Too Big? Too Small? Just Right? An Empirical Perspective on Local Firm Size Distribution and Economic Growth in U.S. Counties and High-Poverty Rural Regions

Abstract

The authors examine the relationship between the distribution of firms across size categories and economic growth, extending Loveridge and Nizalov’s Michigan results to the United States. Using county-level data, including growth and control variables, the authors explore the relationship between the size distribution of firms and 12-year growth patterns for the continental United States and three multistate high-poverty regions. The results of fixed-effect feasible generalized least squares estimation show a connection between employment growth and the distribution of firms across size categories for the continental United States. The results also show a positive link between employment growth and firm size for Lineal America and the Plantation Belt, but no statistically significant relationship for the Borderlands. The results suggest that policies aimed at promoting small business, while important nationally, may differ in impacts across regions and provide an argument for region-level decision making about growth policies.

Keywords

community development business climate industrial location industrial policy

Introduction

This article examines the relationship between the distribution of firms across size categories and economic growth in the United States. Loveridge and Nizalov (2007) explored the validity of the firm size hypothesis for the state of Michigan. They found a strong link between a county’s business size distribution in predicting both job and income growth in Michigan. We expand the geographic scope of Loveridge and Nizalov’s research to examine the role of the local firm size distribution in economic outcomes for the continental United States, with particular attention to three regions with high poverty rates. The results of fixed-effect feasible generalized least squares estimation show a connection between employment growth and the distribution of firms across size categories for the continental United States. The results also show a link between employment growth and firm size for high-poverty regions, while at the same time showing that each region behaves somewhat differently. In each case, a higher proportion of employment in small firms increases employment growth. The results of our estimation suggest that a uniform distribution of firms is not the optimal growth-enhancing distribution for most regions.

First, we summarize the “firm size” debate in economic development literature and provide some insights as to why a range of firm sizes might contribute to overall economic growth in a region. Next, we summarize Loveridge and Nizalov’s (2007) Michigan model, which inspired the current contribution. We then discuss our data, empirical method, and the U.S. regions of interest. Our results follow and show that the firm size distribution is related to both income and employment outcomes. We conclude that analysis specific to the region can provide insights as to which mix of policies—favoring large, small, or “just right” firms—is likely to produce desired outcomes in the region of interest.

The Firm Size Debate in Local Economic Development

Successful job creation strategies can mitigate poverty and raise incomes. These goals have been thought of as key aspects of economic growth and have often been a priority to economic development practitioners (Shaffer, Deller, & Marcouiller, 2006). Job creation is often viewed as a means to those ends. A frequently heard debate in community-level economic development planning sessions has to do with which size firm should get the most attention. Similar to Goldilocks of the classic children’s fable who tried to choose from among three chairs, the competing perspectives seem to boil down to choices of big, small, or “just right.” Some want local job creation efforts to focus on recruiting large firms from outside the area to relocate in the community, whereas others argue for more assistance to smaller existing local businesses or start-ups. This small versus large civic debate has its parallels in the academic literature, so we include perspectives of both research and practice in this section.

Industrial recruitment proponents tend to be on the “large firm” side of the small-versus-large debate. Alongside traditional industrial recruitment through tax incentives, there are strong political pressures that move local decision makers toward high-profile investments in strategic projects. Thus, desperate state and local governments are increasingly moving from supply-side incentives to seeding new large businesses through demand-side initiatives, such as stadiums (Eisinger, 1988). In this worldview, basic entrepreneurship policies aimed at assisting start-ups or smaller firms are deemed to be too slow.

The industrial recruitment job creation strategy enjoyed early successes bringing new industries to the south, and continues to remain popular in many U.S. states (Hodge, 2011) despite various academic criticisms (Loveridge, 1996). Industrial recruitment tactics often include government subsidies and tax breaks, and have been derided as “smokestack chasing” (Bradshaw & Blakely, 1999).

Another popular job creation tactic focuses on retention and expansion (R&E) of existing firms within a region (Morse, 1990). Two basic forms of R&E strategy are found among practitioners—individual and community based (Allanach & Loveridge, 1998). Individual R&E programs rely on a full-time professional who visits firms to solve problems. Due to the limited ability of R&E professionals to visit all firms in a community, this type of program tends to focus on the largest businesses in the community. Thus, it might naturally be supported where the “large firm” partisans hold sway.

Community-based R&E programs use volunteers to visit a wide variety of local firms. The resulting data are used to assess community-wide strategies for becoming more business friendly (Morse, 1990). Community-based R&E programs tend to visit an array of firm sizes, so locations using this variant of R&E can be characterized as being more open to small-firm-type strategies. Community-based R&E practitioners, therefore, occupy a kind of middle ground in the small-versus-large debate.

The Edward Lowe Foundation (ELF) is an example of an organization that follows Goldilocks’s ultimate selection in the small-versus-large debate—the ELF focus is on firms whose size is in the middle: not too small and not too big. The principal ELF job creation strategy is around assistance to second-stage firms. According to the foundation’s website (ELF, n.d.), “Second-stage companies are those that have grown past the startup stage but have not yet grown to maturity.” The ELF website goes on to mention that the second-stage firm size is typically between 10 and 100 employees.

On the end of the small-versus-large continuum are proponents of small business–oriented job creation strategies. An early champion of small business was David Birch (1981) whose empirical study concluded that small businesses are responsible for most job growth. Birch’s pathbreaking work has since been validated by a number of studies focusing on innovation and entrepreneurship (Olfert & Partridge, 2010; Von Bargen, Freedman, & Pages, 2003). This line of inquiry has renewed interest in small businesses as an engine for economic growth (Aquilina, Klump, & Pietrobelli, 2006; Birch, 1987; Robbins, Pantuosco, Parker, & Fuller, 2000). Deller and McConnon (2009) and Shaffer (2006) lay out the theoretical arguments for the role of small firms in regional economic growth. These arguments include Schumpeterian innovation (Schumpeter, 1942, 1961), increased flexibility of small firms in changing environments, and enhancing competition, resulting in greater efficiencies for new and existing firms. However, even proponents of small business–focused job creation policies (Stephens, Partridge, & Faggian, 2011) are quick to point out that many small businesses are not at all innovative. Many small businesses are “mundane” (e.g., hairdresser) or are the result of involuntary unemployment or underemployment. Mundane or involuntary businesses have limited potential for growth, and it can be a tactical error to conflate “small” with “entrepreneurial” in policy development (Goetz, Partridge, Deller, & Fleming, 2010). In a nuanced view on the Goldilocks debate, Low and Weiler (2012) note that communities may select between recruitment or entrepreneurship strategies based on trade-offs across expected growth and risk of volatility. For example, individuals may be more inclined to become entrepreneurs in low- and high-growth areas with volatility than in medium-growth areas without volatility.

