J Econ Finan DOI 10.1007/s12197-011-9189-4

The effect of banking market structure on the volatility of growth of manufacturing sectors in developing countries Indrit Hoxha

© Springer Science+Business Media, LLC 2011

Abstract I explore the effect of banking concentration and banking competition on the volatility of the growth of value added of manufacturing sectors in the developing countries. In this paper, I bring together two strands of literature, one that discusses the effect of financial intermediation on volatility of growth and another one that discusses the effect of banking concentration and competition on credit access. Following the industrial organization literature, I look at the effect of banking competition and banking concentration on the volatility of manufacturing sectors separately. I find that banking concentration has a dampening effect on the volatility of growth of the industries. On the other hand, I find that as banking competition increases, the volatility of the growth of industries increases, also. Keywords Banking · Concentration · Competition · Value Added · Manufacturing Sectors · Volatility JEL Classification D4 · G21 · L11 · O16

1 Introduction The effect of financial intermediaries on real output has generated a heated debate in the literature. In the last decade, a consensus has been reached that higher financial intermediation facilitated higher economic growth, where Rajan and Zingales (1998) have put a cornerstone in the literature. To evaluate

I. Hoxha (B) School of Business Administration, Penn State University Harrisburg, 777 W. Harrisburg Pike, Middletown, PA 17054, USA e-mail: [email protected]

J Econ Finan

the effect of financial intermediaries on real output another important issue is to look whether an increase in financing through financial intermediaries has any effect on the volatility of output. Based on this debate, in this paper, I study the effect of banking competition and banking concentration on the volatility of output at industrial level using data from a group of developing countries. I find empirical evidence that banking competition increases the volatility of the growth of manufacturing sectors, while banking concentration reduces their volatility. From the theoretical point of view, the effect of banking development on the volatility of output is ambiguous. Morgan et al. (2004) suggest that improved access to banking finance allows firms to smooth out their idiosyncratic shocks. However, the effect of banking development on volatility of economic growth can be affected by the stage of the development of the country (Aghion et al. 2004), the type of shocks that the economy faces, such as monetary or real shocks (Bacchetta and Caminal 2000), or whether the economy faces credit demand versus credit supply shocks (Morgan et al. 2004). No consensus has been reached on the effect of banking development on volatility of economic growth in the empirical studies, also. Denizer et al. (2002) find that countries with more developed financial sectors experience less fluctuation in the growth of real per capita output, consumption and investment. On the other hand Easterly et al. (2001) show that between financial development (measured as credit to private sector) and aggregate volatility there is a U-shaped relationship. Comin and Philippon (2005) suggest that the effect of financial development on the volatility of growth depends on aggregation level, and show a positive relationship of financial development to firm level volatility. On the other hand, Correa and Suarez (2008) exploiting the staggered timing of state-level bank deregulation show a negative relationship of US banking deregulation and firm-level volatility. While there is a discussion about the effect of financial development on the volatility of economic growth, the main focus of this paper is to look at the effects of banking competition and banking concentration on the sectoral-level volatility. The novelty of this study is that it brings together two strands of the literature. One strand of the literature looks at the effect of financial intermediation on volatility of economic growth and the second one looks at the effect of banking concentration and banking competition on the credit access. There is another debate on the effect of banking concentration and banking competition on the credit access of firms. Theoretically, on one side Pagano (1993) and Guzman (2000) show that the markets where banking sector is concentrated and less competitive, grow less than their best potential, because firms do not get access to credit. On the other side, Petersen and Rajan (1995) suggest that only in concentrated banking systems, banks have an incentive in investing in relationship banking, which leads to more credit access. Marquez (2002) suggests that an increase in banking competition causes banks to screen firms less rigorously, which worsens the pool of the customers and leads to higher interest rates, resulting in a decreased amount of credit made available to manufacturing sectors.

J Econ Finan

Many empirical studies have shown a relationship between banking market structure and the economic growth. Berger et al. (2004b) and Cetorelli and Strahan (2006) show that higher concentration and more restrictions on competition lead to less new firm creation, and less economic growth. Claessens and Laeven (2005) using a cross-section estimation method for bank competition, find that banking competition is important for the growth of industries dependent on external finance. On the other side, Cetorelli and Gambera (2001) in a cross country study find that industries in need of external finance grow faster in countries where there is more bank concentration, while Mitchener and Wheelock (2010) using historical data for US states find the industries have grown faster in states with higher banking concentration. In a panel study for a sample of 36 countries, Hoxha (2011) finds that industries in need of external finance perform better in markets with more concentrated banking sector, and worse in the markets with higher banking competition.1 The identification strategy used in the study is a similar approach to Larrain (2006) and Raddatz (2006), who find that industrial level volatility decreases with more private credit and financial development. However, the main focus of my study is on banking competition and banking concentration, instead of financial development measured as private credit as a share to GDP. In the literature banking concentration has been used as a measure of banking competition; however, I examine them separately, because they measure two different things. Banking concentration measures the share of the market that is controlled by the largest banks, but does not necessarily measure the competitive environment in the banking sector in a country. There are cases where a nearly competitive environment exists in markets with two or three banks, or where a substantially non-competitive environment has been observed in markets where there are thousands of suppliers, such as the credit card market (Shaffer 2004). To investigate the effect of banking competition on the volatility of growth of sectoral level output, I look at the main effect of banking competition and the effect that it has on specific industries. I measure competition with an index estimated as the sum of the elasticities of bank revenue to input prices. I find that banking competition has a positive effect on the volatility of growth of all industries, and in addition to that, it increases the volatility of industries with higher investment opportunities even more. Using a similar approach, I investigate the effect of banking concentration on the volatility of growth of sectoral level output. In order to measure banking concentration, I use the five-bank concentration ratio and Herfindahl– Hirschman index. I find that the more concentrated is the banking sector the lower is the volatility of growth of manufacturing sectors. However, I do not find enough supportive evidence to show that industries with higher investment opportunities, have a significant reduction in volatility of growth.

