University of Amsterdam, Amsterdam Business School MSc Business Economics, Finance track
Master Thesis
Predicting M&A in the European energy industry Erik Edam, 5983517
December 2013 Supervisor: Dr. J.E. Ligterink
Abstract
This thesis looks for financial variables that contain predictive power to assign potential targets in the energy industry of Europe. The dataset consists of 82 deals and a control group of 153 firms in the period 1990 - 2012. On average there is a significant abnormal return around announcement date of a deal according to the literature. The developed investment strategy tries to outperform the market return by investing in a portfolio with potential targets. The results of the takeover likelihood models suggest that total assets, secured debt, price to book, debt to assets and asset turnover are financial variables that contain predictive power. Debt to assets and total assets seem to be the variables with the highest predictive power regarding significant differences between targets and non-targets, debt to assets predicts the targets very well. The predictor, total assets, has moderate power for assigning targets to portfolios. The best-constructed portfolio has a return of 20.27% within a timeframe of one-year. This finding is significant. The portfolios are not constructed perfectly according to the target ratio in each portfolio. The assignment of targets to the portfolios is not very consistent and most returns are insignificant.
Keywords: Takeovers, predicting, financial variables, abnormal return, target firms and trading strategy
Table of contents 1. Introduction 4 2. Literature review 6 2.1 Energy industry 6 2.2 Takeovers motives 7 2.3 Excessive returns 7
2.4 Predictors and determinants 8
2.4.1 Firm size 8
2.4.2 Undervaluation 9
2.4.3 Inefficient management 9
2.4.4 Secured debt 10
2.4.5 Leverage 10
2.4.6 Short-term technical factors 11
2.5 Country level characteristics 11
2.6 Predicting targets in prior studies 12
3. Methodology 14
3.1 Variables 14
3.2 Model specification 15
4. Data 16
4.2 Descriptive data 17 5. Results 19 6. Trading Strategy 24 7. Conclusion 26 8. Discussion 27 References 29 Appendix 32
1. Introduction
Until the 90’s the government controlled the energy market in most European countries. This way of ownership was historically developed and most preferred in times of war and
economic instability. In the beginning of the 90’s national and European boards of directors decided for liberalisation of the energy industry. The deregulation was intended to change the companies in privately owned businesses and therefore to create more competition. The European Commission improved the cross border regulation and expanded transmission links for better cooperation and competition (Jamasb and Pollitt, 2005).
At the end of the 20th century, there were multiple businesses that tried to compete on the deregulated energy market. The competitive market has led to a lot of mergers and acquisitions (Hooper and Waddams Price, 2010). The biggest European producers of energy try to control parts of the European market to create economies of scale and/or scope. However they know that they are not able to control the market as a monopolist. Deals are still regulated by the government and the European Union. There are multiple institutions that decide to accept or decline the deal regarding monopolies. Energy firms that want to merge with or acquire another energy company need the permission of the national and European board (Jamasb and Pollitt, 2005).
Prior studies have shown that target firms earn a significant percentage in stock return around the announcement date. Song and Walking (2000) and Mulherin and Boone (2000) argue that the short-term abnormal return for target firms is respectively 16.7% and 21.2% on average. For this reason it would be interesting to predict targets for investment strategies. These results are confirmed by the efficient market hypothesis. The hypothesis states that the market is informational efficient, this means that securities reflect the information that is available in the market (Basu, 1977). If there is a strategy that predicts targets based on financial variables this will strike with the efficient market hypothesis. This study discusses mergers and acquisitions within the electricity industry and tries to find firm specific predictors and determinants for potential targets. Recent studies regarding takeovers often exclude the utility sector (Owen and Yawson, 2010; Cai and Tian, 2013). This study tries to fill the gap in the literature due to a lack of information and under research. The research question of this study is: What financial predictors for assigning targets can be used to create portfolios that outperform the market in the European energy industry?
Data is collected from two different databases. The information about deals regarding bids and total takeovers is retrieved from Thomson Reuter’s SDC. The stock return and basic accounting data of these firms is gathered from Datastream. The sample includes deals that
enclose at least 50% of the outstanding stock of the target. The same information for the control group is downloaded from Datastream. The control group is included to discriminate between values of financial variables of the two samples. The following set of
determinants/predictors is included in the model: size, undervalued firm, inefficient
management, secured debt, leverage and short-term technical factors. Merger and acquisition deals started in the industry in the beginning of the 90’s, so therefore data is collected for the period 1990-2012. The data covers European countries and the energy industry.
This thesis gives insights in the market of energy companies in Europe regarding takeovers, however the energy sector is not investigated very extensively. Since most articles disregard the energy industry to control for specific institutional and regulatory constraints’ (Tong and Miao, 2011). The results, the significant determinants and predictors, are used to develop an investment strategy. Investors should consider using the model that is produced to receive a potential gain from target firms in the energy industry.
Paragraph two contains an overview of the energy industry in Europe, motives,
predictors/determinants and predicted targets in prior studies. This is followed by a paragraph about the methodology; variables and the models. Paragraph four takes account for the dataset and the descriptive data. The results of the study and the interpretation are included in the next paragraph. Section six describes the investment strategy that is developed. Followed by the conclusion in paragraph seven. And at last, paragraph eight contains a discussion that reflects the pitfalls and the possible improvements.
2. Literature review
This chapter summarises the literature about mergers and acquisitions. It starts with the energy market in Europe in general. This is followed by the motives for deals and takeovers. Paragraph 2.3 is about abnormal returns around the announcement date of mergers and acquisitions. The next part describes the determinants and predictors of potential targets that are found by prior literature. Section 2.5 gives an overview of the country specific regulation and legislation within Europe. Followed by a section that presents the predictability of developed models for potential targets in prior studies.
2.1 Energy industry
The energy companies supply whole countries, so before deciding for privatization there was need for an economic and a political stable period. Some countries were already privatized before the reform, like Germany and Belgium. Some other countries waited even longer with the deregulation, for instance France. It was hard for foreign firms to participate on these markets (Jamasb and Pollitt, 2005). The European merger and acquisition standards were less severe compared to the United States. Due to these relatively relaxed standards, a lot of mergers and acquisitions were exercised (Newbery, 2007).
It is illegal for European firms to misuse the market power they possess regarding their dominant position. If an acquisition is accepted, the possibility for a more competitive market in the future is excluded because of the decreasing number of firms. Another reason for strictly determining whether an acquisition can be accepted or not is because of high entry costs. A new energy company should make extremely high investments for the power plants, so for this reason new entrants are less likely to appear in the market (Newbery, 2007). Subsequently, concentration is important because this reflects the competitiveness of the market. When a market is highly competitive, firms become more efficient which will lead to a decrease in prices. This market transformation is in consumers’ interest and was the
intention of the entire deregulation (Jamasb and Pollitt, 2005).
