The Determinants of Regional Variation in New Firm Formation*

Catherine Armington ** and Zoltan J. Acs ** ***

** Center for Economic Studies, Washington Plaza II, U. S. Bureau of the Census WashingtonDC 20233-6300.*** Merrick School of Business, University of Baltimore, Baltimore, MD 21201

October, 2000Revised July 2001

Abstract

While much of the literature on new firm formation in the 1980s was motivated byhigh levels of unemployment, much of the focus on new firm start-ups today ismotivated by high-technology. Using a new database we examine the role ofhuman capital, training and education, and entrepreneurial environment on newfirm formation. We find significant differences in new firm formation rates fromindustrial regions to technologically progressive regions. Variations in firm birthrates are explained by industrial density, population and income growth. Theseresults are consistent with thick labor markets and localized knowledgespillovers.

Key Words: New firm formation, spillovers, unemployment, spatial variation

Correspondence to:

Zoltan J. AcsMerrick School of BusinessUniversity of BaltimoreBaltimore, MD 21201E-mail: zacs@ubmail.ubalt.edu

2

INTRODUCTION

A large literature exists on regional variation in firm birth rates. Most of these

studies were published in the early 1990s based on data from the 1980s. These

studies used different indicators, were carried out for different sector, different

countries, and different units of analysis. Many of the resulting papers were

published in a special issue of Regional Studies (1994). They found significant

regional variation in new firm formation, and examined a set of regional

determinants to explain this variation. The explanatory variables that were

generally found to be most important were various measures of unemployment,

population density, industrial restructuring, and availability of financing.

Several recent developments warrant a fresh examination of this subject.

First, theoretical developments in the New Economic Geography (Krugman

1991a, 1991b and 1994), and Endogenous Growth Theory (Romer 1990, and

Nijkamp and Poot 1998) have increased our understanding of spatial

perspectives and economic growth. For tests of these theories see Audretsch

and Fritsch (1994), Audretsch and Feldman (1996) and Anselin, Varga and Acs

(1997, 2000). Second, new and more sophisticated data bases have been

developed that can better identify firm birth rates (Armington 1997 and Acs and

Armington 1998). Third, the recent evolution of the U.S. economy has been

accompanied by a regional shift in economic activity away from traditional

industrial regions to new regional agglomerations of high technology, creating an

3

explosion in entrepreneurial activity and new firm formation (Acs, Carlsson and

Karlsson, 1999).

One measure of the process of change in economic activity is the rate at

which new firms are being established, or what we call the firm birth rate.

Presumably a relatively high regional rate of firm births indicates a process of

resources concentrating within that region, while a relatively low rate of firm births

would indicate the opposite, a deconcentration of economic activity. For

example, in the mid 1990s the average annual rate of new firm formation in the

United States was 3.85 new firms per 1,000-labor force. However, this rate

varied from a high of 5.50 per 1,000-labor force in Colorado down to only 2.91 in

Pennsylvania. The more rapid rates of rapid new firm formation were typically

found in the South and the Southwest.

While much of the literature on new firm formation in the 1980s was

motivated by high levels of unemployment in traditional industrial regions, much

of the focus on new firm start-ups today is motivated by high technology start-ups

that are thought to be driving the new economy. The purpose of this paper is to

re-examine the issue of regional variation in firm birth rates in the United States,

with new data, in light of recent theoretical developments. Today, with

unemployment at an all time low in the United States, there is little interest in the

role of new firms in reducing unemployment.

4

Research focus is shifting to issues of the relationship of human capital,

training and education differences on rates of new firm formation, and their

impact on economic growth. The second section of this paper examines issues

of measurement of new firm formation rates and discusses new sources of data.

The third section presents the theoretical background on some of the

determinants of new firm formation, while the fourth section examines the

empirical results. The conclusions are in the final section.

MEASUREMENT OF FIRM BIRTH RATE

This study utilizes a new database that the Bureau of the Census has

constructed for study of entry, survival, and growth in different types of

businesses. The Longitudinal Establishment and Enterprise Microdata (LEEM)

file has multiple years of annual data for every U.S. private sector (non-farm)

business with employees. The current LEEM file facilitates tracking employment,

payroll, and firm affiliation and (employment) size for the more than eleven million

establishments that existed at some time during 1989 through 1996. This file

was constructed by the Bureau of the Census from its Statistics of U.S. Business

(SUSB) files,1 which were developed from the microdata underlying the

aggregate data in Census’ County Business Patterns. These annual data were

linked together using the Longitudinal Pointer File associated with the SUSB,

which facilitates tracking establishments over time, even when they change

ownership and identification numbers.

5

The basic unit of the LEEM data is a business establishment (location or

plant). An establishment is a single physical location where business is

conducted or where services or industrial operations are performed. The

microdata describe each establishment for each year of its existence in terms of

its employment, annual payroll, location (state, county, and metropolitan area),

primary industry, and start year. Additional data for each establishment and year

identify the firm (or enterprise) to which the establishment belongs, and the total

employment of that firm.

A firm (or enterprise or company) is the largest aggregation (across all

industries) of business legal entities under common ownership or control.