1 For

a detailed information on this debate, please refer to Berger et al. (2004a) and Hoxha (2011).

J Econ Finan

The results of this study show that a higher banking concentration, and/or a lower banking competition would result in lower volatility of growth of industries. These results are in accordance with the previous literature. Many studies (Petersen and Rajan 1995; Cetorelli and Gambera 2001; Zarutskie 2006; Bonaccorsi di Patti and Dell’Ariccia 2004) show that more banking concentration and/or less banking competition leads to more credit access for firms. On the other hand, Larrain (2006) and Raddatz (2006) show that more private credit dampens the volatility of growth. The rest of the paper is organized as follows. A description of data and the estimation of banking competition and banking concentration indices is given in Section 2. In Section 3, I describe the methodology used in the paper. Sections 4–6, explain the empirical results using banking competition, banking concentration and both of them, respectively. Some robustness tests are shown at Section 7 and Section 8 concludes the paper.

2 Data I use firm (bank)-level, industry-level and country-level data from different sources in this paper. In this section, I explain their sources, characteristics and the way I estimated banking concentration and banking competition indices. A summary of the data can be found at Table 1. 2.1 Data on industries Data on value added for each industry are obtained from Industrial Statistics Database (INDSTAT4 2008) which is collected by United Nations Industrial Table 1 Descriptive statistics Variable

Observations

Mean

Std. dev.

Min

Max

Std. dev. of growth of VA of industries Average VA of industries 5-Bank concentration ratio Herfindahl–Hirschman index Competition index Investment opportunity Log of population Log of GDP per capita Trade openness Government spending/GDP

508 508 508 508 508 508 508 508 508 508

0.27 19.59 0.69 0.15 0.60 0.01 3.05 8.05 0.75 0.16

0.22 2.01 0.11 0.05 0.17 0.02 1.67 0.94 0.38 0.04

0.03 13.42 0.50 0.07 0.27 –0.02 –0.24 6.01 0.22 0.10

2.47 24.41 0.92 0.24 0.91 0.06 6.95 9.38 1.50 0.21

Notes: This table reports the descriptive statistics of the main regressions. Value added data is from Industrial Statistics Database (INDSTAT4 2008) collected by UNIDO and it is classified by ISIC. 5-Bank concentration ratio index, Herfindahl–Hirschman Index and Competition index are author’s estimations using the methods explained in detail in data section. The bank-level data used to estimate these indices are from Bankscope. Investment Opportunity measure comes from Ciccone and Papaioannou (2006) and it is estimated as the growth rate of capital stock for each industry using US data for 1980’s. Population, GDP per capita, trade openness, and government spending as share of GDP come from International Financial Statistics (IFS)

J Econ Finan

Development Organization (UNIDO). The industries are classified according to Revision 3 of the International Standard Industrial Classification of All Economic Activities (ISIC). The manufacturing sector data is arranged at 3 and 4 digits of ISIC codes, and include detailed information for 151 industries. The data is originally in national currency and current prices; however, I adjust them by deflating and converting all of them into US dollars. 2.1.1 Investment opportunity The investment opportunity data for each industry comes from Ciccone and Papaioannou (2006). They use US industry data from the NBER Manufacturing Database (Bartelsman and Gray 1996). They proxy investment opportunity with industry level capital growth, which is estimated as the annual log change of the real capital stock in 1980’s.2 In order to use their investment opportunity data, I regroup 151 industries in ISIC Rev. 3 data into 28 industries in ISIC Rev. 2. The ideal measure would be to use a country-industry opportunity index with data from industries in all countries. However using this measure, we would face reverse causality problem. Therefore, using industry level capital growth data from a specific country as a proxy for investment opportunity index seems more reasonable. US seems to be the best possible candidate since its financial markets are well developed and potential of growth of specific industries is not withdrawn due to credit supply of the markets. In this case, we have to assume that investment opportunity is industry specific, that the ranking of investment opportunities of each industry does not vary across countries. 2.2 Data on countries The data on Gross Domestic Product (GDP), government spending, Producer Price Index (PPI), population and the exchange rate are obtained from International Financial Statistics (IFS) collected by International Monetary Fund (IMF). Producer price index is used to deflate the value added data of the industrial sectors. I use the average period exchange rates to convert the data in other currencies into U.S. dollars. The countries included in the regressions are: Argentina, Brazil, Bulgaria, Cyprus, Czech Republic, Hungary, India, Kenya, Korea, Latvia, Panama, Peru, Philippines, Poland, Portugal, Romania, Russia, Slovakia, South Africa, Turkey, and Uruguay.

2 They

use different measures as proxy for investment opportunities, such as sales growth, measured as annual log change of sales in 1980’s, or value added growth, measured as the annual log change of value added for 1980’s. Since they give similar results, I have shown only the estimations using capital growth.