Jamasb and Pollitt (2005) argue that the deregulation led to less concentration on horizontal level in some countries. Actually, in most European countries the concentration increased after the privatization. In 1998 the eight biggest energy companies possessed almost 60% of the market. This percentage increased to 75% in 2002. Jamasb and Pollitt (2005) argue that the energy firms developed faster than European institutions and the legislation. This is partly due to the complex and partial shareholding ownership that is difficult to analyse.
2.2 Takeover motives
The competitive market resulted in a lot of mergers and acquisitions in order to expand the market share and economies of scale (Hooper and Waddams Price, 2010). There are some reasons and motives that clarify these deals in the market. Giammarino and Heinkel (1986) argue that companies could adapt the strength of their strategy by exercising a merger or acquisition. Synergy benefits are an example of strategic motives. Other justifications could be: capitalizing core competences, providing complementary resources to the firm and creating more market power (Hopkins, 1999). Another example of a takeover motive is the financial benefit. Takeovers are more likely to appear in industries when there is cheaper access to capital, a higher tax shield, higher cash flow stability or a lower bankruptcy
probability (Lewellen, 1971). These financial benefits do not affect the acquired firm alone, it affects the whole organization and this makes it more attractive to exercise the deal.
Undervalued firms can act as another economic motive to clarify a takeover. Companies that are temporarily valued below its intrinsic value might become a potential target (Shleifer and Vishny, 2003). The takeover likelihood will also increase when there is a higher growth potential for a certain firm (Hopkins, 1999). When there is a need of becoming a player on a foreign market, the easiest way to enter a different market is to merge with or acquire a company in this specific market. A strategy like this is often used in mature markets (Hopkins, 1999).
2.3 Abnormal return
Companies can gain or lose money in M&A deals. Song and Walking (2000) studied the short-term returns of announcements in M&A for different industries. With a 1-day window around the announcement date of the deal, they find that target firms gain on average 16.7% abnormal return. Mulherin and Boone (2000) investigated the US market between 1990-1999. Their sample contains 1305 firms from 59 industries worldwide and they conclude that target firms earn 21.2%, on average. Besides, they find that abnormal returns from acquisition and divesture announcements are similar. The return from acquisitions for bidder and target together is 3.5% on average, the abnormal return around announcement return for divesture is close to 3% (Mulherin and Boone, 2000).
More relevant to this study, Berry (2000) and Becker-Blease, Goldberg and Kaen (2008) argue that the US target companies in the energy industry experience positive
abnormal returns around the announcement date. Becker-Blease, Goldberg and Kaen (2008) use a dataset of 70 proposed deals between 1992 and 2001. The level of abnormal return for
targets in this industry is lower than for companies active in non-regulated industries. For a two-day window the abnormal announcement returns are close to zero, but still significant (Berry, 2000).
The abnormal return in a one-day window around the announcement date in the European energy industry was 6% between 1990 and 2006 according to Datta, Kodwani and Viney (2013). They agree that the returns for the target firms are smaller but significant in the energy sector compared to the non-regulated industries. An explanation for this relatively low announcement return is the regulation within this industry. Another suggestion is that stable revenues and low levels of risk affect announcement returns (Datta, Kodwani and Viney, 2013). Stable revenues and low levels of risk cause low gains and losses in deals for targets and bidders. For that reason there is a relatively small effect on abnormal returns for targets.
The efficient market hypothesis implies that abnormal returns may occur according the availability of information. The hypothesis states that the market is informational efficient, this means that securities reflect the information that is available in the market (Basu, 1977). This study tries to predict targets based on financial variables and this would strike with the hypothesis.
2.4 Predictors and determinants
The strategy taken by bidders for assigning targets has been investigated frequently in the past and nowadays papers write about it regularly. Numerous economists have tried to discover the determinants and their effect on takeovers (Brar et al., 2009; Palepu, 1986; Hasbrouck, 1984).
2.4.1 Size
Hasbrouck (1984) suggest that the takeover likelihood is affected by the size of companies. The study uses a dataset of 86 targets against 172 non-targets between 1977 and 1982. He shows that the size of the target is substantially smaller than the bidder. Hasbrouck (1984) agues that the bidder needs a certain proportion of equity to invest and finance the deal. He also argues that organizational costs rise with the size of a company and that bigger targets can exercise more harmful defence actions (Hasbrouck, 1984).
Brar et al. (2009) agree in their literature overview with the statements mentioned above. Bigger companies will cause higher adaptation costs in takeovers and smaller firms do not have the power to implement costly takeover defences once they become a target (Brar, 2009). Brar et al. (2009) state that size is a high quality predictor and is negatively related to
the takeover potential. Proxy variables that have been used for size are as follows: market capital, sales and number of employees. The study of Brar et al. (2009) employs a logit model that is developed by Palepu (1986). This model determines the financial variables of interest regarding the takeover likelihood. All these arguments imply that relative small firms become a potential target more often compared to relatively big firms.
(H1) Firm size is negatively related to takeover probability.
2.4.2 Undervalued firms
An undervalued firm is a company that is valued below its true intrinsic value. The takeover potential rises for companies that are relatively undervalued for a certain time period (Shleifer and Vishny, 2003). The likelihood of becoming a target increases even more when there is a higher growth potential (Hopkins, 1999).
Brar et al. (2009) argue that the dividend yield and the earnings yield are the most important measurements for spotting undervalued firms. Targets seem to have lower earning multiples and higher dividend yields. Bidders experience a relative discount when they acquire an undervalued firm (Brar et al., 2009). Smith and Watts (1992) also find that the takeover probability decreases with a smaller dividend payment and higher growth potential.
Shleifer and Vishny (2003) suggest that undervalued companies become targets because of the low valuated equity, in contrast to overvalued firms. Overvalued firms obtain benefits from stock issuing and supply themselves of capital to exercise a merger or
acquisition (Shleifer and Vishny, 2003). All these studies find that acquirers bid on undervalued firms relatively more often.
(H2) Undervalued firm are more likely to be acquired.
2.4.3 Inefficient management
Mergers and acquisitions can act as market mechanism to replace inefficient management into efficient management (Marris, 1964). Barnes (1999) suggests that a management is classified as inefficient when the profits are below average. The current management does not invest in the highest valuable projects and this is not in shareholders’ interest. There are multiple financial variables that can measure efficiency, like: P/E ratio, growth, profit or dividend. When the levels of these parameters fall below average, the entity might become a potential target (Barnes, 1999). Barnes (1999) test 82 targets against 82 non-targets from 1991 till 1993
on the UK market and concludes that firms with inefficient management become a target relatively more often.
Kennedy and Limmack (1996) also test mergers and acquisitions on the UK market regarding inefficient management. The datasets contains data from 1980 till 1989 and creates two subsamples, one with and one without a CEO replacement after a takeover. They have compared these groups and argue that inefficient management is positively related to the likelihood of becoming a target.
(H3) Potential targets have a low level of earnings.