Lichtenstein and Lyons (2006, 2010) offer a subtle critique of the small-versus-large firm policy debate. They propose a conceptual model, which they refer to as the entrepreneurial pipeline theory. Consistent with the perspective of mainstream business literature, but differing from Schumpeter and ELF perspectives, they argue that firms of all sizes can be entrepreneurial. An important point that they add to previous notions of business development is that the distribution of businesses within size classes (the pipeline of firms) may act as an important determinant of economic growth. In developing their model in terms of a “pipeline,” Lichtenstein and Lyons reveal their focus on the dynamics of firm birth, growth, death, and replacement. They suggest, however, that the benefits of a range of business sizes within the region’s economy go beyond mere replacement.

The positive externality benefits of agglomeration economies are well laid out in basic regional economic theory and documented in empirical works (Blair & Premus, 1987; Carlino, 1982; Devereux, Griffith, & Simpson, 2007; Hansen, 1990; Hoover & Giarratani, 1984; Krugman, 1991), but the importance of firms of differing sizes in contributing to regional resilience and growth is a newer concept. Michael Porter (1990, 1998, 2000) advanced the concept of clusters of related businesses around a common theme (e.g., high-tech, wine, tourism, or shoes). Porter’s notion of clusters promotes a form of agglomeration economy, but does not give much attention to positive interactions because of scale differences across firms operating in different sectors. In Porter’s world, the locational cluster advantages come via horizontal (similar firms operating at the same level) or vertical (moving up the value chain of a product) linkages. To expand on Porter’s theory, we posit that diagonal benefits may occur across firm sizes and sectors, and may therefore confer competitive advantages even if a wide distribution of firm sizes in a region does not exhibit the theme-specific characteristics of Porter’s clusters. As Drucker (2012) notes, “Intelligence regarding the mechanisms and spatial scope of agglomeration is crucial for translating empirical research into practical guidance for policy and public decision-making” (p. 1). It is our aim to shed more light on whether agglomeration economies may occur via the mechanism of the distribution of firm sizes. A range of firm sizes in a region could be beneficial to other individual firms at a microeconomic level in several ways, outlined below. Our intent is to provide some possible explanations for mechanisms by which firm size diversity in a region could aid in economic growth. The intensity of each mechanism, or whether it works in this way at all, is beyond the scope of the present work.

Labor market: The presence of small firms can assist in signaling skills for workers who then take positions in larger, higher-paying firms (Baptista, Lima, & Preto, 2012; Felmlee, 1982; Ferrer & Lluis, 2008). Although this signaling relationship might seem as if the large firm is pushing its search and training costs onto the smaller firm, the smaller firm may also gain in some ways. First, workers might be willing to accept lower wages in a smaller firm if they see a career ladder up toward the larger, higher paying firms. We observe this in professional sports, where farm teams offer players opportunities to develop their skills and a chance at the majors. Second, the small firm may benefit from contacts within the larger firm via its former employees, yielding orders or vital industry intelligence.

Contracts: Small firms may provide services or worker amenities that reduce large firm costs, whereas large firms provide a market for the smaller firms. It is also possible that small firms could benefit by repurposing by-products of the larger firm’s main processes. For example, biofuel plants, which average roughly 40 employees (Low & Isserman, 2009), supply animal feed to farms from the distiller’s grain by-product of the refining process. The repurposing of by-products is not necessarily limited to physical goods. There is some evidence that large research and development operations are synergistic with smaller research and development operations, in that an innovation that is useless to the larger company can be purchased/leased and commercialized by a smaller lab focusing on a different sector (Agrawal, Cockburn, Galasso, & Oettle, 2012).

Role modeling: Small, nimble firms may be able to experiment with new process or market development techniques. By observing and interacting with local innovators, larger firms might pick up on ways to improve business without the large transaction costs associated with experimentation in a larger, more complex operation. The benefits may flow in the other direction as second-stage firms start to grow and need to put into place human resource departments and other more formalized management structures.

From a community- or region-wide perspective, a diversity of firms across different sizes might be beneficial in other ways:

Labor market: The presence of a set of larger firms in the region tends to raise overall pay levels, creating income that allows for other kinds of investments (service-sector business, education) in the region. Small innovator firms, with more growth potential, could help sop up new entrants into the labor market, reducing costs of unemployment support and crime that is associated with unemployment (Lin, 2008).

Diverse portfolio: Firm size diversity could buffer the regional economy against the death or departure of a large anchor firm as slightly smaller firms grow. Conroy (1974) and Spelman (2006) offer evidence that a portfolio of industries buffers against risk of a downturn; the same principle might operate in terms of a portfolio of firm sizes. Large firms may be more subject to exchange rate fluctuations (more likely to source or sell in international markets), foreign competition, or resort to outsourcing. While most attention about downturns focuses on mass layoffs by large firms, we note that some macrophenomena may affect small firms more than large firms. For example, more stringent criteria for obtaining credit may constrain small firm growth, as might changes in national policies (e.g., health care) that impose new costs on firms. Thus, a portfolio balancing across small, medium, and large firms could operate to minimize overall risk of widespread downturns in the economy with a diversity of firm sizes.

Political economy: A diversity of firm sizes could produce better political and economic power balance. For example, some towns have been called “company towns” due to the dominance of a single company in the economic landscape. This structure could lead to uncompetitive behavior in the labor demand market. Furthermore, as documented by Solecki (1996), it might also prevent dialog about social issues in which the company’s interests differ from that of the residents. On the other hand, regions without corporate leaders (small firms only) may fail to make needed institutional changes as market conditions change because of the inability to see larger market forces driving economic change.