J Econ Finan

2.3 Data on banks I use bank-level data to estimate bank concentration and bank competition indices. Bank-level data is obtained from BANKSCOPE, a comprehensive, global database that contains information on public and private banks. BANKSCOPE provides information on 28,000 banks around the world and covers years from 1987 to 2006. However, BANKSCOPE includes bank financial statements only for a couple of banks in a limited number of countries before 1994. Therefore, in this study I restrict the empirical analysis to the years between 1994 and 2006. I start with the whole sample of banks that are available in BANKSCOPE, which consists of 124,637 bank-year observations. This sample includes data for the whole period from 1987 to 2006, but dropping the data before 1994 reduces the sample only by 3,359 bank-year observations. BANKSCOPE contains financial statements for a whole group of financial institutions, such as: Bank Holding and Holding Companies, Central Banks, Commercial Banks, Cooperative Banks, Investment Banks/Securities Houses, Islamic Banks, Medium and Long Term Credit Banks, Multi-lateral Governmental Banks, Non-banking Credit Institutions, Real Estates/Mortgage Banks, Savings Banks, and Specialized Governmental Credit Institutions. To estimate banking concentration and banking competition measures, I use data only for commercial banks, savings banks and cooperative banks.3 BANKSCOPE provides consolidated and unconsolidated statements of the banks. I use consolidated statements where they are available, and otherwise use unconsolidated ones, in order to avoid double-counting. 2.3.1 Measuring banking competition and banking concentration Panzar and Rosse (1987) have developed a method to estimate a competition index (“H-statistic”) using bank level data for revenues and factor input prices. This index is calculated as the sum of the elasticities of total revenue of banks to factor input prices, in a reduced form equation for bank revenues. An estimation of H-statistic can be provided by the following reduced form revenue equation: log Rit = α +

J  j=1

β j log wijt +

K  k=1

γk log Xkit +

M 

δm log Z mt + it ,

(1)

m=1

where R is the revenue of bank i at year t; w are the factor input prices of the bank; X are the bank specific variables that affect the bank’s revenue and costs; Z are the macro variables that affect all the banks.

3 The

other institutions either are not regular banks that provide credit in return to interest rate to the private firms for investment, or are banks whose primary objective is not profit maximization. Including them in the bank concentration and/or bank competition estimation would spur the results, as their behavior is different due to several reasons.

J Econ Finan

Using the βˆ j estimates I can calculate H-statistic. H < 0 indicates that we have a monopoly in the market. In case of a monopoly, an increase in any of the input prices increases the marginal costs, which in turn reduces the equilibrium output and total revenue. On the other side H = 1 only in the case of perfect competition. Under perfect competition, an increase in any of the input prices raises both marginal costs and revenues by the same amount. And finally, when 0 < H < 1, we have monopolistic competition. Although an increase in the input prices would raise the revenues, the increase in revenue is somewhat less than the change in the costs. As in the previous studies, I apply some selection criteria to the data before I estimate the banking competition index.4 First of all, I drop the outliers for the main variables used in estimation of competition index (H-statistics), such as data on interest expense and personnel expense. I drop countries that have less than ten observations per year. I estimate H-statistics for the whole period 1994-2006. Following the strategy of Claessens and Laeven (2005), I run the OLS regressions with bank and time fixed effects to estimate the H-statistic. log(Rit ) = αi + β1 log(I Eit ) + β2 log(PEit ) + β3 log(OEit ) +γ1 log(E Ait ) + γ2 log(L Ait ) + γ3 log(Ait ) +δ Dt + it ,

(2)

where Rit is the ratio of gross interest revenue to total assets, which is proxy for the output price of loans; I Eit is the ratio of interest expenses to total deposits and money market funding, a proxy for input price of deposits; PEit is the ratio of personnel expense to total assets, a proxy for input price of labor; and OEit is the ratio of other operating and administrative expense to total assets which stands as a proxy for input price of equipment and fixed capital. E Ait is the ratio of equity to total assets, L Ait , is the ratio of net loans to total assets, and Ait is the total assets (to control for potential size effects). The subscript i denotes bank i, and the subscript t denotes year t and D is a vector of year dummies. H-statistics for each country is estimated as the sum of the coefficients for input prices: price of deposits, price of labor, and price of equipment and fixed capital. Different measures of concentration have been used to estimate the concentration ratio of the banking sector.5 In this paper, I focus on the two most commonly used measures of banking concentration: the k-bank concentration ratio and the Herfindahl–Hirschman Index. The k-bank concentration ratio is

4 Claessens and Laeven (2005) were the first ones to use this index in a cross country study and have

used these criteria to make the estimated H-statistics comparable. I have followed their criteria and strategy in this study. 5 An extensive review of measures of competition and concentration in banking sector used in the literature can be found at Bikker and Haaf (2002).

J Econ Finan

calculated by summing only over the market shares of k largest banks in the market: k C Rk = i=1 si .

(3)

This index gives equal emphasis to the k leading banks, and neglects the rest of banks in the market. However, it is widely accepted that the big banks lead the market and shape it, therefore the role of the small banks is limited. In this study I use the 5-bank concentration ratio.6 This index varies between 0 and 1, where approaching zero means that the market has an infinite number of banks, while approaching one means that the banks included in this ratio own almost all of the market. On the other hand, Herfindahl–Hirschman Index (HHI) is a measure of the size and the distribution of the banks in the market and it is commonly accepted as a measure of market concentration. The HHI is calculated by summing the squares of bank sizes measured as market shares. n H H I = i=1 si2 .

(4)

The advantage of this index is that it does not neglect any bank, meaning that all the banks in the market are included. However, greater weight is given to the larger banks. The range of this index is 1/n and 1, where 1 means that there is only one bank that is monopoly. As the number of banks increases and the disparity of their shares of markets decreases, the index gets smaller in value. An increase in this index means an increase in the market power of the biggest banks. While estimating the concentration indices, I drop the countries which have data available for less than ten banks.

