An empirical analysis of the relationship between employment growth and entrepreneurial activity.

Adrangi, Bahram ; Allender, Mary E. ; Anderson, Robin 等

ABSTRACT

This study investigates the role of entrepreneurial firms in employment creation and business expansion in several industries. Regression results validate the previous research findings that small entrepreneurial firms do have a net positive impact on new employment across all industries in the study. However, the magnitude of job creation may vary across industry types.

INTRODUCTION

The study of entrepreneurship has long recognized entrepreneurial activity as directly related to the creation of employment opportunities. Schumpeter viewed the source of these employment opportunities as "the gales of creative destruction" (Schumpeter, 1934). New industries are born as older ones are replaced. As an example, "In the case of retail trade the competition that matters arises not from additional shops of the same type, but from the department store, the chain store and the mail-order house and the supermarket." (Schumpeter, 1942). Schumpeter's model stood in stark contrast to the more dominant neoclassical general equilibrium model of the time in which the entrepreneur played no role. In fact, in the neoclassical model, markets are always in equilibrium and therefore, by definition, do not include the new innovations and market disrupting forces that entrepreneurs bring. More modern versions of Schumpeter's thesis assert that entrepreneurs are encouraged to undertake the effort because they see an opportunity to develop a new technology. Bill Gates is a classic example of this. Just as the minicomputers produced by firms like DEC overtook mainframes, they declined in the advent of the home computer. Moreover, current research indicates that this process is more prevalent in manufacturing as opposed to service industries (Bednarkzik, 2000, Global Entrepreneurship Monitor, 2001).

While Schumpeter's model of entrepreneurship is based on technological change, an alternative model notes that entrepreneurship is encouraged by economic downturns (Kirzner, 1979). Yusuf and Schindutte (2000) show that in economies where growth is weak, more "survivalist" entrepreneurs as opposed to innovative, growth oriented firms tend to appear. Kirzner's model asserts that entrepreneurs are motivated by profits and substitute self-employment for paid jobs. The question is, are entrepreneurs pulled into self employment by a new idea or pushed in by poor economic circumstances? Are we seeing more service or manufacturing entrepreneurs?

The difference between Schumpeter's model and Kirzner's model of entrepreneurship might be resolved in a long run and short run perspective. In the short run, Kirzner's model is perhaps more applicable. The business cycle will shape the type of employment opportunities and the direction entrepreneurial activity will take. During downturns, small outlets in the service industry will emerge. During economic upturns, larger industries will emerge. A long run perspective in the mode of Schumpeter would direct us to the outcome associated with the death of old industries and the birth of new ones. In a modern context, we might envision the death of an Internet industry or even the decline of a firm that lays skilled workers off, as giving rise to new industries. No doubt the Silicon Valley and Silicon Forest are full of these stories.

Entrepreneurs are usually represented as one of three different types. A gazelle is defined as a firm with at least twenty percent sales growth each year (Birch, 1987). According to Kuratko and Hodgets (1998), "Gazelles are leaders in innovation as shown by the following: New and smaller firms have been responsible for fifty five percent of the innovations in three hundred sixty two different industries and ninety five percent of all radical innovations; Gazelles produce twice as many product innovations per employee as do larger firms; New and smaller firms obtain more patents per sales dollar than do larger firms." Furthermore, between 1990 and 1994, jobs created by gazelles numbered five million and compared to 0.8 million jobs lost by other companies during the same period, thus accounted for a net increase of 4.2 million jobs. Job growth created by gazelles more readily fits Schumpeter's model of entrepreneurship.

A survivalist entrepreneur is defined as one motivated by adverse economic conditions and a lack of paying jobs (Yusuf and Schindebutte, 2000). A lifestyle entrepreneur is motivated largely by the desire for independence and control (Kuratko and Hodgetts, 1998). Many women entrepreneurs fit this description (Scott, 1986). Survivalist and lifestyle entrepreneurs more closely fit the Kirzner model of entrepreneurship.

Statistics support the notion that all types of entrepreneurial firms are important sources of employment and growth of employment in the US economy (Global Entrepreneurship Monitor, 1999, 2000; Kirchhoff, 1994, 1998; Birch, 1979; Baumol, 1993). A recent study showed that during the first half of the 1990s new business startups averaged six hundred thousand per year (The State of Small Business: A Report of the President; Washington, DC, Government Printing Office, 1995). Additionally, small entrepreneurial firms created the most net new jobs in the economy during the period 1977-1990, although the overall percentage share of employment for those firms with fewer than 100 employees remained below that of firms with over five hundred employees (Dennis, 1993). Bednarkzik (2000) shows this to be true as well in the U.S. throughout the 1990s. This statistic is also reflected in a study that predicted which industries will grow the fastest during 1990-2005, in terms of percentage employment. These are all industries in the service sector dominated by small firms (Dennis, 1993).