Establishments are owned by legal entities, which are typically corporations,

partnerships, or sole proprietorships. Most firms are composed of only a single

legal entity that operates a single establishment—their establishment data and

firm data are identical, and they are referred to as “single unit” establishments or

firms. The single unit businesses are frequently owner-operated. Only 4 percent

of firms have more than one establishment, and they and their establishments

are both described as multi-location or multi-unit.

The LEEM data cover all private sector businesses with employees, with

the exception of those in agricultural production, railroads, and private

households. This is the same universe that is covered in Census’ annual County

Business Patterns publications, but establishments with positive payroll during a

6

year and no employment in March of that year are not counted for that year for

this project.

The geographic unit of analysis used for this study is travel-to-work or

Labor Market Areas (LMA’s). They are aggregations of the 3,141 US counties

into 394 geographical regions that contain a high proportion of residential-work

location trips. LMA’s within the U.S. are defined according to the 1990

specification of Tolbert and Sizer (1996) for the Department of Agriculture. Many

of the 394 LMA’s cut across state boundaries, to better represent local economic

areas. Some adjacent smaller Commuting Zones have been grouped together

so that all LMA’s had a minimum of 100,000 population in 1990 (see Reynolds

1994 for further discussion). We code the location of each establishment

according to its most recently specified state and county in the LEEM, assuming

that most of the few location coding changes are corrections. Businesses that

report operating statewide (county = 999) have been placed into the largest LMA

in each state (See Figure 1).

We define six industry sectors in this paper, to better control for aggregation

effects in regions with different distributions of industries. This expands the

industrial detail beyond that of previous studies, most of which were limited to

manufacturing. Industry codes are based on the most recently reported 4-digit

SIC code for the original establishment in each firm. For most firms (single

location firms) this is the only establishment. For most new multi-unit firms, the

7

industry classification of the primary location is the same as that of their

secondary locations. We use the most recently reported SIC code, rather than

the first reported SIC, because the precision and accuracy of the codes tends to

increase over time2. Census often lists new establishments before detailed

industry codes are available for them.

Code Sector Standard Industrial Classification

Dstb Distributive 4000-5199 (transportation, communication,

public utilities, and wholesale trade)

manf Manufacturing 2000-3999

bsrv Business services 7300-7399 and 8700-8799 (incl. engineering,

accounting, research, and management

services)

extr Extractive 0700-1499(agricultural services and mining)

retl Retail Trade 5200-5999

locl Local market 1500-1799 and 6000-8999 excl. Bus. serv.

(construction, consumer and financial services)

These sectors identify industries that might differ in their sensitivity to local

market conditions. For example, local services and construction are more

dependent on regional demand in their supply requirements for new firms, while

manufacturing or distributive services may have greater need forsemi-and

8

unskilled labor. Extractive industries are dependent on the existence of natural

resources and soil conditions.

The relative size of these sectors varies greatly. In terms of share of total

private sector employment, the local market sector is the largest, with 36 percent

of employment. Next largest is retail trade with 21 percent and manufacturing

with 20 percent in 1991. In fact, local market, manufacturing and retail trade

together account for 77 percent of total employment. Business services account

for only 8 percent, distribution for 13 percent and extractive for 1 percent.

Employment in the local market sector grew 44 percent between 1991 and 1996,

and business services grew 26 percent for the same period, dominating the

growth rates.

Firm births include both new single unit firms (establishments, or

locations) with less than 500 employees, and the primary locations of new multi-

unit firms with less than 500 employees, firm wide. Those new firms that had

500 or more employees in their first year of activity appear to be primarily

offshoots of existing companies. Annually, there were somewhat under 150 such

large apparent births of single-unit firms, with an average of about 1500

employees each. About a third of these larger single unit firms were employee-

leasing firms or employment agencies, while the remainder were widely

distributed across industries. However, examination of the new firms with 100-

499 employees in their first year showed that most seemed credible startups,

9

frequently in industries that are associated with large business units, such as

hotels and hospitals. Since this study is not concerned with the employment

impact of startups, there is no danger of the bulk of the data on smaller startups

being swamped by that of a few larger startups that might actually be offshoots of

existing businesses. Therefore, the startups with 100 to 499 employees were

included, if they qualified otherwise.

Single unit firm births in year t are identified on the LEEM as non-affiliated

establishments with a start-year of t or t-1 that had no employment in March of

year t-1, and had positive employment below 500 in March of year t. This avoids

inclusion of either new firms that have not yet actually hired an employee, or

firms recovering from temporary inactivity. The ‘start-year’ is the year that the

establishment entered the Census business register. About 400,000 new firms

generally appear in the business register (with some positive annual payroll) the

year before they have any March employment, and we postpone their ‘birth’ until

their first year of reported employment. An average of 90,000 older firms each

year reduce their March employment to zero and then recover the following year.

