Oil price impact on M&A activity in the U.S. oil and gas industry

Transcription

1 Master thesis: Oil price impact on M&A activity in the U.S. oil and gas industry Name: ANR: Course: Supervisor: Second reader: Date: Martijn Bos Master Thesis Finance Dr. R.G.M. Nijskens Dr. F. Castiglionesi 30 November, 2016

2 ABSTRACT In this study is investigated whether industrial shocks impact M&A activity in the U.S. oil and gas industry. M&A activity is measured by analyzing 6,643 deals in between 1984 and To study industry shocks, oil price variables are included which are a good shock indicator in the oil and gas industry. This study examines these oil price variables besides the existing determinants for M&A activity. The different OLS models found empirical evidence that industrial shocks influence merger waves in the U.S. oil and gas industry. Subsequently, this study compares the same regression between acquirers that are already in the oil and gas industry to non-oil acquirers. No statistical differences were found. 2

3 TABLE OF CONTENTS I. INTRODUCTION... 4 II. LITERATURE REVIEW History of mergers and acquisitions Motives for mergers and acquisitions Oil and gas industry III. HYPOTHESES, METHODOLOGY AND DATA Hypotheses Data Variables Methodology IV. RESULTS V. CONCLUSION Main findings Limitations and further research VI. REFERENCES VII. APPENDIX

4 I. INTRODUCTION Mergers and acquisitions (M&A) are a well-researched topic, with issues such as the creation of shareholder value through takeovers and types of deal breakers often discussed. Furthermore, the reasons why takeovers occur have been investigated. Several theories behind the occurrence of takeovers have been developed in the existing research and will be described in the next chapter. The vast majority of literature agrees that takeovers tend to occur in waves. According to the academic literature, we are now in the 7 th wave. Each of these waves had its own characteristics, but there are three recurring determinants for M&A-activity. The first determinant is the current state of the economy, such as expectations for growth. Second, sufficient capital is needed to accommodate transactions and thus capital markets must be accessible. The last motive takes industrial shocks into account. Though these findings are interesting, they would be even more valuable when applied to specific industries and give empirical evidence. Testing this quantitatively, requires an industry in which the motive for industrial shocks are measured. Mitchell and Mulherin (1996) give examples for shocks, with one of them being shocks due to changes in the oil price. This affects one of the largest industries in the world: the oil and gas industry. Because this industry is seriously affected by changes in the price of oil, the market price is a transparent indicator of the state of the industry and incorporates all the shocks the industry has to deal with. So besides being the biggest industry, it is also one of the most balanced industries worldwide. In addition oil has a significant impact on the world s current economic situation. Some even suggest that this industry is the main driver of economic development, since it is still one of the most important resources. This leads to the following research question: Is oil price positively related to M&A activity in the U.S. oil and gas industry? To investigate this relation is M&A activity measured by the monthly number of deals. Four hypotheses are developed to answer the research question. First, the positive effect of the oil price on the M&A activity will be shown. The second hypothesis found a negative relation 4

5 between oil price volatility and M&A activity. The last two hypotheses could not answer if an acquirer active in the oil and gas industry is more sensitive to oil price and oil price volatility compared to an acquirer who is not active in the oil and gas industry. Concluding, in this study empirical evidence is found that industrial shocks influence merger waves in the U.S. oil and gas industry. Although, the influence is not large. As expected are the best circumstances for M&A activity in this industry a high oil price in combination with low volatility of the oil price. Surprisingly, M&A activity in the oil and gas industry has not been discussed in academic literature for a long period, while it is still one of our largest and most important industries worldwide. This study can be used by oil companies or investment bankers to estimate what to expect from M&A activity. Since oil and gas industry analysts expect lower oil prices with a higher price volatility, less M&A activity in the oil and gas industry is expected. Furthermore, the new combination of control variables seem to be useful in explaining M&A activity. Control variables are conducted from previous literature and all seem significant determinants for measuring M&A activity. The paper is organized as follows. After reviewing the literature on M&A determinants, takeover waves, and the oil and gas industry in Chapter 2, the hypotheses and methodology are presented in Chapter 3. The methodology to test the hypotheses is based on an existing model for M&A activity with the explanatory variable oil price included. The results of the hypotheses tests are shown in Chapter 4 and the last chapter concludes with main findings, the limitations of this study and suggestions for further research. 5

