Banking market structure and its effects on real economic activity: evidence from a post-crisis US market.

(1)

Banking market structure and its effects on real economic activity: evidence from a post-crisis US market.

Master Thesis Finance

Xezonakis Emmanouil (s3300188) Supervisor: Dr Martien Lamers

June 2018

Abstract

This paper tries to investigate the relationship between the banking market structure in the U.S. and real economic activity measured by the change in the number of small and medium establishments (SMEs). There is extensive literature on how branching and bank competition influences the creation of new SMEs. The empirical data although suggest that there is a negative relation between SMEs on the one hand and the availability of bank offices and competition on the other. However, robustness tests confirm that if changes in banking market structure are significant enough, then the hypothesis and existing literature are confirmed.

(2)

2 1. Introduction

It is well established in the economic literature that more in-depth and broader financial markets are associated with better opportunities for financial growth. Since the early 1990s, the liberalization of the US banking sector in the 1960’s has brought forth a branch of economic literature focusing on the effects of banking structure on real economic activity. Most small and medium businesses are inherently dependent on the banking sector and since the financial crisis of 2007, which brought much volatility to the banking system not only in the US but the rest of the western world too, it is increasingly more critical to identify and document the relationship between SMEs and the banking industry than ever. So, how and how much does the banking industry affect real economic activity?

King and Levine (1993), exploring Schumpeter's (1911) idea that services provided by financial intermediaries are essential for technological innovation and economic development in a cross-country setting, find that 4 indicators of financial development are strongly and robustly correlated with growth, the rate of physical capital accumulation and improvements in the efficiency of capital allocation. Then, in 2005, Berger et al. find evidence that the presence of a bank branch in a local area helps induce economic activity and in 2006, Cetorelli and Strahan, further expand by saying that the banking sectors’

structure does affect real economic activity measured by the change in the number of small and medium establishments (SMEs). They find that in markets with less concentrated banking and looser restrictions on branching there are more firms in operation and the overall size distribution is leaning towards smaller firms. In more recent papers, Gilje et al (2016) and Hasan, Kozlowski, Jackowicz, Kowalewski (2017), examining the US and post-crisis Polish market respectively, find that branching helps economic growth by facilitating access to bank financing, lower financial costs and overall fund allocation from wealthier states to ones with greater need of funds.

There is strong support on the hypothesis that bank branching is beneficial to real economic activity, but what about the underlying structure of the banking sector, what about competition among banks? There are two contradicting views on how competition affects real economic activity in the existing literature. The competition-stability and the competition-fragility views were an increase/decrease in competition (respectively) leads to a more stable, less volatile, less distressed and more willing to extend credit banking

(3)

3

system. Boyd and DeNicolo (2005) and Akins, Li, Jeffrey and Rusticus (2016) find that a decrease in the competition results in banks earning greater rents by charging higher loan rates. These higher loan rates translate in a higher default probability on the borrower, who in turn increase theirs. On the other hand, Beck et al. (2013) show that an increase in competition will have a more significant impact on banks’ fragility in countries with more activity restrictions, less fragile systems, more established stock exchanges and more generous deposit insurance schemes. Interestingly, Martinez-Miela and Repullo (2010) find that an increase in competition potentially reduces bank failure and increases stability, but after a certain threshold, an increase in competitions acts as a destabilizing factor among banks.

Based on the literature mentioned before, in this paper I try to define the relationship between the change in the number of small and medium businesses under 500 employees and their size distribution (based on employment) and characteristics of the banking market and competition. More precisely I ask whether the concentration of the banking market and the availability of bank branches in an area influence the creation of small and medium enterprises (SMEs) and their overall size distribution. I contribute to the existing literature by testing this relation in the years during and after the US financial crisis (2006-2015), using a simple measure of bank concentration, Herfindahl–Hirschman Index (HHI), as an independent variable along with population density and unemployment as state-level controls.

To study the relationship between local banking market structures and real economic activity (measured by SMEs’ creation), I study the US market using aggregated data at the State level for the number of SMEs, the number of available bank branches and the measures of competition, all collected for the FDIC and FDIC Summary of deposits.

Moreover, I use data on population density, from the US Census Bureau and data on states unemployment rate taken from the Department of Labor and Training.

I document that small increases in the number of bank branches and competition in an area, contrary to existing literature, have a negative effect on real economic activity, meaning a decrease in the number of firms. This is not only the case for the overall number of firms under 500 employees, but also for the various size categories of firms based on employment size (under 20, between 20 and 99 and between 100 and 499 employees). For these size categories, similar to Cetorelli and Strahan (2006), I find that competition has no relation with SME creation in the group of 100 to 499 employees, meaning larger SMEs are not affected by concentration. For the state controls of

(4)

4

unemployment and density, I find that unemployment has a negative and strong relation with SMEs creation throughout the tests and for overall firms as long as for the size categories. Although density’s effect is insignificant (when controlling for state and year effects) throughout the sample, apart from the group of 20 to 99 employees where density has a strong and positive correlation with the dependent.