We have also included most of the relatively few multi-unit firms (1500 to

6000 per year) that appeared to start up with less than 500 employees in multiple

locations in their first year. We limited multi-unit firm births to those whose

employment in their new primary location constituted at least a third of their total

employment in the first year.3 This rule effectively eliminated the 600 to 1000

10

new firms each year which were apparently set up to manage existing locations -

- relatively small new headquarters supervising large numbers of employees in

mainly older branch locations which were newly acquired, or perhaps contributed

by joint venture partners.

Firm birth rates are calculated for each of the 394 LMAs, based on births

between 1994 and 1995, and between 1995 and 1996.4 There is little contention

that measuring the absolute numbers of new establishments and then comparing

them across regions would be more misleading than revealing. This is because

the economic regions are not homogeneous with respect to size. Two

approaches have been generally adopted in attempting to compare birth rates

across regional markets. The first method standardizes the number of entrants

relative to the number of establishments already in existence, which can be

termed the ecological approach, because it considers the amount of start-up

activity relative to the size of the existing population of businesses.

The second method, which can be characterized as the labor market

approach, is to standardize the number of new firms with respect to the size of

the labor force. The labor force approach has a particular theoretical appeal, in

that it is based on the theory of entrepreneurial choice proposed by Evans and

Jovanovic, (1989). That is, some worker starts each new business. The labor

market approach implicitly assumes that the entrepreneur starting a new

11

business is in the same labor market within which that new establishment

operates.

Regions vary considerably in their average number of employees per

establishment than other regions.5 Therefore, compared to the labor market

approach, the ecological approach would result in relatively higher birth rates in

regions where the establishment size is relatively high and lower in regions

where the mean establishment size is relatively low.

These two approaches are compared in Table 1. This table presents

average annual firm birth rates by U. S. states for 1994-1996, based on both the

labor market approach and the ecological approach. The U.S. average firm birth

rate is 3.85 firms per thousand labor force. This is significantly higher that the

3.20 birth rate reported by Reynolds between 1986 and 1988 (1994, p. 433).

The parallel ecological birth rate is 13.0 firms per hundred establishments. Note

that in this table the states are ordered by their firm birth rates per thousand labor

force. When the corresponding rates per 100 establishments are examined, the

more extreme differences stand out. Thus we see some very low birth rates per

establishment among those states with high birth rates per labor force, in

Wyoming especially, and in Delaware, New York, and the District of Columbia.

Conversely, Illinois and Minnesota stand out as having high fm birth rates per

establishment, but low per labor force.

12

There are several interesting general observations about the variations

across states in Table 1. First, the areas with the highest firm birth rates are all

in the West or South. Colorado, Florida and Montana have the highest birth

rates per labor force, and Nevada and Utah are highest when standardized by

numbers of establishments. Second, the lowest birth rates are in the Northeast

and the Midwest. Third, states both large and small in the West have high rates

of new firm formation. What is surprising here is that many states that are

sparsely populated, such as Montana, Wyoming, Nevada, and Idaho, have

higher firm birth rates than most with denser populations. However, the states

are not very useful as a unit of analysis since substantial local differences are

lost in their aggregation. There is much more variation among LMA’s than

among states.

Table 2 shows the firm birth rates per thousand labor force, along with the

actual numbers of births and base number of establishments in the LMAs with

the highest and the lowest firm birth rates.6 The LMA birth rates range from

10.18 firms per thousand labor force, down to 2.06. These new firm formation

rates appear to be independent of regional size. For example, St. George UT,

which is one-thirtieth the size of Miami FL, has an identical new firm start up rate.

The regional differences appear even more pronounced at the LMA level. The

South and the West have the strongest new firm start-ups rates, while the

Northeast and the Midwest, which were formerly characterized by large scale

manufacturing, continue to lag behind the rest of the country. Dayton Ohio, for

13

instance, was dominated by the large-scale manufacturing of National Cash

Register, and has not yet restructured towards services.7 Finally, although not

shown here, we found very little variation in annual firm birth rates over the time

period studied. For example, the firm birth rate for Miami between 1991-1993

was 6.79 and between 1994-1996 it was 6.49. These figures are remarkably

consistent, given that 1991 was a recession year. For Dayton, Ohio the parallel

numbers were 2.34 and 2.54.

WHY DO BIRTH RATES VARY ACROSS REGION?

It is clear from the previous section that the regional firm birth rate varies greatly

across regions. Alternative explanations for these variations have been explored

in the theoretical and empirical literature. A long tradition of studies of the

determinants of new plant entry has focused on tax rates, transportation costs

and scale economies at the plant level (Bartick, 1989, Harrison and Kanter, 1978

and Kieschnick, 1981). More recently, a growing literature has sought the

determinants of variation in new firm formation on a regional basis (Reynolds

1991, 1994; Keeble and Walker, 1994; Mason, 1994; Audretsch and Fritsch,

1994; Reynolds, Miller and Maki, 1994; Guesnier, 1994; Sutaria, 1999). We

focus on four determinants of regional variation in the firm birth rate: the

existence of regional externalities (or agglomeration or density effects),

unemployment, industrial restructuring and entrepreneurial culture.