6 II. LITERATURE REVIEW This chapter provides a theoretical background, from which the hypotheses will be developed. First, the different merger waves in the last century are discussed. Subsequently the occurrence of takeovers will be explained. This will be followed by an insight into the characteristics of the oil and gas industry. 2.1 History of mergers and acquisitions As described above, mergers and acquisitions tend to cluster in several takeover waves, the first starting in approximately The world is currently in the seventh wave. This study will only consider results observed after 1984, but the first waves will also be discussed briefly below. The different characteristics of, and perspectives about the occurrence of these waves provides a theoretical background in order to determine the factors influencing merger activity. 1 st wave: The first cluster of mergers and acquisitions occurred after a period of economic expansion through industrialization. O Brien (1988) finds that these companies were not specifically aiming to achieve economies of scale, but rather to increase their market power. This within-industry consolidation created the first giants in industries such as the oil industry. The Sherman Antitrust Act was created as a result of weak monopoly regulations, but it was not enough to limit M&A activity (Stigler, 1950). Nevertheless, this first wave was characterized by friendly deals that were financed mostly in cash. The end of this wave was caused by the stock market crash, the beginning of the First World War, the weak banking system of the US and the increasingly rigorous Sherman Antitrust Act. 2 nd wave: After the First World War the economy recovered and the second wave started. This post-war boom provided access to capital, which was driven by low margin requirements and favorable economic conditions. The impact of this wave was significantly lower than the magnitude of the first wave; transactions were mostly vertical takeovers financed by equity. Monopolies could not emerge further because the Sherman Antitrust Act had become stricter. Small companies that wanted to benefit from economies of scale were active and oligopolies emerged instead of monopolies. Finally, The Great Depression brought an end to this wave. 6

7 3 rd wave: The Great Depression, was followed by the Second World War. This period is also known as the conglomerate merger period. Shleifer and Vishny (1991) have explained this tremendous growth as being a result of the Cells Kefuaver Act of This stricter antitrust law aimed to prohibit horizontal and vertical integrations. For this reason, firms chose conglomerate mergers. On the other hand, Sudarsanam (2010) has claimed that the percentage of companies active in unrelated business increased from 9% to 21% because companies wanted to reduce volatility in their income. Just as in the second wave, equity the most used source of financing. What is remarkable about this period is that it was not uncommon for smaller companies to buy larger companies. The takeover activities collapsed completely when the oil crisis began. 4 th wave: The fourth wave was the most spectacular wave of the last century because it was quite different than previous ones on several aspects. While the first three waves takeovers were mostly financed by share issues, firms in the fourth wave financed transactions by issuing debt. The willingness of banks to finance these leveraged buyouts lead to hostility and the largest takeover that had yet been seen. A good example of a large leveraged buyout in this period is the acquisition of RJR Nabisco by KKR. 1 Many conglomerates wanted to eliminate the inefficiencies that were created by the conglomerate mergers of the previous wave. The so-called corporate raiders observed that conglomerates were often worth more if they specialized and thus divided their divisions into small independent firms. This wave is characterized by conglomerates downsizing their businesses, which constituted approximately 20-40% of the M&A activity. Simultaneously, corporations wanted to expand their competitive position through acquisitions (Sudarsanam, 2010). Specific shocks also played a role in this wave. There was active deregulation in the air transport, broadcasting, entertainment, and natural gas and trucking industries (Mitchell & Mulherin, 1996). Research shows that a bid on a firm in the same industry lead to a positive stock market return for the bidding firm during this wave. The opposite holds for unrelated bids, so the market had a negative attitude towards unrelated diversification, a strategy that was 1 Described extensively in Barbarians at the Gate (Burrough & Helyar, 2009) 7

8 appreciated during the third wave (Morck et al., 1990). The stock market crash at the end of the 1980s marked the end of this wave. 5 th wave: The most distinguishing feature of this wave is its international character. The number of crossborder acquisitions increased as a result of technological innovations that led to globalization. In contrast to the previous waves, the amount of divestures declined and corporations actively looked for industry-related targets across their borders to strengthen against global competitors. It was the first time that Europe s M&A activity attained the same level as the market in the US. Large merger deals occurred, that were mostly financed by equity and had a friendly character. The biggest oil merger in history took place during this period between Exxon and Mobil in This deal was valued at approximately $77 billion. The crash on the global stock market in 2000 resulted in a decline in M&A activity (Sudarsanam, 2010). Corporate scandals caused uncertainty and new legislation led to the end of the fifth wave. 6 th wave: This wave began when the global financial markets began to recover after the dot-com bubble. The recovery came quickly because the US Federal Reserve System dropped the interest rates, making debt a popular and cheap source of financing. The number of takeovers increased rapidly during this sixth wave. Harford s (2005) neoclassical theory shows that the abundance of capital liquidity in the market impacted M&A activity. Many of the acquirers were private equity companies during this wave. They bought companies and waited for the market to rise. In contrast to previous waves, private equity became a major player in the M&A market because of the low cost of capital. Martynova & Renneboog (2005) have claimed that the sixth merger wave was initiated by the terrorist attacks of 9/11. At that time the market was highly insecure and investments were postponed. When the market began to rise again, investments exploded and caused a new wave. As in the fifth wave, globalization had a significant impact on the M&A market during the sixth wave. For instance, merger waves began everywhere at the same time. Investors took advantage of their access to global markets, giving this wave a global flavor. In 2007 the sub-prime crisis arrived, making capital markets inaccessible due to the uncertainty. Private equity firms withdrew from the market. 8

