The effect of M&A deals in the banking sector on the mortgage rate spread in the US

(1)

The effect of M&A deals in the banking sector on the mortgage rate

spread in the US

by

Etan Wijnberg

Master Thesis

Submitted to the department of Finance of the University of Amsterdam

in partial fulfilment for the degree of MSc Business Economics: Finance

Student number: 10000652

Date: July, 2015

(2)

1

Statement of Originality

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

I declare that the text and the work presented in this document is 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)

2

Abstract

Bank mergers can either raise or lower the mortgage rate spread and affect the interest rate paid by bank customers on their mortgages. For the purpose of this research, individual mortgage loan contracts were used to create a cross-sectional regression analysis. This paper concludes that for the bank sample as a whole the mortgage rate spread increases in the medium run and decreases in the long run. The results suggest that for the sample as a whole, mergers lead to gains in efficiency in the long run. For the large bank subsample, this paper finds that after a merger has taken place, large banks use their market power and increase prices for consumers in the long run. These results imply that in general, mergers increase market competitiveness and result in lower prices, with the exception of mergers by large acquirers.

(4)

3

In tandem with the increase in market concentration, the mortgage market saw a period of expansion from 2000 to 2006. Indeed, between 2000 and 2003 its growth soared by an astonishing 400 percent (Calabria, 2011). From 2005 mortgage origination saw a marked decline: from 3000 billion to approximately 1800 billion by 2013 (Mortgage Bankers

Association, 2015). Such considerable changes in mortgage origination coupled with financial innovation in fixed-income products at banks makes it an interesting period to analyse. This paper examines the period just before, during and after the financial crisis. The driving factor behind the growth in mortgage origination was financial innovation as a consequence of the liberalization of the financial markets. Fixed-income products such as mortgage-backed securities made it possible to lend to individuals who prior to such innovation were unable to borrow. The Herfindahl-Hirschman Index (HHI) for the banking sector in the US on average increased from 14 to 16 percent in the sample period 2007 to 2014. The trend can be seen in Graph 1.

A bank merger raises a bank’s operational efficiency and at the same time increases its market power (Erel, 2011). For consumers, the bank’s clients, the merger means one of two things: either the mortgage rate spread increases because competition is reduced and banks can exercise greater market power, or the mortgage rate spread declines as a result of lower costs achieved by increased efficiency. The mortgage spread is defined as the interest rate that a consumer pays on his or her mortgage minus the 10 year rate on US treasury notes. The mortgage rate spread is chosen as the measure, because it shows the price consumers pay for their mortgage. At the same time it is an important product for most banks, as mortgage rates are influenced by strategy and pricing decisions (Calabria, 2011).

Research has been done on the conflicting theories, but depending on the industry and the dependent variable, one theory prevails over the other. The most influential research papers also show different periods in which these effects occur. This research shows whether

1

See, for example, Berger and Hannan (1998); Sapienza (2002); Focarelli and Panetta (2003); and Erel (2011) for papers on the effects of mergers on consumer products.

(6)

5

it is the increased market power or the gains in operating efficiency which have the greater impact on the mortgage rate spread and at what stage such an effect occurs. This paper reports several subsamples to support the respective theories and provide this research with

robustness.

M&A deals can be a critical tool in terms of increasing growth rates and increasing market share. At the same time M&A deals influence market concentration and the pricing of consumer products. This research shows that when taking all the bank mergers which took place in the US in the period 2007 to 2013, prices tend to go up in the two years following a merger and begin to go down three years after a merger. For large mergers the first year is positive, after which the second year is negative and the third year is positive again. In this research the one year period after the merger is defined as the short run, the two year period after the merger is defined as the medium run and the three year period after the merger is defined as the long run.

These results contrast with the findings of existing literature. For instance Erel (2011) concludes that during the first two years the effects on the spread of the consumer loans are negative for the bank customers. Focarelli and Panetta (2003) find that in the short run the prices go down but in the long run the banks increase the price on deposits. The difference can be explained by the usage of a different dependent variable, control variables, control group and subsamples. Even though the research conducted here might suffer from omitted variable bias, it is still useful for future research. As the results are significant for the medium and long run. This research also reports on the current situation in the banking sector and market concentration.

The contribution of this paper is that it is the first paper that looks at the relationship between bank M&As and the mortgage rate spread. Therefore it adds to previous research, it adds to the discussion on the conflicting financial theories and it lays a new foundation for further research on M&As in banking. To the best of my knowledge, no other research has been conducted into the relationship between bank M&As and the mortgage rate. There have been publications for different sample periods on other products that banks offer such as deposits (Focarelli and Panetta, 2003, Berger and Hannan, 1998) and consumer loans

(Sapienza, 2002, Erel, 2011), but it is simplistic to assume that M&A deals in banking affect different types of banking services in a uniform way (Kahn et al., 2005).This paper is also relevant for later research on the EU banking sector since the European Union has a more concentrated banking sector than the US and therefore the effects of bank mergers on banking products are most likely stronger. This paper concentrates on the US market because of the

(7)

6

availability of data on mortgage rates in this market. Some papers also write about the Italian market such as Sapienza (2002) and Focarelli and Panetta (2003) because the banking market is geographically delimited within local areas in Italy.

The effect of M&A deals in the banking sector on the mortgage rate spread in the US is analysed in this empirical paper. To this end, the paper is divided into six parts. The first section, the introduction, describes the background and the objectives. The second part, literature, discusses the literature on this topic. The third part, the hypotheses, discusses the different assumptions and hypotheses. Here, the hypotheses of various empirical works of other researchers are also taken into consideration. The fourth section, data and methodology, describes the empirical model. The fifth section discusses the results from the empirical model. Here the analysis of the results of the research is to be found as well as the analysis of the robustness test. The sixth and final section provides the conclusions.

2 Literature review

The impact of bank M&As on the consumer is ambiguous. There are two conflicting theories on this topic and there is a discussion in financial literature as to which theory holds more weight. The first theory is that mergers create an increase in market concentration and therefore increase prices. The second theory is that mergers create gains in efficiency, which in turn enables banks to price their products lower. The effects described in both theories can take place at the same time, but one effect can prevail depending on the product, location and the time period of the sample. Failed mergers, the importance of a control group and the effect of the size of the acquiring bank is discussed later in this paper.

2.1 Market concentration increases prices

The first theory put forward here is that market concentration increases prices (Weiss, 1989). This can be explained by larger firms exercising greater market power by setting higher prices for goods and services without increasing efficiency. A common method applied to measure market concentration is the Herfindahl-Hirschman Index (HHI). This index measures the concentration of the market on a scale of 0 to 1, 1 being a monopolistic market.

Hannan (1991) writes about concentration in the banking sector for loan contracts and Berger and Hannan (1989) wrote about this for deposits; both find a positive relationship between market concentration and higher prices. In both papers the structure-conduct-performance (SCP) hypothesis is used. The SCP hypothesis argues that the concentration of

(8)

7

banks and other competition-related obstacles affect banking in a negative way from a social viewpoint. In the SCP hypothesis the concentration is typically measured with the HHI.