14

The agglomeration effects that contribute to new firm formation can come

from either demand effects, such as increases in population, or from regional

spillovers, such as labor market characteristics. Krugman’s (1991a and 1991b)

theory links firm birth rates to three types of spillovers within a region. The first

emanates from the observation by Marshall (1920) that a pooled labor market

yields increasing returns at a spatial level. Agglomerations are also conducive to

a greater provision of non-traded inputs. Such inputs are provided at both a

greater variety and a lower cost. The third source of convexities emanates from

what Acs, Audretsch and Feldman (1992 and 1994) termed economies of

information flows. Thus, new firms are most likely to be started in regions where

such spillovers are the greatest. This would suggest that both population density

and population growth would be positively related to new firm start-ups (Reynolds

1991). Such agglomerations would also benefit from high personal income

growth.

A rich literature exists in regional economics that sheds some light on how to

capture the extent to which pooled labor markets, non-pecuniary transactions,

and information spillovers exist. One approach suggests that the infrastructure of

services is more developed in regions that are more densely populated.

According to Krugman (1991a, p. 484), “The concentration of several firms in a

single location offers a pooled market for workers with industry-specific skills,

ensuring both a lower probability of unemployment and a lower probability of

labor shortage.” Thus the start-up rate for each industry sector should increase

15

with the existing density of establishments in each sector. Another view is that

localized industries tend to support the production of non-tradable specialized

inputs. Thirdly, informational spillovers give clustered firms a better production

function than isolated producers have. The high level of human capital embodied

in their general and specific skills is another mechanism by which new firm start-

ups are supported. Thus regions that are rich in this resource should have more

start-up activity. University graduates – especially engineers – provide a supply

of labor to local firms. New firm start-ups should be positively related to higher

average levels of education, and negatively related to the levels of unskilled and

semi-skilled workers in the region.

In most studies of new firm formation in the 1980s there was a heavy

emphasis on the explanatory power of unemployment (Evans and Leighton,

1990, Storey, 1991). Unemployment increased significantly in several countries

and stayed at very high levels over an extended period. It was suggested that

when workers are unemployed they might be more likely to start their own

businesses. The formation of new firms, in turn, may reduce the unemployment

rate as such persons start new firm employing not only the owner, but also

others. However this relationship is more complicated. Higher levels of

unemployment might also indicate a reduction in aggregate demand throughout a

regional economy, thereby putting downward pressure on the rate of new firm

formation (Storey and Johnson 1987). Storey (1991, 177) found that, generally

speaking, time series analyses point to unemployment being, ceteris paribus,

16

positively associated with new firm formation, whereas cross sectional, or pooled

cross sectional studies appear to indicate the reverse. Thus even the direction of

the effect of a region’s unemployment rate on new firm formation is

indeterminate. This may be due to differences in sectoral requirements for

startups, with the industry sectors that require relatively small amounts of capital

being more suitable for startups in periods of higher unemployment.

Also associated with studies of new firm formation from the 1980s was the

role of industrial restructuring. Industrial restructuring has been associated with

(1) the shift from manufacturing employment to services, (2) a reduction in both

firm and plant size, and (3) a shift to higher levels of technology. The shift from

manufacturing to services, which are usually less capital intensive than

manufacturing, could increase the rate of new firm formation. Regions that are

dominated by large branch plants or firms will have less new firm formation

(Mason 1994). Most new firm founders have either managerial or skilled labor

backgrounds. Consequently, the occupational structure of a city or region might

also be expected to influence the supply of new firm founders. The spatial

division of labor within multi-site enterprises has resulted in many peripheral

areas being dominated by externally-owned branch plants performing routine

assembly and production services.

An additional factor affecting a region’s firm birth rate is the extent of its

entrepreneurial culture. An entrepreneurial culture is defined as a social context

17

where entrepreneurial behavior is encouraged (Johannisson 1984). This culture

includes two interrelated aspects: first, the entrepreneurial orientation of the local

population, and second, the distribution of entrepreneurial characteristics

amongst local institutions such as the community/regional political leadership,

financial institutions, and educational institutions. Various authors have

emphasized that the strength of local entrepreneurial culture varies spatially,

although empirical testing of the relationship between local cultural and

entrepreneurial activity is difficult, and interpretations of the causes of such

variations differ.

An interpretation of the effect of local culture on entrepreneurial activity is

provided by Illers (1986), who draws upon the work of Danish ethnologists. He

suggests that at least three contrasting ‘life modes’ can be identified: self-

employment, ‘career’ and ‘wage-work’. These life-modes are culturally and

socially determined, and they influence the propensity of individuals to create

new businesses. In the self-employment life-mode the dominant job-related

motivation is to own the means of production and control the production process.

This cultural tradition is carried over from generation to generation. This life style

is most frequently found in rural areas characterized by independent and self-

reliant small-scale farmers, ranchers’ artisans and small business owners. It is

rare in areas dominated by large-scale operations.

18

The dominant value of the ‘wage-earner’ life-mode is the sale of one’s labor

at the highest possible price in order to maximize the utility of one’s leisure time.