9 7 th wave: Although markets appear to be relatively volatile at the moment, it seems that M&A activity has increased once more recently. Merger and acquisition activity has been able to recover, because of changes in the financial sector and divestures around the world created opportunities. Takeovers are a good chance to achieve growth because it is hard for firms to create growth organically. In addition, firms from countries such as the BRIC countries are constantly looking for target firms. Summary All seven merger waves have their own characteristics. Still, recurring patterns exist and three of these patterns are considered to be main factors to characterize a merger wave. First, merger waves start after exogeneous shocks. Second, the economic growth is high during the wave. Last, the wave ends by a stock-market collapse and changes in regulation. These factors are discussed further in section 2.2. The most remarkable differences are the different sources of financing among the waves. An interesting overview of the changing climate in the capital markets is that buyers constantly look for the cheapest source of financing. Furthermore, a new phenomenon appears in the three most recent waves: the size and frequency of mergers and acquisitions increased due to globalization. 2.2 Motives for mergers and acquisitions There are several reasons why companies merge, such as to create more market power, diversify, to benefit from economies of scale or to synergize costs (Andrade et al., 2001). There are three types of takeovers. First, horizontal takeovers combine two or more competitors. Second, vertical takeovers occur when a company merges with its supplier or distributor. Third, a conglomerate takeover takes place when companies from different industries merge. In previous section becomes clear that mergers and acquisitions have come in waves in the past. Different theoretical models have been developed to capture the motives behind these takeover waves. These theories can be classified into two groups: the neoclassical hypothesis sees M&A activity as a response to different industry shocks while the behavioral hypothesis explains the occurrence of takeovers through rational managers who take advantage of consistent pricing errors (Vancea, 2012). 9

10 Neoclassical hypothesis Companies react to the opportunities that accompany different shocks. Neoclassical models suggest that this includes economic, technological and regulatory shocks. Gort (1969) stated that economic disturbance leads to industry reorganization. Firms aim to expand their business when the aggregated demand in their industry grows. In that case, takeovers are a faster way to expand compared to organic growth. Coase (1937) was the first one to identify a connection between an important technological change and M&A activity. Technological shocks can be considered as supply shocks that result in excess productive capacity (Jensen, 1993). So these shocks might either create new industries or generate changes in existing sectors. For industries undergoing a positive demand shock, developed companies are more willing to acquire less productive firms (Maksimovic and Phillips, 2001). Regulatory shocks occur when new laws are implemented or when deregulation occurs. Finally, Harford (1999) has noted that industry or regulatory shocks are a reason for takeovers, but also notes that there must be sufficient capital liquidity to accommodate these transactions. A combination of these factors causes a takeover wave. Behavioral hypothesis The fundamentals of this model have been created by the positive correlation between stock market valuations and M&A activity. Schleifer and Vishny (2003) have argued that overvalued stocks are a trigger for takeovers because this situation gives bidders opportunities to buy undervalued target firms through mergers. However, Rhodes-Kropf and Viswanath (2004) have found that when a firm is overvalued, managers of the target company would not accept the offer unless they too are overvalued, resulting in an overestimation of synergies. In summary, past research has attempted to examine the triggers of takeovers. The neoclassical models seems most interesting to apply for this research, since they investigate changes in the environment of the firms more closely. Furthermore, the studies of Harford (2005) and Mitchell and Mulherin (1996) have found this approach to be more appropriate. The most important finding of this model is that takeovers tend to cluster within industries due to economic, regulatory and technological shocks. However, the state of the capital market at the moment of the takeover is crucial to facilitate the transaction. 10

11 2.3 Oil and gas industry Gort (1969) argues that industries are not affected in the same way by economic shocks. This suggests that there is a dispersion of M&A activity between industries. However, this study will be focusing on one industry to investigate the research question: the oil and gas industry. Many companies are involved in the process of exploring, extracting and refining oil. The growth of this industry started in the first wave during the industrialization period and accelerated during the 20 th century to become the most dominant fuel in the Western world. The dominance of this resource is evident in its impact on different industries, crises, and wars. Since oil is a homogeneous product and used all over the world, is the price a transparent indicator of the state of the industry worldwide. The market price of a barrel of crude oil is set in the oil futures market. Contracts are traded creating a binding agreement that gives the holder the obligation to purchase oil for a certain price on a certain date. There are several factors that impact the price of oil. The main factors are supply, demand, the development of alternative fuel and geopolitical crises. There are two types of traders in the oil future market. First, airlines that want to hedge themselves against the risk of rising oil prices. However, the majority of traders are speculators who try to guess price direction without any intention of actually buying the product. The sentiment of the speculators, such as the expectation that oil demand will increase in the future, results in significant increases in future market prices in the present. The nature of this system makes the price of oil sensitive and volatile. In 2014 the oil price collapsed to historically low levels, from more than $100 per barrel to half the price today. While many industries benefitted from this decline, the oil and gas industry itself did not. Most oil companies experience pressure on their cash flows, which is troubling in an industry that already has to deal with low returns on capital. In response, companies reset their cost base and decrease their investment spending. On the other hand, the shareholders of Royal Dutch Shell recently gave permission for the acquisition of BG Group while the oil price was at a low level. This transaction was valued at approximately $47 billion and became one of the largest deals in the oil and gas industry this decade. However, at least for the next decade, analysts expect that lower prices with a higher price volatility will be the status quo for the oil and gas industry, so further restructuring in the sector will be necessary for it to remain profitable. 11