However, after including the squared term of the measure of offices variable and concentration, I find that although small changes in the underlying variables have results contrary to the hypothesis, when the changes are significant enough then the results confirm the theory and the hypothesis made.

The remainder of the paper is organized as follows. In section 2 I review the existing literature. In section 3 I present my hypothesis and methodology. In section 4 I present my data and variable definitions. In section 5 I discuss my empirical results, and in section 6 I present my concluding remarks and discuss the limitations of the paper alongside policy implications.

2. Literature Review

2.1 Introduction

Economic literature has greatly focused in the past decades on the importance and the overall role played by financial markets and institutions on real economic activity.

Increasingly the day to day of any organization, big or small, is dependent in some form on financial markets. Therefore, it is essential to test and understand the effect that such dependencies have on businesses. Since the late 1960’s the policies preventing interstate branching and acquisitions have largely been liberalized and this has led to the increasingly exciting literature, even today, of how does branching and banking market competition affect the US economy. It has been established by the works of King and Levine (1993), Berger et al. (2005) and Cetorelli and Strahan (2006) that branching has a positive relationship with the increase in real economic activity, although the literature is divided on the matter of how competition affects the stability of the market. So, two contradicting views have emerged on the literature, the competition-fragility, supported by the findings of Keely (1990) and more recently Beck et al. (2013), and the competition- stability view supported by Boyd and De Nicolo (2005) findings along with Cetorelli and

(5)

5

Strahan’s (2006) and more recently Akins, Li, Jeffrey and Rusticus (2016). These papers form the theoretical background and help in formulating the methodology I use to answer the question of how does the banking market structure affect real economic activity in the US market.

2.2 Branching and real economic activity.

In their 1993 paper, King and Levine, explore the idea of Schumpeter (1911), that services provided by financial intermediaries are essential for technological innovation and economic development, in a cross-country setting. They use data on over 80 countries between 1960 and 1989 and test whether increasing levels of financial development are correlated with greater current and future growth rates in the economy, capital accumulation and efficiency improvements. To do that, they define “financial development”, empirically, using four indicators and find that these indicators are strongly correlated with economic growth and the efficiency of capital allocation. In 2005, Berger et al. find that small local banks had an advantage over larger banks in processing

“soft” information and being able to provide more services to local firms, while larger more distant banks had a harder time assisting local businesses. Their paper provides some early evidence that local bank branches help induce economic activity on that area and helps raise the question of whether the presence of bank branches in local areas is a significant drive for real economic activity. Moreover, Cetorelli and Strahan (2006), further expand and support that idea by providing evidence that in markets with concentrated banking, potential entrants face greater difficulty gaining access to credit than in markets where banking is more competitive. By extension, they find that more relaxed restrictions on branching are associated both with more firms in operation and with a smaller size of the firm. They use data on SMEs under 1000 employees in nonfinancial sectors of the US market, primarily from the Census Bureau, along with data on banks and market concentration, mainly from the FDIC. They also find that in concentrated markets, banks have a greater incentive to “protect” the profitability of their existing clients; hence they do not provide as much credit to younger newly created firms resulting in greater barriers of entry for new firms in these concentrated markets.

Also, Rajan and Zingales (2004), argues that already established firms often strongly oppose financial openness, sometimes leading to long-term declines in a country’s growth prospects.

(6)

6

Along the lines of supporting bank branching, Gilje et al. 2016 make the case, after examining how shale booms in one county help increase lending on another county, that banks exposed to liquidity inflows in one area increase their mortgage lending in other non-exposed areas where the banks have branches. Thus, they find that branching helps in distributing wealth which yields a lot of benefits for local economies. In addition, they argue that deregulation of branching (similar to Cetorelli and Strahan, (2006)) has an important part to play in funds allocation and distribution. A role that is hard to substitute even in the most developed, integrated and technologically advanced lending markets.

Also, more recently, Hasan, Jackowicz, Kowalewski and Kozlowski (2017) investigate the effects between local banking and SMEs performance and ability to access debt in the post-financial crisis Polish market. They find, testing the effect cooperative banks have on SMEs, that a strong position of cooperative bank branches facilitates access to bank financing, lowers financial costs, boosts investment and favours growth for SMEs. Thus, yet again, strengthening the idea that branching is positively affecting local real economic activity.

2.2 Competition and its effects on bank market stability.

Competition and its effects on banking systems’ stability have been a subject that has divided the literature for many decades into two opposing views, the competition- stability and the competition-fragility. As the name suggests, the competition-stability related literature finds that increases in competition help stabilize the banking industry, while the competition-fragility related literature suggests the opposite.