2.4.4 Secured debt
Property, plant and equipment can act as collateral and capital can be borrowed based on this secured debt (Stulz and Johnson, 1985). Ambrose and Megginson (1992) state that targets have a higher ratio of tangible fixed assets in their total asset structure and summarize three different explanations. First of all firms should acquire targets with low leverage levels as security to attract debt financing to the company. Secondly, firms in declining markets will experience decreasing growth rates. Asset-rich companies with low growth rates increase the possibility to become a potential target. The acquirer should anticipate on this and restructure the firm to outperform the competitors in the market (Ambrose and Megginson, 1992). The last explanation is that the ratio of secured debt is more valuable than growth opportunities regarding the new management. Acquirers do not have a comparative advantage in managing growth opportunities but they can manage an optimal asset structure (Ambrose and
Megginson, 1992). Ambrose and Megginson (1992) test the difference in secured debt between the targets and control group with the following formula: tangible assets to total assets.
(H4) Secured debt affects the takeover potential in a positive way.
2.4.5 Leverage
Ambrose and Megginson (1992) argue that bidders should acquire firms with low leverage levels to benefit from the capital structure. They have tested that targets have a slightly higher leverage ratio compared to non-targets but this is insignificant (Ambrose and Megginson, 1992). Palepu (1986) confirms the findings regarding the absence of different capital structures. Targets are less often financially distressed and have similar debt levels in their
financial structure compared to non-targets (Brar et al., 2009). Contrarily, liquidity seems to be significantly different between targets and non-targets. Brar et al. (2009) link this
difference to leveraged buyout funds, which is a common strategy in last years. This approach searches for undervalued, cash rich firms and uses leverage to finance the transaction (Brar et al., 2009).
(H5) Liquidity is positively related to the likelihood of becoming a target. (H6) Potential targets have low leverage levels.
2.4.6 Short-term technical factors
Brar et al. (2009) suggest some variables to include in the model. The variables that they add to the model, are: earnings revisions, price momentum and trading activity. It is stated that the share price of companies rises just before the announcement date and therefore affects the short-term price momentum characteristics. In fact, there is proof for pre-knowledge of a month before the announcement date. The 3-month share price momentum and the trading volume variable before announcement date seem to be significantly different between targets and non-targets. Momentum measures the underlying direction of the movement of the stock price. The other variable, trading volume, presents the average volume of stock transactions as a percentage of market capitalization. These findings will improve the timing of the takeover forecast (Brar et al., 2009).
(H7) Momentum in share price increases with the takeover probability.
2.5 Country level characteristics
The paragraphs above describe the different predictors and determinants. In extent of this information there are also country specific laws and regulations regarding mergers and acquisitions. Rossi and Volpin (2004) use a large sample of M&A announcements in the 90’s that were closed before 2002. There are 49 different countries involved in the sample. They conclude that investor protection is positively related to the number of domestic deals and negatively to foreign deals. Higher premiums and deals with higher stock level payments are also associated with better investor protection (Rossi and Volpin, 2004; Brar et al, 2009).
La Porta et al. (2006) studied this topic with a questionnaire and produced and index for the following three country specific characteristics: accounting standards, shareholder protection and liability standards. Rossi and Volpin (2004) suggest that good disclosure of
accounting information increases the takeover probability; potential targets can be identified more efficiently. La Porta et al. (2006) quantify accounting standards with discrete values on a scale from 36 to 83. The next country level characteristic, investor protection, measures the effective right of minority shareholders and the index varies between zero and six (La Porta et al, 2006). Liability standards are positively related to the number of deals that are exercised by the acquirer. The index contains a scale from zero to one, in their study. This characteristic measures the degree of facilitating services for the investor to recover losses, in fact it
measures the protection of the investor (La Porta et al, 2006; Brar et al, 2009).
As summarized above, it becomes clear that the country location of the target firm affects the likelihood of becoming a target. The period dummies control the different
timespans within the dataset and will not act as a timing predictor for takeovers. The country dummies control the specific effects for the different nations.
2.6 Predicting targets in prior studies
Table I shows an overview of the various studies that try to develop predictive models for targets of takeover. These studies try to find some financial variables that can predict the likelihood of becoming a target (Dietrich and Sorensen, 1984; Barnes, 1990). These models assign firms to certain portfolios with the intention of creating portfolios with a high ratio of (potential) targets. This is based on the theory that the abnormal return for targets arises around announcement date of the deal. Some articles describe correct prediction ratios
between 75% and 95% for assigning firms as targets and non-targets. These created portfolios are not tested against the market return. Abnormal returns are not directly related to the predictive power of the models. It is possible that the market will outperform the model (Powell, 2001).
Wansley, Roenfledt and Cooley (1983) claim to have adopted a model that earns a 17% abnormal return on a timespan of 12 to 21 months. This finding is significant but they have their doubts about the robustness due to the potential size effect in the portfolio returns. The securities are not linked according the size matching approach, so the size could bias the returns. They claim to assign targets and non-targets correctly to their groups with 75% accuracy within the sample. They use the following set of predictors: price to earnings ratio, long term debt to assets, net sales, sales growth and market value of equity to total assets. Palepu (1986) corrects the methodology for the size- and industry effect. His findings are in contrast with the results of Wansley et al. (1983) since he claims that abnormal return is absent. Palepu (1986) assigns the target based on the following set of determinants: return on
equity, growth, liquidity, leverage, size, market to book ratio and the price to earnings ratio. This is tested on a sample of 163 firms and a control group of 256 companies between 1971 and 1979. The trading strategy provided by Palepu (1986) is not efficient and will not outperform the market.
Powell (2001) also created a model for predicting targets with a high takeover
likelihood based on the logit model of Palepu (1986). The highest returns can be earned when investors own a portfolio with a maximised ratio of targets. The targets are assigned based on the following financial variables: inefficient management, undervaluation, free cash flows, size, secured debt, growth, liquidity and leverage. The portfolio is created on January 1st 1996 and the results are measured at December 31st 1996. The regression determines whether the model contains quality; significance and explanatory power. The predictability of the model seems to be 84% within the sample. Powell (2001) tests the developed model in the market (population). The results suggest that the model predicts 2% of the targets correctly. Given this, it is highly unlikely that this developed model will earn significant abnormal returns.
Table I Overview predictive models prior studies Related studies Sample
(targets/non-targets)
Sample period
Findings and financial predictors Wansley, Roenfledt and Cooley (1983) 44/101, multiple industries 1975-1976 US
17% abnormal return, not significant. p/e ratio, LT-debt to assets, net sales, sales growth and MV of Equity / Total assets Palepu (1986) 163/256, multiple
industries
1971-1979 US
Model does not outperform the market Abnormal return. ROE, growth, liquidity, leverage, size, market/book and p/e ratio Powell (2001) 471/471, multiple
industries
1986-1995 UK
Unlikely the model outperforms the market. Ineff. Management, undervaluation, free CF, size, secured debt, growth, liquidity and leverage Brar, Giamouridis, and Liodakis (2009) 835/2906, multiple industries 1992-2003 Europe
Significant abnormal returns around 15%. Market cap., dividend yield, sales growth, momentum, trading volume and liquidity
Brar et al. (2009) predicts the firms correctly to the target and non-target group with 72.65%, within the sample. They find a correctly predicted ratio of targets in their validation sample of 0.45. The model is tested within the population (financial firms excluded). The highest top 10% decile, with the highest ratio of (potential) targets, is used as portfolio. The
strategy is called ‘takeover timing portfolio’. This investment strategy results in a return for 1-month, 3-month and 6-month holding of respectively 17.7%, 14.4% and 14.3% on average (Brar et al., 2009).