3 Empirical methodology In order to estimate the effect of banking competition and banking concentration on the volatility of industrial output, I use the following estimation method: Std Dev jc = α + β1 (Bank Competitionc ∗ Investment Opportunity j) + β2 Bank Competitionc + δSize j + γ1 Country Controlsc + γ2 Industry j +  jc ,

(5)

6 There is no clear cut rule for how many banks should be included in this ratio. The cutoff number

of the banks included in the ratio is often arbitrary, such as 3, 4, 5 or 10 of the biggest banks. To prevent arbitrary decision about the cutoff number of the banks included in the concentration ratio, I have used 3-bank and 10-bank concentration ratios and I got similar results. I do not include them in this paper, to prevent overwhelmed tables, however they are available upon request.

J Econ Finan

Std Dev jc = α + β3 (Bank Concentrationc ∗ Investment Opportunity j) + β4 Bank Concentrationc + δSize j +γ1 Country Controlsc + γ2 Industry j +  jc ,

(6)

where j stands for industries and c stands for countries. The dependent variable is the standard deviation of the growth of value added of industries. Different methods can be used to detrend the data before estimating the volatility of the growth of value added of industries. I have detrended the standard deviation data by using the residuals of a regression where I control for any aggregate factors affecting the volatility in country, industry and year levels. The main variables of interest are banking competition and banking concentration, and their interactions with investment opportunity index. The main effects show the total effect of banking competition and banking concentration on the volatility of growth of all industries, while the interactions show whether specific industries are being affected more from the market structure of the banking sector. Another important explanatory variable in the estimation is the average size of the industry. Industries that are small in size, are expected to have higher volatility. The expected sign of this coefficient is negative. I include industry dummies to control for unobservable characteristics at industry level. Since I am interested in looking at the direct effect of banking competition and banking concentration, I do not include country dummies in the main regressions, but I include country level controls, such as per capita GDP, government expenditure as a share of GDP, population size, and trade openness of the country. These control variables should be able to isolate country level effects that can effect the volatility of industrial output, but are not directly related to the market structure of the banking sector. In addition, I run the same regressions where I include both banking concentration and banking competition together. Std Dev jc = α + β2 Bank Competitionc + β4 Bank Concentrationc + β1 (Bank Competitionc ∗ Investment Opportunity j) + β3 (Bank Concentrationc ∗ Investment Opportunity j) + δSize j + γ1 Country Controlsc + γ2 Industry j +  jc ,

(7)

If they are both measures of the banking competition, then due to collinearity the coefficients would lose significance. However, if banking concentration and banking competition coefficients do not lose their significance that means that these two measures are not highly correlated and cannot be used as substitutes to measure banking competition. In addition, Table 2 shows the correlation matrix between all variables used in this study, and the correlation between competition index and any of the concentration indices is very low.

1 –0.23 –0.17 –0.17 0.15 –0.00 –0.13 0.05 0.18 0.05

1 1 –0.14 –0.03 –0.02 0.17 0.63 0.06 –0.41 –0.05

2

1 0.82 0.25 –0.02 –0.40 0.27 0.12 0.32

3

1 0.01 –0.01 –0.15 0.07 0.03 0.30

4

1 –0.02 –0.38 0.53 0.19 0.14

5

1 0.03 –0.01 –0.03 –0.01

6

1 –0.53 –0.63 –0.25

7

1 0.17 0.20

8

1 0.24

9

1

10

Notes: This table reports the descriptive statistics of the main regressions. Value added data is from Industrial Statistics Database (INDSTAT4 2008) collected by UNIDO and it is classified by ISIC. 5-Bank concentration ratio index, Herfindahl–Hirschman Index and Competition index are author’s estimations using the methods explained in detail in data section. The bank-level data used to estimate these indices are from Bankscope. Investment Opportunity measure comes from Ciccone and Papaioannou (2006) and it is estimated as the growth rate of capital stock for each industry using US data for 1980’s. Population, GDP per capita, trade openness, and government spending as share of GDP come from International Financial Statistics (IFS)

1. Std. dev. of growth of VA of industries 2. Average VA of industries 3. 5-Bank concentration ratio 4. Herfindahl–Hirschman index 5. Competition index 6. Investment opportunity 7. Log of population 8. Log of GDP per capita 9. Trade openness 10.Government spending/GDP

Table 2 Correlation matrix of regression variables

J Econ Finan

J Econ Finan

4 Banking competition The estimation results of regressions using banking competition index are presented in Table 3. I start out with the first column, where I include only the main variables of interest such as bank competition, interaction of banking competition and investment opportunity of the industry, and the average size of the industry for the period. The direct effect of an increase of banking competition by 10% would be an increase in the volatility of industries growth rate by 1.5%. The industries with higher opportunities of investment would have an even bigger increase in volatility of growth. The total effect of an increase of 10% in competition would be an increase in volatility of the median industry by 1.9%. The size of the the industry has a significant negative effect to the volatility of the industry’s own growth, which is as expected since the small industries are more volatile. All regressions include industry fixed effects to control for unobservable effects at industry level. In the columns (2)–(5), I add country-level control variables to disentangle the effect of banking competition from other factors which might affect the volatility of the growth of industries. I add the control

Table 3 Competition

Bank competition Interaction (bank competition* investment opportunity) Size of industry

(1)

(2)

(3)

(4)

(5)

0.15*** (3.13) 3.56* (1.74) –0.02*** (–3.60)

0.22*** (3.83) 3.53* (1.72) –0.03*** (–3.77) 0.02** (2.22)

0.19*** (3.63) 3.43* (1.68) –0.05*** (–3.77) 0.04*** (2.77) 0.03** (2.10)