12 III. HYPOTHESES, METHODOLOGY AND DATA This section begins with the deduction of the hypotheses with the theoretical background in mind. The methodologies used will then be discussed, concluding with the choice of variables. 3.1 Hypotheses Several hypotheses were developed to find the impact of the oil price on M&A activity in the U.S. oil and gas industry. First, the effect of the price of oil on M&A activity will be shown. The second hypothesis relates to whether the volatility of the price of oil impacts M&A activity. The last two hypotheses investigate whether there is a difference in explanatory variables for various types of buyers. H1: Oil price is positively related to M&A activity in the U.S oil and gas industry Most oil and gas companies experienced pressure on their cash flows during spells of low oil prices, decreasing spending and changing cost bases. Outsiders would say that mergers and acquisitions will be postponed. Testing this hypothesis will show whether or not there is a positive relationship between the oil price and M&A activity in the oil and gas industry. H2: Increasing oil price volatility has a negative impact on M&A activity in the U.S oil and gas industry The price of oil is volatile in a market where the price is set by the sentiment of speculators. As stated above, analysts expect that higher price volatility will be the standard, so further restructuring in the sector will be necessary. As in every market, uncertainty is never desirable and will thus have a negative impact on investments and M&A activity. H3: The M&A activity with an acquirer active in the oil and gas industry is more sensitive to changes in the oil price compared to the M&A activity with an acquirer who is not active in the oil and gas industry As stated earlier, oil companies cash flows suffer when oil prices are under pressure, which could lead to a demand for extra capital liquidity. Banks are mostly unwilling to lend extra fund to companies in declining markets, thus companies have to look further for other sources of financing. This could be the chance for non-oil companies, such as conglomerates and 12

13 private equity to enter the oil and gas industry. This happened during the sixth merger wave when private equity bought companies that needed financing and waited for the market to rise. H4: The number of deals made by companies active in the oil and gas industry decreases much faster compared to deals made by a non-oil company, when the oil price volatility increases. Unlike companies already active in the oil and gas industry, non-oil companies that enter the oil and gas industry are less dependent on the current state of the industry. Uncertainty about the future means a lower market value of the target firms. New entrants could benefit from this state. 3.2 Data This section presents how the right data was gathered for this research. Database The data for the number of deals was collected from the SDC Platinum database of Thomson Reuters, one of the foremost financial transaction databases. Available information about historical transactions goes back to 1976, but the number of deals in the 1970s and beginning of the 1980s is so low that it is not possible to be certain that every deal was included in the data. Of the concerning deals, the following data was collected: announcement dates, acquirer s; name, industry and nation, target s; name; industry and nation and the percentage of shares. Dreher s (2006) KOF Index provides data which measures globalization for at least three decades. Data for all other variables were collected from Datastream, also a Thomson Reuters database, covering the different economic indicators such as stock market indices and country-specific ratios. Sample The fourth wave of 1984 is used as the start of the sample and ends in Because the research is focused on M&A activity in the U.S. oil and gas industry, only the transactions with an oil company involved have been collected. In table 6 (see appendix) the SIC codes are included which are labeled as oil and gas industry. To be included in the data, either the acquirer, the target or both companies had to be in this industry. This study only included transactions when the acquirer is headquartered in the U.S.. The main reason to verify this choice, is that all control variables refer to U.S. economic conditions. 13

14 The transaction had to be a full acquisition in order to be included in the sample, making it easier to consider similarities in the rationale behind every deal. After analyzing the data in different ways, the 10 th of May 1993 stands out with 36 deal announcements, compared to a maximum of 9 announcements on other days. Focusing on this day in the dataset it becomes clear that Parker & Parsley Petroleum Co. bought 35 different companies from Prudential-Bache. Thus this data point was corrected to one announcement. This results in a sample of 6,643 transactions after excluding the transactions that did not become effective. 3.3 Variables The creation of the models used was based on previous research. The explanatory, dependent, and control variables are described in relation to what previous studies have concluded about the variable. Table 1 contains the key descriptives of each variable. Descriptives # of observations Mean Median Standard deviation Minimum Maximum Explanatory variables Oil price Oil price volatility Dependent variables Deals Deals oil Deals non-oil Control variables Corporate bonds S&P 500-index Unemployment rate Globalization Table 1. This table contains all the descriptives of the variables used. First, the explanatory variables are shown, afterwards the dependent variables. Lastly, the control variables are presented. Explanatory variables Oil price The price of a barrel of West Texas Intermediate (WTI) crude oil was used to measure the price of oil in a certain month because it is most often used as the benchmark for oil pricing. Making it a more appropriate choice than, for example Brent crude oil, because Brent is from the North Sea and WTI is refined mostly in the Midwest and Gulf Coast. 14