When looking at the HHI of a metropolitan statistical area (MSA), the hypothesis is that increased market concentration results in higher prices for consumers. Berger et al. (2004) use a more complex way of measuring competition where they differentiate between banks of different sizes when applying the HHI. They conclude that for banks, competition is good from a social perspective as it increases competition and lowers prices for bank

customers. The results for the market concentration are weak. Berger et al. find that concentration in the banking sector has less effect on MSAs with less regulation. The difference in concentration makes generalization of the results impossible for less regulated MSAs (Berger et al., 2004). Berger et al. deviate from the SCP hypothesis in the sense that they find that the SCP hypothesis needs to be adjusted to take into account the size of bank and type of institution. In this research, this complex HHI calculation has not been applied, because of the limited added value. Instead, the choice has been made to distinguish between large and small acquirers, and to include the size of the acquirer as a variable in the

regression.

Prager and Hannan (1998) investigated large horizontal mergers in the banking sector between 1991 and 1994 and find that when market concentration increases, deposit rates increase significantly. Prager and Hannan find that mergers increase market concentration and therefore market power, but in the case of deposit rates the efficiency effect outweighs the market power effect. When bank M&As overlap geographically, mergers can create more market power for the acquirer and therefore an increase in price for the consumers. In their research, Prager and Hannan also explain the role of antitrust agencies and their contribution in the control of market power.

Sapienza (2002) finds that out-of-market mergers generate less efficiency gains than in-market mergers, defined as banks that merge with other banks in the same province in Italy. Instead of using private loans, Sapienza discusses individual business loans. Sapienza does include time-fixed effects, but only for the first four quarters following the merger. Hence, the paper only reports the short-run impact and not the long-run impact on consumers. The control group in this research is the other banks, as opposed to the failed banks. Sapienza concludes that horizontal mergers with small targets result in lower loan rates in Italy, but when the market share of the acquirer increases, the market power effect outweighs the efficiency gains effect. Sapienza also finds that after a merger there is a reduction in small loans offered.

(9)

8 2.2 Efficiency gains from mergers

The second theory relates to efficiency gains that result from mergers. The increase in efficiency can come from synergies or from redesigning the loan portfolio and risk diversification (Erel, 2011). There are many different types of efficiency gains from bank mergers such as cost reduction, improvements in managerial efficiency and profit efficiency that results from synergies (Akhavein et al., 1997, Berger et al., 1999, and Walter, 2004). Profit efficiency involves a superior combination of inputs and outputs. This term is more inclusive than cost efficiency, because it includes cost and revenue effects (Akhavein et al, 1997).

According to Berger et al. (1993), large banks can benefit most from managerial efficiency when an efficient bank takes over an inefficient bank and then shares its managerial knowledge with the target. They illustrate that increasing scale and scope efficiencies does not influence profit by more than a few percent. Berger et al. look at the X-efficiencies of the banking sector in America and draw their conclusions based on literature, as opposed to empirical research.

Becher (2000) shows that between 1980 and 1997 bank mergers created value through synergies. Becher provides a model for the theory that bank mergers are not only done for the sake of empire building but also for synergistic reasons. The mergers in the 1990s created positive wealth effects regardless of the event window. In his research, Becher looks at stock prices before and after a merger. He finds that targets gain on average by 33 percent, bidders break even and the combination gains by 3 percent. Becher also shows that payment methods do not matter for his research, yet geographic location does. If the acquirer merges with a bank outside of its own state, it affects returns negatively. Overall the wealth effects are positive – this was especially the case in the 1990s according to Becher. This paper explores whether the positive wealth effects, that Becher identifies, are passed on to the consumer.

This forms the second theory, which concludes that efficiency gains reduce prices for consumers. Focarelli and Panetta (2003) show that in Italy for the sample period between 1990 and 1998, the effect of mergers on the price of deposit rates was negative for consumers, but in the long run the efficiency gains outweigh the market power effect. This leads to better prices for consumers. The paper by Focarelli and Panetta differs from other papers on this topic, because it looks at the long run while other papers only examine the short run. Additionally Focarelli and Panetta provide a possible solution to the “merger puzzle”. The merger puzzle questions the usefulness of M&A deals. Focarelli and Panetta find that there

(10)

9

are synergies and they solve the merger puzzle in their paper. In contrast to the paper by Focarelli and Panetta, this paper looks at the American market and the mortgage market and covers a different time-period. In the paper by Focarelli and Panetta the focus lies mostly on the three years after the merger dummy, while in this paper there are three regressions for each year after the merger.

Erel (2011) researched the effect of bank mergers on loan prices. Erel shows that the banks that merged between 1990 and 2000 in the US, on average offered lower interest rates on consumer loans. This means that the efficiency gains outweigh the market power effect and therefore favour the consumers. At the same time Erel (2011) does not find a significant change in the behaviour of loan prices issued in the US. This is contradictory to Sapienza’s (2002) findings. In this paper the change in the loan supply is not discussed, but it is left for future research as it does not address the main question of this research.

Mega-mergers have been trending since the 1980s when regulations on interstate banking changed in the US. The deregulation of large bank mergers resulted in an increase in large bank mergers. This was followed by even more deregulation that made it easier for banks to execute mergers (Akhavein et al., 1997). Akhavein defines mega-mergers as mergers where both parties have assets worth over 1 billion dollars or more for the whole period; this standard is adhered to in this paper as well. In this paper large bank mergers are researched as large banks can impose higher prices as a consequence of market power more easily than small banks. On the other hand there are forces such as antitrust laws that counteract exploitation. The subsample of the large banks is also included in the robustness tests, and also because mergers are not affected by local economic declines.

Akhavein et al. (1997) investigate both profit efficiency and market power effects of mergers which took place in the period 1981 to 1989. They predicted and found that both increased. They show that cost-efficiencies are relatively less for the large banking mergers, but the efficiencies gained by such large mergers are mostly those of managerial efficiency and the sharing of knowledge. They also claim that cost-efficiencies are a poor indicator because they increase as revenues increase. Therefore profit efficiency is a better indicator and they include a profit function in their paper. Akhavein et al. conclude that prices of the loans and deposits do not change significantly as a result of a merger. Akhavein et al. also find that the market concentration measured by the HHI only increases by a small amount in local markets, and therefore conclude that antitrust policy was fairly effective in local markets at that point in time.

(11)

10

Penas and Unal (2004) discuss mega-mergers and how such mergers affected bond prices from the period 1991 to 1998. Penas and Unal find that the returns on bonds are significantly higher when a mega-merger is announced and that banks benefit from the too-big-to-fail status and diversification gains. They also find that synergy gains contribute to a lesser degree in mega-mergers. Most importantly, this paper finds that there are financial benefits for the acquirer post-merger. The financial benefits from mega-mergers on bond prices are expressed in the lower cost of debt.

Erel’s research (2011) is still relevant, but a new trend for banks has emerged. Owing to the US legislative environment and monetary policy mortgage loans have become

increasingly important. From 2000 to 2003 the mortgage market grew by almost 400 percent according to Calabria (2011). Calibria demonstrates that the new mortgage origination increased from 1 trillion dollars to 3.8 trillion dollars in this period. In 2008, new mortgage origination shrank, but since then the mortgage origination has recovered from its 1.5 trillion dollar low point.