0.19*** (3.36) 3.56* (1.70) –0.06*** (–4.22) 0.07*** (4.17) 0.06*** (3.18) 0.14*** (3.29)

0.188 508

0.195 508

0.201 508

0.228 508

0.19*** (3.36) 3.56* (1.71) –0.06*** (–4.41) 0.07*** (4.44) 0.06*** (3.23) 0.13*** (3.16) 0.25 (1.08) 0.229 508

Log of population Log of GDP per capita Trade openness Government spending as share of GDP R2 Observations

Dependent variable: Standard deviation of the growth of value added of industries Notes: Heteroscedasticity robust t-statistics in parentheses. Dependent variable is the standard deviation of the growth rate of value added of industries. Bank competition is estimated using Panzar–Rosse (1987) method. The index is calculated as the sum of the elasticities of total revenue of banks to factor input prices using a reduced form equation for bank revenues for the period 1994–2006. Investment Opportunity measure comes from Ciccone and Papaioannou (2006) and it is estimated as the growth rate of capital stock for each industry using US data for 1980’s. Size of the industry is the average size of the industry for the period 1994–2006, and it is estimation by the author using Industrial Statistics Database (INDSTAT4 2008). Population, GDP per capita, trade openness, and government spending as share of GDP come from International Financial Statistics (IFS). All regressions include industry fixed effects, which are not shown in the table. *, **, *** stand for 10, 5 and 1% significance, respectively

J Econ Finan

variables one by one in each column respectively for population, GDP per capita, trade openness, and government spending as a percentage to GDP. In all regressions the main interest variables remain consistent and significant. According the to most inclusive regression, in column (5), and increase in banking competition by 10% would result in an increase of the volatility of the median industry by 2.3%.

5 Banking concentration I use two different concentration measures to account for the banking concentration. First, I use the concentration ratio of the largest five banks in the country for the period 1994–2006, and then I use the Herfindahl–Hirshcman index. In both cases the shares of the markets are calculated using the assets of the banks. The industrial organization literature brings out advantages and disadvantages of using one or another of these measures, such as the first one takes into consideration only the five largest banks, ignoring the effect of the other banks in the market, etc. (Bikker and Haaf 2002). However, all these issues are out of the scope of this study, that is why I use both measures.

Table 4 Concentration measured as concentration ratio of the five largest banks

Bank concentration CR5 Interaction (bank concentration CR5* investment opportunity) Size of industry

(1)

(2)

(3)

(4)

(5)

–0.33*** (–3.30) –4.25 (–0.65) –0.03*** (–3.88)

–0.42*** (–3.95) –4.04 (–0.63) –0.02** (–1.96) –0.02** (–2.37)

–0.39*** (–3.62) –4.51 (–0.70) –0.03*** (–2.69) 0.01 (0.41) 0.04** (2.32)

–0.32*** (–3.16) –4.88 (–0.77) –0.04*** (–3.23) 0.04** (2.24) 0.06*** (3.19) 0.11*** (2.95)

0.198 508

0.207 508

0.216 508

0.232 508

–0.36*** (–3.66) –4.70 (–0.73) –0.04*** (–3.38) 0.04** (2.35) 0.06*** (3.16) 0.10*** (2.63) 0.52** (2.25) 0.238 508

Log of population Log of GDP per capita Trade openness Government spending as share of GDP R2 Observations

Dependent variable: Standard deviation of the growth of value added of industries Notes: Heteroscedasticity robust t-statistics in parentheses. Dependent variable is the standard deviation of the growth rate of value added of industries. Bank concentration index is estimated as average bank concentration ratio of the five largest banks in the country for the period 1994– 2006. Investment Opportunity measure comes from Ciccone and Papaioannou (2006) and it is estimated as the growth rate of capital stock for each industry using US data for 1980’s. Size of the industry is the average size of the industry for the period 1994–2006, and it is estimation by the author using Industrial Statistics Database (INDSTAT4 2008). Population, GDP per capita, trade openness, and government spending as share of GDP come from International Financial Statistics (IFS). All regressions include industry fixed effects, which are not shown in the table. *, **, *** stand for 10, 5 and 1% significance, respectively

J Econ Finan

Table 4 displays the results of the regressions using 5-bank concentration ratio as a measure of banking concentration. Banking concentration has a negative direct effect and an even more negative effect on the industries that have more investment opportunities, which are the ones that are expected to use the banking sector extensively. However, in all regressions the coefficient of the main effect remains consistently negative and highly significant, while the coefficient interaction term of banking concentration and industry is consistently negative but insignificant. Accounting only for the main effect of the banking concentration, measured as concentration ratio of the five largest banks, an increase in banking concentration by 10% would generate a decrease in volatility between 3 to 4%. In Table 5, I show the results of the regressions where I use Herfindahl– Hirschman index as a measure of banking concentration. In all regressions I control for unobservable effects at industry level. I include different country level controls, which I add one by one in the table. Banking concentration has a dampening effect on the volatility of industrial growth rates. Given the results in Table 5, one cannot say that there is an additional negative effect of banking concentration to industries that might be using the banking system more, however there is a significant and consistent negative direct effect on the

Table 5 Concentration measured as Herfindahl–Hirschman index

Bank concentration HHI Interaction (bank concentration HHI* investment opportunity) Size of industry

(1)

(2)

(3)

(4)

(5)

–0.67*** (–3.46) –4.06 (–0.30) –0.02*** (–3.64)

–0.70*** (–3.65) –3.90 (–0.29) –0.02** (–2.44) –0.01 (–1.06)

–0.64*** (–3.30) –4.73 (–0.36) –0.04*** (–3.10) 0.02 (1.43) 0.04** (2.41)

