Mergers, acquisitions and rival firms: The market power and efficiency view.
*Name: Arnout Kuipers (S2405539) Study Program: MSc Economics
Other programs for which you have submitted this thesis: MSc Finance Course code: EBM000A20
Supervisor: Dr. M. Zaouras 23-6-2017
Abstract: This paper examines the effect of a merger or acquisition on the performance of rival banks using the asset complementarity between banks involved in the deal and the change in their market power. I assess whether prices of rival banks are affected by the merger or takeover and if the change in prices is in accordance with one of the two main views in the existing literature; the market power view and the efficiency view. Under the former view I expect a positive relationship between the Lerner index and rival performance whereas under the latter view I expect a negative relationship between asset complementarity and rival performance. Even though the market power of the banks involved in the deal increases, this paper finds that there is a negative relationship between the Lerner index and rival output prices. However, there is also a negative relationship between the asset complementarity index and rival output prices. This paper therefore finds evidence in support of the efficiency view and in contradiction with the market power view. JEL codes: D430, G210, G250, G340, L110
Field Key Words: Asset Complementarity, Market Power, Rival Performance, Cost efficiency.
* I would like to thank Dr. M. Zaouras for his continuous critical and supporting input and advice that helped me
2
1. Introduction
The inter-connectedness of the current financial system became apparent in the recent financial crisis. Households failed to pay their mortgages resulting in the bankruptcy of multiple banks amongst which Lehman Brothers is a well-known example. These bankruptcies triggered a series of defaults outside the U.S. as well. Financing packages had to be provided by governments to save banks that were deemed “too-big-to-fail”. Other banks failed because they were not large enough or had limited diversification in their asset portfolios and were therefore subject to high
unsystematic risk1.
When determining the appropriate size of banks, there exists a trade-off between banks becoming too large and banks being too small to survive. On the one hand banks need to grow and diversify to limit the unsystematic risk that they are exposed to. Besides, the relative compliance costs that a bank faces decline as it increases in size. On the other hand banks should not be too large that financial difficulties at such a bank will cause problems all across the financial system. One might worry that through mergers and acquisitions banks will become “too-big-to-fail” and exploit their market power. However, these institutions can also engage in mergers and acquisitions to diversify their asset portfolio and to achieve synergy gains. To determine how banks are connected to their financial environment, one can examine how they affect their competitors when they grow. When banks become too large they might take advantage of their market power and sustain a high level of interest rates. However, if they grow to diversify, they might undercut their competitors, resulting in short-term lower profits for these competitors and a motive for them to get more diversified as well.
Banks can have different motives to use a merger or an acquisition as a mechanism to achieve growth. By combining assets with another bank, they can achieve synergy benefits and produce more efficiently. On the other hand, banks might want to increase their market power by combining its bank with a competitor. Having more market power could allow them to increase their prices and sustain higher profits. There exists an extensive amount of literature on the motivations for a merger or acquisition, the gains of a merger or acquisition for the target bank and the bidding bank,
1 For more information about this topics see
3
and the method of payment in such transactions. However, there is less evidence on how competitors are affected by the deal.
Two theories stand out in the literature: the market power view and the efficiency view (Eckbo, 1983). The market power view argues that when two firms merge, the number of firms in a market decreases, causing the competitive pressure in the market to decline, which allows the remaining firms to increase their mark-up and profits. Thus, under the market power view the rivals benefit from the acquisition or merger of two firms and as a result their returns increase (Shahrur, 2005).
The efficiency view however predicts that the rivals will at least not benefit from a merger or
acquisition in the short-run, causing a decrease in returns. The line of reasoning behind this argument is that a merger or acquisition occurs because of synergies or complementarities that result in lower costs for the combined entity, creating a competitive advantage for this firm. The new firm can lower its price due to the lower costs, putting pressure on its rivals to follow and decrease their mark-up. As a result the gains of the rivals will fall. However, competitors might try to lower their costs as well, limiting the effect that the merger or acquisition has on the rival’s returns in the long-run.
Previous research looks into the changes in the returns of rival banks and concludes from these changes whether the merger or acquisition can be categorized under the market power or efficiency view. I however question this methodology. I first test whether a decrease in returns is indeed the result of efficiency gains and subsequently argue that a decrease in returns is in line with the efficiency view. In a similar fashion I test whether an increase in market power results in higher returns for the competitors and thereafter conclude that an increase in returns is in line with the market power view. In sum, this methodology is different from the methodology commonly used in the existing literature because I test the relationship between the rival’s returns and the efficiency and market power view, which is commonly assumed in the existing literature.
4
a good measure to see whether synergy gains are realizable post-merger. Second, I examine the relationships between the market power and asset complementarity on the one hand and rivals’ output prices on the other hand. The results from this analysis will give an indication whether either of the two views is supported by this data. The research questions I try to answer in this thesis are as follows:
Can banks realize synergy gains or do they experience an increase in market power as a result of a merger or acquisition?
And
Do the changes in the rivals’ output prices, either as a result of asset complementarity or a change in market power, give supportive evidence of the market power or efficiency view?
I use of a sample of 136 mergers and acquisitions between commercial banks located in the U.S. during the period 2012 to 2016, to research whether rivals gain or lose as a result of a deal between two banks.
5
cost decreases on to its price. Therefore, as mentioned, I rely on the concept of asset complementarity to assess whether synergy gains can potentially be realized from the merger or acquisition. Finally, to test whether the rival banks are positively or negatively affected by the deal, I regress output prices of these rival banks on the complementarity metric, the difference in the Lerner index and several control variables. I test the significance of the coefficient for the complementarity index and the change in Lerner index and whether they have their expected sign. This paper finds that there are significant positive changes in the Lerner index for the banks involved in the deal as well as for some of the rival banks. These changes appear to be especially present in the years 2014 and 2015. This indicates that there are positive effects on the market power of the banks as are a result of the merger or acquisition. However, the changes in the Lerner index do not appear to positively affect the output prices of the rivals. On the contrary, the coefficients are significantly negative and decrease post-merger, contradicting the market power view. The asset complementarity index has a significantly negative coefficient which is in line with the expectation from the efficiency view.
The remainder of this paper is as follows. Section 2 will discuss the existing literature, my research question and hypotheses, Section 3 entails the data and methodology. In Section 4 I discuss the results followed by a conclusion in Section 5.
2. Literature
This paper focuses on the returns that accrue to competitors of the banks that are involved in the M&A deal. As argued, there are two different views regarding the effects of a merger or acquisition on the returns of their competitors: the efficiency view and market power view. A brief discussion about each view and their respective effects on the returns of the rival firms will be provided in the following sections.
The efficiency view states that rival firms most likely see their returns decline after a merger or
6
a possible cause for such synergies is that the firm can create a more stable revenue stream when merged with a firm that has a complementary asset portfolio.
Mixed evidence is found on the ability of the newly formed firms to actually create these efficiency gains. Rhoades (1993) and Ghosh (2001) do not find evidence that the firms have significantly higher performance or realized synergy gains compared to their peers. Other papers find positive results with respect to an increase in synergetic gains and performance (Healy et al., 1992; Rhoades, 1998; Houston et al. 2001; Cornett et al., 2006). Devos et al. (2009) for example find that there are synergy gains realizable of around 10.03% of the total equity value of the combined firms, of which 80% is attributable to synergies in operations and the remaining 20% due to tax benefits.
The market power view argues that rival firms are expected to gain from a merger if it increases
7
To be able to distinguish between motives, Shahrur (2005) and Fee and Thomas (2004) also calculate the returns for customers and suppliers. They find that in the majority of deals they benefit as well, which is in line with the efficiency view. These results contradict the market power hypothesis since an increase in concentration in the market as a result of the merger caused negative returns for the merging firms and their rivals. The market power hypothesis, on the contrary, states that these returns should be positive. In addition Eckbo (1983) researches the impact on the returns of competitors in the case the merger or acquisition is contested by the government for anticompetitive reasons. His findings show that after the government has contested the merger, the returns of the rivals are still significantly positive. He argues that the returns of the rival firms were expected to decrease since the complaint of the government might result in the merger being restricted or cancelled. Since he finds positive returns, this evidence is another contradiction of the market power hypothesis.
Nonetheless, a significant amount of literature finds that mergers and acquisitions raise the returns of competitors (e.g. Chatterjee, 1986; Farrell and Shapiro, 1990) by increasing the price in the markets. Another explanation for this can be the so-called “purchase probability hypothesis” posed by Song and Walkling (1999). This hypothesis entails that rival firms in a market where a merger or acquisition occurs have an increased probability to be taken over themselves. As mentioned, target firms earn significant abnormal returns upon announcement since it is most likely that the bidder firm will overpay. Hence this increased probability of being taken over is beneficial for the shareholders’ returns. The authors even find that in case a takeover or merger was unsuccessful, the rivals still experience positive abnormal returns. Since my analysis uses the output prices of rival banks instead of stock returns, the purchase probability hypothesis can be ruled out as an explanation for my findings.
8
a proxy for asset complementarity. This asset complementarity index is constructed by making a column vector of the loans in a certain sector as a % of total loans. This column vector looks as follows: Mi = ( 𝑅𝑒𝑎𝑙 𝐸𝑠𝑡𝑎𝑡𝑒 𝐿𝑜𝑎𝑛𝑠 𝑎𝑠 𝑎 % 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠 𝐹𝑎𝑟𝑚 𝐿𝑜𝑎𝑛𝑠 𝑎𝑠 𝑎 % 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠 𝐶𝑜𝑚𝑚𝑒𝑟𝑐𝑖𝑎𝑙 𝑎𝑛𝑑 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 𝑙𝑜𝑎𝑛𝑠 𝑎𝑠 𝑎 % 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠 𝐼𝑛𝑑𝑖𝑣𝑖𝑑𝑢𝑎𝑙 𝐿𝑜𝑎𝑛𝑠 𝑎𝑠 𝑎 % 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠 𝑂𝑡ℎ𝑒𝑟 𝑙𝑜𝑎𝑛𝑠 𝑎𝑠 𝑎 % 𝑜𝑓 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠 ) (1)
Consequently, the MD is calculated as follows:
𝐴𝑠𝑠𝑒𝑡 𝑐𝑜𝑚𝑝𝑙𝑒𝑚𝑒𝑛𝑡𝑎𝑟𝑖𝑡𝑦 = √ (𝑀𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟 − 𝑀𝑡𝑎𝑟𝑔𝑒𝑡)
𝑇
𝑊−1(𝑀
𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟− 𝑀𝑡𝑎𝑟𝑔𝑒𝑡)
, (2)
Where 𝑊−1 is the inverse pooled covariance matrix of the two loan portfolios. Having a high
degree of asset complementarity between two banks can indicate that there are significant synergy gains realizable through diversification. Therefore, the asset complementarity metric can give a good indication of the possible synergy gains that can arise as a result of a merger or acquisition between banks. This thesis uses a similar but slightly simpler variant of the Mahalanobis Distance, namely the Euclidean Distance (ED), which is a commonly used approach to measure organizational differences (see e.g. Berry, Guillén and Zhou, 2010). I use the ED since my dataset
did not yield a proper inverse pooled covariance matrix2. The Euclidean Distance measure follows
a similar computation, but instead of using formula 2, I use the following equation:
𝐴𝑠𝑠𝑒𝑡 𝑐𝑜𝑚𝑝𝑙𝑒𝑚𝑒𝑛𝑡𝑎𝑟𝑖𝑡𝑦 = √(𝑀𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟 − 𝑀𝑡𝑎𝑟𝑔𝑒𝑡)
𝑇
(𝑀𝑎𝑐𝑞𝑢𝑖𝑟𝑒𝑟 − 𝑀𝑡𝑎𝑟𝑔𝑒𝑡) (3) The existing literature finds that the higher the asset complementarity between two firms, the better the post-transaction performance of the combined firm (Kim and Finkelstein, 2009; Hoberg and Philips, 2010; Makri et al. 2010). The success of the post-merger performance partly depends on the ability of the new bank to benefit from the asset complementarities. Supporting this view is the finding by Kim and Finkelstein (2009) that firms that have a broader strategic focus are better able
2 The variance-covariance matrix included very small numbers, making its inverse very large and therefore the
9
to implement different strategies and use different resources. Firm with a too narrow strategic focus do not have the capabilities to adapt to new and different product strategies.