In 1990, Michael C. Keely wrote a paper about the adverse effects of fixed deposit insurance schemes and he found that the increase in bank defaults from 1960 to 1980 was partly attributed to increases in competition (liberalization in restrictions about interstate bank branching and acquisitions). This is because, in more concentrated markets, bank charters would gain value from future rents the charter would gain due to its market power. Thus, banks had little incentive to compete when the gains from competing would be fewer than the losses on charter value. However, Boyd and De Nicolo (2005) argue that, ceteris paribus, a decrease in competition results in banks increasing their risk-taking and earning greater rents by charging higher loan rates. These higher loan rates result in increased risk taking on behalf of the borrower, thus decreasing the stability of the market. Also, Boyd, De Nicolo and Jalal (2006) (2009) find that there is a positive

(7)

7

relationship between bank competition and willingness to lend. Although, Martinez- Miera and Repullo (2010) show that, depending on the degree of correlation in the default rates across loans, the relation between competition and the risk of bank failure could become U-shaped as well. This means that increasing competition initially reduces bank failure, but beyond a certain point more competition has the opposite effect. On the other hand, Beck et al. (2013) findings support the competition-fragility view. They document that an increase in competition will have a more substantial impact on banks’ fragility in countries with “stricter activity restrictions, lower systemic fragility, better-developed stock exchanges, more generous deposit insurance and more effective systems of credit information sharing”. Though, a more recent paper from Akins, Li, Jeffrey and Rusticus (2016), examines the relationship between bank stability and competition in the financial crisis setting. Their findings are consistent with the predictions of De Nicolo (2005) and expand on the setting of the housing market concluding that greater banking competition is associated with higher financial stability.

3.Hypothesis and Methodology development 3.1 Hypothesis

My hypothesis follows the existing literature and it is based on my research question of, how does the US banking sector affect real economic activity. In broad terms, I want to test what and how significant is the effect that the market structure of the banking sector has on real economic activity measured by the creation of SMEs.

H1: Greater access to finance leads to higher economic activity

My first and foremost hypothesis, drawing from the existing literature, is that there should be a positive relation between branching and SME creation and a negative one between market concentration and SMEs. This first positive effect would confirm the existing theory and support the idea of branching deregulation. The secondary negative, in sign, effect would also be in line with existing theory confirming the fact that less competitive markets have a positive effect on SMEs creation. More concentration would mean an increase in the barriers affecting the entry and lower growth prospects for newly

(8)

8

created small and medium businesses, coming from an increase in the market power of individual banks and the adverse selection effects that come with it.

H2: Population density and unemployment measures are significant factors in affecting economic activity.

Other hypothesis relates mostly to the controls I am going to use. I expect a negative and significant relationship between the control variable of unemployment. I expect that a decrease in the rate of unemployment would have a positive effect on the creation of SMEs, as more people enter the workforce. Also, I hypothesize that there is a significant and positive relation between SME creation and population density. This would mean that in areas where the population is denser, like MSAs, there would be far more newly created firms than in more dispersedly populated areas.

3.2 Methodology

In order to test my hypothesis, I need data from various sources. The dataset consists, first of all, of data on the number of small and medium businesses from 2005 to 2015 for all 50 US states. Also, data on the number of available banks branched/offices and the Herfindahl–Hirschman Index (HHI) for each state in the sample period. Next, for the controls, I collect data on population density and reported population estimates per state and the unemployment figures to arrive at the DENSITY and UNEMPLOYMENT independent variables (I will explain this further in Section 4). I am testing the relationship between the dependent variables (the percentage changes in the number of SMEs) and the independent, control variables. The independent variables are, as proxies of a market structure shown in my estimation equation below, the logarithm of the number of existing offices in each state/year (LN.Offices), the Herfindahl–Hirschman Index (HHI) and the logarithm of the state’s population density and the unemployment rate. All relevant data is compiled into a panel series.

ΔSME’s(i,t) = α + β1 Market Structure + β2 Controls + εi,t

The more critical independent variables are the natural logarithm of the branches and the market concentration index (HHI). I expect to find a strong and positive relationship between branches and the creation of new firms, the percentage change in the number of SMEs, and a negative relationship between HHI and the creation of firms. The positive

(9)

9

and negative relations relate to the hypothesis because a strong positive branch relation would mean that bank presence enhances the creation of new firms and the negative relations on HHI would mean that less concentrated and less distraught markets, help with the creation of SMEs. These expectations should hold both in cross-section fixed and time fixed regression. For the other controls, population density and unemployment, I expect a positive relation from density and a negative one from unemployment, again for both cross and time fixed regressions. The positive sign on density would suggest that in fact the more densely populated areas have stronger economic activity and the negative sign on unemployment would intuitively mean that less unemployment leads to more economic activity in the form of new firm creation.

I regress the percentage changes in the number of SMEs on the independent variables controlling for either and both State and year fixed effects to capture any state and year variations. Moreover, I introduce the variable of HHI2, also explained in the data section as being the squared value of HHI. I do this to better capture the effects of HHI on the percentage change of SMEs.

4.Data & Summary

I collect data on the number of SMEs (small and medium businesses) operating in any state i in years 2005 to 2015 form the United States Census Bureau along with data on population estimates in the same time period. Moreover, the FDIC and FDIC's Summary of Deposits provide data on the number of bank branches present at any state and measures of how concentrated a state is, namely the Herfindahl–Hirschman Index (HHI).

Finally, data on states unemployment are collected for the Department of Labor and Training webpage.

