Influence of the interest rate on the volume of mergers & acquisitions in the Eurozone, UK & US in the near-zero interest rate period after the crisis

(1)

1

Influence of the interest rate on the volume

of mergers & acquisitions in the Eurozone,

UK & US in the near-zero interest rate

period after the crisis

Daan Schuurman 10657924

Bsc Economics & Business; Finance & Organization Supervisor: Robin Döttling

Abstract

This research investigates the effect of the near-zero interest rates in the period of 2011-2017 in the Eurozone, UK & US on the aggregate deal value of mergers & acquisitions. Past literature indicated that an increase in the interest rate would have a negative effect on the aggregate deal value. In total, four models have been used to test this hypothesis: two pooled Ordinary Least Squares models, with and without fixed effects and two Fixed Effects models, with and without time-fixed effects. A significant effect is found in the pooled Ordinary Least Squares models but not in the fixed effects models. Thus, there is not enough evidence to infer that the interest rate has an effect on the volume of mergers & acquisitions when the model corrects for time and region fixed effects.

(2)

2

Statement of Originality

This document is written by Student Daan Schuurman who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

1. Introduction

M&A deals are an important driver of economic activity. Golbe and White (1993) were the first to find a pattern in merger activity by using econometric tests using sine waves to prove that merger activity occurs in wave patterns.

According to the Financial Times companies in the EMEA region rather lend money to finance their projects than use their excess cash. This way the cash piles of companies grew from € 700 billion to € 1 trillion in 2015 (Financial Times, 2015). Also, the UK’s 100 largest companies had 40% more cash in 2014 than they had in 2013. This insinuates that companies, rather than investing the excess cash they have, lend money to finance their projects. It has been speculated that this is due to the historically low, near-zero interest rates that we have seen after the crisis. These interest rates would incentivize companies to hoard cash to do M&A transactions when interest rates are raised in the future, in which case target companies with higher values of debt are discounted to a lower price. A consequence of this could be that mergers & acquisitions are postponed during these near-zero interest rate times to a time where interest rates are higher.

There has been some research on the macroeconomic drivers of M&A activity, which will be shown in the literature section. However, most of these papers were made in and focused on different time frames. After the financial crisis, the interest rates have been near zero for an extended period of time to stimulate borrowing and investment. This signals weak confidence in the market and shows that the European Central Bank doesn’t know how to solve the problem of weak demand. Also, these low interest rates sustain investments that would otherwise not be profitable in times of higher interest rates.

All these forces working at the same time pose the interesting question whether the established relationship between the interest rate and the volume of mergers & acquisitions still holds with these post-crisis near-zero interest rates. This paper aims at finding an answer on the question whether interest rates influence the aggregate deal values of mergers & acquisitions in the period of 2011-2017 in the Eurozone, the US & the UK. For simplicity this research only focuses on mergers & acquisitions where the acquirer and the target are in the same region: the Eurozone, the US or the UK. This way cross-border M&A-deals are omitted because there are a lot of other macroeconomic forces determining their volume.

The dependent variable is the aggregate deal value, derived from the Zephyr (Bureau van Dijk, 1990) database, divided by the total GDP of the applicable region. This way the aggregate deal values are scaled to the size of the relevant market. This dependent variable is

(5)

5

regressed in a Pooled Ordinary Least Squares and a panel data regression with fixed effects, both done with and without time-fixed effects. The independent variable in these models is the interest rate: the Federal Funds Rate for the United States, the Refinancing Tender of the ECB for the Eurozone & the Bank Rate of the Bank of England for the United Kingdom. Control variables are industry concentration (measuring whether deals are concentrated in a small number of industries or divided between a lot of industries), GDP % change, HICP index (an indicator of inflation), MSCI index (a stock-market index) and GDP per capita. These control variables are chosen because past literature indicated they may affect the volume of mergers & acquisitions.

In the pooled OLS regressions that don’t control for region fixed-effects the interest rate is significant at the 1% level. However, when the fixed effects models without and with time-fixed effects are used, the interest rate is respectively only significant at the 10% and not significant at all. This means that no significant result of the interest rate on the aggregate deal value has been measured when controlling for time and region fixed effects.

2. Literature review

2.1 Macroeconomic drivers of M&A activity

Nelson (1959) is one of the first to have studied the frequency of M&A. He only focuses on the manufacturing & mining sectors of the economy. This is understandable given the time the research was conducted and the lack of availability of data at that time. In his article Nelson aims to aggregate and augment the small number of previous done studies with time series on merger activity. In his book Nelson finds no evidence for the retardation theory proposed by Watkins (1927). The retardation theory states that producers try to preserve profits while there is fading demand and more competition. This explanation of mergers is still being proposed in our recent era as a driver of merger activity, although the definition of ‘fading demand’ and ‘more competition’ are definitely totally different.

Furthermore Nelson looks at explanations like “the growth of interregional

transportation, and the development of a significant capital market” (Johnson, 1960). Most importantly, in Nelson’s perception, is the development of the capital market. In his book review on Nelson’s Merger Movements in American Industry, 1895-1956 Masten (1960)

(6)

6

states that Nelson finds a better correlation between merger activity and stock prices than between merger activity and the level of industrial production.

A side note to this research is the time when this research is conducted. The first half of the 20th century is most likely not comparable to the economic situation nowadays.

Melicher, Ledolter & D’Antonio (1983) did a time series analysis of aggregate merger activity. They refer to Weston (1961) having done a multiple regression model on the annual changes in merger activity during the interwar period. Weston found merger activity to be significantly related to stock prices but not significantly related to industrial production activity.

Melicher, Ledolter & D’Antonio employed a multiple time series approach to “develop an explanatory model for describing changes in the incidence of mergers relative to changes in macroeconomic variables during the current merger wave”. They propose that merger activity could reflect changes in business activity and changes in the capital markets. This logically makes sense: a better economic outlook and low borrowing costs makes

investing less costly and reversibly a worse economic outlook and high borrowing costs make investing costlier. They argued that their empirical evidence indicated only a weak

relationship between economic conditions and merger activity and a much stronger link between changes in stock & bond prices and merger activity “to the extent that merger negotiations tend to begin about two quarters before consummation, increased merger negotiation activity seems to reflect the expectation of more receptive (and possibly less costly) capital market conditions in the form of higher stock prices and lower interest rates” (Melicher, Ledolter & D’Antonio). This statement is particularly interesting in the light of the recent near-zero interest rate climate, which ensured low borrowing costs for the past 7 years.

Beckenstein (1979) points out that some change in the environment must influence the decision to merge or acquire in the individual sense. Otherwise, he reasons, the merger or acquisition would have occurred earlier. Because of this, this must also be true for the

aggregate merger activity. A side note he places with this statement are that mergers can take quite some time to be completed, in which case the external factors that incentivized the decision of the deal are not present anymore. In accordance to Steiner (1975) Beckenstein concludes that multiple causes would work multiplicatively. One of the factors that

Beckenstein regards as possible influential is the Gross National Product, because a growing economy means more and bigger companies. The reasoning is that the presence of more

(7)

7

companies increases the number of possible, and thus actual, mergers. Next to this, a growing economy also means that the average company is growing, meaning it is in a better position to acquire other companies.

An argument against the growth of the economy stimulating merger activity posed by Beckenstein (1979) is the reasoning that companies, because of their good prospects don’t need to merge or acquire. The theory that firms only resort to mergers & acquisitions when their own growth opportunity has diminished has been proposed before. After this

Beckenstein offers institutional factors that might affect merger activity, with the most convincing one being a change in political activity regarding antitrust activity. While this specific example may not be true for the time-frame applicable to this research, changes in political activity or rules may well have an effect on merger activity.