Such individuals are therefore unlikely to set up new businesses, except possibly

if they were unemployed and unable to find alternative paid employment. This

life-mode is likely to be most common in localities and regions characterized by a

narrow industrial base and dominated by large externally owned firms. The

presence of large firms and secondary branch locations of firms characterize a

region should have a negative effect on the regional birth rate.

Finally, the dominant value of individuals with a ‘career’ life-mode is the

advancement of their career. They are likely to be well-educated and working in

large hierarchical private or public sector organizations. They will start their own

businesses if this becomes the best way in which to benefit from their skills,

knowledge and expertise. These businesses are often technologically advanced,

innovative and with good marketing capabilities. Career mode entrepreneurs are

often concentrated in large metropolitan areas and smaller attractive cities

(Savaage et al. 1988). In fact, the 1990’s saw a high incidence of highly

educated individuals starting new businesses, especially in the technologically

advanced sectors of the economy, like computers, biotechnology, internet

startups. New firm start-ups should be positively associated with higher levels of

educational attainment.

19

EMPIRICAL RESULTS

From the above discussion it should be clear that the major hypotheses

concerning the regional variation in firm birth rates are that (1) higher birth rates

are promoted by regional spillovers; (2) higher unemployment may deter start-

ups in some sectors and increase them in others; (3) industrial restructuring

should promote new firm formation; and (4) the existence of an entrepreneurial

culture should promote start-up activity. To test these hypotheses, we estimate a

regression model where the dependent variable is the 1994-1996 average

annual firm birth rate divided by the labor force (in thousands). This is analogous

to the method used by Keeble and Walker (1994) and Davidsson et al. (1994).

The explanatory (independent or exogenous) variables include the following:

Establishment size is a proxy for the structure of industry in the region. It

is measured as 1994 employment divided by the number of establishments in

1994 in the region. It should be negatively related to regional birth rate since

larger average establishment size indicates greater dominance by large firms or

branch plants.

In order to assess the potential for positive effects from spillovers, many

studies have measured density using the square root of the regional population,

or population per square mile. Such measures, however, do not indicate the

extent of pooled labor markets very well, since they tell us nothing about the

density of similar establishments in the region. Therefore, we introduce a new

20

measure that captures both population density and the number of establishments

in a region. Industry density is the number of establishments in the industry and

region in 1994 divided by the region’s 1994 population. The greater the number

of establishments relative to the population, the more spillovers should be

facilitated (Ciccone and Hall, 1996).

Population growth is the average annual rate of increase in the region in a

previous period (calculating the two-year change from the ratio of the 1994

population divided by 1992 population, and taking the square root of that two-

year change ratio to calculate the annual change ratio). Income growth is the

average annual rate of increase of personal income in the region from 1992 to

1994, similarly calculated. Both of these growth factors from the period

preceding our start-up measurement period are expected to promote new firm

start-ups in the subsequent 1994-96 period.

The unemployment rate is the traditional calculation for the first year of our

start-up measurement period -- the average number of unemployed in 1994

divided by 1994 labor force. Guesnier (1994), Audretsch and Fritsch (1994) and

Reynolds (1994) have used this measure. It is expected to be negatively related

to start-ups overall, but probably positively related to new firm start up rates in

industries with low capital requirements, and negatively related to those with high

capital requirements. The simple correlation between the unemployment rate

21

and the firm birth rate is close to zero, and is not statistically significant.

The share of proprietors in the economy is measured as the number of

proprietors in 1994 divided by the 1994 labor force. Proprietors are members of

the labor force who are also business owners. This measure averages 20.5

percent nationally, and varies from a low of 9.9 percent to a high of 44.8 percent

across LMA’s. It includes both the self-employed who have no employees, and

the owners of unincorporated businesses that have employees. As shown in

Table 3, the simple correlation between the regional birth rate and the share of

proprietors is 0.30, indication a moderately strong relationship between these

variables.

To measure the level of skills in the economy we include two measures of

educational attainment in each region. The first is the share of adults with no

high school degree, defined as the number of adults without a high school

degree in 1990 divided by the number of adults (population 25 years or older).

The lack of a high school degree should be a good proxy for the proportion of

unskilled and semi-skilled labor, and should be negatively related to the birth

rate. The mean percent of the population without a high school degree is 27

percent. In fact, the simple correlation between the percentage of the population

without a high school degree and the birth rate is – 0.19.

22

Finally, share of college graduates is defined as the number of adults with

college degrees in 1990 divided by the total number of adults. This is a proxy

measure of both technical skills needed in the economy, for example engineers

and scientists, and skills needed to start and build a business, like finance and

marketing and complex reasoning. In 1990 an average of 15.9 percent of the

adult population had a college degree. Its simple correlation with the regional

birth rates is positive. Summary statistics for all of the regional variables are

provided in Table 4.