15 The price of Light Sweet Crude Oil Futures contracts have been trading on the New York Mercantile Exchange since Currently WTI is been discounted against Brent, whereas historically it has been traded closely. This anomaly has occurred through a temporary shortage of refining capacity or a surplus of oil in storage. Table 1 shows that the price of oil in the sample has an average price of $42.12, has never been lower than $10.42 and not higher than $140. Figure 2 (see appendix) shows that the price of oil began to increase around 2002, with two steep falls during the subprime crisis in 2008 and the decreased demand for oil in Oil price volatility The volatility of this oil price is also interesting to consider because volatility provides information about the uncertainty in the sector. The Crude Oil Volatility Index, an index based on the 30-day implied oil price volatility, was the first choice for this variable, but the index s data only goes back to For this reason, the variable was constructed specifically for this study by taking the monthly standard deviation of the oil price. As described earlier, the oil and gas industry has dealt with higher oil price volatility in recent years, as confirmed in figure 2 (see appendix). A negative relationship between oil price volatility and the monthly number of deals was expected. Dependent variables M&A-activity Two measures could be used for M&A activity. First of all the total sum of the transactions values of the takeover in that period or the number of deals in that period. The second measure was chosen for this study because deal value is not always communicated and the dependent variable will be too sensitive to large deals. The time frame followed Resende s (1999, 2008) recommendation and worked with monthly data to increase the predictive quality. A smaller time frame was not possible because information for some of the other variables was not available on that scale. Deals were allocated to the period in which they were announced. Three things could occur between the announcement and the effective date: the deal could be completed, the deal could be completed but with adjusted conditions or the deal could be cancelled. Since this variable does not include cancelled deals, this sample only contains completed deals. As described above, conditions such as transaction value are not relevant to our dependent variable, and thus 15

16 it is not important to the sample if a deal was completed or completed after some adjustments to the conditions. Therefore M&A activity is defined as the monthly number of completed transactions that involved an oil company between 1984 and 2015 by an acquirer headquartered in the US. Table 1 shows that there are on average 17.3 deal announcements in the sample per month. As mentioned in the literature review, takeovers cluster over time and this also holds for the U.S. oil and gas industry. Figure 1 shows that the number of deals was structural on a higher level during certain years. This shows that there was a peak in takeovers in the U.S. oil and gas industry during the fifth merger wave between 1996 and Most of the deals in the sixth merger wave took place around More surprising was that the number of deals stayed at a relatively high level after this wave (thus at the start of the subprime crisis) Number of deals Oil related deals Non-oil acquirer Oil acquirer Figure 1. This figure contains the same variable as figure 3 (see appendix), but on a yearly base. The line concerning deals that included at least one oil company shows two peaks in the U.S. oil and gas industry between 1984 and One during the fifth merger wave between 1996 and And one during the sixth merger wave around More surprisingly, the number of deals stayed at a relatively high level after this wave (thus at the start of the subprime crisis). The line concerning number of deals in which both the acquirer and target are oil companies seem to be highly correlated with the first considered line. Lastly, the line concerning deals in which the number of deals for firms entering the oil and gas industry are counted shows an upward trend. 16

17 M&A activity by an oil acquirer The same conditions as our number of deals variable stated above hold for this dependent variable used in regression (2). The only difference is that the acquiring and target firm both have to be active in the oil and gas industry. Deals with a non-oil target were excluded from this variable in order to make the two samples more identical, it left 3,700 deals. In figure 1 plots this variable on a yearly base. This line seem to be highly correlated with the line for overall M&A activity in the U.S. oil and gas industry. This high correlation is confirmed in table 8 (see appendix). Table 1 shows that, on average, there have been 9.6 deals monthly deals where an U.S. oil company bought another oil company between 1984 and M&A activity by a non-oil acquirer Regression (3) considers 1,677 deals of acquirers that have entered the oil and gas industry by buying an oil company. This could be a firm from many other industries such as private equity, business service or the chemical industry. In figure 1 the number of deals concerning firms entering the oil and gas industry is visualized. This line shows an upward trend, implying that there are more and more outsiders entering the industry. On average monthly 4.4 non-oil companies bought an oil company (table 1). Control variables The control variables for measuring M&A activity were mainly extracted from Benzing s (1991) research. The fundamentals of this model were initiated by Steiner (1975) and outline the two basic theories of merger activity. One theory includes the expectations of economic growth, such as business failures, industrial production growth, and stock prices. The other theory considers the state of the capital markets. Although this theory is from more than two decades ago, it was included in Very et al. s (2012) studies of the determinants of M&A activity. Thereby, all the variables have been checked against other studies for their relevance. There have been some adjustments to the variables for previous merger activity, stock market index, and globalization, in Benzing s (1991), to be described briefly below. 17