Shughart & Tollison (1984) discuss what research has been of influence regarding merger activity. They make a distinction between research to “assess the characteristics and motives of one or more of the merger wave episodes” and research on “models to explain the timing of increases in merger activity. Shughart & Tollison find two common problems in the research on merger activity, namely the fact that data on mergers is limited, as was previously discussed and that there hasn’t been enough research on the topic to conclude there is a systematic pattern in merger activity. Subsequently they try to solve the second problem they posed. Shughart & Tollison are not able to reject the hypothesis that merger activity is characterized by white-noise or a first-order autoregressive scheme. Their model doesn’t support the hypothesis that merger activity in a given year is determined by any year beyond the preceding year. This is contrary to popular belief that mergers are clustered in waves. They note that this could be due to their small number of observations in some periods but they also note that their sample data is the same data other researchers used to conclude there is evidence to suggest merger waves exist. Their research focuses only on the existence of merger waves and not on the predictability of merger activity by macroeconomic factors.

Yagil (1996) makes a distinction between horizontal non-conglomerate mergers (NC) & pure conglomerate mergers (PC). He divides the proposed motives for mergers in past literature in two categories: operating and financial where PC have very little operating effects because of the unrelated activities of the two separate firms. Their supposed gains from mergers are only due to more diversification, capital structure changes and less

(8)

8

bankruptcy and tax costs. Because of this distinction he focuses on two macroeconomic variables explaining merger activity: investment level and interest rates. Furthermore he notes that mergers can influence the profitability of a firm and are more beneficial to the

shareholders of the acquiring than the target firm.

Yagil (1996) finds evidence that supports the hypothesis that investment levels and interest rate levels have an influence on the variation of merger activity, in both dollar values and number of deals. The effect of interest rates is bigger than that of investment levels and the effect of interest rates is positive for cash mergers but negative for securities mergers. A higher interest rate means more deals financed with cash and less deals financed with securities.

Choi & Jeon (2011) argued that because mergers frequently occur in groups, so-called waves, there should be macro-economic factors driving them. They looked at the influence of different macroeconomic factors on merger activity. The most important factor on merger frequency they found was real income level while stock-market level and monetary policy most importantly influenced the transaction values of mergers (but also the frequency). Next to this they also found significant values for the bond market and corporate liquidity.

Resende (2008) used a Markov-switching approach to verify whether economic variables that are frequently mentioned in empirical literature do in fact drive merger activity. The conclusion is that real output growth, real growth in money supply and real stock market returns do indeed relate to merger activity in the UK, but only in the low M&A regime.

Concluding, the last 60 years multiple researchers tried to figure out what

macroeconomic factors drive mergers & acquisitions. The most common factors are the stock market behaviour, the interest rate, inflation levels, Gross National Product, investment levels, real output growth and real growth in money supply. However, most of these factors were researched in different economic times ranging from 1950 to 2008. One could imagine that these factors were not constant in this time-frame and their effects may have been bigger or smaller at some moment in time. Furthermore, the near-zero interest rates after the crisis could influence some of these factors and change their effect.

(9)

9 2.2 Near-zero interest rates

Interest rates are an important way for central banks to influence the economy. With it, the central banks are able to control markets. However, the use of interest rates as a

measure to do this is limited to a lower bound: zero. Central banks can’t lower the nominal interest rates below zero because that would mean they would have to pay banks to lend money from them. This would mean central banks can’t influence aggregate demand but that is not totally true as Gürkaynak, Sack, and Swanson (2004) show: asset prices are determined by the current federal funds rate target but also by the communicated policy for the future. Because of this, central banks can have influence while still keeping interest rates on the zero bound by stating what they will do in the near and long future.

About the effects of near-zero interest rates there is excessive research present. The most important and applicable one is probably the paper written by Christiano, Eichenbaum & Rebelo (2011). They try to determine what the government spending multiplier is when there are consistent near-zero interest rates. Based on past models in almost all economic

trajectories they conclude there is little evidence to infer that the multiplier can be

substantially above one. These models are obviously based on higher interest rates that were persistent in the 19th century and the authors think the new near-zero interest rates may give a different outcome. Their economic reasoning behind this is as follows:

Consider now the effect of an increase in government spending when the zero bound is strictly binding. This increase leads to a rise in output, marginal cost and expected inflation. With the nominal interest rate stuck at zero, the rise in expected inflation drives down the real interest rate, which drives up private spending. This rise in spending leads to a further rise in output, marginal cost, and expected inflation and a further decline in the real interest rate. The net result is a large rise in output and a large fall in the rate of deflation. In effect, the increase in government consumption counteracts the deflationary spiral associated with the zero-bound state (Christiano, Eichenbaum & Rebelo, 2011, p.80).

The model the authors constructed supports the hypothesis that the government-spending multiplier can be a very large sustained value when interest rates bind to zero.

(10)

10

Swanson & Williams (2014) show that although the short-term interest rates were constrained to zero by the Fed, the United States Federal Bank, that doesn’t mean the long-term interest rates are constrained. According to the expectations theory, long-long-term interest rates are determined by short-term interest rate by the term structure: long-term rates are an average of the expected future short term rates of interest. This means that forward guidance and asset purchases and the expectations about those two in the future influence longer-term interest rates while the short-term interest rates are bound to zero.

Chen (2017) states that near-zero interest rate policy has a significant expansionary effect but the government multiplier is smaller than previous research indicated. However, he finds a new approach to modeling near-zero interest rate policy, showing that persistence of the decision to keep the interest rate near-zero boosts economic growth.

Eggertsson & Woodford state that open-market operations don’t have an effect when they don’t change the perception of policy changes in the future. This would make a ‘liquidity trap’ possible: a situation in which people expect deflation or insufficient demand and

therefore rather hold their cash. This liquidity trap can however be averted as long the long-term proposed policy is credible and the central bank is delong-termined to follow it.

Concluding, it is not yet clear to what extent near-zero interest rates influence government policy. It seems that not only the current interest-rate is of influence but also the future policy and the present communication of that policy. This seems intuitively: markets try to shape expectations about future circumstances and a clear policy to which a central bank (credibly) commits makes that possible. Furthermore, there are researchers that state that government spending can undo the near-zero interest rate effects because of a large

government multiplier. However, these numbers are heavily based on assumptions and modeling of the future and not on empirical evidence of the past. Thereby, government spending is not unlimited and can become ineffective as well, meaning there are no big instruments left to influence the market sentiment.

2.3 Interest rates and capital structure

Interest rates influences capital choices and this way, some research suggests, influences investment itself. Several theories try to provide a framework about how and why

(11)

11

capital choices are made. Myers (1984) called this “The Capital Structure Puzzle”. The two most important theories are the trade-off theory and the pecking order theory.

The trade-off theory states that the choice between debt and equity is made by trading-off benefits and costs. More debt, for example, increases bankruptcy costs but also increases tax benefits. Bankruptcy costs increase over time while tax benefits, at least most of the time, get less effective with higher values. Thus, there exists an optimal ratio.

The pecking order theory is first proposed by Myers & Majluf (1984) and goes against the idea that there is an optimal ratio. Firms prefer internal financing and will only resort to external financing when internal financing is not sufficient. Accordingly, firms also prefer debt over issuing new equity.

Concluding, interest rates are important in trade-off theory but not so much in the pecking order theory. In that theory firms prefer internal financing over debt regardless of the interest rate. In the pecking order theory we can state that interest rates will influence

investment choices because the optimal ratio is partly determined by the interest rate in the present and the expected interest rate in the future. In the trade-off theory, firms converge towards their optimal ratio and there they are indifferent. This means, that interest rates influence investments.