Of course some of these variables may in fact be endogenous or

correlated with other variables. Although income and population growth were

measured for a previous two-year period, their regional differences are likely to

persist, and future growth differences certainly result from current differences in

startup rates. The share of proprietors and the agglomeration effects measured

by industry density may also be the effect of more firm startups, as well as

contributing factors. In fact, much of the economic geography literature today is

concerned with cumulative growth mechanisms in which cause and effect are

simultaneous. Our regression results should be interpreted carefully, and do not

necessarily imply causality.8

Table 5 shows the results of least squares regression on the 1994-1996

average annual firm birth rates for all industries together, and for each of the six

industry sectors, based on the 394 Labor Market Areas. We present

23

standardized beta coefficients, so that each parameter indicates the sensitivity of

birth rate variation to normalized variation in the corresponding independent

variable. The t-ratios shown for each were calculated from the simple estimated

standard errors. These were recalculated with a correction for

heteroscedasticity, and those were very similar to the uncorrected standard

errors. The estimated coefficients are generally consistent with our expectations,

but with several important exceptions. The explanatory and control variables

together explained two-thirds of the regional differences in firm start-up rates.

Industry density and population growth are both strongly positive and

statistically significant, as predicted by the theory of regional spillovers (Krugman,

1991a and 1991b). In fact, these coefficients are nearly three times as large as

the coefficient on income growth, which is also positive and significant. These

results are consistent with Reynolds (1994), Keeble and Walker (1994) and

Audretsch and Fritsch (1994). When analyzed separately for each of the industry

sectors, we find that both industry density and population growth are positive and

significant for each of the six industry sectors. However, the parameters on

income growth are only significant for business services and local market.

The coefficient for establishment size is negative and statistically

significant, indicating that regions with predominately smaller establishments

have a higher start-up rate than regions with more large establishments. This

supports the thesis that regions that have already restructured away from large

24

manufacturing dominance have a higher start-up rate than regions that have not.

These results are consistent with the findings of Audretsch and Fritsch (1994).

The coefficient for the unemployment rate is positive, although it is tiny

and not statistically significant at the all-industry level. This result is surprising,

given that previous cross-sectional studies have generally found a consistently

negative result (Storey, 1991, and Audretsch and Fritsch, 1994). Furthermore,

the coefficients on unemployment were positive for all of the six sectors, and

significantly so for all but the extractive industries. Perhaps the exceptionally low

levels of unemployment and even shortages of labor in the United States in the

1990’s account for the prevailingly positive relationship between unemployment

and new firm births in this period. The implication here is that as workers shift

from being employed to unemployed, the overall entry rate in the region tends to

go up slightly, although there is no evidence that it is necessarily the unemployed

who are starting the new firms.

The coefficient on the share of proprietors in the region is negative and

statistically insignificant for the all-industry equation, perhaps because the share

of proprietors is strongly negatively correlated with establishment size, -0.63. As

the average establishment size in a region increases there are fewer

opportunities for self-employment, and a smaller proportion of the labor force is

made up of owners. When we drop establishment size from the estimated

regression, the coefficient on self-employment becomes positive and statistically

25

significant, while the other variables remain virtually unchanged. Within several

of the industry sectors – local market, manufacturing, and retail – the share of

proprietors is significantly positively related to firm birth rates. The separate

regressions for the six industry sectors are not very sensitive to the presence of

the establishment size variable. It has positive and statistically significant

coefficients for local market, manufacturing and retail, but its presence somewhat

reduces the t-statistics on other variables.

Finally, the coefficient for human capital, as measured by share of college

graduates, is positive and statistically significant, suggesting that regions that

have higher levels of education will have higher start-up rates. This is consistent

with Savage et al. (1988) and Anselin, Varga and Acs (1997, 2000), who found

that in technologically advanced industry individuals with greater skills,

knowledge and expertise are more likely to start businesses. However, for both

business services and manufacturing this coefficient is only barely positive, and

not statistically significant. Reynolds (1994) found a negative and statistically

significant relationship between college education and the new firm birth rate in

manufacturing. The results do suggest that manufacturing firms may behave

differently than other sectors of the economy.

The positive and statistically significant coefficient for the percentage of

the population without a high school degree is at first surprising, but it is easily

explained. As shown in Table 3, the correlation between the share with no high

26

school degree and the new firm start up rate has a negative coefficient, as

expected. However, it is much more strongly negatively correlated with college

education, with a coefficient of - 0.70. After controlling for the proportion of

adults with college degrees, the additional effect of a greater share of less

educated workers is to facilitate the start-up process by providing cheap labor for

the new firms. Even the most sophisticated businesses need some workers who

are less educated, to do the manual labor. This positive impact of ‘no high

school degree’ after controlling for ‘college degree’ is consistent across most of

the industry sectors, except for business services, with the strongest positive

relationship appearing in the distributive industries.

We use the estimated relationships between regional birth rates of firms

and the regions’ socio-economic characteristics to predict their expected firm

birth rates, given their actual socio-economic characteristics. The difference

between an expected rate and the actual rate is called the “unexplained

variation.” This unexplained variation is a candidate for explanation by regional

factors that have not been included in this analysis. These could include

exceptional resources, particularly effective policies, unusually strong financial

infrastructure for new firms, local research and development centers, and historic

concentrations of rapidly growing or restructuring industries. Further research on

regional variation in firm startup rates should focus on finding which additional

variables are most useful for explaining the remaining unexplained variation, or

residuals.

27

CONCLUSIONS

This paper has reexamined the issue of new firm formation in light of recent

theoretical developments in economic geography and new growth theory. Using

a new longitudinal microdata source, we constructed annual data on firm births

for 384 labor market areas, in six industry sectors, between 1991-1996. We find

considerable variation in the new firm formation rate across regions, but much

less variation over time. The data show significant differences in new firm

formation rates from the industrial northeast to the technologically progressive

southwest. Variations in the firm birth rates are substantially explained by

regional differences in industry intensity, population growth and income growth,

as suggested by the new economic geography.

We also find significant evidence on the importance of human capital on

new firm formation rates. People in regions that have a high percentage of

college graduates are much more likely to start businesses than those in regions

with high concentrations of less skilled workers. The strong negative relationship

between new firm formation rates and establishment size indicates that regions

that have restructured have higher rates of new formation than regions that have

not. We find a little support for a positive impact of unemployment on new firm

formation rates. This is explained in part by the fact that in the 1990’s

unemployment problems were replaced by a labor shortage in the United States.

28

Our findings suggest that the market-size (agglomeration) effects

emphasized by the current generation of new geography models may be less

important than other kinds of external economies. According to Krugman, (1998)

big cities may be sustained by increasing returns that are due to thick labor

markets, or to localized knowledge spillovers, rather than spillovers that emerge

from the interaction of transport costs and scale economies at the plant level.

29

REFERENCES

Acs, Z. J., Audretsch D. B. and Feldman M. P. (1992) Real effects of academic

research: comment, Am. Econ. Rev. 82(1), 363-7.

Acs, Z. J., Audretsch D. B. and Feldman M. P. (1994) R&D spillovers and

recipient firm size, Rev. Econ. Statistics 99, 336-340.

Acs, Z. J., Carlsson B. and Karlsson C. (1999) Entrepreneurship, Small &

Medium-Sized Enterprises and the Macroeconomy. Cambridge University Press,

Cambridge, UK.

Acs, Z. J., Morck R. and Yeung, B. (1999) Productivity growth and firm size

distribution, in Acs, Z. J. et al (eds) Entrepreneurship, Small & Medium-Sized

Enterprises and the Macroeconomy, Cambridge University Press, Cambridge,

UK 367-398.

Anselin, L., Varga A. and Acs Z. J. (1997) Local geographic spillovers between

university research and high technology innovation, J. of Urban Econ., 42, 422-

448.

Anselin, L., Varga A. and Acs Z. J. (2000) Geographic spillovers and university

research: a spatial econometric approach, Growth and Change, 31, forthcoming.

30

Acs, Z. J., Armington C. (1998) Longitudinal Establishment and Enterprise

Microdata (LEEM) Documentation, Center for Economic Studies, U. S. Bureau of

the Census, Washington D. C. CES 98-9.

Armington, C (1995) ‘Deriving Establishment Births from Unemployment

Insurance Data,’ Proceedings of the 1995 Annual Research Conference, Bureau

of the Census, U.S. Dept. of Commerce.

Audretsch, D. B., and Feldman M. P. (1996) R&D spillovers and the geography

of innovation and production, Am. Econ. Rev. 86, 630-640.

Audretsch D. B. and Fritsch M. (1994) The Geography of firm births in Germany,

Regional Studies 28(4) 359-365.

Bartik, T. J. (1989) Small business start-ups in the United States: Estimates of

the effects of characteristics of states, Southern Econ. J. 55, 1004-1018.

Ciccone, C. and Hall, R. E. (1996) Productivity and the Density of Economic

Activity, American Economic Review, 86(1), 54-70.

Davidson, P. Lindmark, L. and Olofsson C. (1994) New firm formation and

regional development in Sweden, Reg. Stud. 38(4), 395-410.

31

Evans, D. and Jovanovic B. (1989) Estimates of a model of entrepreneurial

choice under liquidity constraints, J. Pol. Econ. 95, 657-74.

Evans, D. and Leighton, L. S. (1990) Small business formation by unemployed

and employed workers, Small Bus. Econ. 2, 319-330.

Gerowski, P., (1994) What do we know about entry?, International Journal of

Industrial Organization 13(4), 421-441.

Guesnier, B., (1994) Regional variations in new firm formation in France,

Regional Studies 28, 347-358.

Harrison, B. and Kanter, S. (1979) The political economy of states: job-creation

business incentives, American Institute of Planning Journal 425-35.

Hoover, E. and Vernon, R. (1959) Anatomy of a Metropolis, Harvard, Cambridge,

MA.

Illeris, S. (1986) New firm creation in Denmark: the importance of the cultural

background, in Keeble B. and Wever E. (eds) New Firms and Regional

Development in Europe, Beckerham, Croom Helm.

32

Johannisson, B. (1984) A cultural perspectives on small business – local climate,

Int. Small Business Journal 2(4).

Keeble, D. and Walker, S. (1994) New firms, small firms and dead firms: spatial

patterns and determinants in the United Kingdom, Regional Studies, 28(4), 411-

4427.