–0.55*** (–2.94) –5.47 (–0.42) –0.05*** (–3.62) 0.05*** (3.19) 0.06*** (3.39) 0.12*** (3.17)

0.189 508

0.191 508

0.200 508

0.222 508

–0.66*** (–3.43) –5.23 (–0.40) –0.05*** (–3.81) 0.05*** (3.43) 0.06*** (3.37) 0.11*** (2.90) 0.52** (2.23) 0.228 508

Log of population Log of GDP per capita Trade openness Government spending as share of GDP R2 Observations

Dependent variable: Standard deviation of the growth of value added of industries Notes: Heteroscedasticity robust t-statistics in parentheses. Dependent variable is the standard deviation of the growth rate of value added of industries. Bank concentration index is estimated as average Herfindahl–Hirschman index for the period 1994–2006. Investment Opportunity measure comes from Ciccone and Papaioannou (2006) and it is estimated as the growth rate of capital stock for each industry using US data for 1980’s. Size of the industry is the average size of the industry for the period 1994–2006, and it is estimation by the author using Industrial Statistics Database (INDSTAT4 2008). Population, GDP per capita, trade openness, and government spending as share of GDP come from International Financial Statistics (IFS). All regressions include industry fixed effects, which are not shown in the table. *, **, *** stand for 10, 5 and 1% significance, respectively

J Econ Finan

volatility of all industries. In all regressions, the significance of the coefficient of banking concentration is above 1%.

6 Banking competition and banking concentration together in the same regression Tables 6 and 7 show the results of the regressions including both banking concentration index and banking competition index and their interactions with the investment opportunity of industries. All regressions include industry fixed effects to account for unobservable factors which can affect volatility at industrial level. I include country level control variables such as GDP per capita, population, trade openness, and government spending as a share of GDP. Results of the regressions that use 5-bank concentration ratio as concentration index, and competition index as explanatory variables are shown in Table 6. The last column of the table includes all four country-level control variables. The results indicate that holding everything else constant, an increase in banking competition by 10% has an effect of increasing the volatility of the median industry for investment opportunities by 2.5%, while

Table 6 Competition and Concentration ratio of the five largest banks

Bank competition Interaction (bank competition* investment opportunity) Bank concentration CR5 Interaction (bank concentration CR5* investment opportunity) Size of industry

(1)

(2)

(3)

(4)

(5)

0.22*** (4.44) 4.63* (1.86) –0.41*** (–4.09) –6.33 (–0.92) –0.03*** (–4.07)

0.22*** (3.87) 4.63* (1.85) –0.41*** (–3.92) –6.34 (–0.92) –0.03*** (–3.23) 0.00 (0.00)

0.20*** (3.80) 4.60* (1.84) –0.39*** (–3.68) –6.58 (–0.96) –0.04*** (–3.03) 0.01 (0.96) 0.02 (1.44)

0.19*** (3.57) 4.74* (1.90) –0.32*** (–3.22) –6.99 (–1.03) –0.05*** (–3.51) 0.04*** (2.68) 0.04** (2.43) 0.10*** (2.88)

0.246 508

0.246 508

0.249 508

0.264 508

0.20*** (3.63) 4.73* (1.91) –0.37*** (–3.74) –6.80 (–1.00) –0.05*** (–3.65) 0.04*** (2.77) 0.04** (2.39) 0.09** (2.55) 0.54** (2.34) 0.271 508

Log of population Log of GDP per capita Trade openness Government spending as share of GDP R2 Observations

Dependent variable: Standard deviation of the growth of value added of industries Notes: Heteroscedasticity robust t-statistics in parentheses. Dependent variable is the standard deviation of the growth rate of value added of industries. For information on the regressors look at the data section, or previous tables. All regressions include industry fixed effects, which are not shown in the table. *, **, *** stand for 10, 5 and 1% significance, respectively

J Econ Finan Table 7 Competition and concentration as Herfindahl–Hirschman index

Bank competition Interaction (bank competition* investment opportunity) Bank Concentration HHI Interaction (bank concentration HHI* investment opportunity) Size of industry

(1)

(2)

(3)

(4)

(5)

0.15*** (3.11) 3.67* (1.78) –0.67*** (–3.48) –4.47 (–0.34) –0.02*** (–3.67)

0.19*** (3.42) 3.64* (1.77) –0.63*** (–3.28) –4.77 (–0.36) –0.03*** (–3.51) 0.01 (1.37)

0.17*** (3.28) 3.56* (1.73) –0.59*** (–3.05) –5.26 (–0.40) –0.04*** (–3.40) 0.03** (2.00) 0.03* (1.72)

0.17*** (3.11) 3.67* (1.76) –0.50*** (–2.70) –6.01 (–0.47) –0.05*** (–3.89) 0.06*** (3.62) 0.05*** (2.80) 0.12*** (3.15)

0.217 508

0.219 508

0.223 508

0.245 508

0.17*** (3.08) 3.69* (1.79) –0.60*** (–3.20) –5.77 (–0.44) –0.05*** (–4.05) 0.06*** (3.82) 0.05*** (2.79) 0.12*** (2.90) 0.51** (2.18) 0.251 508

Log of population Log of GDP per capita Trade openness Government spending as share of GDP R2 Observations

