlands Greece Sweden
Observations 2384 726 493 1165 Mean 0.1787 0.2328 0.0791 0.1871 St.Deviation 0.3832 0.4229 0.2702 0.3902 Minimum 0.0000 0.0000 0.0000 0.0000 Median 0.0000 0.0000 0.0000 0.0000 Max 1.0000 1.0000 1.0000 1.0000
Table 4.8. Correlation matrix 2 (response variable: A.ROW)
Correlation Matrix 2 Region Leverage Liquidity Tangibility Size Profitability
A.ROW T.TOTLIA.TOT T.CURRENT T.INT.TOT T.LN.TOT T.EBITDA.TOT
Response variable A.ROW 1.0000 Explanatory variables T.TOTLIA.TOT 0.0460 1.0000 T.CURRENT 0.0577 -0.2746 1.0000 T.INT.TOT 0.0389 -0.0252 -0.0886 1.0000 T.LN.TOT 0.2732 0.0742 -0.1424 -0.1039 1.0000 T.EBITDA.TOT 0.1190 0.1659 -0.0320 -0.0990 0.3604 1.0000
Variables taken from target firms are preceded by “T.” and variables taken from acquiring firms are preceded by “A.”
Table 4.9. Welch’s t-test (A.ROW)
Region Observations Pre-mean Post-mean t-statistic D.freedom
All 1349 0.1794 0.1828 0.8809 564 Netherlands 437 0.1968 0.3229 0.0159** 127 Greece 326 0.0613 0.0769 0.6653 86 Sweden 586 0.2321 0.1500 0.0079*** 402
Page | 26 Model (3.7) runs the data for the financial crisis and Euro-crisis, the aggregate data and the data for the individual countries. Table 4.10 presents the results of the aggregate data during the Euro-crisis. The complete results produced by the second model are presented in the second section of Appendix I. Table 4.11 shows the marginal effects of the explanatory variables for the aggregate data during the Euro-crisis. The Probit analyses of the global financial crisis only find a significant positive effect for leverage in Sweden, thus hardly any support for fire-sales.
Analysis of the individual countries during the Euro-crisis does not provide any conclusive results for the fire-sale variables. This could be attributed to the limited number of observations. A Probit model generally needs more observations than a linear model to arrive at significant results. The limited time-frame and limited acquisitions by firms outside the Eurozone are possibly not enough to find significant results per region.
The aggregate data of the Euro-crisis does provide new significant results. Table 4.10 shows a positive and significant effect for the leverage and tangibility variables on the probability of being acquired by an acquirer from outside the Eurozone. In line with hypothesis two and four, I find that highly leveraged firms and firms with lower asset tangibility are more liable to be targets of outsider acquirers. Contrary to hypothesis one, firm-level liquidity has a highly significant and positive effect on the probability of being acquired by an acquirer from outside the Eurozone. The results show a positive and highly significant effect for the size of target firms. These effects are opposite to the expectations, and liquidity in particular provides interesting results. The estimates and z-values in table 4.10 however, do not provide information about the strength of the effects. To find the strength, a transformation to marginal effects is necessary.
Table 4.10. Probit analysis of model (3.7) with the aggregate data
Hypothesis – Variable Variables Estimate Z-value P>|z|
(intercept) -2.8138 -7.2440 0.0000*** H1 – Liquidity T.CURRENT 0.1101 3.1240 0.0018*** H2 – Leverage T.TOTLIA.TOTASS 0.2907 1.9710 0.0487** H4 – Tangibility T.INT.TOTASS 0.8555 2.1810 0.0292** H5 – Size T.LN.TOTASS 0.2363 4.7730 0.0000*** H6 – Profitability T.EBITDA.TOTASS -0.4068 -1.0960 0.2731
Note: The *, **, and *** placed on the t-statistic scores represent statistical significance at the 10%, 5%, and 1% level.
Page | 27 indicates the expected relation based on the hypotheses and “Observed” is the relation I find. The table shows that if the leverage variable in a target firm increases with one unit at its mean, the probability of being acquired by an acquirer from outside the Eurozone increases by 6.34%. This supports hypothesis two for the aggregate data during the Euro-crisis and is in line with the expectations based on fire-sale theory. Additionally, I find support for hypothesis four, regarding the tangibility variable. The highly significant marginal effect shows that for one unit increase in the INT.TOTASS variable (i.e. decrease in asset tangibility) at its mean, the probability of being acquired by an acquirer from outside the Eurozone increases with 18.65%. During the Euro-crisis, firms with lower asset tangibility and higher leverage are more prone to be targets of outsiders.
Contrary to hypotheses one and five, I find highly significant and positive effects for liquidity and size. During the Euro-crisis, analysis of the aggregate data finds an increase of one unit of the current ratio at its mean increases the probability of being acquired by an acquirer from outside the Eurozone by 2.39%, while one unit increase for size at its mean gives an increase of 5.15%.
Table 4.11. Marginal effects
Hypothesis – Variable Variables Expected Observed Marginal effect Z-value P>|z|
H1 – Liquidity T.CURRENT - + 0.0239 3.1189 0.0018***
H2 – Leverage T.TOTLIA.TOTASS + + 0.0634 1.9617 0.0497**
H4 – Tangibility T.INT.TOTASS + + 0.1865 2.2053 0.0274**
H5 – Size T.LN.TOTASS - + 0.0515 5.0056 0.0000***
H6 – Profitability T.EBITDA.TOTASS -0.0887 -1.0995 0.2715
Note: The *, **, and *** placed on the t-statistic scores represent statistical significance at the 10%, 5%, and 1% level. The “+” indicates a positive marginal effect while the “-” indicates a negative marginal effect on the response variable.