Data on the SMEs consists of data on the overall number of established firms with under 500 employees in each of the 50 US states between 2005 and 2015. Then these numbers are also sorted into three subcategories which are firms with less than 20 employees, firms with 20 to 99 employees and firms with 100 to 499 employees. This allows me to later test if there are different effects depending on the size of the SMEs.

Data on banks branches/offices consists of data on the number of branches in each state from 2005 till 2014 (10 years) and the HHI index for each state between 2005 and 2014. I do not calculate the HHI, but instead, I use the HHI reported in the FDIC's cohesive

(10)

10

market share tables. For the purposes of making the data more presentable and more comfortable to manage, the HHI is scaled down by a factor of 10000.

Other variables used in my regressions are: Density which is the population density in each state for the period of 2006 to 2015. It is collected from the United States Census Bureau and it is used as a control variable, and Unemployment which is the unemployment rate in each state in the same period.

4.1 Variable definitions and Summary statistics

Dependent variables

The primary dependent variable is TOTAL<500, which represents the percentage change in the overall number of firms under 500 employees in any state i between 2006 and 2015. Then the total number of firms is broken down into three more categories which also function as dependent variables. These are SMEs<20, SMEs 20-99 and SMEs 100-499. All of these, as the name suggests, represent the percentage changes in the number of firms with less than 20, between 20 and 99 and between 100 and 499 employees respectively.

Independent variables

The primary two independent variables used in the regressions are Ln(Offices) and HHI.

The HHI is straightforward. It represents the Herfindahl–Hirschman Index for each of the 50 states for each of the 10 years in question (2005 to 2014) and as said before these numbers are scaled down by a factor of 10,000. I also introduce the squared version of the HHI (HHI2) to capture the effects of HHI on the sample. Ln(Offices) is the logarithm of the number of existing offices/branches of all the banks in each state between 2005 and 2014. This variable is transformed to a logarithm due to some extreme values (for example in some states the number of offices exceeds 6,000, while in others it is only 130 as Table 1 shows)

Controls

I include a set of 2 control variables at the state level, UNEMPLOYMENT and DENSITY.

UNEMPLOYMENT is a series of the unemployment rate in each state between 2006 and 2015 and controls for states business cycles. DENSITY represents the density factor for each state in the 10 years in question and controls for the fact that there are substantial

(11)

11

differences in population density between states. The variable is transformed to a logarithm, like Ln(Offices), due to those extreme differences between states density (i.e.

in some states population density goes as high as 1195 while in others like Alaska it is as low as 1.1).

Summary statistics

Looking at the summary statistics in Table 1 of the dependent variables, it can be seen that on average there was a slight decrease in the number of firms with fewer than 500 employees of -0.3% with a standard deviation of 1.6%, meaning that in the period between 2006 to 2015 the average number of small and medium firms was roughly the same. Firms with lower than 20 employees had the same statistics, -0.3% decrease and a 1.5% standard deviation. This is due to the fact that in the sample of SMEs below 500 employees, firms with less than 20 employees represent about 90% of the sample.

Although looking at firms with 20 to 99 and 100 to 499 employees, there was a slight increase in their overall numbers, 0.1% and 0.8% with a standard deviation of 3% and 3.1% respectively. Since the percentage changes and the standard deviations on the four dependent variables are too small, nothing much can be said about the overall change in the number of small and medium US firms other than the fact that the numbers remained roughly the same throughout the sample period.

Moving to the independent variables, on average there were 1919 bank branches in each state in the US with an HHI of 0.11. This means that on average the US banking sector is well established and not concentrated since the average HHI is below 0.18. Although there are states like Alaska, Nevada, South Dakota, Delaware, Hawaii, Minnesota, North Carolina and Rhodes Island that are noticeably more concentrated than others, with the first 3 having HHI of more than 0.3 and as high as 0.64 for South Dakota. The rate of UNEMPLOYMENT was, between 2006 and 2015, on average at 6.4% and in some states, like Michigan, it was as high as 13.7% and in Utah and Hawaii as low as 2.6% with a standard deviation of 2.6%. DENSITY is on average 188 people per square mile with an upper bound of 1195, for New Jersey after 2010, and a lower bound of 1.1 for Alaska between 2000 and 2009.

(12)

12 Table 1 Summary Statistics

This table reports the summary statistics for data used in my panel regressions. Data on SMEs (small and medium businesses) are collected from the Census Bureau. Offices refer to the number of bank branches in a state. HHI is the Herfindahl–Hirschman Index for a state (>0.18 and the state is considered concentrated). UNEMPLOYMENT refers to the unemployment rate of a state in a particular year. DENSITY is the logarithm of the density indicator for each state in each year for 2006 to 2015. TOTAL<500 is the percentage change in the total number of SMEs in a state, SMES<20, SMES 20-99, SMES 100-499, are the percentage changes in the number of SMEs in specific classes in each state (classes are <20, 20 to 99, 100 to 499. Classes are based on the number of employees at the SME)

Mean Median Maximum Minimum Standard

Deviation

Offices 1919 1541 7401 130 1691.07

HHI 0.110 0.084 0.640 0.016 0.092

DENSITY 188.42 97.35 1195.5 1.100 253.2

UNEMPLOYMENT 0.064 0.062 0.137 0.026 0.021

TOTAL<500 -0.003 -0.002 0.056 -0.052 0.016

SMES<20 -0.003 -0.003 0.055 -0.044 0.015

SMES 20-99 0.001 0.007 0.089 -0.112 0.031

SMES 100-499 0.008 0.009 0.104 -0.096 0.030

5. Empirical Results

I start my empirical investigation on the relationship between local banking market structure and the percentage change on small and medium enterprises using a simple test on the creation of new firms. In Table 2, I show whether an increase in the number of small and medium businesses under 500 employees, is correlated with State and banking market traits.

