A. Model I: Industry-Specific Effect

One implication of information-based theories is that a more competitive banking industry should be disproportionately detrimental to the emergence of new firms in sectors where informational asymmetries play a more substantial role.

In a theoretical model considering a continuum of borrowers with varying degree of opaqueness and a continuum of loan markets with varying degree of competition, the above implication is equivalent to saying that the second cross partial derivative of the availability of credit with respect to market power and opaqueness is positive, at least in a given range of opaqueness (a simple theoretical model that shows this result can be found in the Appendix).

In an econometric model, this second mixed derivative can be estimated as the coefficient of an interaction term constructed as the product of a variable describing the degree of opaqueness and one describing the degree of market power in the banking sector.

A setup to estimate the industry-specific effect of bank competition is:

Industry and province dummy variables are included to control for industry-specific and market-specific effects, respectively. Consequently, they will absorb the effect of any variable that does not vary simultaneously across industries and local markets.

The initial share of each industry in the local market takes into account the fact that industries that are already very developed at the beginning of the period are unlikely to grow as fast as those that are at an earlier stage of development and that the entry of new firms is less likely to occur in relatively crowded markets. Hence, we expect a negative sign for its coefficient.

The interaction term is the product of opaqueness in industry i and a proxy for bank competition in market j. The “information channel” predicts a negative sign for its coefficient because bank competition is supposed to be relatively more detrimental to firms characterized by a higher degree of asymmetric information.

In order to take into account the fact that entrepreneurs may use personal wealth to start their activities, we add to the regression another interaction term constructed as the product of a proxy of wealth in market j and the value of average fixed cost in industry i. We expect a negative sign for this coefficient because personal funds may be a reasonable substitute for bank credit only in industries with a relatively small value of fixed costs. While the direct effect of wealth on the rate of birth of firms should be positive, it should also be relatively less important for industries with very large initial investments in fixed assets.

B. Model II: Economy-Wide Effect

The previous setup isolates the differential contribution of bank competition to firm creation with respect to varying degrees of opaqueness, the product of φ₄ and the measure of opaqueness, evaluated at different levels of this variable. In our second model we include directly in the regression the proxies for competition in the local market and drop the market fixed effects. Although econometrically less robust, this setup is useful to assess the first order relationship between bank competition and firm creation.

One first problem determined by the absence of the market fixed effects is a larger potential for biases caused by omitted variables that affect the rate of birth of firms in each local market. A second problem is that some of our competition proxies may have a substantial endogenous component with respect to the rate of birth of firms.

Following the existing literature, we address the first problem by including a set of control variables similar to those employed in standard growth regressions, specifically: initial per capita GDP as a convergence variable; a schooling variable as a proxy for human capital; total credit over GDP as a proxy for financial development; and a proxy for the level of infrastructures.⁹ Structural and policy differences not captured by our regressors that may be relevant, particularly in the case of Southern Italy, are controlled for with geographic area dummy variables.¹⁰ All these variables are described in detail in Section III.F.

We can write Model II as

In the following sections we first motivate our choice of the rate of birth of firms as a proxy for credit availability to informationally opaque borrowers. Then, we illustrate how we measure the degree of opaqueness of the industries and the degree of bank competition in local credit markets.

C. Measuring Credit Availability

Three main arguments have motivated our choice of the rate of birth of firms as an indicator of credit availability to entrepreneurs willing to start a business. First, new firms are likely to be affected by the highest level of asymmetric information problems and have no previous history that may mitigate them. Second, information-related problems most likely get reflected in firms not being able to start rather than not being able to get funds to grow once they have already obtained some bank credit. Consequently, the rate of birth of new firms is where the positive effect of market power suggested by the “information channel”, if empirically relevant, should be stronger. Finally, data on the birth of firms exist with both local market and industry disaggregation, which is not the case for data on growth.

Another advantage is that we implicitly restrict ex ante our firms to be exogenously opaque in a relatively uniform way, except for industry-specific characteristics that are related to the underlying technology or to the type of business, improving the accuracy of our test. Industry growth rates of value added or other measures of production could be strongly influenced by the behavior of large listed firms - that are less opaque and have access to external financing other than local bank credit - and by the size and age distribution of firms in the industry.¹¹

The use of microeconomic data from a sample of existing firms would have allowed us to control for differences in size and age but it would have been intrinsically affected by selection bias since we would have excluded a priori firms never born because too opaque to receive any credit at all.

Empirical support for our strategy is in Rajan and Zingales (1998). In their study the effect of financial development on growth is separated into two components: the effect on the growth of existing establishments and that on the growth in the number of establishments. The positive effect of financial development on growth is found to be stronger for this second variable, suggesting that financial factors affect primarily the creation of new firms rather than the expansion of existing entities.

D. Measuring the Degree of Opaqueness

The literature on financial intermediation measures firm opaqueness with variables that should be correlated to “reputation”, typically firm age or size (Petersen and Rajan, 1995), or the length of relationships between the firm and the bank. Alternatively, opaqueness is associated with “bank dependence”, often measured as the ratio of bank credit to total debt, either at the industry or at the firm level.

In this paper, since we consider the rate of birth of firms, we cannot use firm age because new firms have all age zero. Size is a proxy for transparency only for existing firms because it is related to the history of the firm, and to auditing and disclosure requirements which tend to become tighter with size. In the context of our study bank dependence cannot be used because it is likely to have a large endogenous component, hence we have excluded any measure based on credit volumes.

In order to measure the degree of opaqueness we refer to two basic types of information asymmetries and their consequences: adverse selection - that emerges when lenders cannot observe the “quality” of individual borrowers - and moral hazard - that emerges when lenders cannot monitor the actions of borrowers. Firms in the same industries are likely to share to some extent a similar degree of these information asymmetries, which will instead vary significantly across industries on the basis of some intrinsic characteristic of the production and organization technology.

Our first assumption is that the type of activity of the “perspective” firm influences the relative importance of adverse selection problems. A bank can evaluate more easily the quality of a business plan or project when they are based on a simple technology, with a large predictable component. For example, making nails or paper from raw materials are activities where the effects of the unobservable quality of human capital or effort are less important in determining the outcome than the provision of professional services.

Our second assumption is that these technological factors affect also moral hazard. Again, lenders’ ability to monitor the activities of borrowing firms depends on the technology that characterizes the industry and on the degree of discretion that such technology typically leaves to the manager or entrepreneur. If the technology is complex or there is a large discretionary component, the bank has much to gain from monitoring closely the firm. In addition, some technologies imply naturally the availability of collateral by employing a substantial share of fixed and tangible assets, which reduces moral hazard.¹²

Finally, the severity of hold-up problems is likely to be correlated to the same factors that affect moral hazard. If the activity of the firm is such that monitoring through a relationship has a large value added, the inside bank will be able to extract significant rents from the borrower in the future because of relationship-generated barriers to entry.

Our working assumption is that firm opaqueness varies across industries and is negatively correlated to the relevance of physical capital and other fixed assets in the production process. The larger the share of these assets in the typical balance sheet of firms in the industry the more transparent the industry is.¹³ We have not used intangible assets because they are less homogenous across industries, are often not reported or are equal to zero in some industries. Similarly, we have excluded R&D expenses because reliable information was not available.

