Printer Friendly

The growth of small firm jobs by state, 1984-88.

DURING THE 1980s, there was ample documentation of the growing importance of small firms in creating new jobs, new products and increasing income. A portion of this research also emphasized the major role played by new small firm births, particularly births of new firms with fewer than twenty employees.(1) New and expanding small firms with fewer than twenty employees provided about 19 percent of total employment in 1988 but created more than twice their expected share of new jobs (39.5 percent) between 1984 and 1988. Clearly the smallest firms have become recognized as the most dynamic growers in the economy.

Much of the literature on job creation has, in fact, centered on these rapid growers with fewer than twenty employees, and many state programs have been designed to aid in their formation.(2) In spite of these state programs, however, the share of new jobs created by these smallest firms has also shown great variation. For example, job creation by the smallest firms ranged from 8 percent in Delaware to well over 100 percent in Texas (where large firms lost jobs) between 1984 and 1988.(3) Therefore it is of interest from both a state policy perspective, as well as a national growth perspective, to understand how job creation varies geographically by firms with fewer than twenty employees.

This paper will explain how and why the share of jobs created by very small firms varied among the states between 1984 and 1988. Because per-capita income growth is associated with increased small firm births and expansions of existing small firms, we attempt to quantify the optimal spatial environment in which this activity flourishes.(4) The model below examines the role many of the presumed determinants of growth play on a state-by-state (e.g., cross section) basis in influencing the share of new jobs created by small firms with fewer than twenty employees.

This paper is also of particular interest to larger firms because small firms provide growing markets for the largest companies. In particular, from computers and fax machines in home-based firms to banking services in new firms, small businesses function as both suppliers and purchasers in many newly emerging markets.


There has been considerable disagreement about whether small firms are more likely to enter growing industries or declining industries. Apparently the answer depends upon the life cycle of the industry, and whether the industry is initially dominated by small firms or large firms. For example, as segments of the service sector matured during the 1980s, small firms, while growing absolutely, lost market shares in subsectors that also grew during the 1980s, such as business services. On the other hand, gains in employment and sales shares were recorded by small firms in declining industries, such as the primary metal sector.(5) The former sector, business services, was initially dominated by small firms, while primary metal was historically a big business industry. Some of this recent growth has therefore been counterintuitive.

Other studies have gone further in trying to generalize the relationship between industry growth and small firm growth, on both an absolute and share basis. In a paper recently prepared for the Office of Advocacy of the SBA, David Evans (1991) observed "that there is a weak tendency for industries in which the small business share has increased to be industries in which total employment is declining, and that there is a weak to moderate relationship between increases in the number of small firms and industry expansions." Evans used data from the Small Business Data Base from 1976 to 1986 and found these generalizations lacked significance.(6) Apparently, Schumpeter aside, it is quite difficult to predict empirically how the small firm employment share will change in specific industries based upon historical generalizations.


The following factors are hypothesized to have influenced the share of jobs created by state between 1984 and 1988 (PCTJOBS). The source of data for the dependent variables are the state job creation data based upon the 1984-88 Small Business Data Base (USEEM) files.(7)

The Presence of Large Firms. As summarized above, the presence of large employers is assumed to exert a negative influence on the birth of new small firms; large organizations generally exert a negative influence on entrepreneurial spinoffs, e.g., new firm births. The variable tested in our model below is LFEMP, the percentage of the labor force employed by firms with 500 + employees in 1988, by state. Source: USEEM files.

Small Firm Births. A continuing supply of new small firms is generally a prerequisite for a growing small business sector with continuing employment generation, In addition, a continuing supply of new enterpreneurs is generally assumed to emanate from the births of new small firms, further increasing the share of new jobs created by these smallest firms.(8)

Rather than use the share of jobs contributed by births of all small firms with under twenty employees, we instead chose to use the birth and death rates of small high-technology firms, principally because we expected a stronger relationship between job growth and this rapidly increasing segment of the small firm sector.(9) In addition, using high-tech birth and death rates acts as a proxy to measure the influence of innovation by state. Because almost all new high-technology firms are based upon a new product or process, or a more efficient method of doing an established task, this variable acts as an indirect test of the state innovation rate.