3. Methodology
This thesis presents predictors of potential targets of mergers and acquisitions in the energy industry of Europe. Information about deals regarding bids and takeovers is used in this study. The stock return and basic accounting data is implemented for these firms. This research is quite similar to the study of Brar, Giamouridis and Liodakis (2009). It manufactures a model to predict potential targets in the energy industry because this has not been studied
extensively. The energy industry is often disregarded in prior work because of the regulation within this sector. The model tries to predict targets in this sector exclusively. Portfolios are constructed based on these predictors and tested against the market return.
3.1 Variables
The takeover likelihood is the dependent variable and is determined by a logit model. The variables of interest are specified in table II. These variables are described in the literature of paragraph two.
Table II Summary of variables Variable Description
Firm Size Market value (number of shares outstanding * current stock price)
Sales
Number of employees
Undervaluation Dividend yield (annual dividend per share / price per share)
Price/earnings (market value per share / earnings per share)
Price/book (stock price / total asset–intangible assets & liability)
Inefficient Management Profit margin (net income / sales)
Asset turnover (revenue / assets)
Return on equity (net income / shareholders equity)
Return on sales (net income (before interest and tax) / sales) Return on capital (net income – dividends / total capital) Sales growth ((sales – sales(y-1) / sales(y-1) ) * 100)
Leverage Debt/assets (total debt / total assets)
Debt/equity (total liabilities / shareholders equity)
Liquidity (current assets / current liabilities)
Secured debt (tangible assets / total assets)
3.2 Model specification
The logit model below is developed to measure the effect of firm characteristics on the takeover probability of a certain firm, this is created by Palepu (1986):
p(i,t) = 1/ 1+e-βx(i,t)
where, p is the probability of being a takeover target, i,t is the entity in the related year and β is the vector of the unknown parameters that is estimated (Palepu, 1986).
The model needs some more clarification. When the target receives a bid there can be determined whether to accept or decline the bid. This depends on the amount and type of bid that is received in the particular period. This study only includes completed deals. The bid is based on firm specific characteristics of the target and motives from the acquirer itself. The model includes only quantifiable variables of targets which is described by x(i,t).These variables that affect the target likelihood of the firm are noted as stochastic random variables. The variables are endogenous to the takeover process, the output of the model p(i,t) is a combined function of the probability distribution of the used variables (Palepu, 1986).
The model of Palepu (1986) is used to determine the predictors that affect the takeover likelihood. All the firms in the dataset are implemented in the model and the financial
variables are tested for their predictive power. The control firms are not linked on size and timeline to the targets within the periods. This imperfection could cause biases in the sample construction. As stated in the literature shareholder protection, accounting standards and liability standards affect the likelihood of becoming a potential target. Dummy variables for all these different countries are included in this model to control for these effects. After the 90’s governments improved and changed laws and regulation regarding the energy industry, for that reason a dummy needs to be included for the time-period in the model (Jamasb and Pollitt, 2005). There are a few economic events that might have influenced firm financials. For instance, the merger waves in the 90’s. Within these years a lot of merger deals were exercised in a number of industries. A dummy is set for the period 1990-1997 (Harford, 2005). The dot-com bubble began with high increases in the stock market and high
investments in Internet companies in most cases in 1999-2000. In 2000 the market collapsed and caused a recession, a dummy is created for 1998-2002 (Ljungqvist and Wilhelm, 2003). The years between 2003-2006 were relatively quiet. The origin of the financial crisis was the failure of the financial institutions and this affected all firms in other industries. The
implemented for the period 2007-2012 (Erkens, Hung and Matos, 2012). The results of these takeover likelihood models measure the financial predictors of interest. The period dummies control the different timespans within the dataset and will not act as a timing predictor for takeovers. The country dummies control the specific effects for the different nations. The computed regression model is shown below.
Logit(Takeover)i,t = β0 + β1Xi,t-1 + β2PeriodDummyi + β3CountryDummyi
The model tests the financial factors that affect a takeover by using the whole dataset. Four different portfolios are manufactured each year based on the two best financial predictors. Four different portfolios per year divide the companies by the financial data of last year. So each year the portfolios are constructed again based on the firms in the dataset. The first portfolio contains the firms with the lowest ratio of the specific financial variable. The number of targets is presented in each portfolio and is discussed. The monthly return of each portfolio is tested against the market return to analyse whether the model outperforms the market (Brar et al. 2009). The market model estimates the returns.
Ri,t – Rf = αi + βi(Rm,t – Rf)+ εi,t with 𝐸(𝜀!,!) = 0 and 𝑣𝑎𝑟(𝜀!,!) = 𝜎!!
where, R is the return of the firm, i,t is the entity to the related month, α is the constant and Rm,t is the market return for each month. The alpha is constructed in such a way that the error
term is equal to zero (Blume, 1971). The alpha will act as the difference between the return of the portfolio and the market. If the coefficient of alpha has economic interest and a significant value this means that there is an abnormal return against the market return.
4. Data
This chapter describes the selection of the data and gives the first general insight of the data.
4.1 Data selection
Data is collected from two different databases. The information about deals regarding bids and takeovers is retrieved from Thomson Reuter’s SDC. The accounting data and the stock return data is obtained from the Datastream database. The same information for the control group is downloaded from Datastream. This control group is included in the study to test the potential discriminatory ability of the financial variables between targets and non-targets.
Merger and acquisition deals started in this industry in the beginning of the 90’s, for that reason data is collected for the period 1990-2012 (actually from 1987 to manufacture some legs for certain variables). The data covers the energy industry (SIC code: 4911) in the European Union, where Brar et al. (2009) study Europe and test different industries.
Specifications deals sample:
- Both hostile and friendly takeovers are included in the sample.
- The takeover should contain at least 50% of the outstanding shares of the target. - The target should be a public firm, because of the availability of accounting data.
The sample consists of 82 deals.
The control group contains public firms in the energy sector and is constructed with the Datastream database. Energy firms are included when they were not acquired during 1990-2012, moderated by the Thomson One database. The accounting data is also retrieved from Datastream and firms with too much missing data and unknown Datastream codes are deleted. Data is collected for the year prior to the deal for the target sample. For the control group data is selected one year prior to the randomly assigned date. To create a return
variable, midweek dates are randomly assigned but are dependent on the activity period of the control company. Midweek dates are assigned to control for the end-week biases in returns. By the construction of the dataset the truly matching approach is disregarded. There are 153 entities in the control group.