Since firms in different industries may have specific accounting practices and varying degrees of discretion in determining depreciation, we consider the volume of gross physical assets (Table 1). Our indicator of opaqueness is the ratio of gross total assets to gross physical assets (ASY M_i).¹⁴ The ratios are computed for 22 industries from individual firm data contained in the Centrale dei Bilanci, a large data set of company balance sheets and income statements, referring to the years 1994-98.¹⁵

Table 1.

Indicators of Asymmetric Information

Table 1.

Indicators of Asymmetric Information

Definition	Name	Mean	75° perc.	25° perc.
Gross total assets/Gross physical assets	ASYM	13.678	14.522	5.940
Gross total assets/Gross physical assets for firms less than 5 year old	ASYMY	18.233	22.001	9.975
Percentage of split ratings	SPLIT	55.7	59.2	53.1

View Table

Table 1.

Indicators of Asymmetric Information

Definition	Name	Mean	75° perc.	25° perc.
Gross total assets/Gross physical assets	ASYM	13.678	14.522	5.940
Gross total assets/Gross physical assets for firms less than 5 year old	ASYMY	18.233	22.001	9.975
Percentage of split ratings	SPLIT	55.7	59.2	53.1

We constructed the industry indicators in the following way. To minimize measurement errors and the effect of outlier, we first dropped companies falling in the 5 percent and 95 percent percentiles of the distribution of our ratios, which restricted the sample to a total of 54,360 firms. We took the average of the annual individual firm ratios observed in the period 1994-1998, in order to remove the effects of temporary shocks to specific companies. Then, we computed the mean of such ratios for each industry.¹⁶ As discussed below, for robustness we constructed the opaqueness indicator employing only young firms, defined as firms with age less than 5 years (Table 1).¹⁷

An identification problem may emerge when using the described measure of opaqueness in the regression. It could be the case that firms in industries characterized by a high share of physical capital have also high fixed start up costs because they require a lot of physical capital in absolute terms. According to the standard theory, credit supply is generally restricted in less competitive markets. One possible implication is that firms in large need of funds because of large start-up costs will be those that suffer the most as banks for diversification purposes may tend to issue smaller loans in a given number rather than fund a smaller number of large borrowers.

This hypothesis would yield the same prediction of the “information channel” and the interpretation of the coefficient of the interaction term would not be unique. For example, a positive sign of the interaction term between concentration and opaqueness could mean either that more opaque firms are relatively favored where banks have some monopoly power or that firms that have large start up costs are relatively more constrained in these markets.¹⁸ We resort to two different approaches to distinguish between the two explanations.

The first approach is based on introducing directly a control variable for the initial fixed costs of firms for each industry. This control variable enters the regression interacted with the same measure of bank competition employed to construct the interaction term with the opaqueness measure, purging the second from the potential effect of initial fixed costs. As a proxy of initial fixed costs we employ the average volume of fixed tangible assets of young firms in the industry, defined as those with age less than 5 years (SIZEY_i).

The second approach attempts to solve this problem by using a proxy for opaqueness that should be more independent from the initial size of the investment in each industry. Morgan (1999) suggests that a measure of the opaqueness of a firm is the extent to which rating agencies disagree on its rating. The same reasoning can be applied at the industry level: a large percentage of firms for which raters disagree (split ratings) reflects a greater difficulty in evaluating firms in that industry, hence greater opaqueness. As described below, we compute for each industry the percentage of firms with split ratings and rank industries by this measure.

The split rating variable (SPLIT_i) is defined as the percentage of bond issues for each industry where raters disagree; the data refer to about 1200 bond issues between 1983 and 1993 in the US and the industry classification is based on SIC codes.¹⁹ SIC industry classification can be easily matched with the Italian classification ATECO91, through ISIC classification. In constructing this variable, we first eliminated repetitions from the original data. More specifically, firms with more than one bond issue were counted only once if their rating split did not change across issues; the issues were counted separately if the rating split changed.

Since the rating data that we employ refer to a sample of bond issues in the US, we use the split rating measure as an instrument for our measure based on physical capital. The working assumption remaining that opaqueness is related to technological factors and that the implied industry ranking is persistent across countries.²⁰

Table 2 reports the ranking of the industries based on ASY M and the values of the different opaqueness proxies. A higher rank indicates a greater degree of opaqueness. For example, Construction is a highly opaque industry while Paper and Products, Basic Metal Industries and Wood and Wood Products are all characterized by less information asymmetries. The ranking is quite consitent across proxies and, as expected, young firms tend to be more opaque than the average firm in each industry.

Table 2.

Industry Ranking and Opaqueness Measures

Note: a higher position in the ranking indicates greater opaqueness. Source: Centrale dei Bilanci (1994-98). The table reports the ranking obtained ordering industries by descending values for the indicators listed. *We have grouped Computers and Office Equipment with Non-electrical Machinery to obtain consistency with the US classification and because rates of birth in the first industry had extreme values due to the very small number of firms.

Table 2.

Industry Ranking and Opaqueness Measures

Industry	ASYM	ASYMY	SPLIT
Construction	53.29	67.50	64.3
Wholesale Trade	34.74	41.04	57.4
Vehicle Trade and Repair	22.82	27.82	46.1
Apparel and Finished Textile Products	18.27	22.00	69.2
Printing, Publishing and Allied Industries	17.46	23.03	55.8
Professional Equipment	15.59	19.42	60.0
Retail Trade and Repair Services	14.52	16.14	63.7
Leather and Leather Products	14.37	18.08	.
Non-electrical Machinery and Office Equipment*	13.24	18.37	57.5
Electrical and Electronic Machinery and Supplies	12.76	16.98	59.0
Textiles	11.39	23.35	54.5
Automobiles, Other Vehicles and Parts	8.96	15.19	61.1
Chemicals, Fibers and Allied Products	8.81	12.18	54.3
Furniture, Toys and Miscellaneous Manufacturing	8.78	12.18	53.1
Metal Products	6.95	8.22	53.3
Hotels, Restaurants and Bars	6.36	10.35	58.1
Food and Beverages	5.94	9.97	51.8
Basic Metal Industries	5.72	10.94	52.2
Wood and Wood Products	5.66	7.33	33.3
Rubber and Miscellaneous Plastic Products	5.18	6.61	51.3
Stone, Clay, Glass and Concrete Products	5.10	7.43	55.5
Paper and Allied Products	4.91	6.90	59.2

Note: a higher position in the ranking indicates greater opaqueness. Source: Centrale dei Bilanci (1994-98). The table reports the ranking obtained ordering industries by descending values for the indicators listed. *We have grouped Computers and Office Equipment with Non-electrical Machinery to obtain consistency with the US classification and because rates of birth in the first industry had extreme values due to the very small number of firms.

View Table

Table 2.