In Table 2,3,4 and 5, column (1) and (5) present the results as a pooled sample, while columns (2) and (6), (3) and (7), (4) and (8) present the results with State fixed effects, Year fixed effects and State-Year fixed effects respectively.

The variable representing the natural logarithm of the number of offices in each state (LN_OFFICES) is significant in all eight columns, whether there are or not any fixed effects present. Although, the sign changes depending on year or states fixed effects in play.

When looking at the cross-section in columns (2) and (6), the sign is negative meaning

(13)

13

there is a negative relationship between the increase in the number of firms and the number of available bank branches. This means that by looking between states, the ones with less available branches are the ones with a higher number of firms in creation.

Counter to existing theory, it seems that between states, branching has a negative effect on real economic activity. On the other hand, in columns (3) and (7), where I use year fixed effects, there is a positive correlation between branching and the creation of SMEs.

This confirms the original hypothesis that branching positively affects economic activity, although only in between year estimations. Although the coefficients on LN_OFFICES are statistically significant, their economic significance is quite small. Looking at column (8), a 1% increase in the number of bank offices will result in a 0.055% decrease in the number of SMEs. The percentage change in SMEs is too small to be considered significant in real terms.

The measure of concentration in the banking sector, the HHI variable, is mostly significant apart from column (2) and (4). Throughout the regressions, whether including or not State and/or year fixed effects, it seems that concentration has a positive correlation with SMEs creation. An increase in the concentration in any state at any year increases the number of newly created SMEs. To confirm the effect, I include in the regressions a variable named HHI2, which is the squared term of HHI. From columns (5) to (8) HHI2 is negative and significant and leads to the fact that small increases in the HHI (increases in concentration) positively affect economic activity, although after a certain point of more sigrnificant increases, the effects are reversed. It appears the results reside at the upper slopping part of the reverse U-shaped diagram between HHI (y-axis) and SMEs number increase (x-axis). This result is in line with what Martinez-Miera and Repullo (2010) find in their paper that Increasing competition initially reduces bank failure (as noted by Boyd and De Nicolo (2005)) hence increasing stability, but beyond a certain point, competition increases bank failure. This means that the U.S. market is competitive enough that a decrease in competition will have a beneficial effect, in my case, on SMEs.

As for the economic significance that the coefficients of HHI have, by looking at column (8), a 1% increase in concentration results in a 0.07% increase in SMEs. Again, the effect is quite small to be considered economically significant for small changes in concentration.

Moving to State level controls, the variables measuring the natural logarithm of population density, DENSITY, is insignificant only when both fixed effects are introduced.

It has a highly positive effect on SMEs in between state estimations, confirming that states

(14)

14

with higher population density have higher levels of economic activity, but a negative and rather insignificant economic effect in fixed-year estimations.

The variable representing the unemployment rate, UNEMPLOYMENT, is persistently negative and greatly significant in all estimations. This confirms my hypothesis that an increase in the unemployment rate, either between a year or state or both, has a negative correlation with real economic activity. Moreover, the coefficients on the variable do have economic importance, since a 1% increase in unemployment results in about 0.4%

decrease in the number of SMEs. This is a significant decrease considering economically significant events like the U.S. mortgage crisis, where the volatility in the unemployment rate increases dramatically.

Summarizing Table 2, when state and year effects are present, branching has a negative effect on SMEs creation, concentration has a positive effect, density’s effect is insignificant and unemployment, as expected, has a negative correlation. Although statistically significant, the economic importance of the coefficients, apart from UNEMPLOYMENT, is relatively small even for large-scale changes on the underlying variables.

5.1 Robustness tests

Table 3,4 and 5 report my regressions result for the log of the percentage change in the number of small and medium firms split into three categories defined by employment size. The SMEs under 500 employees are divided into: SMEs with less than 20 employees (SMEs_20), SMEs with 20 to 99 employees (SMEs_20-99) and SMEs with 100 to 499 employees. This allows me to test if there are different effects in the independent variables and controls on the overall size distribution of SMEs in the U.S. market from 2006 to 2015.

The results in Table 3 resemble the ones in Table 2, although this time SMEs_20 is the dependent variable. This is due to the fact that almost 90% of the SMEs in the sample are SMEs with fewer than 20 employees. Looking at column (6), where State fixed effects are present, I find the same relationship between the dependent and the independent variables as in Table 2 again, and also in column (7) where I have time fixed effects.