The exact hypothesis being tested is therefore that the percentage of jobs contributed by small firms with fewer than twenty employees varies directly with small firm high-technology startups (+ relationship) and inversely with the small firm high-tech death rate (- relationship). The mnemonics in the regression model are: SFBIRTHS and SFDEATHS.(10)

Education of the Labor Force. Most studies of entrepreneurship have shown that a weak but positive relationship exists between educational attainment and the probability of starting a business.(11) However the direction of this relationship is also dependent on the state of the economy, the demographic group under study, and the specific industry or industries being studied. Holding constant other influences on business formation other than education, our hypothesis is that the larger the percentage of college graduates in a state's population (CG), the higher the probability of entrepreneurship, and therefore the larger the share of jobs contributed by firms with fewer than twenty employees.

Population 65 + Years Old. Including a measure of the older population 65 + years in each state serves two purposes. First, from the demand side, it proxies a clearly growing market that many new small firms have begun to serve. And from the supply side, it proxies a potential supply of entrepreneurs who take up business ownership after retiring from full-time wage and salary jobs, or who continue to own a side business. We therefore hypothesize that a larger-than-average share of older residents by state (POP65) is positively associated with greater business formation, and hence a larger share of new jobs created by firms with fewer than 20 employees.

Government Assistance. We hypothesize that a differentially higher level of government assistance is likely to result in raising the birth rate of new enterprises by state. In the present case, we expect a positive relationship between state job creation by the smallest firms and BDSTAFF -- the number of full-time business development staff positions in SBA offices in each state in 1988.

Capital Availability. As in the case of government assistance, many ways are possible to proxy the availability of financial resources available to new entrepreneurs. Other things equal, one would expect the share of jobs created in each state by new small firm to be positively related to the availability of these resources, perhaps with a lag. In the present case, the number of venture capital offices by state VCAP was expected to influence job creation positively.(12)

State Business Demographics. Average Firm Age (AGE) is a lagged variable to reflect recent new business formation. States with the youngest average age structure of their component firms would contain a greater proportion of recent small firm births and were therefore more likely to contribute a larger share of new jobs.(13) Lagged New Business Incorporations (INCORPS) are the change in new business incorporations between 1984 and 1988 by state. They also function as a concurrent indicator of new business formation. The variable, also expected to have a positive sign, is computed from Dun and Bradstreet press releases.(14)

Industry Structure Variables. Two variables are used to test for the greater presence of small firm jobs in growing sectors. First, we use the share of small firm employment in growing industries to represent opportunities for new small firms. Specifically,

|Mathematical Expression Omitted~


Growing Sectors are defined as those industries that grew above the national rate between 1984 and 1988. The variables come from the USEEM files of the Small Business Data Base.

The growth of small firms in manufacturing between 1982 and 1987 (SFMFG) was also used to test or proxy demand growth in each state on the assumption that states with growing small firm manufacturing sectors would "spinoff" additional opportunities for small firms in other sectors.(15) The variable was computed from the 1987 Census of Manufacturing and expressed in percentage terms.

The Presence of Military and Government Employment (GOVMIL). In this paper, we hypothesize that the presence of large clusters of military and government employment substitutes for private risktaking in many instances. As such, it is widely believed that entrepreneurship is depressed in states with dominant employment clusters of military and civilian employment. This variable, therefore, will exert a negative influence on the distribution of small firm jobs by state, i.e., fewer new firms are expected in government-dependent areas, lowering the share of jobs contributed by the smallest firms in those states.

Table 1 summarizes the expected directions of casuality on the variables used in the model to explain the jobs created by firms with fewer than twenty employees in each state between 1984 and 1988:
Table 1
Expected Signs on Independent Variables:
Correlates of the Share of Jobs Created by
Firms with Fewer than 20 Employees
Variable Definition Expected Sign
LFEMP Pct. of labor force employed by -
 large firms, 1984
SFBIRTHS Rank of small firm births in +
 high-tech industries, 1984
SFDEATHS Rank of small firm deaths in -
 high-tech industries, 1984
CG Percentage of college graduates +
 by state, 1988
POP65 Percentage of population 65 + +
 yrs. old, 1988
BDSTAFF Number of business develop- +
 ment staff specialists, 1989
VCAP Number of state venture capital +
 offices, 1988
AGE Median age of firms by state, +
INCORPS Change in new business +
 incorporations, 1984-1988
GROWSF Percentage of small firms in +
 growing industries, 1984
SFMFG Growth of small firms in +
 manufacturing, 1982-1987
GOVMIL Percentage of state employment -
 in government and military,

As observed in Table 1, all of the factors except the presence of large firms, an above-average death rate of small firms, and an above-average dependence on military and government employment are expected to influence positively the share of jobs created by the smallest firms in each state.