Dependent variable: Standard deviation of the growth of value added of industries Notes: Heteroscedasticity robust t-statistics in parentheses. Dependent variable is the standard deviation of the growth rate of value added of industries. For information on the regressors look at the data section, or previous tables. All regressions include industry fixed effects, which are not shown in the table. *, **, *** stand for 10, 5 and 1% significance, respectively

the overall effect to all industries is 2%. The direct effect of an increase of banking concentration by 10%, holding everything else constant, is a decrease in volatility of all industries by 3.7%. The median industry for investment opportunities would have an additional 0.7% increase in volatility, however this additional effect is not significant. In Table 7, I present the results of the regressions that include Herfindahl– Hirschman index as a concentration index, and competition index. The coefficient of banking competition varies between 0.15 and 0.19 which is very similar to the results of Table 3, and the coefficient of banking concentration varies between 0.50 and 0.67, which is again very similar to Table 5 results. Including both banking concentration and banking competition measures in the same regression does not change the coefficients and their significance very much, suggesting that they are not highly correlated, otherwise the coefficients would lose significance. The results of Table 7, confirm again the significance of previous tables, since main effects of banking competition and banking concentration remain highly significant. However, it is only the coefficients of interaction of banking competition with investment opportunities of the industries that are significant. The coefficients of interaction of banking concentration with investment opportunities of the industries are negative, but they are insignificant.

J Econ Finan

7 Controlling for other factors that can affect volatility of industries Given the sample of the countries included in the study, one could worry that the results might be biased by the countries that have experienced severe financial crisis. 1990’s and early 2000’s were characterized with many financial crises across the world such as Mexico in 1994, Asian crisis in 1997 (Thailand, Philippines, Malaysia, Singapore, South Korea were all affected), Russia in 1998, Argentina in 1999, and Turkey in 2001. In order to control for any potential bias of the results of this study, I redo the estimations by dropping the countries which were affected by of these crises. I show the results of these regressions in Table 8, where I drop Argentina, Philippines, Russia, South Korea and Turkey. Dropping these countries, reduces my sample by almost 25%, however as seen in Table 8, the significance of the results does not change. Both the main effect of the banking competition and interaction of banking competition and investment opportunities remain significant and positive, showing that an increase in banking competition results in increase of volatility of all manufacturing sectors especially the sectors that have higher

Table 8 Dropping countries that have faced financial crises between 1990–2006 (1) Bank competition Interaction (bank competition* investment opportunity) Bank concentration CR5

Log of population Log of GDP per capita Trade openness Government spending as share of GDP R2 Observations

(3)

0.15** (2.00) 4.18* (1.78) –0.28** (–2.58) –3.52 (–0.43)

Interaction (bank concentration CR5* investment opportunity) Bank concentration HHI Interaction (bank concentration HHI* investment opportunity) Size of industry

(2)

–0.06*** (–3.58) 0.06*** (3.16) 0.04* (1.64) 0.15*** (2.95) 0.41 (1.41) 0.267 386

–0.04*** (–2.80) 0.04** (1.97) 0.05* (1.82) 0.14*** (3.26) 0.50* (1.79) 0.267 386

–0.73*** (–3.17) –6.47 (–0.35) –0.04*** (–2.94) 0.05** (2.47) 0.05** (1.98) 0.15*** (3.37) 0.37 (1.29) 0.274 386

(4)

(5)

0.19** (2.48) 5.72* (1.73) –0.32*** (–2.78) –6.83 (–0.75)

0.13* (1.76) 4.49* (1.86)

–0.05*** (–3.23) 0.04** (2.26) 0.04 (1.44) 0.10** (2.32) 0.61** (2.19) 0.297 386

–0.67*** (–2.95) –8.32 (–0.45) –0.05*** (–3.24) 0.05*** (2.75) 0.04* (1.72) 0.13*** (2.79) 0.43 (1.48) 0.292 386

Dependent variable: Standard deviation of the growth of value added of industries Notes: Heteroscedasticity robust t-statistics in parentheses. Dependent variable is the standard deviation of the growth rate of value added of industries. For information on the regressors look at the data section, or previous tables. All regressions include industry fixed effects, which are not shown in the table. *, **, *** stand for 10, 5 and 1% significance, respectively

J Econ Finan

investment opportunities. On the other hand, the banking concentration has a dampening effect on volatility of all manufacturing sectors, however the coefficient of the interaction of banking concentration and investment opportunity is not significant. There exists a possibility that government subsidies might favor some industries more than some others. In this way one can expect that government spending can be related to volatility of industrial growth. I have tried to see whether government spending has an extra effect on industries with higher investment opportunities, and have not found any significant result. In the same way, I have used and interaction of trade openness and industrial investment opportunities to see whether there is an additional effect of trade openness to the volatility of industries with more investment opportunities, however I could not find a significant result.7

8 Conclusion In this paper, I seek to provide empirical evidence on the effect of banking competition and banking concentration on the volatility of growth of industrial output. An important innovation to the literature is using banking competition and banking concentration separately, and showing that they cannot be used interchangeably. This paper brings together two strands of the literature. The first strand is the literature that looks at the effect of financial intermediaries on the volatility of output at different levels of aggregation, and the second strand is the literature that looks at the effect of banking concentration and banking competition on the credit access of the firms. Using a competition index, estimated as the sum of the elasticities of total revenue of banks to factor input prices, I find that an increase in banking competition increases the volatility of growth of all industries, and it increases even more the volatility of industries that have higher investment opportunities. In addition, I use both banking competition and banking concentration in the same regression and find that the results do not vary much, and are similar in magnitude and significance, showing that concentration cannot be used as a measure of competition. On the other hand, in the regressions that include banking concentration, I find that an increase in banking concentration dampens the volatility of all industries. Banking concentration has an additional negative effect on the industries with higher investment opportunities, however this effect is not very significant. The results of this study show that a higher banking concentration, and/or a lower banking competition would result in lower volatility of growth of industries. These results are in accordance with the previous literature. Many

7 These

results are available, but I choose not to present them here, to prevent overwhelming with tables which have similar results, without adding an extra significant information.