4.2 Descriptive data
This paragraph provides some descriptive information of the data that is used to answer the hypotheses from paragraph two. Table III gives an overview of the descriptive data regarding deals and the control group.
82 Deals are included in the sample and most of these deals have been exercised in 2008 and 2009. At the beginning, the sample contained around 300 deals but due to a lack of financial information a lot of the observations were excluded. As stated before, only deals that contain a minimum 50% of the outstanding shares are included in the dataset. The average percentage of shares obtained at the deal fluctuates between 50% and 100%, showed by the fourth
dataset. From 2000 there are significantly more control companies in the dataset. This can be due to the data availability or to the number of companies that act as public firms.
Table III Descriptive data
Year No. of deals Cumulative acquired by deal Aver. % shares No. of control firms
1990 0 0 / 2 1991 1 1 88 2 1992 1 2 75 2 1993 0 2 / 3 1994 1 3 100 3 1995 2 5 80.6 2 1996 1 6 96.7 0 1997 0 6 / 2 1998 0 6 / 2 1999 1 7 100 3 2000 1 8 85.5 5 2001 2 10 84.1 10 2002 6 16 70.5 3 2003 3 19 77.4 5 2004 2 21 88.7 10 2005 4 25 85.5 2 2006 2 27 62.2 9 2007 6 33 78.6 23 2008 23 56 65.9 8 2009 18 74 79.9 9 2010 7 81 69.4 18 2011 1 82 50.5 19 2012 0 82 / 11
Before the financial variables from the sample and control group are tested against each other a CAR-model is implemented. This model is estimated by an event study. Returns for the event window are predicted based on the so-called estimate window and market returns. The regression estimates the difference between the predicted values and the observed values in the event window.
𝐴𝑅! =
1
𝑁 𝜀!,!
!
where, AR is the abnormal return on day t, ε is the difference between the predicted and observed return. The CAR model stands for cumulative abnormal return and is presented below (Barber and Lyon, 1997).
𝐶𝐴𝑅(𝑡!, 𝑡!) = 𝐴𝑅! !!
!!!!
This event study determines the cumulative abnormal return around announcement date of the deal. The observed return for the 20-day window is compared to the expected return around the announcement. The expected return is based on the market return and the estimate window of -21 to -85 days before the announcement day. The market return is composed by the average return of the dataset (target and non-target group). The literature describes that shareholders of targets earn a significant abnormal return around the announcement date, on average.
Table IV CAR event study (20-day window)
Variable Obs Mean Std. Dev.
CAR 12238 8.577814 39.14267
t-‐stat 0.0210146
The abnormal return around the announcement date according the result of table IV is 8.58%, on average. This means that the literature is confirmed because there is a positive abnormal return. Unfortunately the coefficient of the abnormal return is not significant.
5. Results
If there are some substantial differences between the groups according to the financial variables it might be possible to assign potential targets based on these variables. The differences in the financial variables between the sample and the control group are tested in this paragraph. The variables that show significant difference in mean might contain some power to predict a potential target. The variables of interest are implemented in the logit model of Palepu (1986).
Table V provides an overview of the mean within each group and a test that draws the difference between the means, which is explained by the p-value. Appendix A1 is a table that
contains extended information regarding to table V. The mean of the total dataset is included inclusive all the standard deviations of the means.
Table V Test mean difference between sample and control group
All of the variables from the table are tested with the t-test. Each variable is tested separately, the means of both group are composed and tested against each other. * means a significance level of 10%, ** means a significance level of 5% and *** means a significance level of 1%. The parameters and formulas are specified in table II.
Variables Sample group Control group Difference
Mean Mean P-‐value
Total assets LN 12.89 13.68 0.0057*** Net sales 2044547 2852406 0.2210 Netsales L1 1713774 2630987 0.1669 Netsales L3 2552207 2416608 0.4517 Employees 4130.61 8613.92 0.0476** Market value LN 5.47 5.77 0.1991 Outstanding shares 152.29 95.49 0.3058 Dividend yield 2.57 12.95 0.2574 Earnings yield 3.50 -‐0.03 0.0206**
Price to book value 2.66 1.83 0.0636*
EBIT 314833.3 395554.8 0.3087 Net income 232213.4 208859.7 0.4233 EPSL12M 3.70 -‐2.03 0.0889* Revenue 2044547 2852406 0.2210 Profit margin 4.39 -‐218.34 0.2081 Assettrunover 3.43 0.86 0.0000*** ROE 0.15 -‐0.01 0.0191** ROC 6.48 1.82 0.0003*** ROS 0.09 -‐2.05 0.2230 Sales growth 31.50 65.46 0.2893 Total debt 979262 2286713 0.0546* Long-‐term debt 845442.9 1672215 0.0778* Short-‐term debt 133819.1 614497.2 0.0389** Equity 1206483 2533201 0.0199** Tangible assets 1095890 1867422 0.1069 Current assets 1025654 1742861 0.2019 Current liabilities 804884.8 1924375 0.0999* Debt/assets ratio 0.16 0.28 0.0000*** Debt/equity ratio 0.44 0.57 0.4240
1y-‐Change total debt to assets 0.03 0.04 0.3076
1y-‐Change total debt to equity 0.17 0.09 0.3557
1y-‐Change long-‐term debt to assets 0.015 0.021 0.3482
1y-‐Change long-‐term debt to equity 0.047 0.013 0.4107
1y-‐Change short-‐term debt to assets 0.016 0.019 0.4072
1y-‐Change short-‐term debt to equity 0.12 0.08 0.3973
Liquidity 2.12 1.57 0.0779*
Secured debt 0.43 0.40 0.7251
The number of employees is different between the two groups with a 5% significance level. Total assets of targets show a tendency to be smaller for targets compared to non-targets with a significance level of 1%. These two parameters seem to be potential predictors. Hasbrouck (1984) suggests that size is negatively related to the likelihood of becoming a target, this is confirmed by these first results.
The t-test in the table shows a significant difference for the earnings yield with an alpha of 5%. The difference between both groups is absent for the dividend yield. Another parameter for testing under pricing is the price to book ratio, this is significantly different between the groups with an alpha of 10%. Brar et al. (2009) argue that the dividend yield and the earnings yield are the most important measurements for spotting undervalued firms. Hopkins (1999) suggests that undervalued firms become a target relatively more often than overvalued firms.
Leverage itself is not significantly different between the two groups. The variables debt, equity and current liabilities, which influence leverage, are tested significant different between the groups. Ambrose and Megginson (1992) suggest that bidders should acquirer firms with low leverage levels as security to attract debt financing to the company. The ratio debt to asset is statistically different with an alpha of 1%. Brar et al. (2009) argue that
liquidity levels of targets are relatively high and link this difference to leveraged buyout funds. This approach aims for undervalued, cash rich firms and uses leverage to finance the
transaction. The t-test within this dataset confirms the hypothesis that liquidity differs between target and non-target firms.