Mortgage suppliers have become less dependent on the mortgage interest margin, but instead depend more on fees, gains on sales of mortgages, and the sale of asset-backed securities (ABS) and other financial products (Cetorelli, Mandel and Mollineaux, 2012). Securitization has become increasingly important in the last decade. From 2002 to 2007, the number of mortgages sold soared from 35 percent to an astonishing 60 percent respectively. Past research focused on stock prices or loan spreads to measure the impact of M&A deals. Therefore the dependent variable in this research is the mortgage rate spread.

This research thus differs from the research carried out by Erel (2011) because of the non-overlapping samples, the changed profit structure of banks and the independent variables. This research also looks into large M&A deals specifically as opposed to all M&A deals, because of the demand effect that might occur when looking at all mergers (Garmaise and Moskowitz, 2006). All the theories above yield different empirical results, therefore it is interesting to research the mortgage rate spread and determine the effect of M&A deals on this particular product.

2.3 The ambiguous effect of mergers

The previous section presented the most relevant papers for this research. The most important publications for this paper are the ones by Sapienza (2002), Erel (2011), and Focarelli and Panetta (2003). These papers form the basis of the most recent discussion of the effect of bank

(12)

11

mergers on bank product prices. At the same time, these papers form the basis for the

methodology that is used in this paper. While Focarelli and Panetta (2003) find that efficiency gains need a period of three years to be realized, Erel (2011) finds that a decline in the spreads occurs in the first year and that the decline peaks in the second year. Erel finds that on

average, mega acquirers have lower spreads in the first two years following a merger, but this increases in the third year. The theory relating to this phenomenon is that the largest banks want to increase their market share after a merger and therefore make price-cuts. When the mega-banks have attained their larger market share, they reverse these price reductions. In this paper, the same pattern for mortgage loans is expected as that of Erel’s (2011) with consumer loans.

2.4 Failed mergers

Unlike other papers, this paper uses a control group consisting of failed bank mergers. In this research, bank mergers that were successful are controlled by a group of banks that have had a failed bank merger. The failed bank mergers in this research are planned mergers that are withdrawn at a late stage. Ashton & Pham (2007) in their research use banks that had a bank merger as the treatment group and the banks that did not have a bank merger as a control group. Erel (2011) uses a control-group consisting of banks that did not have mergers as well, the difference between Erel’s paper and the paper by Ashton & Pham is that Erel creates subsamples for size and loan amount to do robustness tests. Garmaise and Moskowitz (2006) use failed mergers to do the robustness test. Germaine and Moskowitz exclude the failed mergers and use them as a placebo test later on in their paper.

In this paper, the decision was taken to include failed bank mergers, because they are the banks that are the closest to the banks that had a successful merger. The only downside is that the size of the control sample of failed mergers in the US is fairly small. A possible explanation for this might be the strict control of bank mergers by the authorities. The strict control and high costs forces merger candidates to carefully prepare the merger at an early stage to reduce the risk of failure. The U.S. Department of Justice and the Federal Trade Commission (2010) apply strict rules and the Fed has the authority to break off a merger for many different reasons. Furthermore, the FDIC needs to give its approval in accordance with the Bank Merger Act.

(13)

12 2.5 Size effect of acquiring bank

Weston & Strahan (1996) show that large banks lend to large businesses while small banks lend to smaller businesses. This can be generalized by stating that large banks provide larger mortgages and smaller banks give out smaller mortgages. One of the reasons for this is that for smaller loans, more control is needed. Smaller banks can collect soft data more easily and have a better relationship with the borrower. This is why the small and large banks differ in their behaviour.

In Sapienza’s paper (2002) the size of a bank is acknowledged to influence many aspects of the bank. Aspects such as the organizational structure, objectives, the pricing of consumer products and many other aspects of the bank. Sapienza finds that size does have an effect on mergers. The effect does not necessarily need to be adverse for the consumers. The effect depends on whether the borrower can switch between lenders and on the costs of switching lender. It is interesting to investigate whether or not this effect is also seen in the mortgage market as it is for business loans.

For these reasons, in this research, a separate variable has been added to measure the size of the acquiring firm when regressing it on the mortgage rate spread. This is done with the variable logarithm of assets. In Sapienza’s paper, sales are used as a measure of size, but in principle this should not yield different results.

According to Patti et al. (2007) another important aspect to consider is that when a merger takes place, the supply of loans by the newly merged bank is likely to go down in the short run and medium run. The effect of the M&A shock lasts as long as three years. Banks reduce credit on average between 8 and 10 percent after consolidating. It is not unlikely that the effect is the same for mortgage loans. It is possible to report the influence of the M&A deals on the mortgage supply, but that is not the focus of the paper and is therefore not discussed here. This effect is larger for smaller banks and is another reason to use large bank mergers instead of just bank mergers in general.

In the paper by Garmaise and Moskowitz (2006) the banks with assets of 1 billion dollars or more have been selected as these are banks that are unlikely to be affected by local economic decline. Germaise and Moskowitz reason that because of the size, scope and health of the banks, the mergers are not driven by the economic decline in the local neighborhood.

(14)

13

3 Hypotheses

When banks merge the mortgage rate decreases and market concentration increases. This could mean that the market power gained from the merger is not used to impose higher prices. Instead the concentration of the market means that the efficiencies can be used to lower the prices for consumers. The Erel (2011) paper shows that for loan spreads, this is indeed the case. This leads to the following hypothesis.

𝐻₁: An increase in market power as a result of M&A deals results in a lower mortgage spread rate.

The market concentration in the banking sector has increased in recent decades. The FDIC collects data for each bank on the deposits. In this research, the data of the FDIC is used to calculate the HHI for each MSA, the HHI ranges from 0 to 1. This hypothesis is rejected if there is a large increase in market power in MSAs in the sample period, over the years in MSAs, measured in HHI. The U.S. Department of Justice and the Federal Trade Commission (2010) has set a guideline in which a competitive market for banks should have a HHI with a maximum of 0.2 so that there is no governmental intervention necessary.

𝐻₂: The average market concentration increases, but not enough to create a concentrated market on average.

When the total assets of the acquirer are 1 billion dollars or more, the acquirer is considered to be large. When the total assets are smaller than 1 billion dollars, the acquirer is considered to be small. According to Berger et al. (1993) the benefit from a large merger comes from management improvements while smaller mergers benefit from cost and profit efficiencies. The bigger efficiencies are expressed in cost, managerial and profit efficiency that result from synergies (Akhavein et al., 1997, Berger et al., 1999, and Walter, 2004).

𝐻3: Large acquirers benefit more from efficiency gains than small acquirers.

If there is less of a decline in mortgage rates for large mergers than for small mergers, the following hypothesis comes to mind. Concentration is measured with the change in HHI for an MSA and the effect of this change on the mortgage spread. A large acquirer is expected to

(15)

14

be able to wield its market power more than a smaller acquirer. Hence, the following hypothesis is formulated.

𝐻4: Large acquirers benefit more from market power efficiency than small acquirers.