J Econ Finan

studies (Petersen and Rajan 1995; Cetorelli and Gambera 2001; Zarutskie 2006; Bonaccorsi di Patti and Dell’Ariccia 2004) show that more banking concentration and/or less banking competition leads to more credit access for firms. On the other hand, Larrain (2006) and Raddatz (2006) show that more private credit dampens the volatility of growth at industrial level. To conclude, given the nature of the banking sector, the policymakers should not aim to decrease the concentration and/or increase the competition in the banking sector. Instead the policies should aim to increase the efficiency of the banks.

Appendix: Concordance table of ISIC Rev. 2 and ISIC Rev. 3 Industrial sector

ISIC Rev. 2

ISIC Rev. 3

Food products Beverages Tobacco Textile Apparel Leather Footwear Wood products Furniture Paper and products Printing and publishing Industrial chemicals Other chemicals Petroleum refineries Rubber products Plastic products Pottery Glass Nonmetal products Iron and steel Nonferrous metal Metal products Machinery Electric machinery Transportation equipment Professional goods Other industries

311 313 314 321 322 323 324 331 332 341 342 351 352 353 355 356 361 362 369 371 372 381 382 383 384 385 390

151, 152, 153, 154 155 16 17 18 191 192 20 361 21 22 2411, 2413, 243 242, 2412 23 251 252 2691 2610 2692, 2693, 2694, 2695, 2696, 2699 2710, 2731 2720, 2732 28 29, 30 31, 32 34, 35 33 369

References Aghion P, Bacchetta P, Banerjee A (2004) Financial development and the instability of open economies. J Monet Econ 51:1077–1106 Bacchetta P, Caminal R (2000) Do capital market imperfections exacerbate output fluctuations? Eur Econ Rev 44:449–468 Bartelsman EJ, Gray W (1996) The NBER productivity manufacturing database. National Bureau of Economic Research Technical Working Paper No. 205

J Econ Finan Berger AN, Demirguc-Kunt A, Levine R, Haubrich JG (2004a) Bank concentration and competition: an evolution in the making. J Money, Credit Bank 36(3):433–452 Berger AN, Hasan I, Klapper LF (2004b) Further evidence on the link between finance and growth: an international analysis of community banking and economic performance. J Financ Serv Res 25:169–202 Bikker JA, Haaf K (2002) Measures of competition and concentration in the banking industry: a review of the literature. Econ Financ Model 9:53–98 Bonaccorsi di Patti E, Dell’Ariccia G (2004) Bank competition and firm creation. J Money, Credit Bank 36:225–252 Cetorelli N, Gambera M (2001) Bank structure, financial dependence and growth: international evidence from industry data. J Finance 56:617–648 Cetorelli N, Strahan PE (2006) Finance as a barrier to entry: bank competition and industry structure in U.S. local markets. J Finance 61(1):437–461 Ciccone A, Papaioannou E (2006) Adjustment to target capital, finance, and growth. CEPR Discussion Paper 5969 Claessens S, Laeven L (2005) Financial dependence, banking sector competition, and economic growth. J Eur Econ Assoc 3:179–207 Comin D, Philippon T (2005) The rise in firm-level volatility: causes and consequences. In: Gertler M, Rogoff K (eds) NBER macroeconomics annual. MIT Press, Cambridge Correa R, Suarez GA (2008) Firm volatility and banks: evidence from U.S. banking deregulation. Finance and Economics Discussion Series 2009-46. Washington: Board of Governors of the Federal Reserve System Denizer CA, Iyigun MF, Owen A (2002) Finance and macroeconomic volatility. Contrib Macroecon 2(1) Article 7 Easterly W, Islam R, Stiglitz J (2001) Shaken and stirred: explaining growth volatility. In: Pleskovic B, Stern N (eds) Annual bank conference on development economics. World Bank, Washington Guzman MG (2000) Research articles: bank structure, capital accumulation and growth: a simple macroeconomic model. Econ Theory 16(2):421–455 Hoxha I (2011) Big banks versus small banks: good or bad for industries? Penn State Harrisburg School of Business Administration Working Paper Series 11-2 INDSTAT4 (2008) UNIDO Industrial Statistics Database Larrain B (2006) Do banks affect the level and composition of industrial volatility? J Finance 61:1897–1925 Marquez R (2002) Competition, adverse selection, and information dispersion in the banking industry. Rev Financ Stud 15:901–926 Mitchener KJ, Wheelock DC (2010) Does the structure of banking markets affect economic growth? Evidence from U.S. state banking markets. NBER Working Papers 15710 Morgan DP, Rime B, Strahan PE (2004) Bank integration and the state business cycles. Q J Econ 110(2):407–443 Pagano M (1993) Financial markets and growth: an overview. Eur Econ Rev 37:613–622 Panzar JC, Rosse JN (1987) Testing for ‘monopoly’ equilibrium. J Ind Econ 35:443–456 Petersen MA, Rajan RG (1995) The effect of credit market competition on lending relationship. Q J Econ 110:407–443 Raddatz C (2006) Liquidity needs and vulnerability to financial underdevelopment. J Financ Econ 80:677–722 Rajan RG, Zingales L (1998) Financial dependence and growth. Am Econ Rev 88:559–586 Shaffer S (2004) Comment on ‘What drives bank competition? Some international evidence’. J Money, Credit Bank 36(3):585–592 Zarutskie R (2006) Evidence on the effects of bank competition on firm borrowing and investment. J Financ Econ 81:503–537