18 Previous merger activity Clark et al. (1988) have supported the hypothesis that past merger activity affects present merger activity. This is an effect of the clustered wave patterns of M&A activity (Golbe & White, 1993). Persons and Warther (1997) have explained that takeover waves are also a result of companies responding to the M&A activity of their competitors. So, successful merger transactions create a knock-on effect in the industry. Similarly, takeover activity declines after a series of unsuccessful takeovers. Benzing s (1991) study has created two variables that look for the number of deals both one year ago and two years ago. Instead of yearly observations, this thesis measures on a monthly base, with an autoregressive model that shows how many lags are present. For regression (2) and (3) the same amount of lags are included as regression (1). Those also measure respectively the number of deals done by an oil acquirer and the number of deals done by a non-oil acquirer. Cost of debt Interest rates are a relevant driver of M&A activity, by providing bargain opportunities for lowleveraged companies or cash from private equity sources. Following Polonchek and Sushka (1987), a rise in the cost of debt increases the reluctance of buyers to make acquisitions. Becketti (1986) and Golbe and White (1993) have also found a negative effect of real interest rates on M&A activity. In contrast, Steiner (1975) and several other authors have found that the interest rate appears to have a positive influence on the dependent variable. This could be because high interest rates are also an indicator for high return on investments, which in turn encourages M&A. The index for U.S. long corporate bonds is included as a measure for the state of the debt capital markets. The mean for these bonds is approximately 101, with a base index of 100. For this variable it is important to note that an increase of the bond price implies a decrease in the cost of debt. Stock market index Both hypotheses explain stock prices as being influential on M&A activity. However, since this thesis follows the neoclassical approach, attention will be given to their arguments to include this variable. The market value of firms was used as an indicator for future economic conditions. Mitchell and Mulherin (1996) have found stock market performance to be the most robust result to takeover activity. The positive influence of the stock market index on the M&A 18

19 activity has been found systematically. One of the first to find this relationship was Weston (1953). Later, Owen (2006) has explained that increasing stock market indices also often mean that the country s current economy is strengthening, leading to increased profits for many firms. In contrast to other literature suggested the country s Gross Domestic Product is excluded, because the stock market index variable already incorporates the country s current economic shocks. Unlike Benzing (1991), this study does not use the absolute value of the S&P 500-index. Instead, stock market growth was used to measure the recent performance. Resende (2008) has also made this decision and argued that, besides the arguments described above, high growth is also associated with high valuations and greater uncertainty about the value of a target firm. This misevaluation argument associates increasing M&A activity with periods of high valuation (Rhodes et al., 2005). The log-returns of the S&P 500-index performance for the past 12 months were used to measure this variable. Table 1 shows that the historical return was between -59.3% and 40.7%, with a mean of 8.0%. Following the literature it is expected that if stock prices are increasing, there is optimism for the future and this would in turn affect merger activity in a positive way. Unemployment rate Another measure of the state of the economy is the unemployment rate in the country. This variable proxies the current state of the economy and business cycle. Beside the industry shocks, Mitchell & Mulherin (1996) have found evidence that employment shocks are positively related to M&A activity. Higher employment means a higher level of production and thus that the economy is in a better state. This is expected to encourage M&A activity. On the other hand, O Shaughnessy & Flanagan (1998) have said that mergers and acquisitions go together with employee layoffs. However, the U.S. unemployment rate has not become higher than 10.0% or lower than 3.8% in the last three decades. The average unemployment rate in the sample is 6.2% (see table 1). Considering the literature background and logical reasoning, it is understood that unemployment and M&A activity are negatively correlated. 19

20 Globalization Many things have changed since Steiner (1975) initiated this model. The world is becoming more connected through technological innovation and globalization has become an important driver for increasing M&A activity because companies have begun to look for similar companies abroad in order to strengthen against global competitors. It is for this reason that it is important to include a variable for globalization. The KOF index was used to measure this indicator. This index also takes social and political dimensions into account as well as the typical economic flow variables. Only the economic dimension concerning variables related to financial and trade restrictions were included for this study because of the expectation of minimal influence of the social and political dimensions. The globalization level in the KOF index is between 0 and 100. In table 1 is stated that this score has been 53.8 and 65.4 for the us, between 1984 and 2015 this score was on average Methodology This paragraph will explain the analyzed and assessed data. The aforementioned hypotheses were tested by an ordinary least squares (OLS) model. An autoregressive model will be used to choose how many lags are used. Ordinary Least Squares Model An OLS model was used to test the statistical significance of the impact of the independent variables on M&A activity (i.e. the dependent variable). The OLS model estimates the effect of the unknown parameters by minimizing the sum of the mean squared error. The significance of this influence is examined at the 1% or 5% significance level. The number of deals per month was used to measure M&A performance. The beta of the explanatory variable (oil price) and the control variables indicate the effect on the number of deals per month. The residual - effects that were not measured in the model are indicated by the error term. Lastly, robust standard errors were used for all regressions, but these standard errors have been corrected to make them consistent. The oil price variable in regression (1) was replaced by the variable of oil price volatility to test the second hypothesis. & = (1) 20