2.4 The effect of interest rates on investments

Berk & Damarzo (2014) explain that “an increase in the interest rate will decrease the investment’s Net Present Value. All else equal, higher interest rates will therefore tend to shrink the set of positive Net Present Value investments available to firms.” Furthermore, they state that the Federal Bank in the United States and other banks use this relationship between interest rates and investments to steer the economy in a certain direction. Because of this generally accepted principle, it is expected that an increase in the interest rate will lower the aggregate deal value. In other words: that there is a negative effect of interest rates on the aggregate deal value.

2.5 Addition to the literature

This study aims to broaden the research on the effect of interest rates on aggregate deal value. The effect of interest rate is measured while controlling for concentration of deals within industries, the GDP % change, the inflation level, stock market level and the GDP per capita. The effect of the interest rate in this period hasn’t been researched before and is

(12)

12

interesting because the near-zero interest rates are a relatively new economic phenomenon. The effect of near-zero interest rates on the economy as a whole is still under heavy debate and thus the discussion whether near-zero interest rates have the same effect on the aggregate deal value of mergers & acquisitions as in the past is relevant.

3. Research question and hypothesis

3.1 Research question

Based on the earlier conducted research on the macro-economic factors influencing merger activity in both number of deals and aggregate deal value, the main research question is as follows:

What is the influence of interest rates on merger activity in the Eurozone, US & UK in the near-zero interest rate period of 2011-2017?

To exclude the crisis and the potential abnormal merger activity that accompanies the crisis, 2011 is chosen as the first year.

3.2 Hypothesis

The hypothesis coupled with this research question is the following:

H0: Interest rates don’t influence the volume of mergers & acquisitions in the

Eurozone, US & UK in the period 2011-2017

H1: Interest rates influence the volume of mergers & acquisitions in the Eurozone, US

& UK in the period 2011-2017

Based on previous research the expected effect is negative, meaning an increase in the interest rate would have a negative effect on the aggregate deal value of mergers & acquisitions.

(13)

13

4. Methodology and data description

4.1 Panel data regression

To answer the research question and test the hypothesis a panel data regression is performed. Panel data analysis is chosen to be able to test between time & regions. The timeframe we focused on is, as earlier mentioned, 2011 till 2017. These years are divided into quarters, giving us 28 quarters from 2011 Q1 till 2017 Q4. The regions chosen are the

Eurozone, United Kingdom & the United States. These regions were chosen because all three faced near-zero interest rates for the first time in this timeframe.

Hsiao (2003) shows that cross-sectional studies and time-series risk not controlling for the heterogeneity of countries and therefore risk giving biased estimators. Panel data analysis controls for this heterogeneity. This way we are able to test whether the effect of interest rates is region- and/or time-invariant in the three regions and 28 quarters that are analyzed.

Figure I is a graph of the output of Zephyr (Bureau van Dijk, 1990) and shows the aggregate deal values in millions of dollars in the three regions. The United States has quite large values in proportion with the Eurozone and the United Kingdom. Furthermore we observe peaks and falls that seem to coalign between the three regions indicating there may be a common value driving them. Also, the drop in value of the United States aggregate deal value near the end of 2017 is very striking. At first, this seemed like a wrong value, maybe because a part of the deals made in the end of 2017 hadn’t been processed in Zephyr yet. However, further research found other sources supporting this decline, for example figure II from the imaa-institute (“M&A in the United States”, 2018)

(14)

14 4.2 Regression models

The models used to research the hypothesis are the following:

Pooled OLS model without time-fixed effects 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑑𝑑𝐴𝐴𝐴𝐴𝑑𝑑 𝑣𝑣𝐴𝐴𝑑𝑑𝑣𝑣𝐴𝐴𝑣𝑣 𝑖𝑖𝑡𝑡

= 𝛽𝛽0 + 𝛽𝛽1(𝐼𝐼𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑣𝑣𝐴𝐴 𝑅𝑅𝐴𝐴𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡 + 𝛽𝛽2(𝐻𝐻𝐴𝐴𝐴𝐴𝐻𝐻𝐻𝐻𝐼𝐼𝑑𝑑𝐴𝐴ℎ𝑑𝑑 − 𝐻𝐻𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡

+ 𝛽𝛽3(𝐺𝐺𝐺𝐺𝐺𝐺 % 𝑐𝑐ℎ𝐴𝐴𝐼𝐼𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡 + 𝛽𝛽4(𝐻𝐻𝐼𝐼𝐻𝐻𝐺𝐺 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡+ 𝛽𝛽5(𝑀𝑀𝑀𝑀𝐻𝐻𝐼𝐼 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖) 𝑖𝑖𝑡𝑡

+ 𝛽𝛽6(𝐺𝐺𝐺𝐺𝐺𝐺 𝑝𝑝𝐴𝐴𝐴𝐴 𝑐𝑐𝐴𝐴𝑝𝑝𝐻𝐻𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡 + 𝜀𝜀𝑖𝑖𝑡𝑡

Where𝐻𝐻 indexes region, 𝐴𝐴 indexes time & 𝜀𝜀_{𝑖𝑖𝑡𝑡} = the error term

Pooled OLS model with time-fixed effects 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑑𝑑𝐴𝐴𝐴𝐴𝑑𝑑 𝑣𝑣𝐴𝐴𝑑𝑑𝑣𝑣𝐴𝐴𝑣𝑣 𝑖𝑖𝑡𝑡

= 𝛽𝛽0 + 𝛽𝛽1(𝐼𝐼𝐼𝐼𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝑣𝑣𝐴𝐴 𝑅𝑅𝐴𝐴𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡+ 𝛽𝛽2(𝐻𝐻𝐴𝐴𝐴𝐴𝐻𝐻𝐻𝐻𝐼𝐼𝑑𝑑𝐴𝐴ℎ𝑑𝑑 − 𝐻𝐻𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡

+ 𝛽𝛽3(𝐺𝐺𝐺𝐺𝐺𝐺 % 𝑐𝑐ℎ𝐴𝐴𝐼𝐼𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡+ 𝛽𝛽4(𝐻𝐻𝐼𝐼𝐻𝐻𝐺𝐺 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡+ 𝛽𝛽5(𝑀𝑀𝑀𝑀𝐻𝐻𝐼𝐼 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡

+ 𝛽𝛽6(𝐺𝐺𝐺𝐺𝐺𝐺 𝑝𝑝𝐴𝐴𝐴𝐴 𝑐𝑐𝐴𝐴𝑝𝑝𝐻𝐻𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡 + 𝜇𝜇𝑡𝑡+𝜀𝜀𝑖𝑖𝑡𝑡

Where𝐻𝐻 indexes region, 𝐴𝐴 indexes time,𝜇𝜇_𝑡𝑡 = time fixed effect & 𝜀𝜀_{𝑖𝑖𝑡𝑡} = the error term

Fixed effects model without time-fixed effects 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑑𝑑𝐴𝐴𝐴𝐴𝑑𝑑 𝑣𝑣𝐴𝐴𝑑𝑑𝑣𝑣𝐴𝐴𝑣𝑣 𝑖𝑖𝑡𝑡

+ 𝛽𝛽3(𝐺𝐺𝐺𝐺𝐺𝐺 % 𝑐𝑐ℎ𝐴𝐴𝐼𝐼𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡 + 𝛽𝛽4(𝐻𝐻𝐼𝐼𝐻𝐻𝐺𝐺 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡 + 𝛽𝛽5(𝑀𝑀𝑀𝑀𝐻𝐻𝐼𝐼 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡

+ 𝛽𝛽6(𝐺𝐺𝐺𝐺𝐺𝐺 𝑝𝑝𝐴𝐴𝐴𝐴 𝑐𝑐𝐴𝐴𝑝𝑝𝐻𝐻𝐴𝐴𝐴𝐴) 𝑖𝑖𝑡𝑡+ 𝜃𝜃𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑡𝑡

Where𝐻𝐻 indexes region, 𝐴𝐴 indexes time, 𝜃𝜃_𝑖𝑖 = region fixed effect & 𝜀𝜀_{𝑖𝑖𝑡𝑡} = the error term

Fixed effects model with time-fixed effects 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑑𝑑𝐴𝐴𝐴𝐴𝑑𝑑 𝑣𝑣𝐴𝐴𝑑𝑑𝑣𝑣𝐴𝐴𝑣𝑣 𝑖𝑖𝑡𝑡

+ 𝛽𝛽3(𝐺𝐺𝐺𝐺𝐺𝐺 % 𝑐𝑐ℎ𝐴𝐴𝐼𝐼𝐴𝐴𝐴𝐴)𝑖𝑖𝑡𝑡 + 𝛽𝛽4(𝐻𝐻𝐼𝐼𝐻𝐻𝐺𝐺 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡 + 𝛽𝛽5(𝑀𝑀𝑀𝑀𝐻𝐻𝐼𝐼 𝐼𝐼𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖)𝑖𝑖𝑡𝑡

+ 𝛽𝛽6(𝐺𝐺𝐺𝐺𝐺𝐺 𝑝𝑝𝐴𝐴𝐴𝐴 𝑐𝑐𝐴𝐴𝑝𝑝𝐻𝐻𝐴𝐴𝐴𝐴) 𝑖𝑖𝑡𝑡+ 𝜃𝜃𝑖𝑖 + 𝜇𝜇𝑡𝑡 + 𝜀𝜀𝑖𝑖𝑡𝑡

Where𝐻𝐻 indexes region, 𝐴𝐴 indexes time, 𝜇𝜇_𝑡𝑡 = time-fixed effect, 𝜃𝜃_𝑖𝑖 = region fixed effect & 𝜀𝜀_{𝑖𝑖𝑡𝑡} = the error term

(15)

15 4.3 Variables

Table I summarizes the variables, their source and a description. The dependent variable is the aggregate deal value.

Table I

Variable Source Description

Aggregate deal value Zephyr In millions of US $, per quarter Interest rate ECB, FRED & Bank

of England

Monthly effective Federal Funds rate in % Herfindahl index Zephyr Number of deals per industry per quarter

divided by the total number of deals, squared

GDP % change OECD Growth rate compared to previous quarter,

seasonally adjusted

HICP index FRED Harmonized Index of Consumer Prices, 2015

= 100, monthly

MSCI index MSCI Equity index tracking largest companies from

the Eurozone, US & UK in US $

GDP per capita World Bank GDP divided by the number of people living in the region/country. Measured in US $ As shown by Stevens (1984), a regression is sensitive to outliers and influence

points. Because of this, of all variables that are used the distribution are checked by a

combination of a Shapiro-Wilk test, a boxplot and a histogram. When needed the variable are changed by either a logarithmic transformation or winsorizing.

4.3.1 Dependent variable

The dependent variable is the aggregate deal value. The data is extracted from Zephyr (Bureau van Dijk,1990) . These values are clustered quarterly and are based on the mergers & acquisitions in the Eurozone, United Kingdom & United States. The timeframe that is applied is on and after 01/01/2011 and up to and including 31/12/2017. The percentage and stake before the merger or acquisition can be up to 49% and after the deal the minimum percentage stake is 50%. The acquirer is listed. This gave 2555 deals in de Eurozone of which 871 deals with known deal values and with an aggregate deal value of $ 445,887 million. In the United Kingdom there were 1716 deals, 1277 with known deal values and with an aggregate deal value of $ 206,996 million. The United States had 7266 deals of which 3519

(16)

16

had a known deal value. The aggregate deal value is $ 2,714.155 million. This sums the total deals to 11537 deals of which 5667 had known deal values and with a total aggregate deal value of $ 3,367,038 million.

Table II shows that the aggregate deal value’s mean is $ 40,238.55 million with deals ranging from $ 451 to $ 231,535 million and quite a high standard deviation in proportion to the mean. When the deal value is divided by the GDP the mean has become quite small. The standard deviation is still quite big in comparison with the mean. This variable showed right skewness which indicated it may benefit from a log transformation. After this, the distribution is indeed more normally distributed. The mean has become negative and the standard deviation has become smaller in comparison to the mean. 4.3.2 Main independent variable

The interest rates chosen are the Federal Funds Rate for the United States, the

Refinancing Tender of the ECB for the Eurozone & the Bank Rate of the Bank of England for the United Kingdom. The Federal Funds Rate is the rate at which banks in the United States are able to obtain small overnight credits. The Main Refinancing Tender of the Eurozone is the lowest rate at which funds are tendered by the ECB. The Bank Rate of the United Kingdom is the rate at which the Bank of England offers to buy stocks from banks with the obligation to sell them back at a certain time. These rates are seen as important rates because the height, direction and changes are quickly followed by other institutions and banks. This way the Central Banks of the Eurozone, United Kingdom & the United States are able to influence the market interest rates. Table II shows that the interest rate has a mean of 0.384 with values ranging from 0 to 1.467.

4.3.3 Control variables

The Herfindahl index, GDP % change & GDP per capita, the Harmonized Index of Consumer Prices (HICP) and the Morgan Stanley Capital International Index (MSCI) are all control variables to control for respectively the industry concentration, Gross Domestic Product, the inflation and the stock market.

Gort and Klepper (1982) and Klepper and Graddy (1990) show that after some time evolving, most products experience so-called “shakeouts” in which the number of producers fall; in extreme cases more than 90%. This would mean that in some industries the aggregate

(17)

17

deal value is expected to be much bigger than in others in some period in time. The timing of these shakeouts has been linked to technological change (Klepper, 2002). Adjacent to this, it can be logically derived that shakeouts may lead to higher equity premiums and thus a higher aggregate deal value. To control for this, a variation of the Herfindahl index is constructed and used. This is done by calculating the number of deals per industry, dividing this by the total number of deals and squaring the result. The total Herfindahl value found per quarter is the sum of these values. In a formula:

𝐻𝐻𝐴𝐴𝐴𝐴𝐻𝐻𝐻𝐻𝐼𝐼𝑑𝑑𝐴𝐴ℎ𝑑𝑑 − 𝐻𝐻𝐼𝐼𝑑𝑑𝐴𝐴𝑖𝑖 = ∑ �𝑛𝑛𝑖𝑖

𝑛𝑛� 2

where 𝐼𝐼_𝑖𝑖: number of deals of industry 𝐻𝐻, and 𝐼𝐼 = ∑𝐼𝐼_𝑖𝑖

This variable measures how much a change of the concentration of deals within industries influences the aggregate deal value.

The HICP index is a standardized indicator of inflation and price stability which makes it possible to compare the inflation and price stability between countries and regions. The MSCI indexes are benchmark indexes that try to replicate the performance of the stock market of the whole world or separate regions or countries.

The GDP per capita for 2017 isn’t available in the World Bank database and thus the values of 2016 are used in 2017 as well. The Herfindahl index and GDP % change are stated in percentages. The HICP and MSCI Index are stated as a number where 100 is the base year for the HICP (2015) and the start year of the MSCI Indexes. The GDP per capita is a dollar value per person.

The variables Herfindahl index, HICP index, MSCI index & GDP per capita were logarithmic transformed because of right skewness. After this the MSCI and HICP Indexes showed some outliers, and to mitigate this the MSCI index is winsorized by 0% & 92% and the HICP index by 9% and 98% thresholds.