The two parameters, return on equity and return on capital, for indicating inefficient management are both significantly different between targets and non-targets. Return on equity is significant with an alpha of 5% and return on capital is significant with a level of 1%. Barnes (1999) states that the level of profits can estimate inefficient management. When profits fall below average, the entity might become a potential target. Kennedy and Limmack (1996) have tested this empirically and their results confirm the statement of Barnes (1999).
With these results several models can be composed for the logit regression however there is a need to control for multicollinearity between the financial variables. A correlation diagram is configured to show the related variables. The correlation table is included in appendix A2. The models are based on the results of table V and the output of the correlation table to avoid multicollinearity. The results in appendix A2 show that the related proxy variables are
turnover and return on capital. In table VI below, several logit models are tested to find variables that contain predictive power to assign potential targets.
Each model is exercised with and without period- and country dummies, separated by model ‘a’ and ‘b’. A few variables are tested significantly different from zero in the logit models. These variables seem to be good predictors for assigning targets. The coefficient of total assets in the first model (a and b) is statistically different from zero. The negative sign implies that the takeover probability rises when total assets decrease. This negative relation confirms the hypothesis. The size of a firm is negatively related to the probability of
becoming a target, this is similar to the findings of Hasbrouck (1984). The models with period- and country dummies contain more explanatory power than the models that disregard these dummies according to the r-squared. But the variables remain significant in both models.
Table VI a Estimates of takeover likelihood models P-values between parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001
Variables Model 1a Model 1b Model 2a Model 2b Model 3a Model 3b
Total assets -‐0.157* -‐0.174* -‐0.0467 -‐0.0812 -‐0.0672 -‐0.102 (0.016) (0.015) (0.586) (0.464) (0.377) (0.233) Dividend yield -‐0.00182 -‐0.00172 0.0670 0.0810 (0.726) (0.739) (0.226) (0.171) Sales growth -‐0.000359 -‐0.000416 0.000657 0.000632 0.00192 0.00126 (0.566) (0.570) (0.902) (0.912) (0.462) (0.640) Momentum -‐0.00426 -‐0.00986 (0.702) (0.481) Liquidity 0.0221 -‐0.174 0.0459 0.0311 (0.719) (0.297) (0.428) (0.609) Secured debt 0.385 1.108 0.674 0.617 (0.591) (0.239) (0.327) (0.399)
Price to book 0.160* 0.172*
(0.045) (0.043) ROE 0.957* 1.075* (0.028) (0.018) Constant 1.346 1.235 0.00841 20.95 -‐0.342 -‐0.727 (0.124) (0.257) (0.994) (0.984) (0.755) (0.582)
Period dummy No Yes No Yes No Yes
Country dummy No Yes No Yes No Yes
N 221 221 121 121 164 164
The probability of becoming a target increases along with the ratio of secured debt. This means that it affects the takeover likelihood positively and confirms the hypothesis. Acquirers bid relatively more on firms with a high level of secured debt. The hypothesis is designed with the article of Ambrose and Megginson (1992). The coefficient of the price to book variable is significant in the third model. The price to book value is positively related to the takeover probability. This result contradicts the hypothesis. According the literature under pricing raises the likelihood of becoming a target (Brar et al., 2009).
Table VI b Estimates of takeover likelihood models P-values between parentheses, * p < 0.05, ** p < 0.01, *** p < 0.001
Variables Model 4a Model 4b Model 5a Model 5b Model 6a Model 6b
Total assets 0.192 0.312 (0.121) (0.077) Dividend yield 0.109 0.00604 0.105 0.0278 0.0542 0.0127 (0.273) (0.963) (0.301) (0.849) (0.528) (0.897) Sales growth -‐0.000139 0.00220 -‐0.000795 0.00148 (0.982) (0.768) (0.907) (0.860) Momentum 0.00167 -‐0.000108 0.00117 -‐0.00154 (0.883) (0.994) (0.919) (0.927) Liquidity 0.0521 -‐0.0443 0.0561 -‐0.0346 (0.418) (0.838) (0.389) (0.875) Secured debt 1.213 3.246* 1.300 3.274* (0.170) (0.019) (0.156) (0.026)
Price to book 0.0537 0.0234
(0.331) (0.678)
Asset turnover 0.218** 0.348*** 0.293** 0.456***
(0.004) (0.001) (0.002) (0.000)
Earnings yield 0.848 0.871 0.802 0.868 1.201 1.830
(0.240) (0.199) (0.276) (0.294) (0.222) (0.103)
Debt to assets -‐0.00218 -‐3.069*
(0.551) (0.020) Profit margin -‐3.262** -‐0.00752 (0.006) (0.257) Constant -‐1.700** 32.17 -‐4.412* 30.01 0.0312 35.78 (0.003) (0.990) (0.019) (0.995) (0.931) (0.994)
Period dummy No Yes No Yes No Yes
Country dummy No Yes No Yes No Yes
N 107 107 107 107 114 114
pseudo R2 0.172 0.422 0.189 0.445 0.144 0.337
The sign of return on equity implies that it is positively related to the takeover potential. The return on equity ratio should have a negative coefficient according the literature. This is because a firm with inefficient management should become a target relatively more often.
The coefficient of debt to assets in the sixth model is also significant. The negative sign means that a lower debt to assets ratio increases the takeovers likelihood. A lower debt to asset ratio affects the takeover likelihood in a positive way, this is confirmed by the study of Ambrose and Megginson (1992).
Asset turnover is the last significant variable in Table VI of the likelihood models. Asset turnover raises the probability of a takeover according to the results. This contradicts the hypothesis because inefficient management affects the likelihood of becoming a target in a positive way (Barnes, 1999). Model five contains more power than model four according the pseudo r-squared. This model explains a higher proportion of the total variation. Since total assets and asset turnover are correlated, model five contains multicollinearity. The logit models imply that the debt to asset ratio, total assets and secured debt are the best predictors to assign potential targets.
6. Trading strategy
This paragraph describes the development of the investment strategy based on the financial predictors. These portfolios are created within the dataset that is used for this study. There is not constructed a validation sample to test last results. This approach might cause some biases due to an increased probability of assigning targets. With the results of table V, table VI and the literature there is suggested that debt to assets and total assets are the best variables to predict a potential target. Since secured debt is not significantly different between targets and non-targets it is not possible to assign targets to a specific portfolio. Below in Table VII sixteen different portfolios are manufactured on the basis of the debt to asset ratio and total asset levels. Each year contains four portfolios with 25% deciles of the variable ratios from low to high that vary by year. The portfolios are constructed over again each year. There are 200 firms used for these portfolios because some of the firms in the dataset miss observations regarding returns.