21 For the third hypothesis, regressions (2) and (3) were used to test if there is a significant difference in the betas for oil price by using two different dependent variables. These regressions were also used to test the fourth hypothesis, only the variable for oil price was replaced by the oil price volatility variable in both regressions. &, = (2) &, = (3) Unit Root Test The Dickey-Fuller (1979) test assesses the variable for the presence of a unit root. If the variable contains a unit root it indicates that there is a non-stationary process and thus the presence of a drift or trend in the variable. In this case a linear trend, the log or first order condition of the variable, can be added. After these steps the null-hypothesis, needed to run an autoregressive model, must be rejected. Autoregressive model Since takeovers tend to come in waves, lags are included in my OLS model. This autoregressive model is used to find the appropriate number of lags of the M&A activity-variable. The test can only be used for variables that are stationary. The autoregressive model provides insight into the random process of a variable since the output variable depends on its own previous values. The number of lags that have a significant influence were shown by running the test for different numbers of lags. Wald test In order to test the third and fourth hypotheses, it is necessary to compare the β1 s of regressions (2) and (3). Assuming that the difference between the two is normally distributed, this Wald test (1943) estimates the difference from 0 between the two betas statistical significance. This was tested by comparing the square of the difference to a chi-squared distribution. 21

22 IV. RESULTS The hypotheses were tested with Stata, to see if the effect of the explanatory variables holds after including the control variables different models are used. H1: Oil price is positively related to M&A activity in the U.S. oil and gas industry Table 2 presents several models that explain the dependent variable, as in Benzing (1991). The explanatory variable oil was regressed on the number of deals in the same month in the first model. As expected, this variable had a positive and even significant influence. This model has a relatively low adjusted R², with only 16.5% of the change in the number of deals explained by the price of oil. Model I II III IV V VI VII Oil price 0.108*** 0.028* * * (9.23) (2.13) (1.44) (1.69) (2.23) (1.34) (2.12) Deals t *** 0.303*** 0.301*** 0.289*** 0.255*** 0.258*** (6.05) (5.38) (5.33) (5.27) (4.51) (4.66) Deals t ** 0.164** 0.162** 0.149** * (3.21) (2.79) (2.77) (2.61) (1.92) (2.01) Deals t *** 0.211*** 0.210*** 0.195*** 0.152** 0.156** (4.26) (3.82) (3.82) (3.56) (2.74) (2.83) Corporate bonds 0.088* 0.079* 0.083* (2.55) (2.31) (2.48) (1.68) (1.88) S&P 500-index * 3.544* (1.33) (1.37) (2.37) (2.20) Unemployment rate (-1.66) (0.97) Globalization 0.624** 0.478*** (3.29) (3.74) Constant *** 3.237*** ** *** (19.04) (4.51) (-1.43) (-1.19) (-0.31) (-3.17) (-4.14) Adjusted R Number of Observations * p<0.05, ** p<0.01, *** p<0.001 Table 2. Different models present the number of deals each month in the U.S. oil and gas industry between 1984 and The variable for previous number of deals, corporate bonds, S&P 500-index and globalization are positive related to the dependent variable. The one for unemployment rate is dropped in our last model, because of economic and statistical insignificance. Almost every model shows a positive and significant beta for the relation between oil price and M&A activity. 22