4 Data & Methodology

4.1 Data

The sample period of the panel data is seven years: from 2007 to 2014. The sample consists of 2,467,572 individual mortgages on which there is data for the mortgage rate spread as can be seen in Panel A of Table 1. Panel B of Table 1 reports that there are 754,335 different

observations for individuals who got a mortgage at banks that had either had a failed or a successful merger over the aforementioned seven-year period. For each individual, the

mortgage rate, individual specifics, bank specifics and HHI were collected. For the purpose of this research, a file was created with individuals that got mortgages at banks that either had a failed or a successful bank merger in the sample period. 566 banks are left in the sample, 484 of which underwent a successful bank merger, 82 had a failed bank merger and eight banks which had both a successful and a failed merger in the same year (see Panel C of Table 3). The mergers that had both a failed and a successful merger in the same year were dropped from the sample. Table 3 shows the difference between banks that had successful mergers (Panel A) and failed mergers (Panel B). The individuals that got a mortgage at an acquiring bank that had a successful merger on average have a higher mortgage rate. The banks that had a failed bank merger on average are larger than the banks that had a successful merger in the sample period.

Using Thomson One, one can find the successful bank mergers in the US for the period from 2007 to 2013. In Thomson One the Acquirer Size and Target Size can be added to create subsamples by the size of the banks and analyse the impact of the merger on the acquirer. Also the size of the merger can be added. The size of the merger has not been used because of the large number of missing values for this variable.

It is possible to gain information on the number of mortgage loans provided in a year from the Federal Financial Institution Examination Council (FFIEC) site. For this paper, the Home Mortgage Disclosure Act (HMDA) data was selected. This data provides information on the home mortgage lending activity of banks. After merging the seven years of data, one file of data is created. Then the data file is merged manually with Thomson One, so that only

(16)

15

the banks that did M&A deals are left in the sample. The files are merged on the basis of NCUSIPs. The NCUSIPs are found through the usage of CRSP where the annual daily stock file gives NCUSIPs and CUSIPs for each bank. This file is merged with the Thomson One file.

In order to calculate the HHI, new data is needed as the MSA data is not given for each institution. Therefore a download is needed from the FDIC site. FDIC is the insurance company of deposits that needs to collect data on deposits. This data has been used as a measure for the HHI on other banking products. The reasoning behind this is that where a bank supplies deposits, it is also able to supply other financial products such as mortgages. This way of measuring the HHI is common in related literature.

The liquidity parameter is constructed for each bank by using the Compustat database. From this database, the cash and total assets are retrieved. This variable is then merged with the master dataset. Compustat also uses data that is supplied by S&P Capital IQ.

The result of combining all of these datasets is a unique database. This unique

database can be used to run cross-sectional regressions and obtain results that are exclusively available via this dataset.

4.2 Methodology

The mortgage spread is the difference between the 10-year US treasury note and the mortgage rate that lenders charge their borrowers. The 10-year US treasury note changes over time, as well as the mortgage rate that lenders offer their customers. Because of the fluctuation in the mortgage rate, the 10-year US treasury note rate needs to be subtracted to compensate for inflation and the economic cycles. The mortgage rate spread is used as the dependent variable in the regression. The mortgage rate spread depends on the individual characteristics, the bank specific characteristics, the area the bank operates called MSA and the time.

The model is a cross-sectional regression model with fixed effects and lagged

variables. Panel data was considered, but because of the loss of information rejected. The only way panel data would work in this case is to take averages for each bank. This creates noise in the output. Using the cross-sectional regression model with fixed effects is considered to be the best fit. To test for the use of fixed or random effects the Hausman specification test has been used. For each regression, a test was carried out to determine whether it would be better to estimate the regression with either fixed or random effects. In all the cases, the usage of the random effect model was rejected and the preferred model is the fixed effects model.

(17)

16 𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒 𝑆𝑝𝑟𝑒𝑎𝑑_{𝑖,𝑘,𝑚,𝑡} = 𝛼 + 𝛽₁(𝑀𝑒𝑟𝑔𝑒𝑟𝐿𝑎𝑔1)_{𝑖,𝑘,𝑚,𝑡}+ 𝛽₂(𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒𝑠𝑖𝑧𝑒)_𝑖,𝑡 + 𝛽₃(𝐴𝑝𝑝𝑙𝑖𝑐𝑎𝑛𝑡𝐼𝑛𝑐𝑜𝑚𝑒)𝑖,𝑡+ 𝛽4(𝑀𝑒𝑑𝑖𝑎𝑛𝐹𝑎𝑚𝑖𝑙𝑦𝐼𝑛𝑐𝑜𝑚𝑒)𝑖,𝑡 + 𝛽₅𝐿𝑜𝑔(𝑆𝑖𝑧𝑒𝐴𝑞𝑢𝑖𝑟𝑒𝑟)_𝑘,𝑡+ 𝛽₆(𝐻𝐻𝐼)_𝑚,𝑡+ 𝛽₇(𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦)_𝑘,𝑡 + 𝑢_𝑚+ 𝑑_𝑡 + 𝜀_{𝑖,𝑘,𝑚,𝑡} 𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒 𝑆𝑝𝑟𝑒𝑎𝑑_{𝑖,𝑘,𝑚,𝑡} = 𝛼 + 𝛽1(𝑀𝑒𝑟𝑔𝑒𝑟𝐿𝑎𝑔2)𝑖,𝑘,𝑚,𝑡+ 𝛽2(𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒𝑠𝑖𝑧𝑒)𝑖,𝑡 + 𝛽3(𝐴𝑝𝑝𝑙𝑖𝑐𝑎𝑛𝑡𝐼𝑛𝑐𝑜𝑚𝑒)𝑖,𝑡+ 𝛽4(𝑀𝑒𝑑𝑖𝑎𝑛𝐹𝑎𝑚𝑖𝑙𝑦𝐼𝑛𝑐𝑜𝑚𝑒)𝑖,𝑡 + 𝛽5𝐿𝑜𝑔(𝑆𝑖𝑧𝑒𝐴𝑞𝑢𝑖𝑟𝑒𝑟)𝑘,𝑡+ 𝛽6(𝐻𝐻𝐼)𝑚,𝑡+ 𝛽7(𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦)𝑘,𝑡 + 𝑢𝑚+ 𝑑𝑡 + 𝜀𝑖,𝑘,𝑚,𝑡 𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒 𝑆𝑝𝑟𝑒𝑎𝑑𝑖,𝑘,𝑚,𝑡 = 𝛼 + 𝛽1(𝑀𝑒𝑟𝑔𝑒𝑟𝐿𝑎𝑔3)𝑖,𝑘,𝑚,𝑡+ 𝛽2(𝑀𝑜𝑟𝑡𝑔𝑎𝑔𝑒𝑠𝑖𝑧𝑒)𝑖,𝑡 + 𝛽3(𝐴𝑝𝑝𝑙𝑖𝑐𝑎𝑛𝑡𝐼𝑛𝑐𝑜𝑚𝑒)𝑖,𝑡+ 𝛽4(𝑀𝑒𝑑𝑖𝑎𝑛𝐹𝑎𝑚𝑖𝑙𝑦𝐼𝑛𝑐𝑜𝑚𝑒)𝑖,𝑡 + 𝛽₅𝐿𝑜𝑔(𝑆𝑖𝑧𝑒𝐴𝑞𝑢𝑖𝑟𝑒𝑟)_𝑘,𝑡+ 𝛽₆(𝐻𝐻𝐼)_𝑚,𝑡+ 𝛽₇(𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦)_𝑘,𝑡 + 𝑢_𝑚+ 𝑑_𝑡 + 𝜀𝑖,𝑘,𝑚,𝑡

The subscripts are i for unique individual, k for unique bank, m for unique MSA and t for years. The three regressions contain a lagged dummy for after the merger. These dummies are either 1, 0 or missing for the years of the sample. MergerLag1 is 1 for one year after the successful merger and 0 for one year after the failed merger. MergerLag2 is 1 for two years after the successful merger and 0 for one year after the failed merger. The third lagged

dummy is 1 for a period of three years after the successful merger and 0 for one year after the failed merger. These variables are also referred to as the lagged variables. The regressions contain a dummy not only for one year, but also for two and three years after the merger to see whether the impact is stronger over a longer time period after the merger. The regressions have to be separated because otherwise collinearity occurs. It is also possible to regress only on the successful mergers and only on the failed mergers as is done in Table 10.