Table II is a summary of the data statistics, showing the number of observations, means, standard deviations and minimum and maximum values. As can be seen, the aggregate deal value divided by the total GDP gives very small values. The Herfindahl index has a mean of 0.363 with values from 0.132 and 0.917. The log transformed value of this variable shows a negative mean and negative minimum and maximum values. The GDP% change shows a positive mean which indicates the country’s GDP’s grew in the total period which in turn

(18)

18

indicates a growth of the total economic performance. The HICP index is winsorized to reduce the outliers present, which lowered the standard deviation with 0.356. The MSCI index is log transformed and winsorized. The GDP per capita is only log transformed.

Table II

Summary statistics N Mean Std. Dev. Min Max Aggregate deal value

Aggregate deal value/GDP

84 84 40,238.55 0.003 51,035 0.004 451 0.0002 231535 0.026 lnAggregate deal value/GDP 84 -6.324 1.209 -8.551 -3.64

Interest rates 84 0.384 0.342 0 1.467 Herfindahl 84 0.363 0.193 0.132 0.917 lnHerfindahl 84 -1.143 0.505 -2.03 -0.087 GDP % change 84 0.442 0.369 -0.421 1.278 HICP 84 99.29 2.396 91.833 104.567 HICP winsorized 84 99.43 2.040 95.8 103.7 MSCI index 84 1,316.79 399.84 791.07 2,505.76 lnMSCI index winsorized 84 7.144 0.266 6.673 7.634 GDP per capita in $ 84 44,755.44 7,577.7 34,353.43 57,638.16 lnGDP per capita in $ 84 10.70 0.166 10.44 10.96

Number of regions 3 3 3 3 3

Tables III, IV & V show summary statistics by region. The first thing that stands out is the large aggregate deal value in the United States in comparison with the Eurozone and the United Kingdom. When divided by the GDP, the value of the United States is still more than 4 times as large as the Eurozone. Furthermore the interest rates in the Eurozone and the United Kingdom are close together and the interest rate of the United States is somewhat lower. The Herfindahl index in the United States is significantly lower than that of the

Eurozone and the United Kingdom, indicating that deals are spread over more industries. This can be explained by the fact that the United States has a lot more deals, expecting a more distributed field. GDP % change and the HICP index are more or less the same among the three regions. The MSCI index, the control variable for the stock market indexes, seems to indicate that the United States stock market performed significantly better than that of the Eurozone and United Kingdom. Lastly, the United States has the highest GDP per capita, followed by respectively the United Kingdom and the Eurozone.

(19)

19 Table III

Sum. Statistics Eurozone N Mean Std. Dev. Min Max Aggregate deal value 28 16388.82 17806.16 2485 77550 Aggregate deal value/GDP 28 .0013189 .0014334 .0002026 .0061821 lnAggregate deal value/GDP 28 -7.069439 .942798 -8.504324 -5.086093 Interest rate 28 .4154844 .4764717 0 1.467391 Herfindahl index 28 .396786 .1863698 .139888 .8114202 lnHerfindahl index 28 -1.033185 .4811836 -1.966913 -.2089693 GDP % change 28 .2997268 .3745586 -.421023 .828669 HICP index 28 99.36643 1.873694 94.52667 102.41 HICP index winsorized 28 99.4119 1.764128 95.8 102.41

MSCI 28 1004.076 112.9512 95.8 1198.85

lnMSCI index winsorized 28 6.905545 0.1150149 6.673392 7.089118 GDP per capita 28 37363.42 2425.299 34353.43 40614.91 lnGDP per capita 28 10.52641 .065002 10.44446 10.61189

Table V

Sum. Statistics United States N Mean Std. Dev. Min Max Aggregate deal value 28 96934.11 49641.44 8342 231535 Aggregate deal value/GDP 28 .005962 .0029806 .0004768 .0138232 lnAggregate deal value/GDP 28 -5.278114 .6535045 -7.648518 -4.281407 Interest rate 28 .2810714 .3257913 .0733333 1.203333 Herfindahl index 28 .2124741 .0669132 .1318853 .4301109 lnHerfindahl index 28 -1.585476 .259829 -2.025822 -.8437121 GDP % change 28 .5197644 .3998173 -.386237 1.278177 HICP index 28 99.64464 1.964304 94.45 102.6667 HICP index winsorized 28 99.69286 1.845025 95.8 102.6667

MSCI 28 1757.428 385.4251 1160.997 2505.761

lnMSCI index winsorized 28 7.429255 .2031387 7.057034 7.634807 GDP per capita 28 54354.29 2928.307 49790.67 57638.16 lnGDP per capita 28 10.90186 .0543359 10.81558 10.96194

Table IV

Sum. Statistics United Kingdom N Mean Std. Dev. Min Max Aggregate deal value 28 7392.714 13546.08 451 65410 Aggregate deal value/GDP 28 .0030175 .0055089 .0001934 .0262775 lnAggregate deal value/GDP 28 -6.624481 1.170852 -8.550692 -3.639041

Interest rate 28 .4553929 .091791 .25 .5

Herfindahl index 28 .4782448 .1890988 .1937389 .9168109 lnHerfindahl index 28 -.81124 .3924037 -1.641244 -.0868541 GDP % change 28 .5059157 .2960446 -.145046 1.151033 HICP index 28 98.85 3.153031 91.83333 104.5667 HICP index winsorized 28 99.18571 2.478839 95.8 103.7 MSCI index 28 1188.873 102.0197 1042.179 1411.938 lnMSCI index winsorized 28 7.07729 .0843998 6.949069 7.252719 GDP per capita 28 42548.61 2178.922 40412.03 46783.47 lnGDP per capita 28 10.65717 .0502003 10.60688 10.75329

(20)

20 4.4 Correlation matrix

The correlation matrix shows that aggregate deal value and Total GDP are positively correlated. This supports Beckenstein’s (1979) theory: a higher GDP means more (striving) companies and thus more (possible) deals.

The interest rate is negatively correlated with the aggregated deal value, which was expected. The log transformed Herfindahl index is the only other variable that negatively correlates with the aggregate deal value. This means that when the Herfindahl index rises, the aggregate deal value falls. This is contrary to the theory proposed before: when deals were more concentrated in certain industries, the aggregate deal value was expected to rise because of the equity premium that shakouts cause.

As argued in past literature the GDP% change, inflation and GDP per capita all have a positive effect on the aggregate deal value. Interestingly, the MSCI index is negatively

correlated with the aggregate deal value while a positive correlation was expected

Another interesting correlation is the one between the GDP per capita and the MSCI index, the stock market proxy. This correlation was also expected to be positive but is

negative. The log transformed Herfindahl variable negatively correlates with all other control variables. This insinuates that positive changes in inflation, stock market levels or GDP means deals are divided between more different industries. The last notable correlation value is the low correlation between GDP % change and the GDP per capita. It was expected that a change in either would correlate a lot with the other. However, a possible explanation is that a high GDP per capita means it is more difficult to have a large percentage growth rate.

Table VI Correlation matrix Aggr. deal value Total GDP Deal value/Tot GDP lnDeal value/Tot GDP Interest rate Industry concentr. GDP % Change HICP Index MSCI Index GDP per capita Aggregate deal value 1 Total GDP 0.62 1 Deal value/Total GDP 0.71 0.19 1 lnDeal value/Total GDP 0.77 0.33 0.83 1 Interest rate -0.31 -0.18 -0.19 -0.30 1 lnHerfindahl index -0.48 0.56 -0.09 -0.25 0.14 1 GDP % Change 0.13 -0.03 0.08 0.14 -0.31 -0.03 1 HICP Index 0.14 0.18 0.99 0.15 -0.25 -0.07 0.25 1 MSCI Index -0.08 0.05 -0.08 -0.20 -0.12 0.04 0.07 0.15 1 GDP per capita 0.58 0.45 0.48 0.62 0.01 -0.51 0.20 0.11 -0.14 1

(21)

21

5. Results & analysis

5.1 Variance Inflation Factors for Multicollinearity

By doing a pooled Ordinary Least Squares (OLS) regression on the logarithm of the aggregate deal value divided by the GDP using the independent variables interest rates, Herfindahl index, GDP % change, HICP index, MSCI index & GDP per capita and doing a variance inflation factor estimation the variables are checked for multicollinearity. This is done by estimating the variance inflation factors (VIF). It measures how much variance (the standard deviation squared) of the estimators is increased because of correlation with other estimators. The VIF then shows how much the standard errors are inflated due to this (multi)collinearity. Table VII shows that all the values are under 10, which is the most common threshold chosen in past literature, for example in Menard (2001).