23 The lags of the number of deals were added in model II, to choose how many lags an autoregressive model was used as a robustness check. First, the variable number of deals was checked for stationarity, without this constant component it is impossible to forecast by definition and to run an autoregressive model. In figure 3 (see appendix) there seems to be a small increasing trend in the mean of the number of deals. To be sure a Dickey-Fuller test is used in table 7 (see appendix). The dependent variable was stationary at a 99% confidence level, so it is not necessary to correct the variable to get a constant. Since the model contains monthly data it is possible to check for seasonality. Thus, an AR(12) is tested in table 9 (see appendix) and a partial correlation regression made in figure 4 (see appendix). Both these models show that only the first three and the fifth lag have a significant impact on the current number of deals. The fourth lag was not significant in the autoregressive model, but was significant in the figure. There is no economic significance that the fifth lag has impact on the dependent variable and the fourth does not. So only the first three lags were added in further models. After adding the lags the adjusted R² immediately increased to over 48%. This occurred because takeovers are clustered over time and are influenced by the contemporary circumstances of the economy and industry. These factors are a good indicator of the number of deals in the coming period because these factors are relatively constant. Therefore, the first month had the most positive and significant impact on our dependent variable, followed by the variables for the number of deals three and two months ago. These findings are in line with conclusions of previous literature on merger waves. The price of U.S. long corporate bonds was the variable used to measure the price of debt in the model. In line with the suggestions of previous literature, an increase in the price of a corporate bond (and thus a lower interest rate) means significantly higher M&A activity. While this variable becomes statistical insignificant in model VI and VII, is it kept because of his economic significance. The S&P 500-variable that measures the growth in the stock market was included in the fourth model. As expected growth is positive related to the number of deals, but insignificant in model IV and V. Nevertheless, in the last 2 models the beta increases and becomes significant after adding other control variables. 23

24 Like Benzing (1991), this study did not observe any statistical significance for the unemployment rate for the entire sample period. Model VI even implies that the unemployment rate is positively related to the number of deals. Furthermore, the variable was not significant, it was decided that it should be excluded from the final model. Before model VII is discussed, closer inspection of the extra added variable that measures the level of globalization is necessary. As expected, this variable had a positive and significant relationship with the dependent variable in the sixth and seventh models. The final model explains 51.0% of the M&A activity according to the adjusted R², since the dependent variable is a stationary series this could be considered as quite high. Nevertheless, by comparing model I and II it becomes clear that most of the explanatory power comes from adding the lags. This is due to clustering effect of M&A activity. As described above only the variable for corporate bonds becomes insignificant in the final model. Table 8 (see appendix) does not show any high correlations between the independent variables, so multicollinearity is not a problem. Between the models the betas for our explanatory variable differ. The impact and significance of the most important variable, the oil price on the number of deals, decreases after adding the control variables. This was in line with expectations, because impact of the oil price was already incorporated in the lags added to the model. As a result, the beta for oil price decreased. The t-value of the coefficient on oil price gives an answer to the hypothesis. In model VII this t-value implies that with a 95% confidence interval that this variable had a positive impact. Thus a higher oil price means more M&A activity in the oil and gas industry. According to this regression, an increase in the price of a barrel of WTI crude oil by $36 will ceteris paribus increase the number of monthly deals by one. This means the oil price has to increase by almost his mean of the last three decades, but is not impossible since it has been $140 in the past. Concluding, although the impact is not large, oil price does have an impact on the M&A activity in the oil and gas industry. 24

25 H2: Increasing oil price volatility has a negative impact on M&A activity in the U.S. oil and gas industry In table 3 the same models were used to test this hypothesis, except the oil price variable was replaced by the oil price volatility variable. The direction of the control variables are the same as in the regressions made for H1. The significance of all the control variables stayed constant or even improved, except the variable for the S&P 500-index, that became insignificant in all of the models. Model I II III IV V VI VII Oil price volatility * * * (-2.03) (-1.85) (-1.76) (-1.58) (-1.74) (-2.45) (-2.46) Deals t *** 0.306*** 0.306*** 0.302*** 0.246*** 0.260*** (6.35) (5.36) (5.35) (5.36) (4.23) (4.64) Deals t *** 0.173** 0.172** 0.169** * (3.49) (2.91) (2.91) (2.86) (1.92) (2.18) Deals t *** 0.222*** 0.222*** 0.217*** 0.155** 0.170** (4.77) (4.11) (4.10) (4.05) (2.80) (3.17) Corporate bonds 0.103** 0.101** 0.108*** 0.065* 0.092** (3.19) (3.07) (3.38) (2.04) (2.84) S&P 500-index (0.32) (0.20) (1.50) (0.81) Unemployment rate * (-0.82) (2.03) Globalization 0.784*** 0.502*** (4.15) (3.89) Constant *** 4.295*** *** *** (21.90) (4.94) (-1.64) (-1.60) (-1.18) (-4.18) (-4.37) Adjusted R Number of Observations * p<0.05, ** p<0.01, *** p<0.001 Table 3. Different models present the number of deals each month in the U.S. oil and gas industry between 1984 and The variable for previous number of deals, corporate bonds, S&P 500-index and globalization are positive related to the dependent variable. The one for unemployment rate is dropped in our last model, because of economic and statistical significance. Almost every model shows a negative significant beta for the relation between oil price volatility and M&A activity. The explanatory variable in this model was volatility of the oil price. According to the adjusted R² of the first model, the variable explains a very small effect in the change of the number of deals in the U.S. oil and gas industry. As expected the variable for oil price volatility is negative related towards number of deals and immediately becomes statistical significant in model I. As in the models presented in the first hypothesis, the beta became closer to zero and even insignificant in model II till V, because the previous number of deals variables had already 25