As shown below, there are 200 firms divided over the periods and portfolios including the 60 target firms. Debt to assets seems to be a good predictor to assign potential targets to portfolios. Total asset does not have the predictive power to assign potential targets to a portfolio according to the last results. Assuming that shareholders receive an abnormal return
Table VII Created portfolios
Portfolio
D/A ratio Low-‐high Period portfolio Deals in Total assets Portfolio Low-‐high Period portfolio Deals in
1 0%-‐25% 1990-‐1997 4 A 0%-‐25% 1990-‐1997 1 2 25%-‐50% 1990-‐1997 1 B 25%-‐50% 1990-‐1997 3 3 50%-‐75% 1990-‐1997 1 C 50%-‐75% 1990-‐1997 1 4 75%-‐100% 1990-‐1997 0 D 75%-‐100% 1990-‐1997 1 5 0%-‐25% 1998-‐2002 4 E 0%-‐25% 1998-‐2002 2 6 25%-‐50% 1998-‐2002 0 F 25%-‐50% 1998-‐2002 4 7 50%-‐75% 1998-‐2002 3 G 50%-‐75% 1998-‐2002 1 8 75%-‐100% 1998-‐2002 0 H 75%-‐100% 1998-‐2002 0 9 0%-‐25% 2003-‐2006 1 I 0%-‐25% 2003-‐2006 0 10 25%-‐50% 2003-‐2006 2 J 25%-‐50% 2003-‐2006 1 11 50%-‐75% 2003-‐2006 1 K 50%-‐75% 2003-‐2006 1 12 75%-‐100% 2003-‐2006 0 L 75%-‐100% 2003-‐2006 2 13 0%-‐25% 2007-‐2012 13 M 0%-‐25% 2007-‐2012 14 14 25%-‐50% 2007-‐2012 17 N 25%-‐50% 2007-‐2012 11 15 50%-‐75% 2007-‐2012 8 O 50%-‐75% 2007-‐2012 7 16 75%-‐100% 2007-‐2012 5 P 75%-‐100% 2007-‐2012 11 Total 60 Total 60
Table VIII Monthly portfolio return against market return
Portfolio
D/A ratio Low-‐high Period Alpha P-‐value
Portfolio
Total assets Low-‐high Period Alpha P-‐value
1 0%-‐25% 1990-‐1997 .1184492 0.837 A 0%-‐25% 1990-‐1997 .1299686 0.862 2 25%-‐50% 1990-‐1997 .0277449 0.952 B 25%-‐50% 1990-‐1997 -‐.1837975 0.722 3 50%-‐75% 1990-‐1997 -‐.0083607 0.990 C 50%-‐75% 1990-‐1997 .4637705 0.325 4 75%-‐100% 1990-‐1997 -‐.3501166 0.463 D 75%-‐100% 1990-‐1997 -‐.6018291 0.181 5 0%-‐25% 1998-‐2002 .3018233 0.623 E 0%-‐25% 1998-‐2002 -‐.5141344 0.490 6 25%-‐50% 1998-‐2002 -‐.1177396 0.933 F 25%-‐50% 1998-‐2002 .9861977 0.374 7 50%-‐75% 1998-‐2002 .2991205 0.695 G 50%-‐75% 1998-‐2002 -‐.1977603 0.846 8 75%-‐100% 1998-‐2002 -‐.6436047 0.181 H 75%-‐100% 1998-‐2002 -‐.5209926 0.252 9 0%-‐25% 2003-‐2006 -‐.4464601 0.611 I 0%-‐25% 2003-‐2006 -‐.9117014 0.530 10 25%-‐50% 2003-‐2006 .8400799 0.398 J 25%-‐50% 2003-‐2006 -‐.7222107 0.431 11 50%-‐75% 2003-‐2006 .0397929 0.970 K 50%-‐75% 2003-‐2006 .7521872 0.250 12 75%-‐100% 2003-‐2006 -‐.5494785 0.470 L 75%-‐100% 2003-‐2006 .7600612 0.087 13 0%-‐25% 2007-‐2012 1.688801 0.002 M 0%-‐25% 2007-‐2012 .6288188 0.947 14 25%-‐50% 2007-‐2012 .3076021 0.569 N 25%-‐50% 2007-‐2012 .8067742 0.132 15 50%-‐75% 2007-‐2012 -‐.1856297 0.719 O 50%-‐75% 2007-‐2012 -‐.4625208 0.294 16 75%-‐100% 2007-‐2012 -‐1.415689 0.884 P 75%-‐100% 2007-‐2012 -‐.3387604 0.370 Total firms 200 Total firms 200
around the announcement date, portfolio 1, 5, 9, 13, A, E, I and M should outperform the market. Most of these portfolios contain a relatively high ratio of targets. Below in table VIII the monthly returns are tested against the market return. The abnormal returns have been tested with the market model. Portfolio 10, 13, F, K, L and M has a higher return than the other portfolios in the same period. All these returns are positive but most of the returns in the table above are statistically insignificant. The monthly return is converted to a one-year period in table IX to analyse the results. The results show that portfolio 10 and 13 earn an abnormal return of, respectively, 10.08% (0.8400799*12) and 20.27% (1.688801*12), on average on a period of one year.
Table IX Average yearly portfolio return against market return
* means a significance level of 10%, ** means a significance level of 5% and *** means a significance level of 1%.
Portfolio Return Portfolio Return Portfolio Return Portfolio Return
1 1.42% 9 -‐5.36% A 1.56% I -‐10.94% 2 0.33% 10 10.08% B -‐2.21% J -‐8.67% 3 -‐0.10% 11 0.48% C 5.57% K 9.03% 4 -‐4.20% 12 -‐6.59% D -‐7.22% L 9.12%* 5 3.62% 13 20.27%*** E -‐6.17% M 7.55% 6 -‐1.41% 14 3.69% F 11.83% N 9.68% 7 3.59% 15 -‐2.23% G -‐2.37% O -‐5.55% 8 -‐7.72% 16 -‐16.99% H -‐6.25% P -‐4.07%
The return of portfolio F and K is 11.83% (0.9861977*12) and 9.03% (0.7521872*12) and the return for portfolio L and M is 9.12% (0.7600612*12) and 7.55% (0.6288188*12) on yearly base. These numbers have economic significance and portfolio 13 and L are statistically significant. Portfolio 13 contains firms with the smallest ratios of debt to assets and a high percentage of targets. Portfolio L contains firms with the highest ratios of total assets and also a high percentage of targets, but this contradicts the literature. Due to the inconsistency of the returns and the insignificant result this investment strategy will not outperform the market.
7. Conclusion
This study examines the energy industry in Europe, this is important since there is a lack of information concerning this topic. The utility industry is often excluded from studies because of the high levels of regulation, this article examines the energy industry exclusively.
The dataset includes (active) firms in Europe between 1990 and 2012. The sample group contains 82 deals of companies that became a target within this period. The control group consists of 153 firms that did not become a target during this period. The data shows a number of financial variables that differ significantly between the sample and control group. The most interesting variables confirmed by the literature are: total assets, dividend yield, the earnings yield, price to book, debt to asset, leverage, debt, equity, return on equity, return on capital, liquidity and secured debt.