Moreover, column (8) variable coefficients have the same sign as column (8) in Table 2, although the significance is lower, apart from UNEMPLOYMENT. This means that an increase in the number of available bank branches shrinks the number of SMEs in a state,

(15)

15

a decrease in competition positively affects SMEs creation (although a significant increase in HHI after a certain point, decreases the number of SMEs as shown by the negative relation between HHI and HHI2, in line with Martinez-Miera and Repullo (2010) findings).

Population density is significant in between state or year estimations but not when both are present, meaning that density’s effect overall is zero.

Table 4 shows the results of the regressions where SMEs with 20 to 99 employees are a dependent variable. In column (6), HHI is insignificant, meaning that when looking at States, concentration has no particular effect on SMEs with 20 to 99 employees. In column (8) the signs on the coefficients are the same as in Table 2 column (8), meaning that opposite to existing theory branching has a negative and concentration has a positive effect, although this time density plays a significant and positive role in the creation of

“medium” sized SMEs.

Looking at the last division of SMEs, the small and medium businesses with 100 to 499 employees, in Table 5, surprisingly in columns (6) to (8) the HHI does not correlate with the dependent variable. This means that larger SMEs are not at all affected by the competition in the banking market. This may come from the fact, also mentioned by Cetorelli and Strahan (2006), that larger firms are not affected by the concentration in their existing market partly because they can get financed from other markets (other states) and party because they might not depend on banks as much as smaller firms do.

Summarizing my finding from Tables 3 to 5, I document that SMEs with fewer than 20 employees are affected by the independent variables the same as SMEs with fewer than 500 employees. The statistical significance of the results is weaker, and the economic significance follows the one on Table 2, where all changes in the underlying variables have minimal effects on the dependent, apart from UNEMPLOYMENT. In Table 4, the economic importance of the coefficients, although not DENSITY, is far greater than Table 3. A one percent increase in LN(Offices), HHI and UNEMPLOYMENT causes a 0.13% decrease, 0.22% increase and an almost 0.7% decrease in the number of “medium” SMEs respectively. Finally, the results on Table 5, concerning competition, are in line with Cetorelli and Stahan (2006) where measures of competition do not affect larger SMEs, while the measure of branching has a statistically and economically negative effect of 0.13%.

(16)

16 Table 2 Regression outputs

Each column in this table reports the coefficients on the independent variables used in the regressions. The dependent variable in this table is the total percentage change in the number of SMEs under 500 employees (TOTAL< 500). The fixed effects used in each of the 8 regressions are reported with a “yes” or “no” on the thirteenth and fourteenth row of the table. Ln_Offices represent the natural logarithm of the number of bank offices in a state for each year (2005-2014). The Herfindahl-Hirschman Index (HHI) is the measure of concentration, and HHI2 represent the squared term of HHI. DENSITY is the natural logarithm of the population density in each state in a year. UNEMPLOYMENT is the unemployment rate for each of the US states in each year. Standard errors are presented in parentheses, and * indicate the statistical significance of the coefficient (***, **, * represent a 1%, 5% and 10% level significance respectively).

(1) (2) (3) (4) (5) (6) (7) (8)

LN_OFFICES 0.004*** -0.124*** 0.004*** -0.054*** 0.005*** -0.126*** 0.005*** -0.056***

(0.001) (0.024) (0.001) (0.019) (0.001) (0.024) (0.001) (0.019)

HHI 0.033*** 0.028 0.019*** 0.014 0.098*** 0.101** 0.082*** 0.079**

(0.008) (0.018) (0.005) (0.014) (0.022) (0.042) (0.015) (0.032)

HHI2 -0.140*** -0.122** -0.134*** -0.106**

(0.044) (0.063) (0.030) (0.048)

DENSITY -0.002*** 0.134*** -0.002*** 0.013 -0.002*** 0.132*** -0.003*** 0.014

(0.001) (0.010) (0.000) (0.013) (0.001) (0.010) (0.000) (0.013)

UNEMPLOYMENT -0.343*** -0.441*** -0.210*** -0.391*** -0.359*** -0.436*** -0.241*** -0.384***

(0.030) (0.040) (0.032) (0.056) (0.030) (0.040) (0.032) (0.056)

No. Observations 500 500 500 500 500 500 500 500

State fixed effects no yes no yes no yes no yes

Year fixed effects no no yes yes no no yes yes

Adjusted R-squared 0.23 0.52 0.64 0.73 0.24 0.52 0.65 0.73

(17)

17 Table 3 Regression output

Each column in this table reports the coefficients on the independent variables used in the regressions. The dependent variable in this table is the percentage change in the number of SMEs under 20 employees (SMES< 20). The fixed effects used in each of the 8 regressions are reported with a “yes” or “no” on the thirteenth and fourteenth row of the table. Ln_Offices represent the natural logarithm of the number of bank offices in a state for each year (2005-2014). The Herfindahl-Hirschman Index (HHI) is the measure of concentration, and HHI2 represent the squared term of HHI. DENSITY is the natural logarithm of the population density in each state in a year. UNEMPLOYMENT is the unemployment rate for each of the US states in each year. Standard errors are presented in parentheses, and * indicate the statistical significance of the coefficient (***, **, * represent a 1%, 5% and 10% level significance respectively).