Table VII

Variance Inflation Factors VIF 1/VIF Interest rate lnHerfindahl index 1.25 1.48 0.8007 0.6752 GDP % change 1.31 0.7638

HICP index winsorized 1.48 0.6765 lnMSCI index winsorized 8.52 0.1173 lnGDP per capita 8.52 0.1174

Mean VIF 3.76

However, when the quarterly time dummies are added, the VIF values get quite high for some variables, as can be seen in Table VIII. The threshold of 10 is violated by the HICP index, MSCI index and the GDP per capita. However, these values are used as control variables. The problem of multicollinearity is that it makes the coefficients unstable and the standard errors big for the variables with collinearity. However, there is no change in the effect of the variable that is tested in this research (the interest rate) and the control variables are still able to perform as controls.

(22)

22 Table VIII

Variance Inflation Factors VIF 1/VIF

Interest rate 1.53 0.655681 lnHerfindahl index 2.17 0.461500 GDP % change 1.82 0.549778 HICP winsorized 22.52 0.044406 lnMSCIC winsorized 18.31 0.054613 lnGDP per capita 17.78 0.056231 2011 Q1 2.05 0.488645 2011 Q2 1.96 0.509179 2011 Q3 1.99 0.501742 2011 Q4 2.08 0.481350 2012 Q1 2.41 0.414290 2012 Q2 2.47 0.405315 2012 Q3 2.91 0.344050 2012 Q4 3.37 0.296682 2013 Q1 4.37 0.228620 2013 Q2 4.79 0.208596 2013 Q3 5.22 0.191653 2013 Q4 5.66 0.176751 2014 Q1 7.63 0.131022 2014 Q2 7.47 0.133865 2014 Q3 6.29 0.158952 2014 Q4 5.53 0.180702 2015 Q1 7.49 0.133591 2015 Q2 6.89 0.145227 2015 Q3 6.53 0.153165 2015 Q4 5.23 0.191133 2016 Q1 7.01 0.142634 2016 Q2 7.56 0.132347 2016 Q3 8.13 0.123030 2016 Q4 9.20 0.108720 2017 Q1 12.28 0.081437 2017 Q2 13.44 0.074391 2017 Q3 14.50 0.068982 Mean VIF 6.93

5.3 Random vs fixed effects

Because of the possibility of unobserved, individual specific, time-invariant factors influencing the outcome and that correlate with the explanatory variables, the use of fixed effects is more applicable in this research. Stock & Watson (2003) state that in the fixed effects model the changes in the dependent variable must be due to influences other than fixed characteristics between the entity. Those fixed characteristics include for example cultural and economic differences between regions.

(23)

23

However, the Hausman test to test whether random or fixed effects is more

applicable resulted in not rejecting the null hypothesis for both the models, indicating that the random effects models should be used. Still, the fixed effects models are used because

theoretically those seem more applicable and to be sure there isn’t any over specification. The results of the Hausman tests and the regression output when random effects are used can be found in tables X and XI in the appendix.

5.4 Regression output

In Table IX the regression results are presented. To control for heteroscedasticity all regressions are done with robust standard errors. The R² of the pooled OLS without time-fixed effects is 51,4%, meaning 51,4% of the variation in the aggregate deal value is explained by the model. The Pooled OLS model with time-fixed effects has a higher R² of 78%. The FE regressions without time-fixed effects has the lowest R² of 20.6% while the FE with time-fixed effects performs quite good with 66.1%. In the fixed effects model without time-fixed effects 16.63% of the variance is due to differences across panels. For the fixed effects model with time-fixed effects that is 44.38% so correcting for time makes the intraclass correlation much bigger.

The main explanatory variable is the interest rate which is significant at the 1% level in both the pooled OLS models. It is only significant in the FE regression without time-fixed effects and there only at the 10% level. The coefficient of the variable interest rate is negative in all models, meaning an increase in the interest rate would mean a decrease in the number of mergers & acquisitions. This is in line with the expected hypothesis: interest rates have a negative effect on merger & acquisitions.

The Herfindahl index is significant at the 10% level in the model with time-fixed effects indicating that correcting for time increases the significance of the Herfindahl index. The interpretation is that when deals are more clustered in less industries, the aggregate deal value rises.

The HICP index is significant at the 10% level in one model: the pooled OLS with time-fixed effects model. The coefficient of the variable is positive, meaning an increase of the inflation index (and thus an increase of the inflation) means an increase in the aggregate deal value.

(24)

24

Lastly the only other variable, besides the quarterly-dummies, that is significant is the GDP per capita. This variable is significant at the 1% level in both the Pooled OLS models and at the 5% level of the FE model with time-fixed effects. The sign of the

coefficient is positive in all models, meaning an increase in the GDP increases the aggregate deal value. This is in line with our expectation. All other variables, besides the time dummies, don’t show any significance.

Several time dummies are significant in models 2 and 4. This indicates that time has an influence on the aggregate deal value. This could be due to the fact that mergers &

acquisitions occur in waves, like Choi & Jeon (2011) found.

Table IX (1) (2) (3) (4)

Regression output Pooled OLS Pooled OLS with time-fixed effects FE regression FE regression with time-fixed effects Interest rate -1.332*** -1.033*** -1.521* -0.654 (0.330) (0.323) (0.386) (0.254) lnHerfindahl index 0.431 0.670* 0.529 0.883* (0.293) (0.337) (0.435) (0.238) GDP % change -0.365 -0.264 -0.310 -0.220 (0.318) (0.223) (0.304) (0.266) HICP index winsorized 0.0305 0.315* 0.0249 0.170 (0.0630) (0.175) (0.0862) (0.147) lnMSCI index winsorized -0.492 -1.019 -0.776 -2.010* (1.048) (1.377) (1.233) (0.636) lnGDP per capita 6.042*** 6.803*** 8.294 5.710** (1.732) (2.011) (4.417) (1.241) 2011 Q1 -0.304 -0.131 (0.871) (1.207) 2011 Q2 -0.765** -0.750 (0.379) (0.478) 2011 Q3 -0.821** -0.804 (0.337) (0.347) 2011 Q4 -1.147*** -1.049 (0.427) (0.722) 2012 Q1 -1.341*** -1.177 (0.459) (0.774) 2012 Q2 -1.498*** -1.202 (0.527) (0.933) 2012 Q3 -0.699 -0.309 (1.057) (0.754) 2012 Q4 -1.947*** -1.517** (0.565) (0.343)

(25)