The expectation is that the first two years after the successful mergers the effect on the price is negative and in the long run the effect on the price is going to be positive. This is because the merger shock wears off after two years and the effect is compensated after three

(18)

17

years. The theory for this is that the banks want to increase their market share and therefore lower the prices in the first two years (Erel, 2011).

Control variables are included on the level of the individual that closes the mortgage, such as the size of the mortgage, the applicant income and the median family income. These controls control for the risk of the borrower. The expected sign for the coefficient of these controls is positive, because if the risk goes up the price should go up. Mortgage size is the size of the mortgage loan and is measured in thousands of dollars. The mortgage size is a common control variable to include as it is included in most academic literature such as Focarelli and Panetta (2003). The applicant income and the median family income give an indication to the bank how risky the loan is. The estimation of risk of a borrower is done by using the FICO code. Given that banks do not share this information publicly, in this research the choice was made to control for credit risk by using these variables. The mortgage size, applicant income, median family income and size of the acquirer are logarithms. Logarithms are used because otherwise the coefficients become small. At the same time logarithms deal with outliers.

Year dummies are also included for each year. In some regressions the beginning years are omitted due to the use of the lagged variable. 𝑢𝑘 and 𝑑𝑡 are respectively the MSA

fixed effects and the year fixed effects. The year fixed effects contain all the macroeconomic developments such as changes in house prices, the financial crisis and changes in the inflation rate. In the papers of Erel (2011), Focarelli and Panetta (2003) and others these are included. In cross-sectional regression analysis it is necessary to include the dummies for year and area.

The bank specific effects are included to deal with the endogeneity of the merger dummy. Mergers are endogenous and depend on bank specific information, such as the liquidity of the bank at the time of the merger or the size of the acquiring bank. To deal with the endogeneity problem the bank specific controls are included in the regression. Erel (2011) uses the size of the acquirer and the non-performing loans of the bank. The size of the

acquirer is the same, but in this paper a different liquidity measure is used that provides a better view of liquidity of the bank as a whole. Bharath et al. (2009) choose an alternative to this approach and use the method of the Instrument Variables (IV) to deal with endogeneity. The Herfindahl-Hirschman Index measures the competitiveness of the banking market as a whole. The HHI-index for each MSA is needed, because not every American can get access to every mortgage rate in the US. The banks that do large mergers in this market have to take anti-trust regulation into account (U.S. Department of Justice and the Federal Trade Commission, 2010). In the regression the HHI-index controls for the differences in market

(19)

18

power of the banks. Combining the MSA with the deposits is a common way to calculate the HHI of banks in the literature.

𝐻𝐻𝐼 = ∑ 𝑠_𝑖2 𝑁 𝑖=1

The liquidity measure used is cash over total assets for each bank. The liquidity measure is included, because it influences the M&A deals made by a bank. If there is more cash on the balance sheet, M&A deals are more likely to occur. The liquidity measure is part of the bank specific variables to control for endogeneity. It would be better to look at the defaulting loans divided by the total amount of loans instead. The reason for using defaulting loans instead of cash over assets is that banks are characterized for not holding cash. Instead of holding cash banks make use of the intra-bank lending. However, the data does not allow for this measure and hence this measure of liquidity is used.

𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦 = 𝐶𝑎𝑠ℎ

𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠

Some of the variables are made logarithms and all the variables with outliers are Winsorised before using them in the regressions. The variables Winsorised are: Mortgage rate spread (Graph 3), Applicant loan amount (Graph 4) and Mortgage size (Graph 5). The first and the last bin are for this reason higher than they would be without the Winsorisation, because the outliers are placed in these bins.

No interaction terms are included in the regression as is done in several other paper like Sapienza (2002) and Erel (2011), because the interactions turned out to be insignificant. Neither do the interactions add to the model.

Furthermore serial correlation has been accounted for, by allowing for clustering of the error term. Petersen (2009) shows that it is necessary to do so in panel data and alternatively in cross-sectional data with time dummies.

(20)

19

5 Results

5.1 Effect of mergers on the mortgage rate spread

This section examines the impact of mergers on the mortgage rate spread. The analysis in this section is based on the results presented in Table 4.

Table 4 reports the three regressions that include all the variables of the entire sample of mortgage loans that either had a successful or a failed merger in the sample period. The number of observations differs due to the usage of lagged variables. Banks that did a

successful or failed merger in one sample period can be excluded in the next. In Column (1) of Table 4 all seven years are included, in Column (2) six years are in the sample period and in Column (3) the sample period is from 2009 to 2014. In Table 4 the year fixed effects were mostly significant and some of the MSA fixed effect dummies were omitted because of perfect multicollinearity. In Table 4 all the controls and all the fixed effects are included for all the regressions. The fixed effects are: time fixed effects, MSA fixed effects and bank fixed effects.

Column (1) of Table 4 shows that the coefficient of interest, the MergerLag1 of 0.918, is insignificant. This means that there is no significant linear dependence of the mortgage rate spread on MergerLag1. In other words, the fact that there was a merger in the year previous to getting a mortgage does not affect the mortgage rate significantly. In Column (1) the controls are all significant for a level of at least 5 percent, except for HHI. In this Column the

logarithm of the mortgage size has a coefficient of -0.739 and is significant at a 1 percent level. This means that if there is a 1 percent increase in the logarithm of the mortgage size, the mortgage rate spread increases by -0.00739. The bigger the size of the mortgage loan, the smaller the mortgage rate spread. Another significant control variable in Column (1) Table 4 is the Log applicant income. The logarithm of the applicant income is 0.363 and is significant at a 1 percent level. This means that with an increase in the Log applicant income of 1

percent, the mortgage rate spread increases by 0.00363. The sign is different from the expected sign. In theory the higher income means less risk for the bank and therefore less costs for the borrower. The logarithm of the mean family income is -0.057 and is significant at a 1 percent level. This means that if there is a 1 percent increase in the logarithm of the mortgage size, the mortgage rate spread increases by -0.00057. Theoretically this is comprehensible. As the borrower’s risk decreases, the rate decreases as well. The control variable for the size of the acquirer is negative and significant for a 1 percent level. This is rational as the larger banks can diversify better and therefore charge lower rates. The

(21)

20

logarithm of the liquidity measure is 56.37 and is significant at a 5 percent level. The increase of 1 percent in the logarithm of the liquidity means a higher mortgage rate spread of 0.56. There are multiple explanations for this phenomenon. The liquidity measure can anticipate a merger or indicate that the bank is not doing so well and needs to hold cash. The adjusted R-square is 0.17. This means that the model has some explanatory value, but that there might be some variables missing.