(1) (2) (3) (4) (5) (6) (7) (8)

LN_OFFICES 0.004*** -0.112*** 0.004*** -0.044** 0.005*** -0.113*** 0.005*** -0.046**

(0.001) (0.024) (0.001) (0.020) (0.001) (0.024) (0.001) (0.020)

HHI 0.032*** 0.029 0.020*** 0.010 0.095*** 0.101** 0.082*** 0.064*

(0.008) (0.018) (0.005) (0.014) (0.021) (0.043) (0.015) (0.033)

HHI2 -0.135*** -0.118** -0.134*** -0.089*

(0.043) (0.064) (0.031) (0.049)

DENSITY -0.002*** 0.119*** -0.002*** 0.008 -0.002*** 0.117*** -0.003*** 0.008

(0.001) (0.010) (0.000) (0.013) (0.001) (0.010) (0.000) (0.013)

UNEMPLOYMENT -0.295*** -0.374*** -0.189*** -0.352*** -0.310*** -0.369*** -0.220*** -0.346***

(0.030) (0.041) (0.032) (0.057) (0.030) (0.040) (0.032) (0.057)

No. Observations 500 500 500 500 500 500 500 500

(18)

Each column in this table reports the coefficients on the independent variables used in the regressions. The dependent variable in this table is the percentage change in the number of SMEs between 20 and 99 employees (SMEs 20-99). The fixed effects used in each of the 8 regressions are reported with a “yes” or “no” on the thirteenth and fourteenth row of the table. Ln_Offices represent the natural logarithm of the number of bank offices in a state for each year (2005-2014). The Herfindahl-Hirschman Index (HHI) is the measure of concentration, and HHI2 represent the squared term of HHI. DENSITY is the natural logarithm of the population density in each state in a year. UNEMPLOYMENT is the unemployment rate for each of the US states in each year. Standard errors are presented in parentheses, and * indicate the statistical significance of the coefficient (***,

**, * represent a 1%, 5% and 10% level significance respectively).

(1) (2) (3) (4) (5) (6) (7) (8)

LN_OFFICES 0.005*** -0.232*** 0.003*** -0.130*** 0.007*** -0.235*** 0.005*** -0.137***

(0.002) (0.051) (0.001) (0.038) (0.002) (0.051) (0.001) (0.038)

HHI 0.043*** 0.008 0.021** 0.037 0.145*** 0.142 0.105*** 0.217***

(0.016) (0.038) (0.009) (0.027) (0.044) (0.090) (0.027) (0.064)

HHI2 -0.218** -0.222 -0.178*** -0.296***

(0.088) (0.136) (0.054) (0.095)

DENSITY 0.000 0.280*** -0.002** 0.066*** -0.001 0.276*** -0.002*** 0.067***

(0.001) (0.022) (0.001) (0.025) (0.001) (0.022) (0.001) (0.025)

UNEMPLOYMENT -0.707*** -0.967*** -0.365*** -0.707*** -0.731*** -0.959*** -0.406*** -0.686***

(0.060) (0.085) (0.056) (0.112) (0.061) (0.085) (0.057) (0.111)

No. Observations 500 500 500 500 500 500 500 500

(19)

Each column in this table reports the coefficients on the independent variables used in the regressions. The dependent variable in this table is the percentage change in the number of SMEs between 100 and 499 employees (SMEs 100-499). The fixed effects used in each of the 8 regressions are reported with a “yes” or “no” on the thirteenth and fourteenth row of the table. Ln_Offices represent the natural logarithm of the number of bank offices in a state for each year (2005-2014). The Herfindahl-Hirschman Index (HHI) is the measure of concentration, and HHI2 represent the squared term of HHI. DENSITY is the natural logarithm of the population density in each state in a year. UNEMPLOYMENT is the unemployment rate for each of the US states in each year. Standard errors are presented in parentheses, and * indicate the statistical significance of the coefficient (***, **, * represent a 1%, 5% and 10% level significance respectively).

(1) (2) (3) (4) (5) (6) (7) (8)

LN_OFFICES 0.004*** -0.183*** 0.003** -0.123*** 0.005*** -0.182*** 0.003** -0.124***

(0.002) (0.053) (0.001) (0.047) (0.002) (0.053) (0.001) (0.047)

HHI 0.023 0.042*** 0.007 0.069** 0.057 -0.016 0.019 0.100

(0.015) (0.040) (0.011) (0.033) (0.042) (0.095) (0.032) (0.079)

HHI2 -0.074 0.096 -0.028 -0.051

(0.083) (0.143) (0.064) (0.117)

DENSITY -0.002** 0.138 -0.004*** 0.038 -0.003** 0.139*** -0.004*** 0.039

(0.001) (0.023) (0.001) (0.031) (0.001) (0.023) (0.001) (0.031)

UNEMPLOYMENT -0.699*** -0.878*** -0.291*** -0.527*** -0.707*** -0.882*** -0.297*** -0.524***

(0.057) (0.090) (0.066) (0.136) (0.058) (0.090) (0.068) (0.137)

No. Observations 500 500 500 500 500 500 500 500

(20)

20 6.Conclusions

I find that less competition in the local US market, meaning more concentration and less available bank branches, increases the overall number of small and medium businesses.