Industry Ranking and Opaqueness Measures

Industry	ASYM	ASYMY	SPLIT
Construction	53.29	67.50	64.3
Wholesale Trade	34.74	41.04	57.4
Vehicle Trade and Repair	22.82	27.82	46.1
Apparel and Finished Textile Products	18.27	22.00	69.2
Printing, Publishing and Allied Industries	17.46	23.03	55.8
Professional Equipment	15.59	19.42	60.0
Retail Trade and Repair Services	14.52	16.14	63.7
Leather and Leather Products	14.37	18.08	.
Non-electrical Machinery and Office Equipment*	13.24	18.37	57.5
Electrical and Electronic Machinery and Supplies	12.76	16.98	59.0
Textiles	11.39	23.35	54.5
Automobiles, Other Vehicles and Parts	8.96	15.19	61.1
Chemicals, Fibers and Allied Products	8.81	12.18	54.3
Furniture, Toys and Miscellaneous Manufacturing	8.78	12.18	53.1
Metal Products	6.95	8.22	53.3
Hotels, Restaurants and Bars	6.36	10.35	58.1
Food and Beverages	5.94	9.97	51.8
Basic Metal Industries	5.72	10.94	52.2
Wood and Wood Products	5.66	7.33	33.3
Rubber and Miscellaneous Plastic Products	5.18	6.61	51.3
Stone, Clay, Glass and Concrete Products	5.10	7.43	55.5
Paper and Allied Products	4.91	6.90	59.2

Note: a higher position in the ranking indicates greater opaqueness. Source: Centrale dei Bilanci (1994-98). The table reports the ranking obtained ordering industries by descending values for the indicators listed. *We have grouped Computers and Office Equipment with Non-electrical Machinery to obtain consistency with the US classification and because rates of birth in the first industry had extreme values due to the very small number of firms.

E. Measuring Bank Competition

The next task is to select a suitable measure of the degree of competitiveness in the banking industry at the local market level. Most empirical studies have employed structural measures, typically the Herfindahl index or other concentration ratios.²¹ The ability of these structural measures to capture the degree of competition in a market depends crucially on the particular form of strategic interaction that actually takes place.²² Because we do not have information on this aspect, we employ six alternative measures.

First, given its widespread use in policy and empirical analysis, we use the Herfindahl index calculated on deposits market shares.

Second, under the assumption that significant changes in the market structure of the industry affect banks’ expectation to extract future rents from borrowers,²³ we construct two measures of structural change in the banking sector: the absolute variation of the Herfindahl index between 1996 and 1998 and the sum of the absolute variation of the market shares of individual banks.²⁴

An important aspect of competitiveness that may not be captured by the measures described above is potential competition. Unfortunately, reliable data on interest rate margins are not available, particularly because it is difficult to control for the risk component embedded in the observed spreads. We then resort to measures of entry.

We consider the market share of banks that were originally chartered within the local market as proxy of the degree of closure with respect to entry by outside banks in neighboring markets. This measure is likely to be related also to the importance of soft information possibly precluded to banks originating in different markets.²⁵

As a direct measure of entry, we calculate the share of branches opened in each province in year t by banks that were not present in that province in year t – 1. Finally, we employ a measure that includes entry inclusive of the effects of mergers and acquisitions. These measures may be affected by more endogeneity problems than structural measures of competitiveness because banks are likely to enter markets where there is higher growth and, possibly, a higher rate of birth of firms. While this problem is mitigated in Model I, it is likely to be serious in the average industry model. Hence, we do not estimate Model II with entry-based variables.

F. Data and Variables Description

To implement our test we employ a panel of data referring to 22 industries and 103 Italian local credit markets, defined as provinces, in the period 1996-1999.²⁶ Yearly observations are averaged over time so each observation for our dependent variable is the annualized rate of birth of firms in industry i and province j in the entire period.

The source of data on the number of firms in each industry and province is the database Movimprese, published by InfoCamere. Movimprese contains aggregate information collected from local firm registries, including the number of firms that register in each year by province, type of legal entity, and by activity based on a detailed classification. End of year stocks of registered firms, of operating firms, and the number of firms that get cancelled are available as well. The information collected by Infocamere has undergone significant changes over time and homogeneous data is available only since 1996.

Three variables have been computed for non-financial firms. The dependent variable in our regressions (BIRTH_ij) is the average annual rate of birth of firms in industry i and province j over the years 1997-99, where for each year t such rate is defined as the ratio of newly registered firms in each province in year t divided by the number of total registered firms at the end of year t – 1. An important advantage of our data is the availability of information on gross flows. We compute the actual rate of birth of firms rather than deriving it from beginning and end of period stocks, which would include the potential effect of exogenous factors on the “death” of firms.²⁷

Second, as in standard growth regressions, we control for the relative size of each sector since there could be a convergence effect such that industries with fewer firms at the beginning of the sample period have relatively higher birth rates and because new entry is more likely to occur in less “crowded” markets; we include the variable REGSHARE_ij defined as the number of registered firms in industry i and province j divided by the total number of registered firms in province j at the end of year 1996.

In some of the specifications, we include as a control variable the number of firms that have been cancelled from the registry (DEATH_ij), defined as the yearly average of the annual ratio of canceled firms in year t divided by total registered firms at the end of year t – 1. Our concern was that new firms replaced dead ones in markets where there was a high turnover due to exogenous factors varying both across markets and industries (whose effects would not be captured by the industry or the province fixed effects). In addition, we wanted to control for the fact that firms that change name would be counted as one firm cancelled and one new firm.

The measures of the degree of competition in the local banking sector are constructed from the statistics on commercial banks collected by the Bank of Italy. The Herfindahl index (HERF_j) is the average of the yearly sum of the squares of the market shares of deposits in province j, based on the location of the branch. The variable LOCBANKS_j is the share of deposits in province j held by banks chartered in province j, again as the average of yearly observations. The absolute variation of Herfindahl (ABSVHERF_j) in the period considered is the sum of the absolute values of the yearly changes in the index. The index of mobility of market shares (MOBSHARES_j) is the sum of the absolute values of the changes in individual bank deposit market shares between year t and year t – 1. Yearly values have been averaged over the period considered. The variable ENTRY _j is the sum over t = 1996, 1997,1998 of branches in year t that belong to banks that were not present in that market in year t – 1 divided by the number of branches in the province in 1996. Finally, (ENTRY M&A_j) is the same variable not corrected for entry due to mergers and acquisitions. Table 3 reports the descriptive statistics for the panel.

Table 3.

Variable Definition and Descriptive Statistics

Observations are stacked as in the estimation and are averages over time of the original data for the period 1996-1998 except where specified. The number of observations is 2266 (22 industries and 103 provinces).

Table 3.

Variable Definition and Descriptive Statistics

Observations are stacked as in the estimation and are averages over time of the original data for the period 1996-1998 except where specified. The number of observations is 2266 (22 industries and 103 provinces).