A related question is the relationship between the strength of the economy and business start-ups. Results are mixed here. One study notes that entrepreneurship in the U.S. is declining and cites the strength of the American economy. "Jobs are plentiful, particularly for those with enough skill to form their own business. Paid employment is relatively more attractive, sometimes neutralizing the 'pull' toward ownership. At the same time, the number of people who experience 'negative pushes', such as unemployment, falls. Fewer people are forced to make life-changing decision. The result is fewer starts." (ICSB, 2000). This study suggests that a strong economy reduces the level of entrepreneurial activity because workers have many alternatives in paid work.

Kirchhoff (1994, 1998) shows too that during periods of economic recessions, small firms are major sources of job creation. He notes too that when the economy expands following recessions, the share of small firms accounting for net new jobs fell. However, he notes that firms with over five hundred employees account for fifty percent of the private-sector workforce and during recessions these firms experience net declines in employment. This makes the overall share of small firms look larger. His analysis concludes that "small firms are the primary job creators in the United States."

Other work, however, shows that the level of entrepreneurial activity increases with the strength of the economy (Kauffman Center, 2001; Zacharikis, Reynolds, Bygrave, 1999). Perhaps the answer to the difference in these results lies in the type of entrepreneurial activity looked at. We certainly know that during the 1990s when the economy was experiencing the largest expansion in history, dot-com starts were up considerably and the venture capital money was plentiful. Over the past year, that money has not been as plentiful and many of those industries have failed. In fact, the average length of employment at Internet companies is eight months. (Foote, 2000). We know less about manufacturing although some research suggests that it plays a lesser role in employment than do service industries (Bednarzik, 2000). This may be due to the typically lower costs associated with starting a service as opposed to a manufacturing business.

Data clearly show a direct relationship between access to capital and the longevity of the firm (Bates, 1990; Global Entrepreneurship Monitor, 2000). This is in turn directly related to the question of whether entrepreneurial activity increases or declines with economic growth. Then we can look at the extent to which entrepreneurial activity creates employment. This paper examines these relationships both theoretically and empirically. We provide a theoretical model describing the relationship between availability of capital to the entrepreneur and employment. We then test the empirical relationship between employment and types of industries; manufacturing, service, distribution, and other productive services. We test also the relationship between employment and firm size. Our findings suggest that employment and entrepreneurial activity are positively related. We also find that employment and firm size are negatively correlated and the service industries account for the majority of employment created by small firms. These findings support our theory. It is important to note that our findings are consistent with the previous research discussed above. What we bring to the discussion is a theoretical foundation and an econometric analysis that strengthens the conclusions of previous work.

The organization of the rest of the paper is as follows. Section II explains the economic theory of employment creation given resource constraints. Section III describes the data, their source, empirical models, and the findings of the research. A brief summary and conclusions comprise the last section of the paper.

THE THEORETICAL FRAMEWORK

Every firm operates subject to a budget constraint. For an entrepreneurial firm, this budget constraint will include venture capital. A basic cost function can show how the availability of venture capital will affect the firm's ability to produce output with the resulting implication for employment. We can write this cost function as

min w * x s.t. f(x) = y

where x represents a vector of the factors of production, specifically capital and labor and w represents the prices of the factors of production, the cost of capital and wages; f(x) is the production function that combines the factors of production to produce output y. The availability of venture capital directly affects the cost of capital for the entrepreneur. The less venture capital is available to the entrepreneur, the higher is the effective cost of capital. Thus, the entrepreneur's budget constraint becomes more binding on its ability to produce and ultimately, its survivability.

We can look at this process graphically and show the effect a decline in the availability of capital will have on employment. If availability of capital is directly linked to the health of the economy, then we can extrapolate to the relationship between the health of the economy and employment at the firm level.

When that source of capital shrinks the budget constraint becomes more binding on the small firm and shifts to the left. This means that less capital and labor are required, i.e., employment by these firms declines. This is consistent with the research discussed above illustrating the importance of capital to the longevity of entrepreneurial firms. As our empirical results following indicate, small entrepreneurial firms are the major source of net job creation in the United States. This is also consistent with the research. Together, these parts of the model clearly point to the importance of venture capital to the birth of entrepreneurial firms and employment in the economy. Flows of venture capital may also be uneven across types of industries and hence affect the birth of new firms in those industries.

We now turn to an empirical investigation of employment creation due to the birth or expansion of new firms, as well as employment loss due to the death of firms. Specifically, we are using regression analysis to estimate the effect on job creation contributed by small firms compared to medium and large firms. Our theory suggests that small firms have the largest impact on employment. We also use regression analysis to examine the impact of the type of industry that new firms enter on employment birth. Specifically, we look at employment birth as a function of whether the firm is in manufacturing, distribution, other productive industries, or service. As discussed above, initial research results are mixed on where we might expect the largest impact, but all research suggests that entrepreneurial firms will have a net positive impact on employment across all types of industries.

DATA AND EMPIRICAL FINDINGS

The data set for this study is provided by the Databases for the Study of Entrepreneurship. This database provides information on employment creation by entrepreneurial firms classified by firm size as well as the industry type. Manufacturing, "Other Productive," Distributive, and Service industry firms with numbers of employees less than twenty, between twenty and four hundred ninety nine, and more that five hundred are considered. The data set spans 1989-1996, allowing for reliable statistical inferences. Given the nature of the pooled time series and cross sectional database, the Ordinary Least Squares method (OLS) may not be appropriate for the estimation of the models under study. Therefore, we employ an estimation method that addresses potential problems presented by our type of data. In order to avoid spurious regression estimates and inferences, all variables are initially tested for stationarity by unit root tests. The following is a brief description of the objectives and methods for these tests.

Table 1 reports the findings of the ADF (Dickey and Fuller (1979)) and PP (Phillips (1987)) tests of unit roots. Panel A and B present unit root test results for level series and their percentage changes, respectively. The ADF entails estimating [DELTA] [x.sub.t] = [alpha] + [beta] [x.sub.t-1] + [[summation].sup.L.sub.j=1] [[gamma].sub.j] [DELTA][x.sub.t-j] + [v.sub.t] and testingsa the null hypothesis that [beta]=0 versus the alternative of [beta]<0, for any x. The lag length j in the ADF test regressions are determined by the Akaike Information Criterion (AIC). The PP test estimates [DELTA] [x.sub.t] = [alpha] + [beta] [x.sub.t-1] + [v.sub.t] and tests the null hypothesis that [beta]=0 versus the alternative of [beta]<0. Three variations of the ADF and PP regressions are estimated: with intercept, trend and intercept, and neither trend nor intercept. The purpose of this approach is to insure that the test results are robust in the presence of drifts and trends. The PP test may be more appropriate if autocorrelation in the series under investigation is suspected. The statistics are transformed to remove the effects of autocorrelation from the asymptotic distribution of the test statistic. The formula for the transformed test statistic is given in Perron (1988). The lag truncation of the Bartlett Kernel in the PP test is determined by Newey and West (1987). In both the ADF and PP tests the MacKinnon (1990) critical values are used. Accepting the null hypothesis means that the series under consideration is not stationary and a unit root is present.

Following stationarity tests, two sets of regression models are proposed to determine the effects of size and industry on employment birth, death, business expansion, and business contraction. In each regression equation, the dependent variable (employment birth or death, for example) is regressed on a set of dummy variables that capture the size or industry effects. The following regression models are therefore estimated:

Y= [alpha] + [[beta].sub.2] I[P.sub.t] + [[beta].sub.3] I[D.sub.t] + [[beta].sub.4] I[S.sub.t] + [u.sub.t]. (1)

In equation (1) the dependent variable (employment birth, for example) is a function of the type of industry. Variables IP, ID, and IS, represent productive, distribution, and service firms. The objective is to test whether the type of the industry in which an entrepreneurial firm is operating has any effect on the employment creation. For example, a positive and statistically significant $3 would indicate that an entrepreneurial firm in the distribution industry is contributing to the employment creation. Variables IP, ID, IS assume values of one for productive, distributive, and service sector firms, respectively, and zero otherwise. The parameter " captures the effects of manufacturing entrepreneurial firms.

Similarly, to examine the size effects of entrepreneurial firms on employment and business expansion, we estimate equation (2).

Y= [gamma] + [[lambda].sub.1] [M.sub.t] + [[gamma].sub.2] [L.sub.t] + [w.sub.t]. (2)

In equation (2) the dependent variable (employment birth, for example) is a function of the firm size. Variables M and L represent medium and large entrepreneurial firms, while (captures the effects of small entrepreneurial firms on the dependent variable. For example, a positive and statistically significant (would indicate that a small entrepreneurial firm is contributing to the employment creation. Both equations are estimated by Newey-West heteroscedasticity and autocorrelation consistent method (NWHAC) (Newey and West (1987)). This method allows for a general covariance matrix estimator that takes into account both the possibility of serially correlated and heteroscedastic residuals in our pooled time series and cross section data.

The summary statistics indicate a slight deviation from normality. These types of deviations may stem from the nature of our pooled time series and cross sectional observations and may suggest estimation methods that adjust for non normality of underlying variable distributions. Results of the ADF and PP stationarity tests are reported in Table 1. It is shown that variables in level are stationary by the PP test. When variables are measured in first difference, then both the PP and ADF test suggest that variables are stationary. Because interpreting regressions on first differences of variables are hard to interpret, we report our regressions in levels of variables. However, the first difference regressions produced qualitatively the same results and are available from the authors upon request.

The Newey-West autocorrelation and heteroscedasticity adjusted parameter estimates of equation (1) are presented in Table 2. According to Table 2, entrepreneurial firms in all industries contribute significantly and positively to employment creation. Also, the same is true regarding employment loss. A notable difference between the first two columns is that coefficient sizes are smaller for employment loss. This indicates that entrepreneurial firms contribute more to employment generation than loss of employment. Considering business expansions and contractions, our findings in the last two columns indicate that entrepreneurial firms contribute to expansion in manufacturing and productive sectors. The coefficients of service and distributive sectors are statistically insignificant showing that entrepreneurial activities in these sectors during the study period were quite limited. This finding is plausible because the raw data indicates that entrepreneurial firms already had a sizable presence in distributive and services sectors and perhaps were running out of room to expand. Indeed, the rate of expansion in manufacturing and other productive sectors is consistently higher than distributive and service sectors for all the study years and firm sizes. In regards to business contraction, entrepreneurial firms show contraction in manufacturing and other productive activities and the coefficient sizes are comparable to those of expansion.

Table 3 presents overwhelming evidence that small entrepreneurial firms are largely responsible for job creation and losses with roughly equal coefficient sizes. It is notable that the coefficients of small size effect are positive and statistically significant in employment birth and expansion equations and in job loss and contraction equations. However, the coefficient sizes indicate a much larger expansion effect than contraction. Thus, it is safe to deduct that small entrepreneurial firms are active in expanding businesses and creating employment.

The statistical results for medium and large entrepreneurial firms are mixed. The first two columns indicate that medium and large size entrepreneurial firms do not contribute to either employment birth or to employment loss, while the last two columns show that these firms were not expanding during the period under study. These findings suggest that the key to employment creation is entrepreneurial firms of small size.

SUMMARY AND CONCLUSIONS

This paper has examined the relationship between entrepreneurial activity and job creation. We also look at employment created by entrepreneurs as a function of the type of industries they enter. Although there exist different models of entrepreneurship, notably Schumpeter's and Kirzner's, both predict that small firms are a source of net employment growth in the economy. We use regression analysis to estimate these relationships. The data set spans 1989-1996, allowing for reliable statistical inferences. Our findings validate the research prediction that small firms do have a net positive impact on job creation. We also find that this is true over all types of industries although the strength of that impact varies among types of industries. This last result may be due to the fact that flows of venture capital are uneven across types of industries and will therefore affect the birth of firms across industries. This is an interesting question for further research.

REFERENCES

Bates, T. (1990). Entrepreneur human capital inputs and small business longevity. The Review of Economics and Statistics, 72(4), 551-559.

Baumol, W.J. (1993). Entrepreneurship Management and the Structure of Payoffs. Cambridge, MA: MIT Press.

Bednarzik, R. (2000). The role of entrepreneurship in U.S. and European job growth. Monthly Labor Review, 123-127.

Birch, D. (1979). The Job Generation Process. Cambridge MA: MIT Program on Neighborhood and Regional Change.

Birch, D. (1987). Job Creation in America. New York: The Free Press.

Dennis, Jr., W. (1993). A Small Business Primer. Washington D.C.: The NFIB Foundation.

Dickey, D. A. & W. A. Fuller. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American Statistical Association, 427-431.

Foote, D. (2000). Some preventive medicine for cases of dot-com-it-is. Computerworld, 34(12), 34.

Global Entrepreneurship Monitor, Kauffman Center for Entrepreneurial Leadership, 2000.

International Council for Small Business Bulletin, 3rd/4th quarter, 2000, 32.

Katz, J. (2000). Advances in Entrepreneurship, Firm Emergence, and Growth, 4, New York: JAI Publishing.

Kirchhoff, B. A. (1994). Entrepreneurship and Dynamic Capitalism. Westport, CT: Quorum Books.

Kirchhoff, B. A. (1997). Entrepreneurship Economics. In W. Bygrave, (Ed.), The Portable MBA in Entrepreneurship. New York: John Wiley.

Kirzner, I. (1979). Perception, Opportunity, and Entrepreneurship. Chicago: University of Chicago Press.

Kuratko, D. & R. Hodgetts. (1998). Contemporary Entrepreneurship, (4th Ed.) New York: The Dryden Press.

MacKinnon, J. G. (1990). Critical values for cointegration tests, In R.F. Engle & C.W.J. Granger (Eds.) Long-run Economic Relationships: Readings in Cointegration Oxford: Oxford University Press, 267-76.

Newey, W. K. & K. West. (1987). A simple positive-definite heteroscedasticity and autocorrelation consistent covariance matrix. Econometrica, 55, 703-708.

Perron, P. (1988). Trends and random walks in macroeconomic time series. Journal of Economic Dynamics and Control, 12, 297-332.

Phillips, P.C.B. (1987). Time series regression with a unit root. Econometrica, 55, 277-301.

Schumpeter, J. (1934). The Theory of Economic Development. Cambridge, MA: Harvard University Press.

Schumpeter, J (1942). Capitalism, Socialism, and Democracy. New York: Harper and Brothers.

Yusuf, A. & M. Schindehutte. (2000). Exploring entrepreneurship in a declining economy. Journal of Developmental Entrepreneurship, 5(1), 41-57.

Zacharakis, A., P. Reynolds & W. Bygrave. (1999/ 2000). National Entrepreneurship Assessment: United States of America, Executive Report of the Global Entrepreneurship Monitor.

Bahram Adrangi, The University of Portland

Mary E. Allender, The University of Portland

Robin Anderson, The University of Portland

Table 1: Summary and Stationarity Statistics

Level ADF PP M s S K

Birth -2.03 13.19 *** 0.064 0.027 0.17 1.92
Loss -2.25 -12.02 *** -0.059 0.023 -0.53 2.39
Expansion -1.87 -10.68 *** 0.122 0.046 0.79 3.07
Contraction -1.29 -3.21 ** -0.110 0.033 -1.40 5.03

 First Difference

Birth -7.63 *** -23.82 ***
Loss -8.29 *** -24.15 ***
Expansion -5.64 *** -22.72 ***
Contraction -4.60 *** -14.25 ***

*** indicates significance at 1 percent level.

Notes: M, s, S, and K stand for mean, standard deviation, skewness,
and kurtosis. ADF and PP regressions are estimated with intercepts
and no trends. The lag structure for the ADF test is determined based
on the AIC criterion, while for the PP test Newey--West criterion is
used.

Table 2: Newey-West HAC Regression Results Equation (1)

 Birth Loss Expansion Contraction

Manufacturing 0.045 *** 0.048 *** 0.108 *** 0.088 ***
 (8.26) (0.08) (11.72) (21.25)
Other Productive 0.065 *** 0.020 *** 0.047 *** 0.068 ***
 (11.81) (2.94) (3.58) (11.59)
Distributive 0.075 *** 0.014 *** -0.004 0.011
 (13.22) (2.13) (-0.31) (1.87)
Service 0.070 *** 0.008 *** 0.011 0.007
 (12.71) (1.25) (0.87) (1.30)
[R.sup.2] 0.168 0.106 0.190 0.680
F 5.418 *** 3.192 *** 6.291 *** 56.717 ***
LL 191.36 202.78 148.03 215.35

Notes: *** indicates significance at 1 percent level.

Table 3: Newey-West HAC Regression Results Equation (2)

 Birth Loss Expansion Contraction

Small 0.095 *** 0.087 *** 0.175 *** 0.111 ***
 (31.29) (42.60) (35.07) (18.12)
Medium -0.046 *** -0.037 *** -0.071 *** 0.009
 (-10.82) (12.89) (-10.06) (1.03)
Large -0.047 *** -0.046 *** -0.087 *** -0.011
 (-10.89) (16.06) (-12.42) (1.32)
[R.sup.2] 0.660 0.781 0.682 0.064
F 78.65 *** 144.81 *** 87.04 *** 2.80 ***
LL 228.91 261.90 187.32 170.28

Notes: *** indicates significance at 1 percent level. The
classification of firms as small, medium, and large is based on the
number of employees <20, 20-499, and >500.