Column (2) of Table 4 shows the regression with a lagged variable of two years. This time MergerLag2 is 1.155 and is significant at a 1 percent level. This means that if there is a merger two years before getting a mortgage rate, the spread goes up by 1.155 percentage points. In other words, it becomes more expensive to get a mortgage two years after a bank was involved in a merger. The control variables Log mortgage size, Log applicant income and log median family income have the same signs as in Column (1) and are all significant for a 1 percent level. HHI is -1.37 and is significant at a 5 percent level. This means that an increase in the competitiveness of the market results in a lower mean mortgage spread two years after the merger. Log acquirer size also has the same sign as in Column (1) and is significant at the 5 percent level. The logarithm of the liquidity measure is 37.76 and is significant at the 5 percent level. The adjusted R-square is 0.2 meaning that the model has some explanatory value, but that there might be some variables missing.

Column (3) of Table 4 reports the regression with a three year lagged variable. The variable of interest, MergerLag3 is -0.9 and is significant at a 1 percent level. The control variables Log mortgage size, Log applicant income and log median family income have the same signs as in Column (1) and (2) and are all significant for a 1 percent level. The logarithm of the acquirer size is now smaller, but still significant for a 1 percent level. The variable liquidity is omitted. The adjusted R-square is 0.21 meaning that the model has some explanatory value but that there might be some variables missing.

Table 9 reports the correlations between the different variables of this sample. The Table shows that the highest significant correlation is 0.37 and is between Log mortgage size and Log applicant income. It is understandable that there is a positive correlation between the two. As none of the correlations go above 0.5 there seems to be no correlations issues with the regressions.

Table 4 provides evidence in support of hypothesis 1. In the long run, prices go down for consumers. This can be found in Column (3) of Table 4, where variable Mergerlag3 is negative.

(22)

21 5.2 Failed mergers as a control group

Table 5 reports the different regressions with the lagged variable of one year. The Table shows the effect of different regressions on the adjusted R-square when the insignificant controls are dropped. The failed and successful mergers are separate variables in these regressions. This makes it possible to compare the treatment with the control group.

Column (1) of Table 5 reports the regression with the lagged variable of one year and all the controls, and shows insignificant variables for both the failed and the successful variable. All the individual controls are significant, namely the logarithm of mortgage size, the logarithm of applicant income and the logarithm of the mean of the family income. The adjusted R-square is 0.19 and does not change for the following years. Column (2) of Table 5 reports the regression with the lagged variable of two years and all the controls. This shows a significant variable of interest at a 5 percent level for the successful mergers, but the failed mergers are still insignificant. Column (3) of Table 5 shows the regression with the lagged variable of three years and all the controls. Here the failed mergers are significant at a 1 percent level. This is the only year in which this variable is significant. Failed mergers are -0.51 and significant at a 1 percent level. This means that in the third year the failed mergers had a lower mortgage rate spread than the other mergers. As the variable is only significant in the third year it is likely that there is a bias of some sort. This is discussed in section 5.3.

Table 6, 7 and 8 show another robustness check for the regressions. These tables show the regressions with and without the insignificant controls. Column (1) of Table 6 reports the regression with the lagged successful variable of one year on the mortgage rate spread with controls. Column (2) of Table 6 is the same, but for failed mergers instead of successful ones. Then in Column (3) (4) (5) and (6) of Table 6 controls are dropped, but the variables of interest stay insignificant for the first year.

Column (1) of Table 7 reports the regression with the lagged successful variable on the mortgage rate spread with controls. Column (2) of Table 7 is the same, but for failed mergers instead of successful once. Then in Column (3) (4) (5) and (6) of Table 7 controls are

dropped. Only in Column (1) of Table 7 the variable of interest is significant, meaning that dropping insignificant controls in this case does not make the variable more significant.

Column (1) of Table 8 reports the regression with the lagged successful variable on the mortgage rate spread with controls. Column (2) of Table 8 is the same, but for failed mergers instead of successful ones. Then in Column (3) (4) (5) and (6) of Table 8 controls are

(23)

22

mergers are significant in the third year for a 1 percent level and none of the successful mergers are significant for the third year. This shows that three years after the failed merger the influence on the price is significant and negative.

The tables presented in this section indicate that in the short run the regressions seem to be insignificant. In the medium and long run they seem to be significant. It is possible that the effect on the mortgage rate spread does not occur in the short run, but that is not very likely. Dropping insignificant controls from the regressions does not affect the significance of the variables of interest in a substantial way. In the next section, several explanations are given for the behavior of the variables.

5.3 Possible explanations for insignificance of certain variables

A number of factors have contributed to the insignificance of the variables of interest in some of the regressions. First of all the clusters might have caused the sample size to become too small with the lagged variables. The clustering occurs when the cross-sectional fixed effect regression is generated. The second explanation for the insignificance is that the dataset is unbalanced, because of the exit of individual banks and bank products. Some of the banks went bankrupt or stopped existing in the sample period or were founded during the sample period causing gaps in the data. At the same time it is unbalanced because the banks do not report mortgage rates for each year, but only for some years. Most likely the discontinuous reporting causes a bias as banks only report the lower mortgage rates; if they are in the position to do so. The third possible explanation for the insignificant results lies in the master dataset. Not many banks report the mortgage rate spread and almost none of the banks in the sample report it for all the years consistently. The missing years cause gaps in the data and this makes the data partially insignificant.

The fourth reason for the insignificance is the likeliness of the omitted variable bias causing the regressions to yield less than logical results. Important causal factors that are left out could be the size of the merger, the defaulting loans and the loan covenants. As explained earlier in this paper, the size of the merger is excluded because of the large number of missing values and the defaulting loans are excluded because the dataset does not allow for this variable. The dataset does not allow for the loan covenants either. Bharath et al. (2009) show that there is a significant relationship between the loan covenants and the loan spread.

Therefore it is not unlikely that this variable causes omitted variable bias. Instead, in this research, the choice was made to fulfil all the risks for the individual with the following

(24)

23

controls: the variables size of the merger, the income of the applicant, and the median family income.

The fourth problem the regression might suffer from is an econometric problem that can arise when the dependent and the explanatory variables are jointly determined. The problem is called simultaneity. A possible solution for future research for this problem is the Instrumental Variable (IV) approach (Bharath et al., 2009).

Both the omitted variable bias and simultaneity can cause endogeneity. If this is the case, there is a correlation between the error and the dependent variable.

5.4 Effect of mergers on the mortgage rate spread for large and small banks

Thus far, the results have shown an economically significant result for the variable of interest and on the mortgage rate set by banks for consumers for the medium and long run. The results are all significant for a level of at least 10 percent. For further evidence of the nature of the relationship between the mortgage rate and the acquiring bank, the differences in acquirer size were examined.