Page | 28
5. Discussion
This chapter summarises the findings of the different analyses. Furthermore, it includes the discussion of the findings and compares the results with other papers. The second section of the chapter covers the implications for professionals and identifies issues of potential academic interest. The concluding section of this chapter covers the limitations of this research.
5.1. Comparing the findings with existing literature
To correctly interpret the results, I first combine the findings of the univariate analyses in Table 5.1. The graphical presentation in Appendix F, would support the hypotheses by showing clear diverging patterns between targets and acquirers during the crises. The statistical univariate analysis supports the hypotheses if the acquisition targets on average have significantly lower liquidity, higher leverage and higher short-term debt. Furthermore, the hypotheses expect lower asset tangibility, smaller size and lower profitability in the post-crisis period. Nevertheless, these findings do not allow me to draw conclusions about firm-level determinants of fire-sale acquisitions. It shows whether the firm-level variables in targets change in line with the general expectations based on fire-sale theory (i.e. make firms more liable to be targets). Initial support of the hypotheses based on the graphs and t-tests is summarised in Table 5.1. It features the hypotheses and shows whether the graphs or t-tests find supporting evidence. Table 5.1 shows for all regions which firm-level variables act in line with the hypothese. This indicates which variables make firms more prone to be targets during the financial crises.
Page | 29
Table 5.1. Initial support for the hypothesis indicating susceptibility to crisis effects
Note: “Yes” indicates that the graphical analysis or Welch’s t-test provides evidence in support of the hypothesis. 5.1.1. Main findings
Combining the results presented above in Table 5.1 with the findings of Probit model (3.7) allows me to extract the main findings related to fire-sale theory. The main findings are related to liquidity, leverage, and asset tangibility. These are discussed first.
H1: Firms with lower liquidity are more prone to be a target (of an outsider) during a
financial crisis.
The t-tests find no evidence that targets have lower liquidity positions in the post-crisis period compared to the pre-crisis period. Furthermore, the graphs provide no clear evidence that targets have substantially worse liquidity positions than acquirers during the crisis. This does not support hypothesis one and is contrary to Alquist et al (2014), and Brunnermeier and Pedersen (2009). Alquist et al. (2014) research liquidity-driven FDI in emerging countries and find that when the financial system fails, liquidity is often the driver of inward FDI. The lack of evidence for hypothesis one can possibly be explained by Desbordes and Wei (2012) who argue that the concept of liquidity-driven transactions is only feasible in emerging countries with less developed financial markets.
On the contrary, Brunnermeier and Pedersen (2009) argue that enough liquidity can shield a firm from the effects of a crisis. Thus a liquidity shield against decreasing asset prices, forced liquidation of assets, and the deterioration of the balance sheet. I find no evidence that firm-level liquidity has a more prominent role in acquisitions during the post-crisis period. Although the fire-sale effects of the liquidity spirals are mainly based on leverage and intangibility, Probit analysis (3.7) even finds evidence against
Hypothesis 2 Hypothesis 3 Hypothesis 4 Hypothesis 5 Hypothesis 6 All countries
Graphical analysis - Yes Yes - -
Welch's t-test - - Yes Yes Yes
The Netherlands
Graphical analysis - Yes Yes - -
Welch's t-test Yes - Yes Yes Yes
Sweden
Graphical analysis - - Yes - -
Welch's t-test - Yes Yes Yes Yes
Greece
Graphical analysis - Yes - - -
Page | 30 hypothesis one. Contrary to the expectations built on fire-sale literature, liquidity has a small positive effect on the probability of being acquired by a firm outside the Eurozone. This shows that targets with better liquidity were more liable to be targets of outsiders during the financial crises. Even though liquidity seems directly related to fire-sales, the current ratio seems of little importance to target firms. It can be argued that the contagion or domino effect of fire-sales is particularly related to structural problems resulting from intangibles and leverage. When an organisation has to secure more funding to overcome such structural problems, leverage and asset tangibility are probably more important than the current ratio. Financial institutions would review leverage and intangibles instead of the current ratio before granting additional financing.
H2: Firms that are highly leveraged are more prone to be a target (of an outsider) during a financial
crisis.
The t-tests find that in both the Netherlands and Greece, targets are higher leveraged during the post-crisis period. This supports hypothesis two and is in accordance with Brunnermeier and Pedersen (2009), who find that highly leveraged firms are affected more if funding liquidity and market liquidity disappear. The lack of evidence for Sweden suggests that the penalising of unsustainable leverage positions (implemented after their banking crisis in the 90s) worked. Focussing on fire-sale effects, the Probit analysis shows leverage has a positive effect on the response variable. The results show that highly leveraged firms were more liable to be targets of acquirers from outside the Eurozone. This is in line with hypothesis two and expectations based on fire-sale theory. The deteriorating balance sheets and inability to get more funding result in fire-sales and affect highly leveraged firms more. During both financial crises, highly leveraged firms were more probable targets, and during the Euro-crisis the highly leveraged firms were more prone to be targets of outsiders. Contrary to Weitzel et al. (2014), I do find evidence of fire-sale acquisitions during the Euro-crisis.
H4: Firms with lower asset tangibility are more prone to be a target (of an outsider) during a financial
crisis.