The results of the takeover likelihood models suggest that total assets, secured debt, price to book, debt to assets, ROE and asset turnover are financial variables that contain predictive power. Unfortunately the effects of price to book, ROE and asset turnover do strike with the literature. Due to a combination of literature and results, debt to assets and total assets are assigned as the most reliable predictors of the financial variables. The dataset is divided in four different portfolios for each financial predictor for each year. The first portfolios in each period hold the 25% firms with the lowest debt to asset ratio and the last portfolios in each period hold the 25% firms with the highest debt to asset ratio. The
portfolios for total assets are composed in the exact same manner. The output shows that the most takeovers are included in the first portfolio this is confirmed by the literature. Portfolio 1, 5, 10 and 14 contains most takeovers based on debt to assets ratios this is in line with the literature. According to the literature portfolio A, E, I and M should contain most targets and portfolio D, H, L and P the smallest ratio of targets. These portfolios do not contain a regular slope regarding the ratio of targets within the portfolios.
The descriptive data shows that targets receive a return of 8.58% around the announcement, on average. This means that there is a positive abnormal return but this coefficient is not significant. The created portfolios have been tested against the market. Portfolio 13 and L are statistically significant with a return of respectively 20.27% and 9.12% on average on a one-year period. In conclusion, some of the portfolios outperform the market but the coefficients are inconsistent and often statistically insignificant. The trading strategy shows some imperfections regarding the construction and return of the portfolios.
8. Discussion
Most of the data is retrieved from Datastream. For some firms there was no availability of data and for some variables data was missing. The size of the dataset is limited due to missing financial data. Therefore, it is recommended for future research to choose the dataset carefully based on the data availability.
Another pitfall of this study is that the number of deals fluctuates over the years. For instance, most of the years contain one, two or three deals against 2008 and 2009 when, respectively 23 and 18 deals were exercised. To improve the predictive power, control firms should be similarly distributed over years regarding the sample group. Besides, the truly matching approach is disregarded in this study. This study recommends including this in the methodology.
This article focuses on the energy industry, however it might be better to study the utility industry as a whole. This is recommended since more data is available for the entire industry. However, precaution should be taken here since the different laws and regulations of the entire industry should be taken into account.
References
Ambrose, B. W. and Megginson, W. L. (1992). ‘The role of asset structure, ownership structure, and takeover defenses in determining acquisition likelihood’, Journal of Financial and Quantitative Analysis, Vol. 27, pp. 575–89.
Barber, B.M. and Lyon, J.D. (1997). Detecting long-run abnormal stock returns: The
empirical power and specification of test statistics, Journal of Financial Economics, Vol. 43, pp. 341-372.
Barnes, P., (1999). ‘Predicting UK takeover targets: some methodological issues and an empirical study’, Review of Quantitative Finance and Accounting, Vol. 12, , pp. 283– 301.
Basu, S. (1977). Investment performance of common stock in relation to their price-earnings ratios: a test of the efficient market hypothesis. The Journal of Finance, Vol. 32, No. 3, pp. 663-682.
Becker-Blease, J. R., Goldberg, L. G. and Kaen, F. R. (2008) Mergers and acquisitions as a response to the deregulation of the electric power industry, Journal of Regulatory Economics, 33, 21–53.
Berry, S.K. (2000). Excess Returns in Electric Utility Mergers During Transition to Competition. Journal of Regulatory Economics, vol. 18, pp 175-188.
Blume, M.E. (1971). On the assessment of risk, The Journal of Finance, Vol. 26, pp 1-10. Brar, G., Giamouridis, D. and Liodakis, M. (2009). Predicting European takeover targets. European Financial Management, Vol. 15 Iss. 2, pp 430-450.
Cai, Y. and Tian, X. (2013). Does firm’s geographic location affect its takeover exposure? SSRN working paper.
Datta, S., Kodwani, D. and Viney, H. (2013). Shareholder wealth creation following M&A: evidence from European utility sectors, Applied Financial Economics, 23:10, 891-900 Dietrich, J.K. and Sorensen, E. (1984). An application of logit analysis to prediction of Merger targets, Journal of business research, vol. 12, pp. 393-402
Erkens, D.H., Hung, M. and Matos, P. (2012). Corporate governance in the 2007-2008 financial crisis: Evidence from financial institutions worldwide, Journal of Corporate Finance, Vol. 18, pp. 389-411
Giammarino, R.M. and Heinkel, R.L. (1986). A model of dynamic takeover behavior. The Journal of Finance, Vol. 41, pp. 465-480.
Harford, J. (2005). What drives merger waves? Journal of Financial Economics, Vol. 77, pp. 529-560.
Hasbrouck, J. (1984). The characteristics of takeover targets; q and other measures. Journal of banking and finance, Vol. 9, Iss. 3, pp. 351-362
Hooper, E. & Waddams Price, C. (2010). Innovation and competition in generation and retail power markets, Institute of Economic Affairs.
Hopkins, H.D. (1999). Cross-border mergers and acquisitions: Global and regional Perspectives, Journal of International Management, 5, pp. 207-239
Jamasb, T. & Pollitt, M. (2005). Electricity Reform in the European Union: Review of Progress toward Liberalization & Integration.
Jensen, M. (1986). Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers. The American Economic Review, Vol. 62, pp. 323-329.
Kennedy, V. A. and Limmack, R. J., ‘Takeover activity, CEO turnover, and the market for corporate control’, Journal of Business Finance and Accounting, Vol. 23, 1996, pp. 267– 85.
La Porta, R., Lopez-de-Silanes, F., and Shleifer, A., (2006) What works in securities laws? Journal of Finance, Vol. 61, pp. 1–32.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., and Vishny, R., (1998) Law and finance, Journal of Political Economy, Vol. 101, pp. 678–709.
Lewellen, W. (1971). A pure financial rationale for conglomerate merger. Journal of Finance, 26, 521-545.
Ljungqvist, A. and Wilhelm, W.J. (2003). IPO Pricing in the Dot-com bubble, The Journal of Finance, Vol. 58, pp. 723-752.
Marris, R. (1964) The Economic Theory of Managerial Capitalism, The Free Press, Glencoe, Il,
Martynova, M. and Renneboog, L. (2006). Mergers and acquisitions in Europe. Social Science Research Network Working Paper, no 114/2006.
Matthew T. Billett, M.T., King, T.H.D. and Mauer, D.C. (2004). Bondholder Wealth Effects in Mergers and Acquisitions: New Evidence from the 1980s and1990s
The Journal of Finance, Vol. 59, No. 1, pp. 107-135
Mulherin, J.H. and Boone, A.L. (2000). Comparing acquisitions and divestures. Journal of corporate finance, Vol. 6, pp. 117-139
Newbery, D. (2007). What are the issues in merger and acquisitions arising from electricity Market restructuring? working paper
Owen, S. and Yawson, A. (2010). Corporate life cycle and M&A activity. Journal of Banking & Finance, Vol. 34, Iss. 2, pp 427-440
Palepu, K. G., (1986). Predicting takeover targets: a methodological and empirical