Table 10 reports the regression on the subgroup of assets that are larger than 1 billion dollars. This group is called the large group. This group is particularly interesting because of its different behaviour when compared to the small group and the lack of the demand effect that can increase the noise in the results. Table 10 shows that all the variables and controls are significant for a 5 percent significance level at least, except for the HHI control variable. All the controls have the same sign as the full sample in Table 4. From the odd numbered Columns in Table 10, one can conclude that in the first year after a merger the average mortgage price went up by 0.353 percent, after which it went down by 0.542 in the second year, and after three years the price of mortgages went up again by 0.6 percent for the banks that were involved in the merger.

The results for the first year run contrary to expectations. It was expected that the first year would see prices go down, as they do in the second year, to gain a larger market share directly after the merger. Then in the third year the prices were expected to go up again. These expectations were partly borne out, but only for the medium and long run. For the short run, the increase in price can be explained by different factors. First of all, the sample is yearly based, because of the mortgage data. Therefore a merger taking place at the beginning of the year could have a different effect to a merger taking place at the end of the year and can

(25)

24

therefore have a different effect on the outcome for the following year. Given this potential inconsistency in the data, it is better to examine the medium and long run.

The failed mergers were omitted for the second and third year due to collinearity. In the first year the variable is 0.7, meaning that the group of mortgage borrowers that borrowed at a bank that had undergone an unsuccessful merger the previous year has to pay more than the group of mortgage borrowers borrowing at a bank which had a successful bank merger.

Table 11 reports the results from the regressions on the sample group for banks with assets less than 1 billion dollars. For the variables of interest the only significant ones are in Columns (4) and (6). Two years after the failed merger, the small group has had a significant increase in prices of 0.2 that is significant at a 10 percent level. Three years after the failed merger the prices have gone down by 0.55. Unfortunately the results for the successful mergers are not significant. Therefore the comparison cannot be made for the small subsample. What is of interest in the small subsample is that the control variable HHI is significant and negative for each year. This is only the case for the banks that underwent a bank merger and have less than 1 billion dollars in assets. This means that if the HHI goes up in each of the three years, the effect on the smaller banks is that they lower their prices significantly.

At the end of this analysis we can conclude that hypothesis 4 can be accepted. Large acquirers benefit more from market power efficiency than small acquirers one year after a merger. The small successful and failed acquirers are unable to exploit the market

concentration. In fact the concentration of the market causes the smaller firms to lower their prices. The efficiency gains from the merger overcome the possibility to exploit the market power for banks with less than 1 billion dollars in assets.

In the light of Sapienza’s paper (2002), which also examines the period one year after the merger, the results make sense. Sapienza also finds insignificant results for the

concentration measure, but significant results for the merger dummy. The insignificance of the concentration measure for large acquirers can be explained by the reasoning that large banks have clients with larger loans who therefore can apply to many banks. The increase in the size of the loan makes the borrower less dependent on the bank. At the same time, a large bank is usually stationed in many MSAs and is therefore not influenced by the HHI as much as a smaller bank.

(26)

25 5.5 Effect of mergers on the mortgage rate spread for different sizes of mortgage loans

Another robustness check is illustrated by Tables 12 and 13. In Table 12 the sample has been reduced to mortgages in excess of 100,000 dollars. Mortgage loans equal to or higher than 100,000 dollars are considered to be large loans, and the loans that are less than 100,000 dollars are labelled small mortgage loans.

Table 12 shows that the large loans are not significantly influenced by the risk of the borrower, which is not the case for the whole sample. For the large loans the adjusted R-square is very low, indicating that there might be different variables that influence this group of loans. In this sample there should be no significant change in the mortgage rate after a merger because a large mortgage loan is not affected much by direct competition, as clients who loan such large amounts can choose to go to a number of different banks.

Table 13 reports the effect of mergers on the mortgage spread for small mortgage loans. It shows that all the years are significant for the failed mergers variable. In the first two years the effects are positive and in the third year the effect is negative. The adjusted R-square is somewhat higher than in the other regressions, but the first and third year do not have the significant variable successful mergers. In this case only the second year sees a rise in the mortgage rate. This cannot be explained by academic literature. It is likely that this regression suffers from omitted variable bias.

Based on the available literature, one would expect that small loans are more affected by mergers than large loans. The small borrowers usually cannot apply to all banks, and are usually limited to local banks. The large borrowers can apply to most large banks. Consequently, the HHI should have more influence on the small loans (Garmaise and

Moskowitz, 2006). This paper does not reach the same conclusion. In fact, these results reveal that only the size of the mortgage matters for large loans, while for small loans all the

individual characteristics of the mortgage borrower matter.

6 Conclusion

This paper analyses the effect of bank mergers on the mortgage rate spread. Strong evidence is found that in the medium run prices are affected negatively; banks on average set higher prices for consumers. In the long run the prices are affected positively for consumers and go down. This suggests that in the long run mergers create efficiency gains.

In a subsample, similar research was carried out, yet exclusively on large banks. For large banks the opposite effect was seen. For the large acquirers, prices go up in the long run

(27)

26

after going down in the medium run. Hence, mergers by large acquirers lead to worse prices for consumers. For large banks this means that following a merger they use their market power and increase prices in the long run. Sapienza (2002) also discusses the long run effect.

There are many papers that try to explain the effect of bank mergers on all sorts of products offered by the banks such as Sapienza (2002) and Focarelli and Panetta (2003). When the background literature to this question is examined, the question that arises is which financial theory prevails when analysing M&A deals in the banking sector. There are two leading theories that are dealt with in this paper. The first theory put forward is that mergers create increased market concentration and therefore lead to higher prices. The second theory is that mergers create gains in efficiency, which in turn enables banks to lower the product prices.

The first theory that states that market concentration increases prices is supported by empirical literature on the banking sector and loan contracts by Hannan (1991) and by Berger and Hannan (1989) on deposit rates. These empirical papers demonstrate a positive

relationship between market concentration and prices. Prager and Hannan (1998) analyse the effect of horizontal mergers and concentration on the deposit rates. They find a significant increase in price that can be explained by the increase in concentration and the increase in market power. Sapienza (2002) finds that horizontal mergers with small targets result in lower loan rates in Italy, but when the market share of the acquirer increases, the market power effect outweighs the efficiency gains effect. Sapienza also finds that after a merger there is a reduction in small loans offered. Similar to Berger (2004), Sapienza (2002) uses different subsamples that produce different outcomes. The difference in the results between the subsample for acquirer size and the subsample for in-market mergers is particularly marked.

This paper concludes that although the HHI and market concentration increase, the result is not the same for all subgroups. The HHI is of particular significance in the case of smaller banks and an increase in concentration results in prices going down. These results show that there is no support for the theory that an increase in market concentration causes prices to go down for either the small bank sample or the sample as a whole. For the large bank subsample on the other hand, the theories are supported. This outcome is in line with expectations as larger banks can use their position more easily than smaller banks.