Density has negative and positive correlations with SME creation when year or state fixed effects are included respectively, but when both are active then density is insignificant.

Although branching and competition have different effects from the hypothesis made in this paper and existing literature, the variable of unemployment does confirm the theory and the hypothesis in having a strong and negative correlation with SMEs creation.

After testing the effects of banking market structure on the size distribution of SMEs I document that all 3 categories of SMEs, based on employment numbers, are negatively and significantly affected by branching and unemployment alone. While density only significantly affects SMEs with 20 to 99 employees. Concentration, measured but HHI, positively relates to SMEs creation in the smaller and medium SMEs (less than 20 and 20 to 99 employees) when it does not affect “larger” 100 to 499 employees firms. So, an increase in the number of bank offices would shift the size distribution more towards smaller SMEs, while an increase in concentration would have a more significant positive impact on medium SMEs leaving small and large SMEs number mostly unaffected.

Although the primary results do not confirm all the hypothesis made, after including the squared terms of LN Offices and HHI in the regressions, I find that large enough increases on both underlying variables would result in the effects predicted by the literature. This deviation in my results and the literature can be attributed to the available data pool. I use aggregated data on all business sectors regardless of bank dependence when calculating the number of SMEs and aggregated data when measuring the number of existing bank branches. Moreover, I mainly focus on the period after the financial crisis that hit the U.S. market while most existing literature focuses on earlier time frames.

These are the main limitation of the paper.

My findings, although small in economic significance, have some policy implications.

They suggest that the U.S. banking market is not concentrated enough, meaning that competition is so intense that it adverse effects on real economic activity since an increase in concentration has a positive effect on SMEs creation. Further research needs to be conducted to test the effects of competition on economic activity after the crisis of 2007, preferably at a county level and in a disaggregated data level.

(21)

21 References

Akins, B., Li, L., Ng, J., Rusticus, T.O., 2016. Bank competition and financial stability:

evidence from the financial crisis. Journal of Financial and Quantitative Analysis 51(1), pp.

1-28.

Beck, T., De Jonghe, O., Schepens, G., 2013. Bank competition and stability: Cross-country heterogeneity. Journal of Financial Intermediation 22(2), pp. 218-244.

Berger, Allen, Nathan Miller, Mitchell Petersen, Raghuram Rajan, and Jeremy Stein, 2005, Does function follow organizational form? Evidence from the lending practices of large and small Banks, Journal of Financial Economics 76, 237–269.

Boyd, J.H., Gianni De Nicolò, and Abu M. Jalal, 2006. Bank Risk-Taking and Competition Revisited: New Theory and New Evidence. IMF Working Paper

Boyd, J.H., Gianni De Nicolò, and Abu M. Jalal, 2009. Bank Competition, Risk, and Asset Allocations. IMF Working Paper

Boyd, J.H., Gianni De Nicolò, G., 2005. The theory of bank risk-taking and competition revisited. J. Finance 60 (3), 1329–1343.

Cetorelli, N., Strahan, P.E., 2006. Finance as a barrier to entry: Bank competition and industry structure in local US markets. The Journal of Finance 61(1), pp. 437-461.

Gilje, E.P., Loutskina, E., Strahan, P.E., 2016. Exporting liquidity: Branch banking and financial integration. The Journal of Finance 71(3), pp. 1159-1184.

Hasan, I., Jackowicz, K., Kowalewski, O., Kozłowski, Ł., 2017. Do local banking market structures matter for SME financing and performance? New evidence from an emerging economy. Journal of Banking & Finance 79, pp. 142-158.

Martinez-Miera, D., Repullo, R., 2010. Does competition reduce the risk of bank failure?

Rev. Finan. Stud. 23 (10), 3638–3664.

Michael C. Keeley, 1990. Deposit Insurance, Risk, and Market Power in Banking. The American Economic Review, Vol. 80, No. 5 (Dec., 1990), pp. 1183-1200

Rajan, Raghuram G., and Luigi Zingales, 2004, Saving Capitalism from the Capitalists:

Unleashing the Power of Financial Markets to Create Wealth and Spread Opportunity.

(22)

22

Robert G. King and Ross Levine 1993. Finance and Growth: Schumpeter Might be Right.

The Quarterly Journal of Economics, Vol. 108, No. 3 (Aug., 1993), pp. 717-737.

Schumpeter, Joseph A., 1912, Theorie der Wirtschaftlichen Entwicklung. Leipzig: Dunker

& Humblot (The Theory of Economic Development, translated by Redvers Opie, 1934.

Harvard University Press, Cambridge, MA).