25 2013 Q1 -1.372** -0.730 (0.632) (0.885) 2013 Q2 -1.182 -0.428 (0.845) (1.416) 2013 Q3 -1.857** -1.010 (0.800) (0.715) 2013 Q4 -2.639*** -1.805 (0.822) (0.883) 2014 Q1 -1.090 0.0236 (0.989) (1.336) 2014 Q2 -1.883* -0.724 (1.037) (1.779) 2014 Q3 -0.843 0.279 (0.854) (1.222) 2014 Q4 -0.454 0.383 (0.784) (0.690) 2015 Q1 -0.562 0.427 (1.283) (0.611) 2015 Q2 -0.845 0.116 (0.998) (1.504) 2015 Q3 -0.936 -0.0495 (0.879) (1.017) 2015 Q4 -1.977** -1.251 (0.749) (0.843) 2016 Q1 -1.441 -0.486 (1.015) (1.426) 2016 Q2 -2.251** -1.311 (0.971) (1.311) 2016 Q3 -1.651 -0.637 (1.013) (1.347) 2016 Q4 -1.724 -0.638 (1.100) (0.912) 2017 Q1 -1.961 -0.841 (1.295) (1.981) 2017 Q2 -2.943* -1.767 (1.486) (1.599) 2017 Q3 -3.827** -2.575 (1.453) (1.294) Constant -69.30*** -100.4*** -90.63 -67.90 (15.62) (15.30) (46.29) (28.14) Observations 84 84 84 84 R-squared 0.514 0.780 0.206 0.661 Number of regions

Robust standard errors Yes Yes

3 Yes

3 Yes Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1 Base quarter: 2017 Q4

(26)

26

6. Conclusion

In this paper there is an attempt to figure out what the effect of interest rates on M&A volume is in the period 2011-2017. The. In this period, companies have issued more debt and hoarded cash. Because of past literature the expected relationship was a negative influence of interest rates on the aggregate deal value. This is tested by multiple models: the pooled OLS with and without time-fixed effects and the fixed effects model with and without time-fixed effects.

In the pooled OLS models the variable of interest, the interest rate, is significant at the 1% level. However, in the fixed effects model that only corrects for region fixed effects, this interest rate is only significant at the 10% level. In the fixed effects model that corrects for region and time fixed effects the interest rate is not significant anymore. This indicates that a lot of the explanatory power that the interest rate has in the first two models is because of the fact that region fixed effects are not excluded. Because of this, we can’t conclude that the interest rate has an effect on the aggregate deal value. However, near-zero interest rates haven’t been around for longer than 7 years in the Eurozone, UK & US which makes it difficult to find a significant relationship because of the lack of available data.

One limitation that should be noted is the lack of a variable on regulation changes. After the crisis there have been extensive changes in regulation regarding investment and the financing of investments and it seems logically acceptable that those are of influence on the aggregate deal value. However, because of the relatively new near-zero interest rates of which nobody really knows all the macroeconomic effects, more research on the topic is warranted.

(27)

27

References

Baltagi, B. H. (2008). Econometric analysis of panel data. Chichester, UK: John Wiley & Sons.

Bank of England (2018). Statistical Interactive Database – official Bank Rate history. Retrieved from https://www.bankofengland.co.uk/boeapps/iadb/Repo.asp, June 4, 2018.

Beckenstein, A. R. (1979). Merger activity and merger theories: An empirical investigation. Antitrust Bulletin, 24(2), 105-128.

Berk, J. B., & DeMarzo, P. M. (2014). Corporate finance : the core. Third edition. Boston, MA: Pearson.

Bureau van Dijk Electronic Publishing. (1990). Zephyr. Amsterdam: Bureau van Dijk Electronic Pub. Retrieved from https://zephyr.bvdinfo.com/version

2018619/home.serv?product=zephyrneo&loginfromcontext=ipaddress, June 1, 2018. Board of Governors of the Federal Reserve System (US), Effective Federal Funds Rate [FEDFUNDS], retrieved from FRED, Federal Reserve Bank of St. Louis;

https://fred.stlouisfed.org/series/FEDFUNDS, June 2, 2018.

Choi, S.H. & Jeon B.M. (2010) The impact of the macroeconomic environment on merger activity: evidence from US time-series data, Applied Financial

Economics, 21(4), 233-249, doi: 10.1080/09603107.2010.528365

Christiano, L., Eichenbaum, M., & Rebelo, S. (2011). When Is the Government Spending Multiplier Large? Journal of Political Economy, 119(1), 78-121.

doi:10.1086/659312

Eggertsson G. B. & Woodford, M. (2003). Optimal Monetary Policy in a Liquidity Trap, NBER Working Papers 9968, National Bureau of Economic Research, Inc.

Eurostat, Harmonized Index of Consumer Prices: All Items for Euro area (19 countries) [CP0000EZ19M086NEST], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/CP0000EZ19M086NEST, June 3, 2018.

Eurostat, Harmonized Index of Consumer Prices: All Items for United States [CP0000USM086NEST], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/CP0000USM086NEST, June 3, 2018

Eurostat, Harmonized Index of Consumer Prices: All Items for United Kingdom [CP0000GBM086NEST], retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/CP0000GBM086NEST, June 3, 2018)

(28)

28

Golbe, D., & White, L. (1993). Catch a Wave: The Time Series Behavior of Mergers. The Review of Economics and Statistics,75(3), 493-499. doi:10.2307/2109463

Gordon, S. (2015, July 7). European companies stash more cash. The Financial Times. Retrieved from https://www.ft.com, May 20, 2018.

Gort, M., Klepper, S., 1982. Time paths in the diffusion of product innovations. Economic Journal 92 (3), 630 – 653.

Greene W.H. (2000). Econometric analysis. Upper Saddle River, N.J: Prentice Hall. Gurkaynak, R. S. and Sack, B. P. and Swanson, E. T. (2004). Do Actions Speak Louder than Words? The Response of Asset Prices to Monetary Policy Actions and Statements. FEDS Working Paper No. 2004-66. http://dx.doi.org/10.2139/ssrn.633281

Heath, D., Jarrow, R., & Morton, A. (1992). Bond Pricing and the Term Structure of Interest Rates: A New Methodology for Contingent Claims Valuation. Econometrica, 60(1), 77-105. doi:10.2307/2951677

Hsiao, C. (2003). Analysis of Panel Data, Vol. 34 of Econometric Society monographs. Cambridge University Press, Cambridge, 2nd ed.

Johnson, A. (1960). Merger Movements in American Industry, 1895–1956. By Ralph L. Nelson. National Bureau of Economic Research, Number 66, General Series, Princeton, 1959. Pp. xxi 177. $5.00. Business History Review, 34(2), 256-258. doi:10.2307/3111556

Klepper, S. (2002). Firm Survival and the Evolution of Oligopoly. The RAND Journal of Economics, 33(1), 37-61.

Klepper, S., Graddy, E., 1990. The evolution of new industries and the determinants of market structure. RAND Journal of Economics, 21 (1), 27 – 44.

Masten, J. (1960). Southern Economic Journal, 27(1), 69-69. doi:10.2307/1055798 Melicher, R., Ledolter, J., & D'Antonio, L. (1983). A Time Series Analysis of Aggregate Merger Activity. The Review of Economics and Statistics, 65(3), 423-430. doi:10.2307/1924187

Menard, W. S. (2001). Applied Logistic Regression Analysis. Quantitative Applications in the Social Sciences. 07-106.

Myers, S.C., Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have, Journal of Financial Economics, Volume 13, Issue 2, Pages 187-221, ISSN 0304-405X, https://doi.org/10.1016/0304-405X(84)90023-0.

MSCI. End of day index data search. Retrieved from https://www.msci.com/end-of-day-data-search, June 1, 2018).

(29)

29

M&A in the United States. (2018, June 22), Retrieved from https://imaa-institute.org/m-and-a-us-united-states/#m-and-a-waves, May 15, 2018.

OECD (2008), “GDP percentage change”, OECD National Accounts Database (Accessed on June 3, 2018).

OECD (2008), “Historical GDP – Expenditure Approach”, OECD National Accounts Database (Accessed on June 3, 2018).

Resende, M. (2008). Mergers and acquisitions waves in the UK: a Markov-switching approach, Applied Financial Economics, 18:13, 1067-1074, DOI:

10.1080/09603100701408155

Shughart, W., & Tollison, R. (1984). The Random Character of Merger Activity. The RAND Journal of Economics, 15(4), 500-509.

Statistical Data Warehouse (2006). ECB/Eurosystem policy and exchange rates. Official interest rates. Retrieved from http://sdw.ecb.europa.eu/browse.do?node=9691107 (Accesed on June 3, 2018).

Stock, J. H., & Watson, M. W. (2007). Introduction to econometrics. Boston: Pearson/Addison Wesley.

Swanson, E.T. & Williams, J.C. (2014). Measuring the Effect of the Zero Lower Bound on Medium- and Longer-Term Interest Rates. American Economic Review, 104 (10): 3154-85.DOI: 10.1257/aer.104.10.3154

Watkins, M. Industrial Combinations and Public Policy, Cambridge, 1927, Woodford, Michael. 2011. Simple Analytics of the Government Expenditure Multiplier. American Economic Journal: Macroeconomics, 3 (1): 1-35.

Smith, A. (2014, October 6). UK’s largest companies sit on £53bn mountain of cash. The Financial Times. Retrieved from https://www.ft.com

Stevens, J. P. (1984). Outliers and influential data points in regression analysis. Psychological Bulletin, 95(2), 334-344.

Weston, J. Fred, The Role of Mergers in the Growth of Large Firms (Berkeley and Los Angeles: University of Cali- fornia Press, 1961), chapter 5.

Worldbank. GDP per capita (current US$). Retrieved from https://data.worldbank.org/indicator/NY.GDP.PCAP.CD, June 8, 2018.

Yagil, J. Mergers and macro-economic factors, Review of Financial Economics, Volume 5, Issue 2, 1996, 181-190,

(30)

30

Appendix

A.1. Hausman tests

To decide between fixed or random effects the Hausman specification test was performed. The Hausman specification test tests for misspecification by comparing the estimators when there is no misspecification with the estimators when there is

misspecification. When there is no misspecification the estimators should be close to one another.

This is done by running both a fixed and a random effects regression and saving the estimates. The null hypothesis is that random effects are preferred over fixed effects. If so, the unique errors are not correlated with the regressors. The tests on the model with and the model without time-fixed effects both failed to reject the null hypothesis, which indicates the use of the random effects model.

Hausman output without time-fixed effect:

Table X (b) (B) (b-B) sqrt(diag(V_b-V_B))

Hausman test fixed random Difference S.E.

Interest rate -1.521265 -1.332061 -.1892038 .2618101

lnHerfindahl index .5286851 .4311893 .0974958 .1155932

GDP % change -.309985 -.3648192 .0548342 .0716567

HICP index winsorized .0248738 .0304874 -.0056136 .0066069

lnMSCI index winsorized -.7764944 -.4918466 -.2846477 .2245468

lnGDP per capita 8.293715 6.042198 2.251517 2.367287

b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic

chi2(6) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 1.75

(31)

31 Hausman output with time-fixed effect:

Table XI Hausman test

(b) (B) (b-b) sqrt(diag(V_b-V_B))

fixed Random Difference S.E.

Interest rate -.6542984 -1.033254 .378956 .4203776

lnHerfindahl index .883109 .6703193 .2127897 .1120747

GDP % change -.2204487 -.2640576 .043609 .

HICP index winsorized .1702516 .3145882 -.1443365 .0783644 lnMSCI index winsorized -2.01043 -1.019289 -.9911411 .4404333

lnGDP per capita 5.709847 6.802882 -1.093035 2.693516 2011 Q1 -.1313445 -.3042287 .1728842 . 2011 Q2 -.7499326 -.7648071 .0148746 . 2011 Q3 -.8044629 -.8209548 .0164919 . 2011 Q4 -1.048524 -1.146803 .098279 . 2012 Q1 -1.176802 -1.341213 .1644106 . 2012 Q2 -1.201924 -1.497718 .2957946 .0967919 2012 Q3 -.3085455 -.6989599 .3904144 .1579743 2012 Q4 -1.517147 -1.946898 .4297514 .2671251 2013 Q1 -.7297906 -1.371952 .6421614 .3843995 2013 Q2 -.4277992 -1.18199 .7541903 .4439969 2013 Q3 -1.009957 -1.856829 .8468712 .4950325 2013 Q4 -1.805355 -2.639421 .8340664 .6291302 2014 Q1 .0236183 -1.090026 1.113644 .7458937 2014 Q2 -.7239405 -1.883493 1.159553 .7650343 2014 Q3 .2793591 -.843466 1.122825 .7365405 2014 Q4 .3828144 -.4542549 .8370693 .5081205 2015 Q1 .4274164 -.5617894 .9892058 .592448 2015 Q2 .1164801 -.8450599 .96154 .5720936 2015 Q3 -.0495318 -.9355849 .8860531 .540086 2015 Q4 -1.251253 -1.977124 .7258706 .4055916 2016 Q1 -.4859166 -1.440649 .9547326 .5206008 2016 Q2 -1.311309 -2.251491 .9401822 .5372302 2016 Q3 -.6372343 -1.650814 1.01358 .5746947 2016 Q4 -.6380007 -1.723707 1.085706 .5902205 2017 Q1 -.8412565 -1.960679 1.119422 .6138713 2017 Q2 -1.766647 -2.94262 1.175973 .6260008 2017 Q3 -2.574593 -3.826672 1.252079 .6447091

b = consistent under Ho and Ha; obtained from xtreg

B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic

chi2(33) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 4.17

Prob>chi2 = 1.0000

(32)

32 A.2. Regression output random effects

Table XII (1) (2)

Regression output RE regression RE regression with time-fixed effects

Interest rate -1.332*** -1.033*** (0.178) (0.148) lnHerfindahl index 0.431 0.670*** (0.377) (0.184) GDP % change -0.365 -0.264 (0.317) (0.273) HICP index winsorized 0.0305 0.315*** (0.0898) (0.0838) lnMSCI index winsorized -0.492 -1.019 (1.044) (0.797) lnGDP per capita 6.042** 6.803*** (2.353) (0.973) 2011 Q1 -0.304 (1.111) 2011 Q2 -0.765* (0.463) 2011 Q3 -0.821** (0.378) 2011 Q4 -1.147** (0.559) 2012 Q1 -1.341*** (0.488) 2012 Q2 -1.498** (0.609) 2012 Q3 -0.699 (1.115) 2012 Q4 -1.947*** (0.126) 2013 Q1 -1.372*** (0.358) 2013 Q2 -1.182 (0.867) 2013 Q3 -1.857*** (0.462) 2013 Q4 -2.639*** (0.414) 2014 Q1 -1.090 (0.681) 2014 Q2 -1.883* (1.043) 2014 Q3 -0.843 (0.513)

(33)

33 2014 Q4 -0.454** (0.213) 2015 Q1 -0.562 (0.374) 2015 Q2 -0.845 (0.975) 2015 Q3 -0.936* (0.548) 2015 Q4 -1.977*** (0.419) 2016 Q1 -1.441* (0.831) 2016 Q2 -2.251*** (0.743) 2016 Q3 -1.651** (0.719) 2016 Q4 -1.724*** (0.339) 2017 Q1 -1.961 (1.367) 2017 Q2 -2.943** (1.251) 2017 Q3 -3.827*** (0.595) Constant -69.30*** -100.4*** (26.15) (7.115) Observations 84 84 Number of regions Robust standard errors

3 Yes

3 Yes Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1 Base quarter: 2017 Q4

Influence of the interest rate on the volume of mergers & acquisitions in the Eurozone, UK & US in the near-zero interest rate period after the crisis