The second theory deals with the efficiency gains that result from mergers. This theory asserts that a merger leads to lower prices in the long run. The two possible reasons for the price decline in the long run are synergies from redesigning the loan portfolio and risk diversification (Erel, 2011). Synergies result in different types of efficiency gains, namely:

(28)

27

cost, managerial and profit efficiency (Akhavein et al., 1997, Berger et al., 1999 and Walter, 2004). Focarelli and Panetta (2003) show that in Italy the effect of mergers on the price of deposit rates is negative, but in the long run the efficiency gains outweigh the market power effect. This leads to better prices for consumers in the long run. Erel (2011) discusses the effect of bank mergers on loan prices. In his research he finds that banks that merged between 1990 and 2000 in the US, on average, offer lower interest rates on consumer loans. This demonstrates that the efficiency gains overcome the market power effect and therefore favour the consumers. Erel (2011) finds that a decline in the spreads occurs in the first year and that the decline peaks in the second year. Erel shows that on average, mega acquirers have lower spreads in the first two years after the mergers, but they are positive in the third year. The idea behind this is that the largest banks attempt to increase their market share after the merger and therefore make price-cuts. In this paper the size of the acquirer has been taken into

consideration, because of the demand effect that can occur (Garmaise and Moskowitz, 2006). Even though the mortgage rate spread in the medium run goes up, the spread goes down in the long run. Therefore, the conclusion can be drawn that in the long run mergers cause the prices for consumers to go down on average.

Most papers use banks that did not undergo a merger as a control group (Ashton & Pham, 2007). This paper uses failed bank mergers as a control group. Another different approach has been chosen with respect to the size of the acquirer. The size of the acquirer influences many aspects of the bank (Sapienza, 2002). Therefore subsamples were created for large and small acquiring banks. The subsample of large banks is especially interesting since it yields opposite results.

This paper relies on a unique database, an alternative methodology and a traditional model. The HMDA database is incomplete for the purpose of this research. Not every bank included in this database reports the mortgage rate spread every year, causing gaps in the data. Moreover, the bank does not report the mortgage rate spread for each individual, only for some. Therefore the mortgage rate spread is probably not the best instrument to measure the effect of mergers on consumer prices. The methodology differs from previous research for the following reasons. The size of the merger is excluded because the dataset does not allow for this variable, there were too many missing values in Thomson One. The FICO measure of borrower risk is unavailable, as are loan covenants. These variables have therefore been replaced by income, median family income and size of the mortgage. Also the liquidity measure is an alternative to the defaulting loans measure. Therefore the results might include omitted variable bias, simultaneity and selection bias.

(29)

28

From a consumer’s perspective it would be beneficial if smaller banks merged in the US, as this results in a reduction in the mortgage rate spread. The opposite effect can be expected if large banks merge, as they could use their market power to increase prices. The regulatory authorities should consider the potential impact of large mergers and limit the potentially negative impact of such mergers. The conclusions that can be drawn for the US could possibly be extended to certain markets in Europe, for example to the Italian banking market. The Italian banking market is generally not concentrated, if measured with the HHI for each MSA. However, this does not apply to all European sub-markets. Further research is necessary before conclusions can be drawn on this subject.

This paper creates a foundation for future research in the form of a useful literature review, a transparently constructed dataset and a clear methodology. As the mortgage rate spread is not a commonly used variable to research the effect of M&A deals, this research lays the foundations for doing so. This paper highlights some of the obstacles and problems that might be encountered in future research. The methodology can be used and variables can be replaced and added to make a comprehensive analysis. When this is done, other

explanations might be found for the insignificant variable in the first year and more importantly the effect of bank mergers on the mortgage rate spread.

(30)

29

References

Akhavein, Jalal D., Allen N. Berger, and David B. Humphrey (1997). The effects of megamergers on efficiency and prices: Evidence from a bank profit function, Review of

Industrial Organization 12, 95-139.

Ashton, J. K., & Pham, K. (2007). Efficiency and price effects of horizontal bank mergers.,

working paper, University of East Anglia

Becher, D. A. (2000). The valuation effects of bank mergers. Journal of Corporate Finance,

6(2), 189-214.

Berger, A. N., Demirgüç-Kunt, A., Levine, R., & Haubrich, J. G. (2004). Bank concentration and competition: An evolution in the making. Journal of Money, Credit and Banking, 433-451.

Berger, Allen N., Rebecca Demsetz, and Phillip Strahan (1999). The consolidation of financial services industry: Causes, consequences and implications for the future, Journal of

Banking and Finance 27 (2), 404-31.

Berger, Allen N., and Timothy H. Hannan, (1989). The price-concentration relationship in banking, Review of Economics and Statistics 71, 291-299.

Berger, A. N., Hunter, W. C., & Timme, S. G. (1993). The efficiency of financial institutions: a review and preview of research past, present and future. Journal of Banking & Finance,

17(2), 221-249.

Bharath, S. T., Dahiya, S., Saunders, A., & Srinivasan, A. (2009). Lending relationships and loan contract terms. Review of Financial Studies, hhp 064.

Calabria, M. (2011). Fannie, Freddie, and the subprime mortgage market. Cato Institute. ISO 690

(31)

30

Cetorelli, N., Mandel, B. H., & Mollineaux, L. (2012). The evolution of banks and financial intermediation: framing the analysis. Federal Reserve Bank of New York Economic Policy Review, 18, 1-12.

Di Patti, E. B., & Gobbi, G. (2007). Winners or losers? The effects of banking consolidation on corporate borrowers. The Journal of Finance, 62(2), 669-695.

Erel, I. (2011). The effect of bank mergers on loan prices: Evidence from the United States.

Review of Financial Studies, 24(4), 1068-1101.

Focarelli, D., & Panetta, F. (2003). Are mergers beneficial to consumers? Evidence from the market for bank deposits. American Economic Review, 1152-1172.

Garmaise, M. J., & Moskowitz, T. J. (2006). Bank mergers and crime: The real and social effects of credit market competition. The Journal of Finance, 61(2), 495-538.

Hannan, T. H. (1991). Bank commercial loan markets and the role of market structure: Evidence from surveys of commercial lending. Journal of Banking and Finance, 15(1), 133-149.

Kahn, C., Pennacchi, G., & Sopranzetti, B. (2005). Bank Consolidation and the Dynamics of Consumer Loan Interest Rates. The Journal of Business, 78(1), 99-134.

Mortgage Bankers Association (2015), accessed 23rd of June 2015 <https://ycharts.com/indicators/mortgage_originations>

Penas, M. F., & Unal, H. (2004). Gains in bank mergers: Evidence from the bond markets.

Journal of Financial Economics, 74(1), 149-179.

Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. Review of financial studies, 22(1), 435-480.

(32)

31

Prager, R. A., & Hannan, T. H. (1998). Do substantial horizontal mergers generate significant price effects? Evidence from the banking industry. The Journal of Industrial Economics,

46(4), 433-452.

Sapienza, P. (2002). The effects of banking mergers on loan contracts. The Journal of

Finance, 57(1), 329-367.

U.S. Department of Justice and the Federal Trade Commission (2010). Horizontal Mergers Guidelines

Walter, I. (2004). Mergers and Acquisitions in Banking and Finance: What Works, What Fails, and Why. (Oxford University Press, New York).

Weiss, L. W. (1989). Concentration and price. Cambridge, Mass. : MIT Press.

Weston, J., & Strahan, P. E. (1996). Small business lending and bank consolidation: Is there cause for concern?. Current issues in Economics and Finance, 2(3).

(33)

32

The effect of M&A deals in the banking sector on the mortgage rate spread in the US