Variable		Mean	S. Dev.	Min	Max
BIRTH	Percentage Rate of Birth of Firms	5.604	2.660	0	32.326
REGSHARE	Share of Total Registered Firms	0.031	0.051	0	0.310
DEATH	Percentage Rate of Cancellation of Firms	5.749	2.090	0	19.978
HERF	Herfindahl Index of Deposits by location of branches	0.182	0.080	0.053	0.538
LOCBANKS	Share of Deposits held by Local Chartered Banks	0.404	0.230	0	0.939
ABSVHERF	Absolute Change of Herfindahl between 1996 and 1998	0.020	0.022	0.0003	0.134
MOBSHARES	Sum of absolute values of annual changes in market shares, yearly average	0.077	0.028	0.031	0.175
ENTRY	Number of branches of banks not present in the market in year t-1 divided by the total number of branches in year t, cumulated	0.058	0.092	0	0.516
ENTRYM&A	As ENTRY but includes banks that enter by acquisition or merger	0.076	0.101	0	0.516
WEALTH	Per capita Value Added in 1996, million ITL	29.098	7.889	15.707	48.056
SIZEY	Average industry value of fixed assets for firms with less than 5 years (billion ITL)	2.127	2.274	0.277	11.473
HCAPITAL	Share of population with high school or college degree	21.266	3.128	14.40	33.20
BANKDEV	Ratio of Bank Credit to Value Added in 1996	0.551	0.159	0.214	1.572
ROADS	Kilometers of roads divided by the area of the province (square Kilometers/100)	108.36	25.800	51.799	159.0
NEAST	Equal to 1 if the province is in the North East, 0 otherwise	0.213	0.409	0	1
NWEST	Equal to I if the province is in the North West, 0 otherwise	0.233	0.422	0	1
CENTER	Equal to 1 if the province is in the Center, 0 otherwise	0.155	0.362	0	1
SOUTH	Equal to 1 if the province is in the South, 0 otherwise	0.223	0.416	0	1
ISLAND	Equal to 1 if the province is in Sicily or Sardinia, 0 otherwise	0.126	0.332	0	1

View Table

Table 3.

Variable Definition and Descriptive Statistics

Observations are stacked as in the estimation and are averages over time of the original data for the period 1996-1998 except where specified. The number of observations is 2266 (22 industries and 103 provinces).

Variable		Mean	S. Dev.	Min	Max
BIRTH	Percentage Rate of Birth of Firms	5.604	2.660	0	32.326
REGSHARE	Share of Total Registered Firms	0.031	0.051	0	0.310
DEATH	Percentage Rate of Cancellation of Firms	5.749	2.090	0	19.978
HERF	Herfindahl Index of Deposits by location of branches	0.182	0.080	0.053	0.538
LOCBANKS	Share of Deposits held by Local Chartered Banks	0.404	0.230	0	0.939
ABSVHERF	Absolute Change of Herfindahl between 1996 and 1998	0.020	0.022	0.0003	0.134
MOBSHARES	Sum of absolute values of annual changes in market shares, yearly average	0.077	0.028	0.031	0.175
ENTRY	Number of branches of banks not present in the market in year t-1 divided by the total number of branches in year t, cumulated	0.058	0.092	0	0.516
ENTRYM&A	As ENTRY but includes banks that enter by acquisition or merger	0.076	0.101	0	0.516
WEALTH	Per capita Value Added in 1996, million ITL	29.098	7.889	15.707	48.056
SIZEY	Average industry value of fixed assets for firms with less than 5 years (billion ITL)	2.127	2.274	0.277	11.473
HCAPITAL	Share of population with high school or college degree	21.266	3.128	14.40	33.20
BANKDEV	Ratio of Bank Credit to Value Added in 1996	0.551	0.159	0.214	1.572
ROADS	Kilometers of roads divided by the area of the province (square Kilometers/100)	108.36	25.800	51.799	159.0
NEAST	Equal to 1 if the province is in the North East, 0 otherwise	0.213	0.409	0	1
NWEST	Equal to I if the province is in the North West, 0 otherwise	0.233	0.422	0	1
CENTER	Equal to 1 if the province is in the Center, 0 otherwise	0.155	0.362	0	1
SOUTH	Equal to 1 if the province is in the South, 0 otherwise	0.223	0.416	0	1
ISLAND	Equal to 1 if the province is in Sicily or Sardinia, 0 otherwise	0.126	0.332	0	1

In all the specifications we have included the interaction term between a proxy of wealth in the province and a measure of average size of young firms in the industry, to control for personal wealth as an alternative to bank credit to fund a new activity. The proxy for wealth is per capita nominal value added in 1996, the initial year of our sample (WEALTH_j). Size is measured by the variable SIZEY_i, computed as the average of the volume of physical capital of young firms in the industry (with age less than five years), with the same procedure described for ASY M_j and ASY MY_j.

In some of the specifications we have included the interaction between a proxy of start up costs and competition measures for the reasons discussed in section 3.4. For this purpose we have employed the variable SIZEY_i defined as the average volume of physical capital for firms with age less than five years in the Centrale dei Bilanci sample in each industry i.

For the purpose of the estimation of Model II, we have computed several control variables that enter the regression in place of the province fixed effects. As a proxy for human capital we employ the percentage of the population with a high school degree or a college degree (HCAPITAL_j) obtained from Italian Census data. The variable BANKDEV_j is a standard proxy for development of the banking system, calculated as the ratio of total credit to GDP in 1996, the initial year of the period examined. The variable ROADS_j, a proxy for the level of infrastructures, is the number of kilometers of roads per 100 squared km in the region the province belongs to. A set of standard geographical dummy variables, based on the classification of the Bank of Italy statistics, control for structural differences bewteen areas of the country. These variables are NEAST_j, CENTER_j, SOUTH_j and ISLAND_j and are equal to 1 if the province is part of the North East, Center, South of Italy or if it is in Sicily or Sardinia, respectively, 0 otherwise (North West is the excluded category).

A. The Industry-Specific Effect

In the base regression BIRTH_ij is a function of industry and market fixed effects, and of the interaction term between ASY M_i and bank competition in market j. We employed alternatively each of the measures of competition. Control variables that are included in the basic specification are the proportion of firms in the industry i and in market j in 1996 (SHARE_ij) and the interaction between per capita GDP in 1996 and the average size of fixed assets of firms in industry i, to control for personal wealth as an alternative source of funds for start-ups (WEALTH_j* SIZEY_i). Coefficients for the industry and province fixed effects are not shown.

This basic regression (Table 4-A) shows that all the interaction term coefficients have the expected sign and that five out of six are statistically significant.²⁸ The coefficient of the proportion of registered firms also has the expected negative sign and is statistically significant. The coefficient for the second interaction term has the expected negative sign but is not significant.

Table 4-A.

Bank Competition and Firm Birth Rate (ASYM)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 4-A.

Bank Competition and Firm Birth Rate (ASYM)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHAREij	-10.267 *** 2.503	-10.290 *** 2.499	-10.226 *** 2.504	-10.273 *** 2.499	-10.253 *** 2.481	-10.281 *** 2.479
Competition*ASYM	0.058 0.039	-0.360 ** 0.150	-0.201 * 0.119	0.026 * 0.014	-0.101 *** 0.034	-0.117 *** 0.035
WEALTH*SIZEY	-11.026 ** 4.809	-10.437 ** 4.853	-10.856 ** 4.804	-10.293 ** 4.917	-9.856 4.855	-9.698 4.887
Adj. R2	0.355	0.356	0.356	0.356	0.357	0.357
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 4-A.