The same results are found by Hoberg and Philips (2010). What distinguishes this study from other studies regarding asset complementarity and competition, is the methodology used to identify complementary firms. Instead of using SIC codes to match firms in the similar industry, the authors use 10-K product descriptions to match firms. The main benefit of this methodology according to the authors is that: “it can jointly capture firm similarity relative to other firms within and across markets”. Their main finding is that the higher the degree of complementarity between firms, the better the after merger performance. Besides, they find that firms that have a large amount of patents are more likely to become a takeover target. This is due to the fact that patents indicate that the firm has several unique technologies that become available to the acquirer when the firm is taken over. An additional finding of this paper is the fact that the less competitive a market is, the more benefits the combined firm can retain from the product complementarities. Moreover, the more different are the merging firms from its competition, the higher will be the returns to the combined firm after the merger. This paper uses a similar method to match banks with their competitors as Hoberg and Philips (2010). However, I use the location of a bank instead of 10-K product descriptions to determine the rival banks.
Hypotheses and Research Question
Most literature, including the papers mentioned above, use the change in the returns of the rivals post-merger to observe the motive of the merger or acquisition. In this paper I firstly examine the potential motive of the deal by using the concept of asset complementarity and the Lerner index, where the former might indicate synergy gains post-merger and the latter show changes in market power of the banks involved. Consequently I research whether this motive is correctly incorporated into the returns of the rivals. So the research questions I try to answer in this thesis are:
Can banks realize synergy gains or do they experience an increase in market power as a result of a merger or acquisition?
And
10
To answer these research questions, I test the effects on rival performance as a result of a merger or acquisition. This paper uses the methodology from the existing literature on the Lerner index and asset complementarity to make a distinction between two views, the market power and efficiency view. Whereas the relationship between rival performance and the two views is commonly assumed in the existing literature, this paper explicitly tests the relationships. From these tests I subsequently conclude which view is most supported in this paper.
As mentioned, the Lerner index measures market power and therefore is used to assess the market power view, whereas asset complementarity is used to test the efficiency view. According to the market power view, I expect to find a positive relationship between the prices of rival banks and the change in the Lerner index due to the merger or acquisition. This change in the Lerner index is based on the difference between its pre-and-post-merger values. This view relies on the theory that
in a market with an 𝑛 amount of banks, there will be 𝑛 − 1 banks as a result of the acquisition or
merger. Due to this decrease in the amount of banks active in the market, the market power of each remaining bank will increase. Being able to take advantage of this increase in market power, banks will increase their prices (e.g. Perry and Porter, 1985). This line of reasoning results in my first hypothesis:
H1: An increase in the Lerner index of the merged banks implies a higher level of market power for the merged banks.
If the first hypothesis holds, the higher market power might induce them to set higher prices. To benefit from this, their competitors will follow by setting higher prices as well, potentially increasing their profits. Consequently, my second hypothesis follows:
H2: An increase in the Lerner index of the merged banks will result in higher prices of the rival banks.
11 H3: A high value of the asset complementarity between the acquiring and target bank will result in a decrease in prices of rival banks.
3. Data & Methodology
From the Zephyr database provided by the Bureau van Dijk, I obtained a sample of commercial banking deals including 210 mergers and acquisitions. These mergers and acquisitions are executed by commercial banks located in the United States between 2012 and 2016. This time period is specifically chosen to reflect the post-crisis effects of a bank merger or acquisition on the interdependence of banks. This is done by looking at the impact of these deals on the performance of rival banks. Besides I have chosen to use yearly data in my analysis. This is done for several reasons. Firstly, when using quarterly data, some of the variables used to calculate the input variables for the Lerner index were not available anymore in the Call Reports. Besides, when using these input variables to estimate the Lerner index on a quarterly basis, the relative number of missing observations for some variables almost doubled, for one variable even from 30% to 60% of the total number of observations. The combination of these two effects resulted in very high Lerner indices, mainly in the range of 0.9 and 1. Having much more credible estimations of the Lerner indices with yearly data, I decided to use these values in my further regression.
13
different sectors, namely: Real Estate, Farm, Commercial and Industry, Individuals and Others. Again, these values are taken at the end of the year before the merger took place and are measured in book values. Looking at Panel B of table 1, one can observe that most of the loans outstanding are in the real estate sector followed by the commercial and industry sector. The farmers sector has the lowest stake in the loan composition of these banks. The sample of rival banks consists in total of 847 banks, with 402 of them being competitors of the acquiring bank and the remaining 445 competitors of the target banks. Table 2 displays the summary statistics of the group of rival banks.
Differences can be observed when comparing the group of target and acquiring competitors, though the differences are very small. The rivals of the acquirer are on average equally sized in terms of assets but have a slightly higher operating income. The difference between expenditures is also slightly in favour of the acquirer’s rivals.