The Lichtenstein and Lyons (2006, 2010) framework essentially advocates for business-related economic development policies to consider the total distribution of firms within a region, and for policies to focus on gaps in the pipeline. Thus, it advocates for a more tailored look at the small-versus-large debate with policies based on missing elements in the local economy, rather than a one-size-fits-all policy favoring small or larger businesses. The Lichtenstein and Lyons framework was developed over years through hands-on regional development and business assistance for entrepreneurs. However, they did not empirically test the framework or consider how the presence of a diversity of firm sizes—even if firms in the continuum are not entrepreneurial or growth oriented—might still be beneficial to the local economy.

Fotopoulos (2012) makes an important point about the role of small businesses in economic growth that supports Lichtenstein and Lyons. He examines the relationship between self-employment rates and per capita income growth in 197 European regions across 15 countries.¹ Fotopoulos finds an L-shaped relationship between self-employment rates and per capita income growth, indicating that in some European Union regions growth might be faster with less self-employment. The Fotopoulos result provides an important caveat to those who promote small business as an economic development strategy: It may not be appropriate for all regions.

A Summary of the Loveridge and Nizalov Model and Results

Our empirical approach follows the method developed by Loveridge and Nizalov (2007), but expands the analysis to the continental United States. We provide a summary of the Loveridge and Nizalov contribution here. To overview the Loveridge and Nizalov approach, regression analysis controlled for local factors and estimated the effect of business size distribution on income and job growth. The data were annualized variables from Michigan’s 83 counties from 1988 to 2000. They used two measures of business size distribution:

Share of employment in each size category

Deviations from equal distribution of employment across sizes

Information on the number of establishments (firms) in nine employment size categories is provided by the U.S. Census Bureau’s County Business Patterns. The employment categories range from 1 to 4 employees to 1,000 or more employees. To take into account the potentially disproportionate influence that firms with several hundred employees can have on the local economy, they used the middle of each range to weight the number of establishments in each size category by employment size.² Loveridge and Nizalov (2007) also computed an index akin to a Gini coefficient to represent the employment share distribution in a single variable. The index measures one half ³ the sum of absolute deviations of employment shares in each category (x_i) from a uniform distribution (11.1%).

Index = \frac{1}{2} \sum_{i = 1}^{9} | x_{i} - 11.1 % |

A larger value of the index indicates a more unequal firm size distribution. Thus, if the uniform distribution is optimal, then a larger value of the index would correspond negatively to economic growth and vice versa. To estimate the connection between the distribution of firm sizes and annual regional economic growth, Loveridge and Nizalov used the model shown in Equation (2):

\begin{array}{l} G R_{i, t, t + 1} = β_{0} + β_{1} {size distrib}_{i t} + β_{2} {educ}_{i t} \\ + β_{3} {manuf}_{i t} + β_{4} {farm}_{i t} + \\ β_{5} {gov}_{i t} + β_{6} {denst}_{i t} + β_{7} {year}_{t} \\ + e_{i} + u_{i t} \end{array}

The dependent variable in Equation (2), ( $G R_{i, t, t + 1}$ ) consists of the growth rate of two measures of economic conditions: per capita income and jobs per 1,000 residents in county i at time t. The size distribution is either captured via the index described in Equation (1) or via eight variables describing employment shares of the bottom eight firm sizes (1000+ is the omitted category). Two dependent variables and two formulations for size distribution produce four estimated equations. The other independent variables are identical across the four equations. To control for the level of education (educ), they included the share of the population in county i with a bachelor’s degree. Year binary variables (year) help control for time-specific shocks common to all counties across the country. To control for the local industry structure they include the jobs per capita in manufacturing (manuf), farming (farm), and government (gov). The number of business establishments per square mile (denst) takes into account nonemployer businesses and any agglomeration effects due to the degree of concentration and urbanization. Finally, e_i represents time invariant conditions in county i and u_it is a well-behaved error term.

They estimated Equation (2) with a fixed-effects generalized least squares method. This provides an efficient way to control for the time invariant heterogeneity while also controlling for relevant factors related to regional economic growth. The fixed-effects generalized least squares estimation takes advantage of within-county variation to show how changes in the business distribution affect annual per capita income and job growth rates.

Equation (2) was estimated for income and job growth for counties in the state of Michigan and separately for metro and nonmetro counties. The business size distribution was found to be significant, both in the shares equations and in the index/index-squared equations. They concluded,

Our major finding is that there are strong links between business-size distribution and regional economic growth. The results indicate that the growth-optimizing distribution of employment by business size is not uniform. We found also that optimal job and income growth−enhancing distributions would have a higher share of the smallest businesses than is currently the average. However, a larger share of businesses with more than 1,000 employees would also facilitate income growth in some counties. (Loveridge & Nizalov, 2007, p. 258)

A limitation of the Loveridge and Nizalov study is that it covered only one state, and that particular state is heavily invested in large-scale manufacturing via its historic position as the headquarters for the nation’s largest auto makers. This article addresses the single-state limitation and expands the analysis to explore the relationship between economic growth and firm size distribution in three of the nation’s most consistently poor rural regions (Figure 1).

Figure 1.

Regional delineations.

U.S. Data

We adapt the Loveridge and Nizalov study by first incorporating U.S. data. Figures 2 and 3 provide information about the firm size distribution in the United States. Figure 2 illustrates that firms are heavily concentrated in the smaller employee categories for U.S. counties in 1990. Figure 3 shows that the employment-weighted size distribution is much more uniform for U.S. counties in 1990. By inspection of Figure 3, one can discern that the average index value for the United States is nonzero. Basic descriptive statistics for our variables are found in Table 1.

Figure 2.

Distribution of establishments by employee size categories, U.S. counties 1990.

Figure 3.

Distribution of employment across business size classes, U.S. counties 1990.

Table 1.

Summary Statistics.

	Continental United States		Lineal America		Borderlands		Plantation Belt
Variables	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Income growth	1.79	5.09	1.61	2.55	1.75	7.26	1.79	3.24
Job growth	0.82	3.66	0.56	3.22	0.60	4.13	0.49	3.92
Education	14.60	6.96	11.51	5.33	13.89	5.63	12.64	5.87
Manufacturing	70.41	55.17	91.05	60.43	23.85	23.40	90.21	56.33
Government	81.89	43.96	63.99	22.55	92.51	29.45	80.09	53.72
Farming	49.33	50.88	30.51	28.51	56.22	68.02	27.32	19.18
Business density	5.77	68.36	2.36	3.81	2.65	10.62	2.59	7.92
No. of counties	3,046		413		103		471

Empirical Method

Our basic estimation method follows Loveridge and Nizalov. Our variables of interest, employment-weighted size shares and index, also coincide with the previous study. Each model is estimated using a panel of U.S. county-level data from 1988 to 2000 (12 annual observations). This time period was selected to be parallel with Loveridge and Nizalov. Selecting this set of years also avoids the possibly artificial boom and resulting crash that took place in years subsequent to our analysis time period. Some analysts have concluded that the boom and crash were housing related (Cooley & Rupert, 2010). The construction sector is composed of many small firms, so including this anomalous post-2000 time period could bias the results of our estimations. In particular, cheap, high-risk credit may have inflated the success of small businesses during that era beyond what might be experienced in a more ordinary credit regime. It should also be noted that the housing-fueled boom did not affect all regions equally, so time dummies would not adequately control for the boom/bust effect (Carruthers & Mulligan, 2012).

The data are county level aggregates for the continental United States. Data from the County Business Patterns were used to create the business establishment density (denst) and the measures of firm size distribution. The employment and income variables (job and income growth, manuf, farm, gov) were obtained from the Local Area Annual Estimates from the Bureau of Economic Analysis. Finally, the share of population with a bachelor’s degree was downloaded from the U.S. Census Bureau and imputed using linear interpolation over time for each county.⁴

The Regions

In a modification of Loveridge and Nizalov’s basic approach, we explore the implications of the distribution of firm sizes on economic growth for three multistate cultural zones that carry particular policy interest. Deller (2010) shows evidence of significant spatial heterogeneity in the effect of microenterprises on economic growth. We focus on high-poverty regions having distinct cultures, and historically poor development outcomes.⁵ Rather than following officially drawn borders or state lines, we follow Loveridge, Yi, and Bokemeier (2009), and rely on Census county-level modal or majority ethnicity reports to develop nonoverlapping regional boundaries in an approach similar to that taken by Nostrand (1970). The regions, mapped in Figure 1, are described in the following paragraphs.

The Borderlands area is the contiguous set of counties near the border with Mexico, where the modal response to the 2000 Census ethnicity question indicated Hispanic descent. The area’s economy can be characterized by low-skill manufacturing, extensive agriculture, and natural resource extraction, and it serves as a type of “first stop” for immigrants from Latin America. Immigration would be expected to exert downward wage pressure on the more established households responding to the Census. The Plantation Belt lies in former slave states and is defined by contiguous counties where the majority reported African American status in the 2000 Census.

Lineal America is defined as the contiguous set of counties in the southeast where Census respondents gave their ethnicity as “American” as the modal response in 2000.⁶ They thus seemingly consider themselves American by lineage. The region has about the same number of counties (413 vs. 420) as the politically-defined Appalachian Regional Commission, and overlaps considerably with Appalachia but does not include Appalachian Pennsylvania or New York, and extends further south and west than the officially designated Appalachian region.⁷ Much of the area was initially dominated by mining in the 1900s. Over time, the area has diversified somewhat by default because of increasing capital intensity in the mining sector. Today mining employment is important but not dominant in many Lineal America counties. About 6% of the region’s counties show mining as nondisclosed in Bureau of Economic Analysis employment data in 2000. Mining averages roughly 3% of total employment in the remaining counties. The region is characterized by European settlement in the prerevolutionary period, followed by long-term stagnation, resulting in little in-migration. We can speculate that it is perhaps long ties to the region that caused so many residents to list ethnicity as “American.” In areas with more fluid migration patterns, people may tend to cleave more closely to country-of-origin ties in forming ethnic identity. Whatever the cause of so many individuals declaring “American” as their ethnicity, the region would seem to form a cultural unit.

In all three regions, a few counties did not display the filtering characteristic, but were included if they are surrounded by counties displaying the filtering characteristic on the grounds that employment or income growth may be influenced by regional cultural spillovers as well as county-level characteristics (Pereira & Andraz, 2012). The Plantation Belt and Lineal America share a border but do not overlap because in counties where the majority report African American ethnicity, that is the modal ethnicity by definition.

Appendix Table A1 shows that the regions differ from the rest of the United States in generally accepted outcomes of concern in poverty studies: median household income, poverty rates, and obesity. Drewnowski and Specter (2004) suggest that obesity and poverty are interrelated. The obesity rate can also serve as a proxy for health. In their 2001 article, Sturm and Wells found obesity contributes more to morbidity than smoking, drinking, or poverty. The three regions included in our study thus merit special consideration in national policy analyses.

Regression Results and Discussion

Tables 2 through 5 show the FEGLS estimates for both the job growth and income growth equations over 12 years.⁸ We also estimated models using per capita wages and salaries from the County Business Patterns. These results can be found in Tables A2 and A3 of the appendix.⁹ FEGLS is the most efficient estimation technique among feasible estimators; nonetheless, it does not control for all forms of bias (Besley & Case 2000). Therefore, we also present a comparison of alternative estimation techniques in Tables A4 and A5 of the appendix. We first discuss results for the control variables before turning to the results for the employment shares and firm size distribution index.

Table 2.

Employment Shares and Income Growth.

	(1)	(2)	(3)	(4)
Variables	Continental United States	Lineal America	Borderlands	Plantation Belt
Education	7.609*** (0.696)	0.166 (9.022)	10.200*** (3.326)	−2.507 (7.156)
Manufacturing	−0.011*** (0.002)	−0.013*** (0.003)	−0.020 (0.038)	−0.012*** (0.003)
Government	−0.020*** (0.003)	−0.016** (0.007)	−0.031 (0.028)	−0.023*** (0.005)
Farming	−0.040*** (0.005)	0.017* (0.009)	−0.049 (0.032)	0.038*** (0.013)
Business density	−0.005 (0.009)	−0.201** (0.079)	0.028 (0.430)	0.006 (0.053)
Employment share in 1-4–employee businesses	0.009 (0.020)	0.197*** (0.055)	−0.127 (0.096)	0.131** (0.067)
Employment share in 5-9	0.003 (0.014)	−0.018 (0.032)	−0.141 (0.099)	0.048 (0.031)
Employment share in 10-19	0.040*** (0.013)	−0.059** (0.024)	0.226** (0.108)	−0.042* (0.024)
Employment share in 20-49	0.001 (0.013)	0.044** (0.020)	0.155* (0.090)	−0.027 (0.020)
Employment share in 50-99	0.001 (0.011)	0.005 (0.015)	−0.074 (0.081)	−0.000 (0.017)
Employment share in 100-249	0.008 (0.010)	0.005 (0.012)	0.101 (0.089)	0.003 (0.014)
Employment share in 250-499	−0.001 (0.009)	0.007 (0.010)	−0.026 (0.087)	−0.004 (0.011)
Employment share in 500-999	−0.005 (0.008)	−0.002 (0.009)	0.076 (0.086)	−0.011 (0.010)
Constant	−1.668*** (0.082)	−1.699*** (0.114)	1.019 (0.633)	−2.258*** (0.137)
Joint significance employment shares	F = 13.04	F = 25.83	F = 27.38	F = 15.19
p Value	.1106	.0011	.0006	.0555
Observations	35,121	4,890	1,091	5,559
Number of counties	3,046	413	103	471
Log likelihood	−1,01,340	−10,635	−3,462	−13,466
Wald χ²	5,400	1,557	159.2	1,430

Note. Standard errors in parentheses.

p < .1. **p < .05. ***p < .01.

Table 3.

Employment Shares and Job Growth.

	(1)	(2)	(3)	(4)
Variables	Continental United States	Lineal America	Borderlands	Plantation Belt
Education	3.604*** (0.518)	−7.242 (12.048)	−0.151 (1.540)	9.976 (8.992)
Manufacturing	−0.031*** (0.002)	−0.031*** (0.004)	−0.079*** (0.017)	−0.034*** (0.004)
Government	−0.039*** (0.002)	−0.028*** (0.009)	−0.026** (0.013)	−0.051*** (0.007)
Farming	−0.026*** (0.004)	0.045*** (0.012)	−0.067*** (0.015)	−0.030* (0.016)
Business density	0.011 (0.007)	−0.086 (0.105)	0.025 (0.199)	0.088 (0.067)
Employment share in 1-4–employee businesses	0.262*** (0.015)	0.157** (0.074)	0.041 (0.045)	0.317*** (0.084)
Employment share in 5-9	0.057*** (0.011)	0.095** (0.042)	0.012 (0.046)	0.066* (0.039)
Employment share in 10-19	0.039*** (0.010)	−0.065** (0.033)	0.011 (0.050)	0.000 (0.031)
Employment share in 20-49	0.025*** (0.009)	0.060** (0.026)	0.015 (0.042)	0.037 (0.026)
Employment share in 50-99	0.042*** (0.008)	0.007 (0.021)	0.004 (0.037)	0.053** (0.022)
Employment share in 100-249	0.022*** (0.007)	0.030* (0.017)	0.045 (0.041)	0.032* (0.018)
Employment share in 250-499	0.002 (0.006)	0.004 (0.014)	−0.062 (0.040)	0.040*** (0.014)
Employment share in 500-999	−0.002 (0.006)	−0.004 (0.012)	0.032 (0.040)	−0.002 (0.012)
Constant	0.626*** (0.061)	1.054*** (0.152)	1.031*** (0.293)	0.770*** (0.172)
Joint significance employment shares	F = 416.78	F = 30.7	F = 10.72	F = 44.42
p Value	.000	.000	.2182	.000
Observations	35,121	4,890	1,091	5,559
Number of counties	3,046	413	103	471
Log likelihood	−90,955	−12,049	−2,621	−14,736
Wald χ²	3,443	736.5	206.6	745.5

Note. Standard errors in parentheses.

p < .1. **p < .05. ***p < .01.

Table 4.

Index and Income Growth.

	(1)	(2)	(3)	(4)
Variables	Continental United States	Lineal America	Borderlands	Plantation Belt
Education	7.627*** (0.696)	−0.871 (8.972)	9.781*** (2.944)	−1.302 (7.123)
Manufacturing	−0.012*** (0.002)	−0.016*** (0.002)	−0.005 (0.036)	−0.014*** (0.003)
Government	−0.021*** (0.003)	−0.016** (0.007)	−0.032 (0.028)	−0.023*** (0.005)
Farming	−0.040*** (0.005)	0.020** (0.009)	−0.055* (0.031)	0.042*** (0.013)
Business density	−0.005 (0.009)	−0.189** (0.078)	−0.061 (0.435)	0.010 (0.053)
Index	−0.007 (0.026)	0.092** (0.044)	0.259 (0.181)	−0.065 (0.048)
Index squared	0.000 (0.000)	−0.002* (0.001)	−0.004 (0.002)	0.001 (0.001)
Constant	−1.670*** (0.082)	−1.637*** (0.113)	0.616 (0.627)	−2.193*** (0.135)
Observations	35,121	4,890	1,091	5,559
Number of counties	3,046	413	103	471
Log likelihood	−10,1343	−10,645	−3,474	−13,473
Wald χ²	5,394	1,529	131.5	1,413

Note. Standard errors in parentheses.

p < .1. **p < .05. ***p < .01.

Table 5.

Index and Job Growth.

	(1)	(2)	(3)	(4)
Variables	Continental United States	Lineal America	Borderlands	Plantation Belt
Education	3.585*** (0.521)	−8.686 (11.993)	−0.703 (1.349)	11.810 (8.964)
Manufacturing	−0.038*** (0.002)	−0.036*** (0.003)	−0.070*** (0.017)	−0.041*** (0.004)
Government	−0.040*** (0.002)	−0.028*** (0.009)	−0.027** (0.013)	−0.051*** (0.007)
Farming	−0.027*** (0.004)	0.045*** (0.012)	−0.057*** (0.014)	−0.026 (0.016)
Business density	0.012* (0.007)	−0.098 (0.105)	0.022 (0.199)	0.089 (0.067)
Index	0.017 (0.020)	0.026 (0.059)	0.201** (0.083)	0.194*** (0.060)
Index squared	0.000 (0.001)	0.000 (0.001)	−0.002* (0.001)	−0.004*** (0.001)
Constant	0.745*** (0.061)	1.109*** (0.151)	0.983*** (0.287)	0.910*** (0.170)
Observations	35,121	4,890	1,091	5,559
Number of counties	3,046	413	103	471
Log likelihood	−91,159	−12,064	−2,623	−14,750
Wald χ²	2,996	702.6	203.4	713.1

Note. Standard errors in parentheses.

p < .1. **p < .05. ***p < .01.

Education was consistently significant with a positive sign, as we would expect a priori for the United States. However, the positive effect of education breaks down for Lineal America (income and job growth) and the Plantation Belt (income and job growth). In Lineal America, the region’s dependence on extractive industries and history of attracting college-educated people wishing to reconnect to the land (Salstrom, 2003) may play roles in influencing this relationship. Job growth in the Plantation Belt may be in lower skill arenas.

Unsurprisingly, given the shrinkage in the manufacturing sector over the period of analysis, Manufacturing was statistically significant and negative for the United States in all four equations. This relationship held up for the regions as well, with the exception of the Borderlands (income growth only), where it was statistically insignificant. The Borderlands income growth equations were also the only equations where the Government and Farming sector variables were not significant (and negative).

The business density variable was (weakly) significant (with positive sign) for the United States only in the job growth/index equation, but significant and negative for both Lineal America income growth equations. Scholars (Audretsch & Thurik, 2000; Oberhauser, 1995) have noted that part of the country especially focused on low-paying “homework” or small home-based or craft enterprises. It could be that these employment strategies impede overall income growth in the region, providing subsistence wages while holding labor out of sectors that can achieve economies of scale.

Turning to our employment share variables, we first discuss the equations with the size category explicitly modeled by employment shares.¹⁰ We test whether there are differential effects of the employment share variables across the regions by estimating the model in Equation (2) using interaction terms for each region. The results, available from the authors on request, show statistically significant employment share differences for the high-poverty regions for both income and job growth.¹¹ This supports the need to examine each region. The F tests of joint significance of the employment shares variables are shown in both Tables 2 and 3. The employment shares are jointly different from zero at the 10% level for all regions except income growth in the United States and job growth in the Borderlands.

For job growth in the United States (Table 3), each of the smallest six firm-size categories were statistically significant and positive. In general, the magnitudes of employment shares also decrease as the size categories increase. This suggests that the employment shares in the smaller size categories have the biggest influence on job growth. Only the 10- to 19-employee size category, however, was statistically significant and positive in the U.S. income growth equation (Table 2). Table 1 also shows that increased share in the 10- to 19-employee size category is significant across all our regions, but the sign varies, indicating that some may be overinvested in this size class. In Lineal America, the negative relationship of the 10- to 19-employee size category also carries forward into job growth.

The results for the index of employment shares created from Equation (1) are found in Tables 4 and 5.¹² The index variable was not significant in either U.S. income or job growth equations; however, the index was positive (with expected opposite squared sign) and statistically significant for income growth in Lineal America. This implies that a more unequal distribution of firms is income-growth enhancing in Lineal America. The index is also positive and statistically significant for job growth in the Borderlands and the Plantation Belt, suggesting that deviation from a uniform distribution increases job growth.

Conclusion and Policy Implications

Our results demonstrate that the size distribution of firms in a local economy does seem to matter, especially with respect to job growth outcomes. When employment is weighted toward the smaller end of the distribution, a county does better than its national counterparts. On average and after controlling for other variables, nationally, income growth has a positive relationship with only one relatively narrowly defined small-size category (10-19 employees). However, with neutral or positive effects on the smaller end of the spectrum, the implications are relatively clear—focusing on smaller firms seems to pay off for most regions.

Results from our three high-poverty regions imply that region-specific analysis is advisable in understanding the results of the interplay among firm sizes. In the Plantation Belt, for example, contrary to the national story, more employment by firms in the relatively large 200-250 category is associated with job growth. A 200-employee firm is the type of scale that seems to be attractive to state agencies offering incentive packages. So in the Plantation Belt, old-style industrial recruitment might still be an important strategy, if the goal is job growth. However, our analysis provides evidence that expansion of the number of this size firm is likely to influence income growth in that region. Another contrarian region is Lineal America, where employment share in some of the smaller firm-size categories comes up negative, indicating overreliance on small firms, both for job and income growth. Here we return to an observation made earlier—not all firms are entrepreneurial. Perhaps a disproportionate share of the small firms in Lineal America are lifestyle firms, where neither owner nor workers wish to gain additional income through growth. Much of the region has a history of attracting “back-to-the-land migrants” seeking this type of lifestyle (Salstrom, 2003). It could also simply be that Lineal America simply has too many firms on the small end of the continuum, similar to the Fotopoulos (2012) finding for some European Union regions. Our findings for the multistate high-poverty regions argue for a continued role for organizations such as the Appalachian Regional Commission, but with the caveat that the mission and regional coverage of such organizations should be reviewed from time to time as needs change, and the nature of the policies needed for addressing disparities evolve.

The Loveridge and Nizalov employment share index did not perform well at the national level, but was significant for three of our four underperforming regions. In particular, the index was statistically significant in the income growth model for Lineal America and in the job growth model for the Borderland and Plantation Belt regions. The index is in some ways a “straw man” in that it measures deviations from equal employment shares. It could well be that the ideal national distribution is skewed toward smaller or larger firm sizes. Future research on deviations from alternative distributions might produce more significant results in national modeling exercises. Such results might inform national policy, but one of our principal findings is that optimal firm size distributions should be explored and acted on by subnational regions rather than via a nationally set standard. The optimal distribution will likely vary by region.

The present study uses publically available data that do not provide information on firm mobility between employment size classes. Work by Haltiwanger, Jarmin, and Miranda (2010), using the U.S. Census Bureau’s Longitudinal Business Database, suggests that the effect of firm size on economic growth is sensitive to controlling for firm age. This highlights an area of future research that examines firm graduation between size categories, neighboring firm sizes, and economic growth. Thus, one could examine more rigorously the agglomeration effects of the firm size distribution and look at whether job growth comes from small businesses being formed or whether the small businesses grow in size.

Footnotes

Appendix

Table A5.

Alternative Estimation Methods: Job Growth.

	(1)	(2)	(3)	(4)	(5)	(6)
	POLS	POLS	FE	FE	FD	Abond
Education		0.496 (0.353)		4.709*** (0.648)	−0.468 (3.594)	2.063 (1.492)
Manufacturing		−0.005*** (0.001)		−0.041*** (0.002)	−0.141*** (0.025)	−0.143*** (0.026)
Government		−0.003*** (0.001)		−0.052*** (0.003)	−0.147*** (0.033)	−0.175*** (0.036)
Farming		−0.000 (0.001)		−0.028*** (0.005)	−0.174*** (0.017)	−0.128*** (0.014)
Business density		−0.000 (0.000)		0.009 (0.007)	−0.003 (0.004)	−0.031 (0.025)
Index	−0.026 (0.023)	−0.015 (0.025)	−0.033 (0.022)	0.023 (0.023)	0.078 (0.066)	0.083 (0.061)
Index squared	0.001* (0.000)	0.000 (0.001)	0.001* (0.000)	−0.000 (0.000)	−0.002 (0.001)	−0.001 (0.001)
Job growth 1-year lag						0.065* (0.034)
Constant	0.897*** (0.286)	1.586*** (0.322)	1.081*** (0.323)	7.032*** (0.465)	0.238*** (0.063)	30.085*** (4.715)
R² log likelihood	.002	.051	.000	.082	.122

Note. Robust standard errors in parentheses. Time dummies are suppressed for display purposes. POLS = pooled ordinary least squares; FE = fixed effects; FD = first difference; Abond = Arellano-Bond.

p < .1. **p < .05. ***p < .01.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors gratefully acknowledge support for this research from U.S. Department of Agriculture, National Institute of Food and Agriculture competitive grant number 2009-35900-05935.

Notes

Author Biographies

Timothy Komarek is an assistant professor in the Department of Marketing and Economics at Valdosta State University. His interests include regional economic development models, forecasting, labor, environment, and housing.

Scott Loveridge is the director of the North Central Regional Center for Rural Development (NCRCRD) and professor of Agricultural, Food, and Resource Economics at Michigan State University. As director of the NCRCRD, he provides leadership for university involvement in a wide array of rural development issues in a 12-state region of the upper Midwest.

References

Agrawal

Cockburn

Galasso

Oettle

(2012, June). Why are some regions more innovative than others? The role of firm size diversity. Paper presented at the Fifth Annual Conference on Entrepreneurship and Innovation, Searle Center on Law, Regulation, and Economic Growth, Chicago, IL. Retrieved from http://www.law.northwestern.edu/faculty/programs/searlecenter/workingpapers/documents/Oettl_Firm_Size_Diversity_Regional_Innovation.pdf

Allanach

Loveridge

(1998). An assessment of maximum-training business visitation programs. Economic Development Quarterly, 12, 125-136.

Aquilina

Klump

Pietrobelli

(2006). Factor substitution, average firm size and economic growth. Small Business Economics, 26, 203-214.

Atkinson

A. B.

(1970). Measurement of inequality. Journal of Economic Theory, 2, 244-263.

Audretsch

D. B.

Thurik

A. R.

(2000). Capitalism and democracy in the 21st century: From the managed to the entrepreneurial economy. Journal of Evolutionary Economics, 10, 17-34.

Baptista

Lima

Preto

M. T.

(2012). How do former business owners fare in the labor market? Job assignment and earnings. European Economic Review, 56, 263-276.

Besley

Case

(2000). Unnatural experiments? Estimating the incidence of endogenous policies. Economic Journal, 110(467), 672-694.

Birch

D. L.

(1981). Who creates jobs? Public Interest, 65(3), 3-14.

Birch

D. L.

(1987). Job creation in America: How our smallest companies put the most people to work. New York, NY: Free Press.

10.

Blair

J. P.

Premus

(1987). Major factors in industrial location: A review. Economic Development Quarterly, 1, 72-85.

11.

Bradshaw

Blakely

(1999). What are “third-wave” economic development strategies? From incentives to industrial policy. Economic Development Quarterly, 13, 229-244.

12.

Carlino

G. R.

(1982). Manufacturing agglomeration economies as returns to scale: A production function approach. Papers of the Regional Science Association, 50, 95-108.

13.

Carruthers

Mulligan

G. F.

(2012). The plane of living and the precrisis evolution of housing values in the USA. Journal of Economic Geography, 12, 739-773.

14.

Conroy

(1974). Alternative strategies for regional industrial diversification. Journal of Regional Science, 14, 31-46.

15.

Cooley

T. F.

Rupert

(2010, April 21). The great housing recession continues. Forbes.com. Retrieved from http://www.forbes.com/2010/04/20/housing-foreclosure-unemployment-opinions-columnists-thomas-cooley-peter-rupert.html

16.

Deller

(2010). Spatial variations in the role of microenterprises in economic growth. Review of Regional Studies, 40, 71-97.

17.

Deller

McConnon

J. C.

(2009). Microenterprises and economic growth: A panel study of the US states 1977-1997. Applied Economics Letters, 16, 1307-1312.

18.

Devereux

M. P.

Griffith

Simpson

(2007). Firm location decisions, regional grants, and agglomeration externalities. Journal of Public Economics, 91, 413-435.

19.

Drewnowski

Specter

S. E.

(2004). Poverty and obesity: The role of energy density and energy costs. American Journal of Clinical Nutrition, 79, 6-16.

20.

Drucker

(2012). The spatial extent of agglomeration economies: Evidence from three U.S. manufacturing industries (Center for Economic Studies Paper 12-01). Washington, DC: U.S. Census Bureau.

21.

Edward Lowe Foundation. (n.d.). 2nd Stage growth companies. Retrieved from http://edwardlowe.org/who-we-serve/secondstage/

22.

Eisinger

P. K.

(1988). The rise of the entrepreneurial state: State and local economic development policy in the United States. Madison: University of Wisconsin Press.

23.

Felmlee

D. H.

(1982). Women’s job mobility processes within and between employers. American Sociological Review, 47, 142-151.

24.

Ferrer

Lluis

(2008). Should workers care about firm size? Industrial & Labor Relations Review, 62, 104-125.

25.

Fotopoulos

(2012). Nonlinearities in regional economic growth and convergence: The role of entrepreneurship in the European Union. Annals of Regional Science, 48, 719-741.

26.

Goetz

S. J.

Partridge

Deller

S. C.

Fleming

D. A.

(2010). Evaluating U.S. rural entrepreneurship policy. Journal of Regional Analysis & Policy, 40, 20-33.

27.

Hansen

E. R.

(1990). Agglomeration economies and industrial decentralization: The wage-productivity trade-offs. Journal of Urban Economics, 28, 140-159.

28.

Haltiwanger

J. C.

Jarmin

R. S.

Miranda

(2010). Who creates jobs? Small vs. large vs. young (NBER Working Paper No. 16300). Cambridge, MA: National Bureau of Economic Research.

29.

Hodge

T. R.

(2011). Inventory of state job creation and job retention incentives, 2010 (Policy Brief). East Lansing: North Central Regional Center for Rural Development, Michigan State University.

30.

Hoover

Giarratani

(1984). An introduction to regional economics (3rd ed.). New York, NY: Alfred A. Knopf.

31.

Krugman

(1991). Geography and trade. Leuven, Belgium: Leuven University Press.

32.

Lichtenstein

G. A.

Lyons

T. S.

(2006). Managing the community’s pipeline of entrepreneurs and enterprises: A new way of thinking about business assets. Economic Development Quarterly, 20, 377-386.

33.

Lichtenstein

G. A.

Lyons

T. S.

(2010). Investing in entrepreneurs: A strategic approach to building a stronger regional and community economy. Santa Barbara, CA: Praeger/ABC-CLIO.

34.

Lin

M.-J.

(2008). Does unemployment increase crime? Evidence from U.S. data 1974-2000. Journal of Human Resources, 43, 413-436.

35.

Loveridge

(1996). On the continuing popularity of industrial recruitment. Economic Development Quarterly, 10, 151-158.

36.

Loveridge

Nizalov

(2007). Operationalizing the entrepreneurial pipeline theory: An empirical test of the optimal size distribution of local firms. Economic Development Quarterly, 21, 244-262.

37.

Loveridge

Bokemeier

(2009). Why are they moving away? Comparing attachment to place in the Great Plains to the rest of the nation. Online Journal of Rural Research & Policy, 4(1). doi:10.4148/ojrrp.v4i1.56

38.

Low

S. A.

Isserman

A. M.

(2009). Ethanol and the local economy industry trends, location factors, economic impacts, and risks. Economic Development Quarterly, 23, 71-88.

39.

Low

S. A.

Weiler

(2012). Employment risk, returns, and entrepreneurship. Economic Development Quarterly, 26, 238-251.

40.

Morse

(Ed.). (1990). The retention and expansion of existing businesses. Ames: Iowa State University Press.

41.

Nostrand

(1970). The Hispanic-American borderland: Delimitation of an American culture region. Annals of the Association of American Geographers, 60, 638-661.

42.

Oberhauser

A. M.

(1995). Gender and household economic strategies in rural Appalachia. Gender, Place and Culture, 2, 51-70.

43.

Olfert

M. R.

Partridge

M. D.

(2010). Best practices in twenty-first-century rural development and policy. Growth and Change, 41, 147-164.

44.

Parker

S. C.

(2004). The economics of self-employment and entrepreneurship. Cambridge, England: Cambridge University Press.

45.

Pereira

A. M.

Andraz

J. M.

(2012). On the regional incidence of highway investments in the USA. Annals of Regional Science, 48, 819-838. doi:10.1007/s00168-010-0409-1

46.

Porter

(1990). The competitive advantage of nations. New York, NY: Free Press.

47.

Porter

(1998). Clusters and the new economics of competition. Harvard Business Review, 76(6), 77-90.

48.

Porter

(2000). Location, competition, and economic development: Local clusters in global competition. Economic Development Quarterly, 14, 15-34.

49.

Robbins

D. K.

Pantuosco

L. J.

Parker

D. F.

Fuller

B. K.

(2000). An empirical assessment of the contribution of small business employment to US state economic performance. Small Business Economics, 15, 293-302.

50.

Salstrom

(2003). The neonatives: Back-to-the-land in Appalachia’s 1970s. Appalachian Journal, 30, 308-323.

51.

Schumpeter

J. A.

(1942). Capitalism, socialism and democracy. New York, NY: Harper & Row.

52.

Schumpeter

J. A.

(1961). The theory of economic development. Cambridge, MA: Harvard University Press.

53.

Shaffer

Deller

Marcouiller

(2006). Rethinking community economic development. Economic Development Quarterly, 20, 59-74.

54.

Shaffer

(2006). Establishment size and local employment growth. Small Business Economics, 26, 439-454.

55.

Solecki

(1996). Paternalism, Pollution, and Protest in a Company Town. Political Geography, 15(1): 5-20.

56.

Spelman

(2006). Growth, stability, and the urban portfolio. Economic Development Quarterly, 20, 299-315.

57.

Stephens

Partridge

Faggian

(2011). Innovation, entrepreneurship and economic growth in lagging regions (Swank Program Papers). Columbus: Ohio State University. Retrieved from http://aede.osu.edu/sites/aede/files/publication_files/Innovation%2C%20entrepreneurship%20.pdf

58.

Sturm

Wells

K. B.

(2001). Does obesity contribute as much to morbidity as poverty or smoking? Public Health, 115, 229-235.

59.

Von Bargen

Freedman

Pages

E. R

. (2003). The rise of the entrepreneurial society. Economic Development Quarterly, 17, 315-324.