Bank Competition and Firm Birth Rate (ASYM)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHAREij	-10.267 *** 2.503	-10.290 *** 2.499	-10.226 *** 2.504	-10.273 *** 2.499	-10.253 *** 2.481	-10.281 *** 2.479
Competition*ASYM	0.058 0.039	-0.360 ** 0.150	-0.201 * 0.119	0.026 * 0.014	-0.101 *** 0.034	-0.117 *** 0.035
WEALTH*SIZEY	-11.026 ** 4.809	-10.437 ** 4.853	-10.856 ** 4.804	-10.293 ** 4.917	-9.856 4.855	-9.698 4.887
Adj. R2	0.355	0.356	0.356	0.356	0.357	0.357
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

For robustness we repeated the exercise using ASY MYj, the opaqueness measure computed only for young firms (Table 4-B). The estimates of the interaction terms for these robustness checks support the conclusions of our base regression.²⁹

For all three specifications and all interaction terms, we also tested for the relevance of the potential effect of the “administrative turnover” of firms, by introducing the rate of cancellation of firms (DEATH_ij). We find that this variable is significant and has a positive coefficient (see Tables 5-A and 5-B) indicating that higher rates of creation of firms tend to be correlated to higher rates of exit from the industry in local markets. Results from these further robustness check confirmed our previous findings since the coefficients of the interaction term maintained their expected sign, with a slight loss in significance. Further analysis is required since the cancellation rate may be partly endogenous if banks in less competitive markets encourage the birth of lower quality firms that will have a greater mortality.³⁰

Table 4-B.

Bank Competition and Firm Birth Rate (ASYM for young firms)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYMY is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 4-B.

Bank Competition and Firm Birth Rate (ASYM for young firms)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYMY is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-10.292 *** 2.504	-10.337 *** 2.497	-10.263 *** 2.507	-10.313 *** 2.497	-10.375 *** 2.485	-10.337 *** 2.486
Competition*ASYMY	0.053 0.032	-0.254 ** 0.122	-0.177 * 0.096	0.019 * 0.011	-0.083 *** 0.029	-0.075 *** 0.027
WEALTH*SIZEY	-11.020 ** 4.808	-10.571 ** 4.841	-10.862 ** 4.805	-10.438 ** 4.889	-10.028 ** 4.868	-10.103 ** 4.847
Adj. R2	0.355	0.356	0.356	0.356	0.357	0.356
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 4-B.

Bank Competition and Firm Birth Rate (ASYM for young firms)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYMY is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-10.292 *** 2.504	-10.337 *** 2.497	-10.263 *** 2.507	-10.313 *** 2.497	-10.375 *** 2.485	-10.337 *** 2.486
Competition*ASYMY	0.053 0.032	-0.254 ** 0.122	-0.177 * 0.096	0.019 * 0.011	-0.083 *** 0.029	-0.075 *** 0.027
WEALTH*SIZEY	-11.020 ** 4.808	-10.571 ** 4.841	-10.862 ** 4.805	-10.438 ** 4.889	-10.028 ** 4.868	-10.103 ** 4.847
Adj. R2	0.355	0.356	0.356	0.356	0.357	0.356
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 5-A.

Bank Competition and Firm Birth Rate (ASYM, controlling for DEATH)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 5-A.

Bank Competition and Firm Birth Rate (ASYM, controlling for DEATH)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-9.014*** 2.436	-9.039*** 2.432	-8.977*** 2.436	-9.022*** 2.429	-9.037*** 2.414	-9.007*** 2.413
Compctition*ASYM	0.041 0.036	-0.303** 0.141	-0.185* 0.109	0.032** 0.0134	-0.112*** 0.032	-0.100** 0.031
WEALTH*SIZEY	-3.913 2.616	-3.464 2.644	-3.782 2.616	-3.113 2.676	-2.755 2.656	-2.861 2.643
DEATH	0.375*** 0.053	0.374*** 0.053	0.375*** 0.053	0.376*** 0.053	0.373*** 0.053	0.374*** 0.053
Adj. R2	0.409	0.409	0.409	0.410	0.410	0.410
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 5-A.

Bank Competition and Firm Birth Rate (ASYM, controlling for DEATH)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-9.014*** 2.436	-9.039*** 2.432	-8.977*** 2.436	-9.022*** 2.429	-9.037*** 2.414	-9.007*** 2.413
Compctition*ASYM	0.041 0.036	-0.303** 0.141	-0.185* 0.109	0.032** 0.0134	-0.112*** 0.032	-0.100** 0.031
WEALTH*SIZEY	-3.913 2.616	-3.464 2.644	-3.782 2.616	-3.113 2.676	-2.755 2.656	-2.861 2.643
DEATH	0.375*** 0.053	0.374*** 0.053	0.375*** 0.053	0.376*** 0.053	0.373*** 0.053	0.374*** 0.053
Adj. R2	0.409	0.409	0.409	0.410	0.410	0.410
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 5-B.

Bank Competition and Firm Birth Rate (ASYM for young firms, controlling for DEATH)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

***Statistically significant at the 1%, **statistically significant at the 5%, *statistically significant at the 10% level.

Table 5-B.

Bank Competition and Firm Birth Rate (ASYM for young firms, controlling for DEATH)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-8.698 *** 2.484	-8.728 *** 2.479	-8.673 *** 2.486	-8.724 *** 2.476	-8.744 *** 2.468	-8.776 *** 2.468
Competition*ASYMY	0.042 0.029	-0.176 0.114	-0.161 * 0.086	0.020 * 0.011	-0.070 *** 0.025	-0.074 *** 0.026
WEALTH*SIZEY	-10.934 ** 4.761	-10.615 ** 4.784	-10.797 ** 4.755	-10.337 ** 4.839	-10.084 ** 4.807	-10.054 ** 4.820
DEATH	0.375 *** 0.052	0.374 *** 0.052	0.375 *** 0.052	0.375 *** 0.052	0.375 *** 0.052	0.374 *** 0.053
Adj. R2	0.4132	0.413	0.413	0.413	0.414	0.414
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%, *statistically significant at the 10% level.

View Table

Table 5-B.

Bank Competition and Firm Birth Rate (ASYM for young firms, controlling for DEATH)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-8.698 *** 2.484	-8.728 *** 2.479	-8.673 *** 2.486	-8.724 *** 2.476	-8.744 *** 2.468	-8.776 *** 2.468
Competition*ASYMY	0.042 0.029	-0.176 0.114	-0.161 * 0.086	0.020 * 0.011	-0.070 *** 0.025	-0.074 *** 0.026
WEALTH*SIZEY	-10.934 ** 4.761	-10.615 ** 4.784	-10.797 ** 4.755	-10.337 ** 4.839	-10.084 ** 4.807	-10.054 ** 4.820
DEATH	0.375 *** 0.052	0.374 *** 0.052	0.375 *** 0.052	0.375 *** 0.052	0.375 *** 0.052	0.374 *** 0.053
Adj. R2	0.4132	0.413	0.413	0.413	0.414	0.414
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%, *statistically significant at the 10% level.

A third set of estimates addresses the issue of a potential identification problem related to our measure of firm opaquenss. As discussed previously, firms with a larger share of fixed assets might have also large fixed start up costs. If, as from standard theories, credit supply is restricted in less competitive markets, these same firms would be the ones to suffer the most from the lack of competition. The interaction term would have the same sign as that predicted by the “information channel”, but for a different reason.

The first robustness check we have performed to rule out this explanation is to include in the regressions an interaction term between the same competition measure in each specification, and the inverse of the volume of fixed assets of young firms in each industry. We employed the inverse of the variable SIZEY_i because this way the sign of the coefficient predicted by the start-up costs hypothesis should be the same as that of the interaction term between ASY M_i and the competition measure.

Results of the estimation with this control variable show that the main interaction term retains its sign and significance (Table 6). The interaction term between competition and the inverse of SIZEY_i has instead the opposite sign, suggesting that competition in the banking sector is relatively more beneficial for firms that have small start-up costs, holding constant other characteristics of the industry.

Table 6.

Bank Competition and Firm Birth Rate (ASYM, controlling for start-up costs)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 6.

Bank Competition and Firm Birth Rate (ASYM, controlling for start-up costs)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-10.317 *** 2.508	-10.513 *** 2.481	-10.447 *** 2.489	-10.322 *** 2.494	-10.388 *** 2.476	-10.557 *** 2.473
Competition*ASYM	0.074 * 0.043	-0.631 *** 0.157	-0.480 *** 0.143	0.042 *** 0.016	-0.184 *** 0.038	-0.202 *** 0.039
WEALTH*SIZEY	-10.999 ** 4.817	-10.916 ** 4.869	-11.014 ** 4.815	-10.838 ** 5.002	-11.031 ** 4.848	-10.889 ** 4.910
Competition*(1/SIZEY)	-0.339 0.554	6.037 *** 2.258	6.150 *** 1.886	-0.377 * 0.208	1.885 *** 0.574	1.947 *** 0.652
Adj. R2	0.355	0.357	0.357	0.356	0.359	0.359
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 6.

Bank Competition and Firm Birth Rate (ASYM, controlling for start-up costs)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998 for young firms. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients. The basic regression is reported below. DEATH is given by the ratio between the number of firms canceled and the number of registered firms at the end of the preceding year. Annual data for the period 1997-1999 are averaged over time. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
REGSHARE	-10.317 *** 2.508	-10.513 *** 2.481	-10.447 *** 2.489	-10.322 *** 2.494	-10.388 *** 2.476	-10.557 *** 2.473
Competition*ASYM	0.074 * 0.043	-0.631 *** 0.157	-0.480 *** 0.143	0.042 *** 0.016	-0.184 *** 0.038	-0.202 *** 0.039
WEALTH*SIZEY	-10.999 ** 4.817	-10.916 ** 4.869	-11.014 ** 4.815	-10.838 ** 5.002	-11.031 ** 4.848	-10.889 ** 4.910
Competition*(1/SIZEY)	-0.339 0.554	6.037 *** 2.258	6.150 *** 1.886	-0.377 * 0.208	1.885 *** 0.574	1.947 *** 0.652
Adj. R2	0.355	0.357	0.357	0.356	0.359	0.359
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

As a second approach, we instrumented the measure ASY M_i with the variable SPLIT_i. As described in section III.D, the variable SPLIT_i represents the percentage of firms with split ratings in industry i, from a sample of rating data of US firms. This particular specification of our empirical model was estimated under the assumption that there is some technological reason behind opaqueness and that the industry ranking implied by such technological difference persists across countries. However, the only statistical condition required for this specification is correlation between the variable SPLIT_i and ASY M_i, and independence of SPLIT_i from the start-up cost in industry i.

Results for the instrumental variables estimation are reported in Table 7. In the estimation we have instrumented directly the interaction term with an interaction term constructed multiplying the split rating variable with each proxy of competition. Again, the results from this specification are consistent with our main findings, with four interacted terms coefficients out of six with expected sign and significant.

Table 7.

Bank Competition and Firm Birth Rate (ASYM, Instrumental Variables Estimation)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. Competition*ASYM is instrumented with Competition* SPLIT. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 7.

Bank Competition and Firm Birth Rate (ASYM, Instrumental Variables Estimation)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. Competition*ASYM is instrumented with Competition* SPLIT. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
SHARE	-11.353 *** 2.476	-11.195 *** 2.513	-11.337 *** 2.495	-11.299 *** 2.492	-11.128 *** 2.505	-11.091 *** 2.507
Competition*ASYM	-0.031 0.104	-1.008 ** 0.392	-0.885 ** 0.357	0.019 0.035	-0.187 * 0.109	-0.219 * 0.120
WEALTH*SIZEY	-11.524 ** 4.898	-11.435 ** 4.903	-11.497 ** 4.888	-11.496 ** 4.898	-11.397 ** 4.896	-11.378 ** 4.902
Adj. R2	0.352	0.353	0.353	0.352	0.353	0.353
Observations	2163	2163	2163	2163	2163	2163

***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 7.

Bank Competition and Firm Birth Rate (ASYM, Instrumental Variables Estimation)

The dependent variable is the ratio between the number of newly registered firms and the number of registered firms at the end of the preceding period. Annual data for the period 1997-1999 are averaged over time. HERF, LOCBANKS and MOBSHARES are averages of annual data. ABSVHERF is the sum of absolute values of yearly changes. ENTRY and ENTRYM&A are the ratio of branches opened by outside banks in the period divided by end of 1996 number of branches. ASYM is the inverse of gross physical assets divided by gross total assets and is the industry average for the period 1994-1998. REGSHARE is the industry’s share of the total number of registered firms is computed dividing the 1996 number of registered firms of the industry by the total number of registered firms in that year in each local market in 1996. Competition*ASYM is instrumented with Competition* SPLIT. The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients.

Variables	Dependent Variable: Rate of Birth of Firms in industry j and market i
Competition proxy:	HERF	ABSVHERF	MOBSHARES	LOCBANKS	ENTRY	ENTRYM&A
SHARE	-11.353 *** 2.476	-11.195 *** 2.513	-11.337 *** 2.495	-11.299 *** 2.492	-11.128 *** 2.505	-11.091 *** 2.507
Competition*ASYM	-0.031 0.104	-1.008 ** 0.392	-0.885 ** 0.357	0.019 0.035	-0.187 * 0.109	-0.219 * 0.120
WEALTH*SIZEY	-11.524 ** 4.898	-11.435 ** 4.903	-11.497 ** 4.888	-11.496 ** 4.898	-11.397 ** 4.896	-11.378 ** 4.902
Adj. R2	0.352	0.353	0.353	0.352	0.353	0.353
Observations	2163	2163	2163	2163	2163	2163

***Statistically significant at the 1%, **statistically significant at the 5%.

A final robustness check was performed, by including simultaneously multiple measures of competition in the regression. Our concern was that static and dynamic measures captured different aspects of how the structure of the banking industry is related to competition. In particular, the static market power measures such as HERF_j and LOCBANKS_j do not take adequately into account factors like contestability or rivalry. On the other hand, measures of structural change alone do not take into account initial conditions. We tried to supply to this problem by estimating the model including more than one competition measure in the same specification. The results reported in Table 8 broadly confirm our previous findings. In addition, the inclusion of static and dynamic measures of competition tends to increase the significance of the static measures without significantly changing their coefficients.

Table 8.

Bank Competition and Firm Birth Rate: Robustness with All Indicators

The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

***Statistically significant at the 1%, **statistically significant at the 5%.

Table 8.

Bank Competition and Firm Birth Rate: Robustness with All Indicators

The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Denendent Variable: BIRTH
REGSHARE	-10.317 *** 2.499	-10.774 *** 2.480	-10.315 *** 2.481	-8.728 *** 2.462	-10.829 *** 2.461	-9.179 *** 2.450
HERF*ASYM	0.113 *** 0.041	0.166 *** 0.455	0.100 ** 0.039	0.073 * 0.037	0.142 *** 0.043	0.114 *** 0.041
ABSVHERF*ASYM	-0.518 *** 0.153	-0.862 *** 0.167	-0.463 *** 0.155	-0.340 ** 0.148	-0.746 *** 0.156	-0.552 *** 0.146
ENTRY*ASYM	- -		-0.078 ** 0.035	-0.067 ** 0.031	-0.146 *** 0.039	-0.138 *** 0.036
LOCBANKS*ASYM	- -		0.018 0.014	0.020 0.013	0.026 * 0.015	0.024 * 0.013
WEALTH*SIZEY	-10.356 ** 4.846	-10.870 ** 4.865	-9.106 * 4.972	-9.250 * 4.920	-10.890 ** 5.063	-10.813 ** 5.014
HERF*(1/SIZEY)	- -	-1.163 * 0.600	- -	- -	-0.911 0.591	-0.890 0.569
ABSVHERF*(1/SIZEY)	- -	7.654 *** 2.485	- -	- -	6.435 *** 2.395	4.836 ** 2.247
ENTRY*(1/SIZEY)	- -	- -	- -	- -	1.579 *** 0.584	1.618 *** 0.548
LOCBANKS*(l/SIZEY)	- -	- -	- -	- -	-0.214 0.209	0.137 0.199
DEATH	- -	- -	- -	0.371 *** 0.0523	- -	0.370 *** 0.052
Adj. R2	0.357	0.358	0.358	0.414	0.360	0.416
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 8.

Bank Competition and Firm Birth Rate: Robustness with All Indicators

The coefficients for the constant terms and industry and market indicator variables are not reported. Robust standard errors are below coefficients

Variables	Denendent Variable: BIRTH
REGSHARE	-10.317 *** 2.499	-10.774 *** 2.480	-10.315 *** 2.481	-8.728 *** 2.462	-10.829 *** 2.461	-9.179 *** 2.450
HERF*ASYM	0.113 *** 0.041	0.166 *** 0.455	0.100 ** 0.039	0.073 * 0.037	0.142 *** 0.043	0.114 *** 0.041
ABSVHERF*ASYM	-0.518 *** 0.153	-0.862 *** 0.167	-0.463 *** 0.155	-0.340 ** 0.148	-0.746 *** 0.156	-0.552 *** 0.146
ENTRY*ASYM	- -		-0.078 ** 0.035	-0.067 ** 0.031	-0.146 *** 0.039	-0.138 *** 0.036
LOCBANKS*ASYM	- -		0.018 0.014	0.020 0.013	0.026 * 0.015	0.024 * 0.013
WEALTH*SIZEY	-10.356 ** 4.846	-10.870 ** 4.865	-9.106 * 4.972	-9.250 * 4.920	-10.890 ** 5.063	-10.813 ** 5.014
HERF*(1/SIZEY)	- -	-1.163 * 0.600	- -	- -	-0.911 0.591	-0.890 0.569
ABSVHERF*(1/SIZEY)	- -	7.654 *** 2.485	- -	- -	6.435 *** 2.395	4.836 ** 2.247
ENTRY*(1/SIZEY)	- -	- -	- -	- -	1.579 *** 0.584	1.618 *** 0.548
LOCBANKS*(l/SIZEY)	- -	- -	- -	- -	-0.214 0.209	0.137 0.199
DEATH	- -	- -	- -	0.371 *** 0.0523	- -	0.370 *** 0.052
Adj. R2	0.357	0.358	0.358	0.414	0.360	0.416
Observations	2266	2266	2266	2266	2266	2266

***Statistically significant at the 1%, **statistically significant at the 5%.

In summary, the empirical evidence in this section is consistent with the idea that a higher degree of competition in the banking industry is more detrimental (or less favorable) to firms operating under more severe asymmetric information conditions.

However, until now we have not taken any view with regard to the total effect of competition on firm creation. We present some evidence in the next section.

B. The Economy-Wide Effect

In Model II the potential for an endogeneity problem of entry-based competition measures with respect to the rate of birth of firms is virtually unsolvable. Indeed, we expect that measures closely related to entry are significantly influenced by the dynamics of economic activity in the local market, hence by the rate of birth of firms. It is reasonable, instead, that structural measures whose cross-sectional dispersion varies more slowly over time should be less affected although not entirely immune. Thus, we have estimated the economy-wide effect only for the proxies that are more closely related to structural factors, specifically HERF, ABSV HERF and LOCBANKS_j³¹. In addition, the robustness tests reported in Table 8 have suggested to employ jointly concentration and the change in concentration since there are no longer the fixed effects and it is reasonable that the two variables capture different relevant features of the banking industry. We will briefly discuss the results for various specifications that include HERF and ABSV HERF (Table 9). Similar results were obtained with the variable LOCBANKS but will not be discussed for the sake of brevity.

Table 9.

Bank Competition and Firm Birth Rate: Total Effect

The coefficients for the constant and the industry fixed effects are not reported. Robust standard errors arc below coefficients.

Note: two provinces were dropped because no information was available for HCAPITAL. ***Statistically significant at the 1%, **statistically significant at the 5%.

Table 9.

Bank Competition and Firm Birth Rate: Total Effect

The coefficients for the constant and the industry fixed effects are not reported. Robust standard errors arc below coefficients.

Variables	Dependent Variable: BIRTH
REGSHARE	-8.690 *** 1.948	-8.706 *** 1.948	-6.491 *** 2.001	-6.517 *** 1.998	-6.900 *** 1.996
HERF	13.504 *** 2.294	11.942 *** 2.445	11.491 *** 2.214	10.354 *** 2.358	10.715 *** 2.416
HERF SQUARED	-28.534 *** 4.425	-28.535 *** 4.451	-23.298 *** 4.322	-23.331 *** 4.338	-23.357 *** 4.341
ABSVHERF	-11.354 *** 3.034	-4.148 4.532	-9.377 *** 2.866	-4.136 4.280	-6.126 4.551
HERF*ASYM	- -	0.114 *** 0.042	- -	0.084 ** 0.039	0.135 *** 0.044
ABSVHERF *ASYM	- -	-0.526 *** 0.178	-	-0.384 **	-0.668 ***
			-	0.169	0.203
WEALTH*SIZEY	-10.494 ** 4.868	-9.957 4.906	-10.840 ** 4.822	-10.447 ** 4.858	-10.816 ** 4.886
BANKDEV	-0.102 0.391	-0.102 0.390	0.565 0.378	0.561 0.377	0.558 0.377
WEALTH	-0.218 0.167	-0.229 0.167	-0.169 0.154	-0.178 0.154	-0.171 0.154
ROADS	-0.003 0.002	-0.003 0.002	-0.004 ** 0.002	-0.004 ** 0.002	-0.004 ** 0.002
HCAPITAL	-0.043 ** 0.020	-0.043 ** 0.020	-0.052 *** 0.019	-0.052 *** 0.019	-0.052 *** 0.019
NORTHEAST	0.364 ** 0.143	0.364 ** 0.143	0.314 ** 0.136	0.314 ** 0.136	0.314 ** 0.136
CENTER	0.155 0.164	0.155 0.164	0.121 0.150	0.121 0.150	0.122 0.150
SOUTH	0.533 ** 0.210	0.533 ** 0.210	0.546 ** 0.192	0.546 *** 0.192	0.546 *** 0.192
ISLANDS	-0.149 0.269	-0.149 0.269	-0.152 0.253	-0.152 0.253	-0.151 0.253
DEATH	- -	- -	0.405 *** 0.055	0.403 *** 0.055	0.401 *** 0.055
HERF*(1/SIZEY)	- -	- -	- -	- -	-2.764 * 0.626
ABSVHERF*(1/SIZEY)	- -	- -	- -	- -	6.305 ** 2.579
Adj.R2	0.242	0.245	0.318	0.319	0.320
Observations	2222	2222	2222	2222	2222

Note: two provinces were dropped because no information was available for HCAPITAL. ***Statistically significant at the 1%, **statistically significant at the 5%.

View Table

Table 9.

Bank Competition and Firm Birth Rate: Total Effect

The coefficients for the constant and the industry fixed effects are not reported. Robust standard errors arc below coefficients.

Variables	Dependent Variable: BIRTH
REGSHARE	-8.690 *** 1.948	-8.706 *** 1.948	-6.491 *** 2.001	-6.517 *** 1.998	-6.900 *** 1.996
HERF	13.504 *** 2.294	11.942 *** 2.445	11.491 *** 2.214	10.354 *** 2.358	10.715 *** 2.416
HERF SQUARED	-28.534 *** 4.425	-28.535 *** 4.451	-23.298 *** 4.322	-23.331 *** 4.338	-23.357 *** 4.341
ABSVHERF	-11.354 *** 3.034	-4.148 4.532	-9.377 *** 2.866	-4.136 4.280	-6.126 4.551
HERF*ASYM	- -	0.114 *** 0.042	- -	0.084 ** 0.039	0.135 *** 0.044
ABSVHERF *ASYM	- -	-0.526 *** 0.178	-	-0.384 **	-0.668 ***
			-	0.169	0.203
WEALTH*SIZEY	-10.494 ** 4.868	-9.957 4.906	-10.840 ** 4.822	-10.447 ** 4.858	-10.816 ** 4.886
BANKDEV	-0.102 0.391	-0.102 0.390	0.565 0.378	0.561 0.377	0.558 0.377
WEALTH	-0.218 0.167	-0.229 0.167	-0.169 0.154	-0.178 0.154	-0.171 0.154
ROADS	-0.003 0.002	-0.003 0.002	-0.004 ** 0.002	-0.004 ** 0.002	-0.004 ** 0.002
HCAPITAL	-0.043 ** 0.020	-0.043 ** 0.020	-0.052 *** 0.019	-0.052 *** 0.019	-0.052 *** 0.019
NORTHEAST	0.364 ** 0.143	0.364 ** 0.143	0.314 ** 0.136	0.314 ** 0.136	0.314 ** 0.136
CENTER	0.155 0.164	0.155 0.164	0.121 0.150	0.121 0.150	0.122 0.150
SOUTH	0.533 ** 0.210	0.533 ** 0.210	0.546 ** 0.192	0.546 *** 0.192	0.546 *** 0.192
ISLANDS	-0.149 0.269	-0.149 0.269	-0.152 0.253	-0.152 0.253	-0.151 0.253
DEATH	- -	- -	0.405 *** 0.055	0.403 *** 0.055	0.401 *** 0.055
HERF*(1/SIZEY)	- -	- -	- -	- -	-2.764 * 0.626
ABSVHERF*(1/SIZEY)	- -	- -	- -	- -	6.305 ** 2.579
Adj.R2	0.242	0.245	0.318	0.319	0.320
Observations	2222	2222	2222	2222	2222

Note: two provinces were dropped because no information was available for HCAPITAL. ***Statistically significant at the 1%, **statistically significant at the 5%.

In all the specifications reported concentration enters the regression with a linear and a second order term, both having significant coefficients. The linear term has a positive coefficient while the squared terms has a negative coefficient, indicating an inverted U-shape average relationship between HERF and BIRTH. This parabola reaches its maximum around a value of HERF of 23 per cent in the basic specification (column 1). On average, very low values of concentration are associated to low firm birth rates, but high concentration is detrimental to the creation of firms. In other words, some competition is good, too much may be a problem.

This finding is consistent with the results of other studies. Cetorelli and Gambera (2001) find a relationship of a similar shape between growth and concentration in the banking industry. Jackson and Thomas (1995) find a positive effect of bank concentration on employment growth in new firms and a negative effect for mature firms, consistent with the view that the effect of concentration is not homogeneous across firms with different opaqueness.

We have then estimated the total effect of concentration allowing for differences based on opaqueness, by including the interaction term between ASY M and HERF. These estimates confirm the signs and significance of the interaction term (Table 8, column 2, 4 and 5) and are robust to the inclusion of the cancellation rate and the start-up costs variable. Similarly, ABSV HERF and its interaction terms are robust.

Taking the estimated parameters of column 6 we can evaluate the effect of concentration on the rate of birth of firms for different industries based on the most complete specification (Figure 1). This derivative is a function of the values of HERF and ASY M if we set SIZEY at its mean value. For example, for firms with a relatively low opaqueness (25th percentile), the derivative is equal to 0.44 at the mean value of HERF (18.2 per cent). It turns negative when HERF is larger than 19.1 per cent). If we consider a relatively high level of opaqueness (75th percentile), the derivative of BIRTH with respect to HERF is equal to 1.6 at the mean value of HERF, and turns negative when HERF is around 21.7 per cent. When opacity is equal to the maximum value observed in our sample of industries, the derivative remains positive for values of HERF below 32 per cent.

Derivative of BIRTH with respect to HERF, as a function of HERF

(evaluated for 25th, 50th and 75th percentile of ASYM)

Citation: IMF Working Papers 2001, 021; 10.5089/9781451843910.001.A001

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Derivative of BIRTH with respect to HERF, as a function of HERF

(evaluated for 25th, 50th and 75th percentile of ASYM)

Citation: IMF Working Papers 2001, 021; 10.5089/9781451843910.001.A001

• Download Figure
• Download figure as PowerPoint slide

Derivative of BIRTH with respect to HERF, as a function of HERF

(evaluated for 25th, 50th and 75th percentile of ASYM)

Citation: IMF Working Papers 2001, 021; 10.5089/9781451843910.001.A001

• Download Figure
• Download figure as PowerPoint slide

The purpose of this simple exercise is to show that HERF has a relatively larger effect for more opaque industries on BIRTH when the derivative is positive, and a negative effect smaller in absolute value for these same industries. The differential effect is represented by the vertical shift between the lines depicted in Figure 1. Clearly, the derivative itself is positive for relatively low values of concentration and turns negative for high values, suggesting that high concentration is detrimental to the creation of firms. In addition, Figure 1 is suggestive of a dominance of the economy-wide effect of bank competition over its differential effect depending on opaqueness, as the second is relatively small in magnitude.

The estimation of Model II confirms the other findings concerning the effect of WEALTH and start-up costs. In local markets with a higher per capita GDP firms with smaller start-up costs are relatively favored, suggesting that personal wealth is a substitute for banks credit. Finally, we find evidence in the data of a sort of convergence effect on the creation of firms, since provinces with relatively more registered firms and more infrastructures have lower rates of birth. The general level of bank development is not significant but has a positive sign. Human capital, instead, appears to be negatively correlated with the creation of firms perhaps because of lower opportunity costs of high level education in environmnets with less business opportunities.