Lerner index:
My analysis consists of two parts. In the first part I construct the Lerner index of the banks before the merger and of the combined bank after the merger and test the significance of the difference in these Lerner indexes. This gives me an indication whether there is a significant change in market power due to the merger or acquisition. The Lerner index is often used as an indicator for market power of firms and banks (e.g. see Maudos and Guevara, 2007; Ariss, 2010). A positive change in this index suggests that a deal is for market power purposes. Higher market power will likely result in a higher price ceteris paribus, increasing the Lerner index. On the other hand, the effect on the
Table 2 Summary Statistics
This table contains the summary statistics of the competitors of the banks in the merger and acquisition data sample. This sample includes 847 competitors from the U.S. commercial banking industry during the period of January 1st,
2012 until and including December 31st , 2016. All values in this table are log-transformed and denominated in U.S.
dollars. Mean 13.09 13.09 10.14 10.13 9.52 9.51 Median 12.88 12.85 9.85 9.89 9.22 9.21 Std.Dev 1.89 1.89 1.76 1.89 1.69 1.69 Minimum 8.01 8.18 7.13 1.10 6.56 5.77 Maximum 21.24 21.17 18.16 18.10 17.43 17.25
Total Assets Operating Income Total Expenditures
Acq Targ Acq Targ Acq Targ
14
Lerner index is ambiguous if a merger is for efficiency purposes since the price can decrease as a result of cost decreases, which are offsetting forces in the numerator of the Lerner index. The change in the Lerner index depends on the ability of the bank to lever the cost decreases onto its price.
The FDIC has a database of call reports for each member bank in the US. These call reports contain all sorts of information about the banks, e.g. their loans outstanding, total assets or wage expenses. From these call reports I obtained data for the Lerner indices and the price data of the competitors. To calculate the Lerner index for the sample of banks, I have used the methodology presented by Spierdijk and Zaouras (2016). At first, the authors construct a total cost function for each bank, both the banks involved in the deal and their competitors, where total costs are regressed on three input price variables, a total asset variable and a trend variable accounting for technological progress.
The input price variables are the price of purchased funds, the core deposit rate, the wage rate and the price of physical capital. The price of purchased funds is obtained by dividing the expenditures on purchased funds by the amount of purchased funds. The core deposit rate is similarly calculated by dividing the expenditure on core deposits by the amount of core deposits. The wage rate is calculated as the expenditure on labour services divided by the number of full-time equivalent employees. The price of physical capital consists of the expenditures on physical capital divided by the amount of physical capital in that bank. The total cost variable is the sum of expenditures on purchased funds, core deposits, personnel expenses and expenditures on physical capital. This variable together with the core deposit rate, wage rate and price of physical capital are normalized with the price of purchased funds to achieve linear homogeneity. Moreover, these variables and the total asset variable are transformed by taking their logarithm. The total cost regression consequently looks as follows:
log (𝐶̃𝑖𝑡) = 𝛼 + ∑ 𝛽𝑗,𝑝log(𝑃̃𝑗,𝑖𝑡) + (1 2) ∑ ∑ 𝛽𝑗𝑘,𝑝𝑝log(𝑃̃𝑗,𝑖𝑡) log(𝑃̃𝑘,𝑖𝑡) + 4 𝑘=2 4 𝑗=2 4 𝑗=2 ∑4 𝛽𝑗,𝑝𝑦log(𝑃̃𝑗,𝑖𝑡) log(𝑌𝑖𝑡) 𝑗=2 + 𝛽𝑦log(𝑌𝑖𝑡) + ( 1 2) 𝛽𝑦𝑦log(𝑌𝑖𝑡) 2+ 𝛽 𝑡𝑖𝑚𝑒,𝑦𝑡𝑙𝑜𝑔(𝑌𝑖𝑡) + 𝛽𝑡𝑖𝑚𝑒2,𝑦𝑡2𝑙𝑜𝑔(𝑌 𝑖𝑡) + 𝛽𝑡𝑖𝑚𝑒𝑡 + 𝛽𝑡𝑖𝑚𝑒2𝑡2+ 𝜀𝑖𝑡 , (4)
where 𝐶 is the total cost variable, 𝑃 indicates the different input price variables and 𝑌 is the total
zero-15
mean error term that is orthogonal to the regressor variables for bank 𝑖 at year 𝑡. Having estimated this total cost function, I can calculate the marginal costs for each bank. The marginal cost function is the first derivative of the total cost function. This function is as follows:
𝑀𝐶𝑖𝑡 = 𝐶𝑖𝑡
𝑌𝑖𝑡
𝛿𝑙𝑜𝑔𝐶𝑖𝑡
𝛿𝑙𝑜𝑔𝑌𝑖𝑡 (5)
This equation can be rewritten into the following to use the estimates of regression 4 to estimate the marginal costs of each bank:
𝑀𝐶𝑖𝑡 = 𝐶𝑖𝑡
𝑌𝑖𝑡 ( 𝛽𝑦+ ∑ 𝛽𝑗,𝑝𝑦log(𝑃̃𝑗,𝑖𝑡) + 𝛽𝑦𝑦log(𝑌𝑖𝑡) + 𝛽𝑡𝑖𝑚𝑒,𝑦𝑡 + 𝛽𝑡𝑖𝑚𝑒2,𝑦𝑡
2 ) 4
𝑗=2 (6)
Using the marginal costs values obtained from equation 6, I can calculate the Lerner index for each bank by using the following formula:
𝐿 =
𝑃−𝑀𝐶𝑃 (7)
Where P, the output price, is proxied by dividing operating income by total assets, i.e. average revenue.
Asset Complementarity
16 Source: FDIC Bank Find
I use the Euclidean distance (ED) to measure the asset complementarity between banks. This is a slightly simpler variant of the Mahalanobis Distance concept presented Kim and Finkelstein (2009). The advantage of the Mahalanobis Distance (MD) compared to the Euclidean distance (MD) measure is that the MD incorporates the fact that the different loans in a bank’s portfolio might be correlated which each other which is not captured by the ED. In my sample of banks, the variance-covariance matrix between the acquiring banks and rival banks includes very small negative numbers. This means that there is hardly any correlation between these two portfolios. Moreover, these small values result in a very large inverse of the variance-covariance matrix. As a result, the formula presented by Kim and Finkelstein (2009) to calculate the MD yields too large
values, i.e. above 1. Hence, I decided to leave out the inverse variance-covariance matrix (W-1 in
formula 2) and instead calculate the Euclidean Distance.