The linear regression model results may be summarized as follows. The explanatory variables accounted for about one-third of the variance in the share of jobs created by the smallest firms. In particular, three variables contributed the most to the explained variance and were significant at the .95 level: the birth rate of small high-tech firms (SFBIRTH), the level of business development assistance (BDSTAFF), and the small firm share of growth industries (GROWSF). The education variable (CSG) was significant at the .90 level, while the variables measuring small firm death rates and the unlagged change in new business incorporations were not significant.(16)

It is not surprising that greater small firm high-tech births are associated with greater-than-average job-creation shares by the smallest firms. Apparently state programs to attract high technology firms do indeed bring disproportionate results if successful. As other research has shown, (Phillips, 1990b), states with higher innovation rates and clusters of high-technology firms do indeed contribute disproportionate job growth: high-tech firms expand faster than non high-technology firms.(17)

Nor is it surprising that new small firms create a larger share of new jobs when they form a larger segment of growing industries in a state. Moreover, as hypothesized, states with above-average concentrations of college graduates also tend to have higher small firm births and greater job creation, other things equal. Although significant, the CSG variable contributed less than 10 percent to the total explained variance (simple correlation of .26). Therefore, as the literature indicates, the education effect, taken by itself, is positive but not very strong.

Two other hypothesized effects were insignificant: the change in new business incorporations (INCORPS), and the small firm death rate (SFDEATH). The lack of significance of the new incorporations variable may be due to misspecification; either the variable should be lagged, or new incorporations may not accurately reflect the changing business climate in a state because of changes in the legal form of a business or changes in business locations. In addition, as observed in other studies, new incorporations and new job growth are not simultaneous events; there is not much job creation by a business until it has survived for at least four to six years. On the other hand, death rates do not exhibit much variance either longitudinally or cross sectionally and therefore may not be correlated with above-average rates of job creation either (simple r = -.13).(18)

Another set of equations introduced the separate effect of the government/military variable (GOVMIL). Again, we hypothesized that large concentrations of military and government employment tend to reduce differentially the birth and job creation rate of the smallest firms.

The variable, while being significant and having the expected negative sign, did little to increase the explanatory power of the models. Although the variable was significant when interacted with the rest of the variables, the simple correlation with the dependent variable was only -.14 and insignificant. Therefore while high levels of military/government employment clearly do not increase employment growth by the smallest firms, the statistical significance of the effect appears small.

Table 2 evaluates each of our hypotheses at the means of the respective independent variables. The table indicates the effect of a 1 percentage point increase in the respective dependent variable given a 1 percentage point increase in the respective independent variable:
Table 2
Elasticities of the Independent Variables
Evaluated at Their Means
Variable Elasticity Y or N
GOVMIL - 1.19 Y
CSG 1.11 Y
LFEMP 0.97 N
POP65 - 0.56 N
AVAGE - 0.20 N
SFDTH - 0.14 N

Several interesting observations emerge from Table 2. First, the largest effect on the share of jobs created by the smallest firms is caused when small firms make up a larger than average share of growth industries in the state (GROWSF). For each 1 percent increase in the share of small firms in a state's growing sectors, the share of jobs contributed by these smallest firms rises 1.68 percent, a fairly elastic result. The role of college graduates (CSG) is next in importance. Each increase of 1 percent in a state's college educated population results in a 1.2 percent increase in jobs created by the smallest firms, on average. In addition, as observed above, the presence of state business development assistance professionals (BDSTAFF) and a prior record of significant small firm births also add modest influences to the job creation process.(19)

On the negative side, a significant influence is caused in each area by the presence of military/government employment, with an elasticity of -1.2. Clearly this can cause a downward effect on small firm job creation, and areas such as the Washington, DC, MSA probably do indeed have a smaller share of new jobs contributed by the smallest firms for just this reason. The implication for such areas is that it is especially difficult to build an appropriate entrepreneurial climate because of the risk-averse effect of such large institutions. Notice that the pure effect of large firm employment, by itself, was not significant, implying that the presence of large organizations do not necessarily depress a state's entrepreneurial climate. Rather, it may be the presence of large business, combined with the presence of large scale military/government employment, that causes a depressing effect on entrepreneurship.