Kieschnick, M. D. (1981) Taxes and growth: business incentives and economic

development Washington. D. C. Council of State Planning Agencies.

Krugman, P. (1991a) Increasing returns and economic geography, Journal of

Political Economy 99, 483-99.

Krugman, P. (1991b) History and industry location: the case of the

manufacturing belt, American Economic Review, 81, 80-83.

Krugman, P. (1994) First nature, second nature, and metropolitan location,

Journal of Regional Science, 33, 129-144.

Krugman, P. (1998) Space: The final frontier, Journal of Economic Perspectives,

12, 161-174.

Marshall A. (1920) Principles of Economics 8th edition. Macmillan, London.

33

Mason, C., (1994) Spatial variations in enterprise: the geography of new firm

formation, in R Burows, Ed, Deciphering the Enterprise Culture, N.Y. Routledge,

pp 74-106.

Nijkamp, P., Poot, J. (1989) Spatial perspectives on new theories of economic

growth, The Annals of Regional Science, 32(1), 7-38.

Reynolds, P. (1991) Regional characteristics affecting business growth:

Assessing strategies promoting regional economic well-being, submitted to the

Ford Foundation, Grant No. 900-013.

Reynolds, P. (1994) Autonomous firm dynamics and economic growth in the

United States, 1986-1990, Regional Studies 28(4) 429-442.

Reynolds, P. Miller, B. and Maki, W. R. (1994) Regional characteristics affecting

business volatility in the United States, 1980-1984, in Karlsson C. et al (eds.)

Small Business Dynamics: International, National and Regional Perspectives,

London, Routledge, 78-115.

Reynolds, P. (1999) Creative destruction, in Acs, et al (eds.) Entrepreneurship,

Small & Medium-sized Enterprises and the Macroeconomy, Cambridge

University Press, Cambridge U.K.

34

Romer, P. (1990) Endogenous technological change, Journal of Political

Economy 98, S71-S102.

Savage, M., Dickens, P. and Fielding, T. (1988) Some social and political

implications of the contemporary fragmentation of the service class in Britain, Int.

J. of Urban and Reg. Res., 12(3).

Storey, D. J. (1991) The birth of new firms—does unemployment matter? a

review of the evidence, Small Bus. Econ. 3(3) 167-78.

Storey D. J. and Johnson A. M. (1987) Job Generation and Labour Market

Change, Macmillan, London.

Sutaria, V. (forthcoming) The Dynamics of New Firm Formation, Ashgate,

London.

Tolbert, C.M. and M. Sizer, (1996) ‘U.S. Commuting Zones and Labor Market

Areas: a 1990 Update,’ Rural Economy Division, Economic Research Service,

U.S. Department of Agriculture, Staff Paper No. AGES-9614.

U.S. Small Business Administration (1999) Statistics of U.S. Business –

Microdata and Tables of SBA/Census Data, Office of Advocacy, Washington DC.

35

This research was initiated and supported by the Kauffman Center for Entrepreneurial Leadershipat the Ewing Marrin Kauffman Foundation, as the first step of a larger project to analyze thecauses of regional differences in new firm formation rates in the United States. The research wascarried out at the Center for Economic Studies (CES), U. S. Bureau of the Census Washington D.C. under the title, “U. S. Geographical Diversity in Business Entry Rates.” We would like to thanktwo anonymous referees and the editor of this Journal for helpful comments. Research resultsand conclusions expressed are those of the authors and do not necessarily indicate concurrenceby the Bureau of the Census or the Center for Economic Studies. All errors and omissions areour responsibility.

1 The SUSB data and their Longitudinal Pointer File were constructed by Census under contractto the Office of Advocacy of the U.S. Small Business Administration. For documentation of theSUSB files, see Armington (1998).

2 There is a small number (10,000 to 16,000) of new firms each year for which no industry code isever available. Most of these are small and short lived. These have been added to the Localmarket category, which is, by far, the largest of our sectors.

3 We tested a similar rule using one-half, and found that the primary difference was in quite smallmulti-unit firms, where the smaller share was credible for the first year.

4 In fact, birth rates were calculated for each annual period from 1990 through 1996, but thesewere found to be quite consistent in their rank ordering across LMA’s, so the average of the twomost recent years were used for most of this analysis. Using two years of births, rather than asingle year, minimized problems with very small numbers of births for the smaller sectors andLMAs.

5 In particular, manufacturing establishments generally are larger than others, so areas with largershares of manufacturing have relatively fewer establishments for their labor force size.

6 The complete list of LMAs ranked by the firm birth rate is reported in Appendix A.

7 It is a well-documented regularity that both plants and firms in large cities tend to be smallerthan those in small cities (Hoover and Vernon, 1959). This would suggest that smaller industrialcities might have the most difficulty restructuring.

8 We have also abstained from considering financial variables and regional knowledge factorssuch as research and development expenditures. The availability of adequate financial resourcesto fund new firms is an important determinant of new firm formation, which we hope to take intoaccount in subsequent research. Both university-based and industrial research and developmentactivity are probably important contributors to regional new firm start-up rates.