This ED will still give a good estimation of the asset complementarity, however slightly less accurate than the MD. Instead of using equation 2 as has been done in Kim and Finkelstein (2009), I will use the formula 3 presented above to calculate the asset complementarity index. The distribution of the asset complementarity index is given in figure 1 and a histogram of the asset complementarity is displayed in figure 2. One can observe that most distance values lie between 0.0 and 0.6, where the mean value is 0.186 and the median value is slightly lower at 0.151. This means that for quite some deals there is a small or modest level of asset complementarity between the target and acquiring bank.
Figure 1: Distribution of Euclidean Distance
17
Figure 2: Histogram of Euclidean Distance Source: FDIC Bank Find
Some synergies can therefore be expected after the merger or acquisition. Further analysis is needed to conclude whether these values will have a significant impact on the competitors of the acquiring and target banks.
Rival Prices
For my second phase it is important to find data for rival banks. The existing literature usually uses the 4-digit SIC codes of banks to define their competitors in the market (e.g. Shahrur, 2005). Since this paper focusses on commercial banks which all have the same 4-digit SIC code, I look into the location of the banks to define their competitors. Rival banks are selected that have their headquarters or a subsidiary in the same city as the target or acquiring bank. A limitation of this methodology is that potential competitors outside the city where the target or acquirer is located, are not included. On the contrary, the SIC-codes methodology might include banks as competitors which are not actual competitors of each other. A small bank in the east of the U.S. might have the same 4-digit SIC code as a small bank in the west of the U.S. but the banks are probably not competitors due to the large distance between them.
The number of competitors for each bank differs, ranging from zero or just 1 competitor in the same city to more than 10 for a bank in the larger cities. To determine whether the merger or acquisition has any impact on the rival firms, I look at the impact on the prices of the rival banks. The existing literature often uses the abnormal stock returns as an indicator of the impact of a
18
merger or acquisition on the returns of their rivals. Due to the fact that my sample focusses on commercial banks, the number of listed competitors is limited. Therefore, I chose to use the output prices of banks instead of stock returns as a measure for rival performance. This output price is obtained in a similar way as the output price required in the Lerner index.
To get a clear picture of the effect of the Lerner index on the output prices, I have divided the Lerner index into two variables, namely Lerner index before and Lerner index after the merger. The “Lerner index before” has been constructed by taking the value of the Lerner index of an acquiring or target bank for the years before the merger or acquisition happens and setting it zero afterwards. These values are used for the competitor of the acquiring or target bank involved in that deal. For instance a bank that is involved in a deal in 2016, will have data for the years 2011 up to including 2015 and will be zero for 2016. Each competitor for a particular bank will be given the same values of the Lerner index in the years 2011 to 2016 as that bank to clearly see how the value of the “Lerner index before” affected the prices of the competitors. The “Lerner index after” variable follows the same method, however in this case the years before the merger are given a value of zero whereas for the years after the merger the estimated values of the Lerner index are used. As stated, hypothesis 2 concerns the change in the Lerner index. Therefore I will test if the change from the “Lerner index before” to the “Lerner index after” is significant.
I have specifically chosen to take the indices before and after separately to be able to observe their individual effects on the rivals’ prices as well. This will allow me to draw some conclusions from the individual relationships of the “Lerner index before” and “Lerner index after” variables with the rivals’ output prices, in addition to the result obtained from the significance test of the change in the Lerner indices. This would not be possible if I instead would use only a single Lerner index variable that only measures the change after the merger or acquisition compared with its pre-merger value. This distinction might be beneficial by providing more information in case there is no evidence to support the expected positive effect of the change in the Lerner index on output prices. Control Variables
Competitor’s own Marginal Cost: Costs partially determine the price set by a bank. The costs have
19
the rival banks as control variable in the regression. Marginal costs are estimated with the use of formula 6.
Bank Size: Larger and older banks usually have more experience and resources to successfully
complete a merger or acquisition. Moreover, they may be better able to exploit their increase in market power or benefit from their synergy gains compared to smaller banks. This however depends as well on the number of (large) banks in the market. Hence I will also control for the level of competition in the market and the cross-product variable of competition and bank size. The size of each bank has been measured in terms of the total assets of the bank.
Relative Size: Next to the absolute value of bank size, I have also included the relative acquisition
size as a control variable. A merger or acquisition where the acquirer is much larger compared to the target might have less impact on the competitors of the acquirer compared to a merger where the size is relatively equal. Relative size is measured as the ratio of a target bank’s assets to the acquirer bank’s assets.
Competitor’s own Lerner index: In this regression I try to capture the effect of the Lerner indices
of the acquiring and target banks on the prices of their competitors. However the price of the competitors is also determined by the value of its own Lerner index. Therefore I used the lagged value of the competitors’ Lerner index as a control variable in the regression.
0 50 100 150 200 250 300 350 0 5 10 15 20 25 30 1 2 3 4 5 6 7 8 9 10 >10
Ba
nk
s
F
requency
Number of rival banks in the same city as the acquiring bank
20 Market Structure: I use a dummy variable to control for the amount of competitors that a bank has.
Figure 3 displays the frequency of an acquiring bank having a certain amount of rivals in the same city and the total number of rival banks in each group. The frequency is displayed on the left vertical axis in combination with the bar graph. One can see from the bar graph that this distribution is fairly equally distributed among the different acquirers although a large group of these banks have 1 or 2 competitors. The right axis and the line graph display the total number of banks in each category. Looking at the line graph, one can observe that a lot of the rival banks in this sample are located in the group where acquiring banks have more than 10 rivals in the same city. This means that the sample is biased towards acquiring banks which have a lot of rivals. In other words, a few deals with banks that have a lot of rivals in the same city can largely influence the results compared to deals with banks that have just a few competitors in the same city.
Figure 4 displays the same distribution but for the rivals of the target banks. The frequency of rival banks is less equally distributed compared to figure 3. One can observe from the bar graph that most target banks have a low number of competitors in the same city. More striking is the result from the line graph. The total number of competitors in the group of target banks that have 10 or less competitors is very low. The group of target banks that have more than 10 competitors contains the most rival banks. This implies that also in the group of rivals from the target bank the results might be biased to the few deals where the target banks have more than 10 competitors.
Figure 4: Distribution of Rivals Target Bank Source: FDIC Bank Find
0 20 40 60 80 100 120 140 160 0 5 10 15 20 25 30 35 1 2 3 4 5 6 7 8 9 10 >10
Ba
nk
s
F
requency
21
Regression
To get an indication how the merger or acquisition between the two banks affects the performance of their rivals, I regressed the output prices of the rival banks on the Lerner index before and after the merger, the asset complementarity index and several control variables to be able to test my second and third hypothesis. This regression looks as follows:
𝑃𝑟𝑖𝑐𝑒𝑠 = 𝛽1+ 𝛽2 𝐿𝐼𝐵𝑒𝑓𝑜𝑟𝑒𝑖𝑡+ 𝛽3 𝐿𝐼𝐴𝑓𝑡𝑒𝑟𝑖𝑡+ 𝛽4 𝐴𝑠𝑠𝑒𝑡𝐶𝑜𝑚𝑝𝑙𝑒𝑚𝑒𝑛𝑡𝑎𝑟𝑖𝑡𝑦𝑖 +
𝛽5 𝐿𝑜𝑔(𝑀𝑎𝑟𝑔𝑖𝑛𝑎𝑙𝐶𝑜𝑠𝑡)𝑖𝑡 + 𝛽6 𝐿𝑜𝑔(𝑆𝑖𝑧𝑒)𝑖𝑡+ 𝛽7 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑆𝑖𝑧𝑒𝑖𝑡+ 𝛽8 𝐿𝑒𝑟𝑛𝑒𝑟𝑖,𝑡−1+
𝛽9 𝐶𝑖𝑡𝑦𝑖+ 𝛽10 𝑌𝑒𝑎𝑟𝑖 + 𝛼𝑖 + 𝜐𝑖𝑡, (8)
where 𝐶𝑖𝑡𝑦 is a dummy variable equal to 1 if an acquiring or target bank has more than 10
competitors in the same city and 𝑌𝑒𝑎𝑟 is a dummy variable for each year in the data sample except for the base year 2011. I tested equation 8 for autocorrelation following Drukker (2003) and for stationarity by using a Fisher-type unit-root test in combination with the dickey fuller option. The price variable is stationary, however there appears to be an autocorrelation problem. To deal with the autocorrelation, I have estimated the equation with robust standard errors. Moreover, the marginal cost and size variable have been log-transformed since they were in dollars while the other values are mainly ratios. As mentioned, I expect a positive relationship between the output prices and the Lerner index before and after the merger and that their difference is significantly positive. Moreover I expect a negative relationship between the output prices and the asset
complementarity index. This means that I expect that the 𝛽2 and 𝛽3 coefficients will be positive
and 𝛽4 coefficient will be negative.
22
omitted heterogeneity is uncorrelated with the regressors. This should be kept in mind when interpreting my results.
4. Results
A summary of the estimation results of the Lerner index can be found in table 3 for the banks involved in the deal and in table 4 for their competitors. Table 3 displays summary statistics for both the acquiring and target banks before the merger or acquisition, after the merger or acquisition and for the difference between these periods. For the banks involved in the deal, one cannot calculate the Lerner index for the target after the merger or acquisition, since it no longer exists. During the merger or acquisition, the target bank is usually taken over or merged into the acquiring bank. Therefore, the value of the Lerner index of the target bank after the merger is given the same value as the value measured for the acquiring bank after the merger or acquisition. This allows me to make a comparison between the change in Lerner index for the acquiring and target banks. Finally, the difference in each third column is calculated by subtracting the Lerner index value before the merger from the value after the merger.
24
Looking at the estimation results year by year, one can conclude that this increase in market power is most likely in the later years, where the difference in the mean and median Lerner index is higher and positive whereas in 2012 and 2013 for example these values are lower and for the acquirer even negative. Moreover, the mean values in 2014 and 2015 are statistically significant. An explanation that the other years are not significant might be that there are relatively less deals in 2012 and 2013 compared to 2014 and 2015. Even though 2016 had a comparable number of mergers and acquisitions as the preceding years, the after-merger period is shorter compared to the other years. This shorter period might be too small to show a significant increase in the Lerner index. Another reason why the results are not statistically significant or why the changes in the Lerner index are not larger might be due to regulatory issues. Mergers or acquisitions that might limit competition severely are likely to have been prohibited by the regulator, decreasing the likelihood of mergers with large increases in market power.
26
Effects on Prices
Table 5 depicts the results from my final regression, where I have regressed the output prices of the rival firms on the value of the Lerner index (LI) before and after the deal together with the value of my asset complementarity index. In addition the regression includes a set of control variables, including rivals’ own marginal costs, size, relative size, a lagged value of the Lerner index of the rivals themselves, market structure, and year dummies.
The main results are displayed in the column labelled Total. One can observe that the LI Before coefficient is significantly negative related to the output prices of the rivals and the same holds even more so for the LI After coefficient. The asset complementarity index is also significantly negative related to output prices. Finally looking at the control variables, all except the size variable have the expected signs, although relative size and the city dummy are insignificant. Moreover, I observe that in 2013 and 2014 the values are significantly lower compared to my base year of 2011. In sum, from these results I can conclude that my third hypothesis cannot be rejected, since the asset complementarity index has a significantly negative effect on the output prices of the competitors. To confirm or reject the second hypothesis, I tested whether the difference in the effect of the Lerner index before and after the merger is significant with a simple F-test. This test points out that the difference is indeed significant at the 10% level. However, the second hypothesis stated that there should be a positive relationship between the Lerner index change and output prices of rival firms, which is clearly not the case. Hence my second hypothesis is rejected. As seen in the estimation result of the Lerner index in Table 3 and 4, the results and significance might be different for the target and acquiring rivals. Therefore, I split the sample into rivals from acquiring banks and rivals from target banks and I have performed the same regression for both samples. The results from this estimation can be found in the second and third column of Table 5.
28
insignificant as also observed in the acquiring banks’ sample. The asset complementarity index’ coefficient is still negative and significant and slightly larger compared to the acquiring banks’ sample. One can conclude from this analysis that probably the impact of the merger and acquisition is more present in the markets of the target banks than in the markets of the acquiring banks. There is another characteristic of my dataset which might have an impact on the estimation results presented in the column total, namely the number of competitors in each city (As discussed more extensively above). I have included a dummy variable to account for this in my analysis but, this does not provide the full picture. Therefore as a robustness check I have again split my sample. However this time the sample is split based on the criteria that a bank has either more than 10 competitors or it has 10 or less competitors. The results for both samples are shown in the last two columns of Table 5. Looking at the column of the 10 or less competitors group, one can observe that the “Lerner index before” variable has no effect on the output prices, where the “Lerner index after” variable is still negatively related although insignificantly so. Their difference is also not significant. The relationship between the asset complementarity variable and output prices as remained negative and highly significant.
The story is different for the group of banks that have more than 10 competitors; the coefficients for the before and after merger Lerner index are both negative and significant at the 1% level. In contrast, testing for their difference shows however that the “Lerner index after” coefficient is not significantly different from its before merger value in this case. The asset complementarity index is still significantly negatively related to the output prices, however only at the 5% level in this sample. One can thus observe that the effect of market power on rival’s output prices is more visible in the markets where there are more than 10 competitors.
As a final robustness check, I have looked into a different way of calculating the value of the Lerner index before the merger. As I have also done in Table 3, I have calculated the “Lerner index before” as a weighted average of the Lerner indices of the acquiring and target bank, where the weights are based on total assets. Since the value of the Lerner index after the merger is a combination of the two Lerner indices as well, this might give a better comparison of the value of the Lerner index before the merger compared to its after merger value. The results are shown in Table 6.
30
observe from the first column that the coefficients have slightly changed. Their individual significance has remained the same but the difference between the two Lerner indices became insignificant. Looking at the sample split up in target and acquiring banks, the results have not changed much either, except for the coefficients of the Lerner indices for the target banks. The values have become (more) significant, especially the LI before coefficient. The difference between the before and after value remains insignificant. The asset complementarity index is slightly less significant in the sample of the acquiring banks.
Finally the last two columns again depict the results for the sample that is divided by the number of competitors, with a cut-off value of 10. The results look fairly similar to the ones presented in Table 5. Interestingly though is the fact that in the sample with more than 10 competitors the asset complementarity index has become insignificant. As shown by the comparison between the acquiring and target banks, the effect of the asset complementarity is more negative and more significant in the sample of target banks. The acquirers are on average larger and therefore have a larger weight in the joint before Lerner index. It is possible that therefore the effect of the asset complementarity is lower and insignificant compared to the sample where the Lerner index before the merger or acquisition is unweighted.
31
To test the argument of Boone (2008) on my sample, I have looked briefly into the relationship between the rivals’ own Lerner index and those of the acquiring and target banks. This regression does not try to explain the Lerner index of the competitors by regressing them on the Lerner index of the regular banks. I merely want to see the effect of the latter on the former by looking at the sign of the coefficient of the regular banks’ Lerner index. The results of this relationship can be found in table 7. Looking at the first two columns, the results in this table show that there is a negative and significant relationship with the Lerner indices before and after the acquisition or merger and the Lerner index of the rivals themselves. This could in fact be in line with the output reallocation effect discussed by Boone (2008) since as observed from table 3, the Lerner indices of the banks involved in the merger or acquisition has on average significantly increased post-merger.
Table 7: Random Effects Estimation Result of Rival's Own Lerner index This table contains the estimation results of the regression relating the Lerner index of the rivals to the Lerner indices of their corresponding acquiring or target bank. This sample contains data on around 830 banks during the years 2011 until and including 2016. Column 1 contains the results for the sample with the before Lerner index being calculated for the target and acquiror separately. Colum 2 contains the results for the sample where the before Lerner index value is a weighted average of the target and acquiror bank's Lerner index value, where the weights are based on total assets. Significance at the 10%, 5% and 1% is indicated by *, ** and *** respectively.
Constant 0.814 *** (.021) 0.818 *** (.021) LI Before -0.014 ** (.007) -0.033 *** (.008) LI After -0.019 ** (.008) -0.024 *** (.008) Size -0.010 *** (.002) -0.010 *** (.002) Relative Size 0.000 (.000) 0.000 (.000) City 0.007 (.006) 0.006 (.006) 2012 -0.007 (.006) -0.009 (.006) 2013 0.018 *** (.006) 0.015 ** (.006) 2014 0.007 (.007) 0.005 (.007) 2015 0.042 *** (.006) 0.041 *** (.006) 2016 0.048 *** (.009) 0.046 *** (.009) R-squared within 0.043 0.044 between 0.043 0.049 overall 0.044 0.048 N 4864 4854 Total
32
It is therefore interestingly to see whether this difference between the pre-merger and post-merger values is significant, which would indicate that the merger or acquisition has potentially had a negative effect on the market power of the rival banks. Using the same simple F-test as I have done with the results above, I have found that this difference is not significant. So there is no significant change in the Lerner index of the competitors caused by the increase in market power observed for the regular banks. Moreover, this implies that the negative relationship between the Lerner index and prices cannot be explained with the argument of Boone (2008). One can conclude that my second hypothesis is still rejected based on the results of table 7. In the last two columns of table 7 the before merger Lerner index is calculated as a weighted average. One can observe that both Lerner indices have a negative significant relationship with the rivals’ own Lerner index. This difference however is again not significantly different from zero. So with this computation of the pre-merger Lerner index there is also no negative significant post-merger effect on the market power of the rivals’ Lerner index.
In sum, based on the results presented above I reject my second hypothesis, but fail to reject my first and third hypothesis. One can therefore state that there is evidence to support the efficiency view but no evidence supporting the market power view, a result that is also mainly found in the existing literature. Further research has to be done to explain why there is lacking evidence to support the market power view.
5. Conclusion
33
competitors, which do not have the same synergies and as a result they see their prices decline together with their performance.
A first look at the summary statistics of the Lerner indices for both the banks involved in the mergers or acquisition, as well as their competitors, showed that especially for the banks involved in the deal the mean of the Lerner index increased significantly post-merger, and this increase mainly occurred in the years 2014 and 2015. These results were evidence to support my first hypothesis. For the competitors there was no overall significant increase in the mean Lerner index, only for the rivals of the acquirer there was a significant increase in 2015 and for the rivals of the target in 2013.
In the final estimation results in table 5 and 6 I did not see the expected positive effect of the before and after Lerner index on output prices. On the contrary, I found that there exists a negative relationship between these variables and that this value was significantly more negative after the merger or acquisition had occurred. These results contradict my second hypothesis, stating that as a result of an increase in market power, output prices of the rival banks should increase. The results do support my third hypothesis. Hypothesis 3 stated that there should be a negative relationship between the asset complementarity index and output prices. Indeed this relationship in the results is negative and significant. The performed robustness checks did not alter the relationships between the Lerner indices, asset complementarity index and output prices. However the results showed that the significantly negative relationships of the Lerner indices were mainly present for the rivals of the target banks and for the rivals of banks that had 10 or less competitors in the same city. For the asset complementarity, the effects were strongest in the sample of target banks and the sample of banks with 10 or less competitors.
34
found for the total sample of banks both when the Lerner index was calculated separately and as a weighted average. The difference however before and after the merger or acquisition was not significant. This result is therefore inconsistent with the output reallocation effect, ruling this out as a possible explanation. Therefore the second hypothesis is still rejected. On the contrary, I fail to reject my third hypothesis, indicating that there is evidence to support the efficiency view. Besides the negative relationship between the Lerner indices and output prices, there are some more limitations to my research that can be covered in future research. Taking into account more years and more deals for each year might give different outcomes. As observed for the years 2014 and 2015, where there are more mergers and acquisitions compared to the preceding years, the difference in the Lerner index was significantly positive. Perhaps due to the financial crisis the amount of deals in the early years of my sample was limited. I have also discussed other possible methods to define competitors: by SIC codes or 10-K product descriptions. One could apply the 10-K product descriptions as a substitute for the competitor’s city as a method to define competitors. My analysis could also be extended to other financial institutions, making it possible to use SIC codes to define competitors.
35
6. References
Ariss, R. 2010. On the implications of market power in banking: Evidence from developing countries. Journal of Banking & Finance. 34, 765-775.
Berry, H., Guillén, M., Zhou, N. 2010. An institutional approach to cross-national distance. Journal of International Business Studies. 41, 1460-1480.
Boone, J. 2008. A new way to measure competition. The Economic Journal. 188, 1245-1261.
Chatterjee, S. 1986. Types of synergy and economic value: the impact of acquisitions and mergers on rival firms. Strategic Management Journal. 7, 119-139.
Cornett, M., McNutt, J., Tehranian, H. 2006. Performance changes around bank mergers: revenue enhancements versus cost reductions. Journal of Money, Credit and Banking. 38, 1013-1050
Devos, E., Kadapakkam, P., Krishnamurth, S. 2009. How do mergers create value? A comparison of taxes, market power, and efficiency improvements as explanations for synergies. Review of Financial Studies. 22, 1179-1211.
Drukker, D. 2003. Testing for serial correlation in linear panel-data models. The Stata Journal. 3, 168-177.
Eckbo, B., 1983. Horizontal mergers, collusion and stockholder wealth. Journal of Financial Economics 11, 241-273.
Farrell, J., Shapiro, C. 1990. Horizontal mergers: an equilibrium analysis. The American Economic Review. 80, 107-126
Fee, C., Thomas, S. 2004. Sources of gains in horizontal mergers: evidence from customer, supplier, and rival firms. Journal of Financial Economics. 74, 423-460.
Ghosh, A. 2001. Does operating performance really improve following corporate acquisitions? Journal of Corporate Finance 7, 151-178
Hausman, J. 1978. Specification tests in econometrics. Econometrica. 46, 1251-1271
Healy, P., Palepu, K., Ruback, R., 1992. Does corporate performance improve after mergers? Journal of Financial Economics. 31, 135-175
36 Houston, J., James, C., Ryngeart, M., 2001. Where do merger gains come from? Bank mergers from perspectives of insiders and outsiders. Journal of Financial Economics. 60, 285-331.
Kim, E., Singal, V. 1993. Mergers and market power: evidence from the airline industry. The American Economic Review. 83, 549-569.
Kim, J., Finkelstein, S., 2009. The effects of strategic and market complementarity on acquisition performance: evidence from the U.S. commercial banking industry. Strategic Management Journal 30. 617-646.
Makri, M., Hitt, P., Lane, P. 2010. Complementary technologies, knowledge relatedness and invention outcomes in high technology mergers and acquisitions. Strategic Management Journal 31, 602-628.
Maudos, J., Guevara, J. 2007. The cost of market power in banking: Social welfare loss vs. cost inefficiency. Journal of Banking & Finance. 31, 2103-2125.
Perry, M., Porter, R. 1985. Oligopoly and the incentive for horizontal mergers. The American Economic Review. 75, 219-227
Prager, R., Hannan, T. 1998. Do substantial horizontal mergers generate significant price effects? Evidence from the banking industry. The Journal of Industrial Economics. 46, 433-452.
Rhoades, S. 1993. Efficiency effects of horizontal (in-market) bank mergers. Journal of Banking and Finance. 17, 411-422
Rhoades, S. 1998. The efficiency effects of bank mergers: An overview of case studies of nine mergers. Journal of Banking and Finance. 22, 273-291
Shahrur, H., 2005. Industry structure and horizontal takeovers: analysis of wealth effects on rivals, suppliers, and corporate customers. Journal of Financial Economics 76. 61-98.
Song, M., Walkling, R. 1999. Abnormal returns to rivals of acquisition targets: A test of the ‘acquisition probability hypothesis. Journal of Financial Economics. 55, 143-171.