At the beginning of this paper, we set out to explain why the share of jobs created by the smallest firms varied by state from 1984 to 1988, and we advanced a number of hypotheses that were empirically tested to examine this relationship. We can now report that some of our hypotheses were confirmed. In general our models explained about one-third of the total variation in the percentage of jobs created by the smallest firms by state -- a significant, but nonetheless not a completely satisfactory result.

We have had a number of indications that new small firms contribute more jobs when they are part of growing, expanding, or emerging industries. In particular, the significance of the high technology birth variable (SFBIRTHS) combined with the significance of the variable that measured the presence of small firms in growing industries (GROWSF) indicate that new small firms contribute a larger share of new jobs when they are in high-tech related sectors. Therefore our results confirm that state programs that encourage the formation of locally grown high-technology industries will have an expected payoff in raising the overall job creation share contributed by the smallest firms, even if the high tech development itself represents a small part of total job growth. Put in other words, the spinoffs effects from these activities have large and expected effects.

Our results also indicate that continued investments in our education system, in terms of increasing the college trained population in each state, have the effect of increasing job creation in that state. Therefore local educational investments clearly have a significant effect, although not a major one, in increasing locally based entrepreneurship, other factors equal.

Our third major finding has to do with the negative influences on job creation shares. In particular, our results have shown that the presence of large military/government complexes tends to depress the entrepreneurial spirit. Therefore states with large components of government or military employment must be especially diligent in maintaining a wide variety of programs to stimulate locally based entrepreneurship. Of course, it is possible that the research and development expenditures of the military/government complex tend to dominate in these areas and substitute for private research and development expenditures. The regression results indicated that a 1 percentage point increase in military/government employment in a state tended to lower the share of jobs created in that state by the smallest firms by almost 1.7 percent, the largest effect observed, but of only marginal significance statistically.

Some of our other hypothesis that were not significant also deserve mention. The presence of older persons, by itself, exerted no particular influence on entrepreneurship. Perhaps this reflects the notion that while a large elderly population provides a good consumption base, it is less clear that such areas, on average, cause a large increase in new firms. Similarly, the separate influence of large firms in depressing job creation shares, an expected result, turned out to be insignificant. Therefore, it is perhaps the presence of government and military employment, influencing larger scale operations, in the Middle Atlantic states that acts to lower the share of jobs created by the smallest firms, rather than the presence of large firms by themselves. Finally, the presence of venture capital did little to explain job creation shares by the smallest firms, perhaps because of the concentrated nature of venture capital.

We note finally the exploratory nature of this research. Cross-section models are useful to indicate tendencies and directions, rather than absolute truths. However our models do suggest further directions for explaining how and why the jobs created by the smallest firms vary across states.

Bruce D. Phillips is Director, Data Base Development, U.S. Small Business Administration, Office of Advocacy, Washington, DC. An earlier version of this paper was presented at the 34th Annual Meeting of the National Association of Business Economists, Dallas, TX, September 13-16, 1992. Comments are those of the author and do not necessarily represent views of the U.S. Small Business Administration.


1 See, for example, Executive Office of the President, The State of Small Business: A Report of the President (Washington, D.D., Government Printing Office, editions of 1982-1991); Bruce D. Phillips, "The Increasing Importance of Small Firms in the High Technology Sector: Evidence from the 1980s," Business Economics, 26 (1), January, 1991, pp. 40-47; Bruce A. Kirchoff and Bruce D. Phillips, "Employment Growth in the Decade of the Entrepreneur," unpublished paper prepared for the Babson Entrepreneurship Research Conference, Pittsburgh, Pa., April, 1991; Paul D. Reynolds and Wilbur A. Maki, "Business Volatility and Economic Growth," prepared under contract for the Office of Advocacy of the U.S. Small Business Administration, May, 1990.

2 A good summary is found in David Osborne, State Technology Programs: A Preliminary Analysis of Lessons Learned (Council of State Policy and Planning Agencies, Washington, DC, November, 1989).

3 U.S. Small Business Administration, Office of Advocacy, Small Business Data Base, unpublished data, USELM file version 8, 1990.

4 See Reynolds and Maki, op. cit. See also Paul D. Reynolds, Brenda Miller, and Wilbur R. Maki, "Regional Characteristics Affecting Business Volatility: U.S. 1980-1984," unpublished paper presented at a conference on the Formation, Management, and Organization of Small and Medium Sized Enterprises, Jonkoping, Sweden, September 22-27, 1991.

5 Based upon Table 3 from U.S. Bureau of the Census, 1987 Enterprise Statistics, ES-87-1, "Company Statistics," (Government Printing Office, 1991). Note that on an absolute basis, small firms grew in both declining and expanding industries during the 1980s. However, on a relative basis, their market dominance seemed to increase in areas that lost absolute jobs on a national basis.

6 David S. Evans, "Industry Dynamics and Small Firms in the United States." Prepared under contract SBA-3069-AER-88 for the Office of Advocacy of the U.S. Small Business Administration, September, 1991.

7 In the discussions in this and subsequent sections, the terms Small Business Data Base and U.S. Establishment and Enterprise Microdata (USEEM) files are used interchangeably. While the model discussed uses the share of jobs created by firms with fewer than twenty employees between 1984 and 1988 as the dependent variable, two other variable forms were also tested. These were the shares of jobs created by state by firms with fewer than twenty employees between 1984-86 and between 1986-88, respectively. The majority of the latter results were insignificant, and are not discussed further here. Further details are available from the author.

8 See, in particular, Michael Fritsch, "Regional Differences in New Firm Formation: Evidence from West Germany," Regional Studies, 26 (3), 1992, pp. 233-241.

9 See, in particular, Bruce D. Phillips, "The Increasing Importance of Small Firms in the High Technology Sector: Evidence from the 1980s," Business Economics, 26 (1), January, 1991, pp. 40-47.

10 Operationally, the ranks of state high-tech birth rates between 1980 and 1984 are used in the model, implying a four-year lag between firm births and the creation of jobs. The exact variable is taken from William J. Dennis and Bruce D. Phillips, "The Synergism of Independent High-Technology Business Starts," Enterpreneurship and Regional Development, 2 (1990), p. 1-14. Other research by Phillips and Kirchoff has shown that most new small firms do not create many jobs before their fourth or fifth year of operations, so the assumption of a four-year lag on the birth rate variable seems a reasonable assumption. See Bruce D. Phillips and Bruce A. Kirchoff, "Formation, Growth and Survival; Small Firm Dynamics in the U.S. Economy," Small Business Economics, 1 (1), 1989, pp. 65-74. High-tech death rates in the above are also used, derived from the 1976-84 files of the Small Business Data Base.

11 For a good summary of this literature see David S. Evans and Linda S. Leighton, "Some Empirical Aspects of Entrepreneurship," American Economic Review, 79 (3), 1989, pp. 519-535; see also Richard J.Boden, "Gender Differences in Entrepreneurial Selection and Performance," Ph.D. dissertation, University of Maryland, 1990.

12 This variable was kindly provided by Professor Richard Florida of Carnegie-Mellon University.

13 In 1988, Florida and Arizona had, on average, the youngest firms in the country at ten years each on the Small Business Data Base files.

14 One variable omitted from this section is the influence of the state's unemployment rate. This is because there is no agreement on the direction of influence of the unemployment rate on starting businesses in the short run. For two good examples of this literature see Richard Highfield and Robert Smiley, "New Business Starts and Economic Activity: An Empirical Investigation ," International Journal of Industrial Organization, 5 (1986), pp. 51-66; David J. Storey, "The Birth of New Firms -- Does Unemployment Matter? A Review of the Evidence," Small Business Economics, 3 (3), September 1991, pp. 167-179.

15 This was the same argument made with the use of the large firm variable above.

16 Detailed regression results are available from the author directly.

17 See Bruce D. Phillips, "The Importance of Small Firms in High-Technology Industries," unpublished paper prepared for the Western Economic Association Meetings, San Diego, California, July, 1990.

18 See, in particular, Catherine Armington, "Removing Business Restructuring From Data on Startups and Closures," Proceedings of the American Statistical Association, 1986, pp. 260-265.

19 It is possible that the effects of the latter two variables will be larger when alternative lead/lag structures of the model can be tested.
COPYRIGHT 1993 The National Association for Business Economists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Phillipa, Bruce D.
Publication:Business Economics
Date:Apr 1, 1993
Previous Article:Banking conditions and the credit crunch: implications for monetary growth and the economy.
Next Article:The business economist at work: an economist's work in a city planning department.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters