Received: 7 November 2018 Revised: 30 July 2019 Accepted: 10 September 2019

DOI: 10.1111/pirs.12482

FULL ARTICLE

Economic growth through entrepreneurship:

Determinants of self‐employed income across
regional economies

David B. Willis1 | David W. Hughes2 | Kathryn A. Boys3 |

Devin C. Swindall4,5

1
Department of Agricultural and
Environmental Sciences, Clemson University, Abstract
Clemson, SC 29634‐0309, USA Knowledge of the determinants of self‐employment income
2
Department of Agricultural and Resource
is critical to entrepreneurial development strategies if the
Economics, University of Tennessee, Knoxville,
TN 37996‐4518, USA development goal is to increase incomes not just employ-
3
Department of Agricultural and Resource ment. Using American Community Survey data, uncondi-
Economics, North Carolina State University,
Raleigh, NC 27695, USA
tional quantile regression is used to investigate differences
4
Clemson Institute for Economic and in the relationship between entrepreneurial income and an
Community Development, Columbia, SC array of individual, industry, and regional characteristics
29229, USA
5 across the self‐employment income distribution. Personal
IMPLAN Group LLC, 16905 Northcross
Drive, Suite 120, Huntersville, NC 28078, USA attributes, such as education, race, age, and gender, both
explain differences in self‐employment income and vary in
Correspondence
Dr. David Willis, Department of Agricultural importance across the income distribution. Regional
and Environmental Sciences, Clemson
agglomerative effects are significantly positive and stronger
University, 239 McAdams Hall, Clemson
University, Clemson, SC 29634‐0312, USA. at the upper end of the self‐employed income distribution.
Email: [email protected]

KEYWORDS

economic growth, entrepreneurship, income, self‐employed

JEL CLASSIFICATION
J24; J31; O18; R11

-------------------------------------------------------------------------------------------------------
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
© 2019 The Authors. Papers in Regional Science published by John Wiley & Sons Ltd on behalf of Regional Science Association
International

Pap Reg Sci. 2019;99:73–95. wileyonlinelibrary.com/journal/pirs 73

74 WILLIS ET AL.

1 | I N T R O D U CT I O N

State and local government officials have traditionally focused on major industry and/or firm recruitment as a driver
of job and income growth in regional economies. Some researchers have criticized this approach to economic devel-
opment because it often fails to consider the public cost of providing additional services, such as roads, schools,
police protection, water and sewer services, and the potential negative environmental costs that often accompany
industrial growth (Leistritz & Sell, 2001). As an alternative entrepreneurial led growth facilitated through a support
system, may provide a more cost‐effective means to engender regional economic growth than an industrial
recruitment‐led approach (Shrestha, Goetz, & Rupasingha, 2007; Spindler, 1994; Yenerall, 2008).
While entrepreneurial activity is an intuitively promising vehicle for fostering economic development (Edmiston,
2007; Henderson, 2002), the extent to which self‐employment facilitates local economic growth is unclear. The
appeal of entrepreneurial based regional growth strategies is derived from entrepreneurial theory that view entrepre-
neurs as change agents who bear the risk associated with introducing innovations (Evans & Jovanovic, 1989;
Formaini, 2001). In the new growth theory, entrepreneurs are innovators that possess the broad skill set necessary
to utilize new technologies to meet new economic opportunities (Lazear, 2004; Romer, 1994).
Prior U.S. entrepreneurial research has focused on regional growth in the number of self‐employed individuals
rather than self‐employment income (Edmiston, 2007; Gallardo & Scammahorn, 2011; Henderson, 2002; Low &
Weiler, 2012; Van Praag & Versloot, 2007), or on the relationship between the number of self‐employed individuals
and various measures of regional economic activity (Acs, 2007; Bunten, Weiler, Thompson, & Zahran, 2015; Fleming
& Goetz, 2011; Glaeser & Kerr, 2010; Low, 2009; McGranahan, Wojan, & Lambert, 2010). However, if entrepreneurs
earn less than their salaried employment alternatives, increased entrepreneurship may not be an efficient regional
development strategy.
In a national county‐level study Low, Henderson, and Weiler (2005) examined the relationship between per job
entrepreneurial income1 and income determinants. County differences in average proprietor income was positively
related to urban status, percent of county population with a college education, average county bank deposits, high
speed internet access, and having an interstate highway. Counties located in the U.S. West and Midwest had lower,
and Northeastern counties had higher, per job entrepreneurial income. In a study of Appalachian counties, Stephens
and Partridge (2013) estimated total county income increased by 0.65% for a 1% increase in county proprietor
income. These studies, however, did not investigate the determinants of per capita income within a regional context.
Other studies have focused on the importance of entrepreneurial characteristics and/or personal goals in
explaining entrepreneurial income at the national level. A sampling of these studies includes those conducted by:
Aldrich & Cliff, 2003; Evans & Jovanovic, 1989; Evans & Leighton, 1989; Gurley‐Calvez & Hammond, 2010;
Hamilton, 2000; Holtz‐Eakin, Joulfaian, & Rosen, 1994; Kusmin, 2010; Moskowitz & Vissing‐Jorgensen, 2002; Shane
& Venkataraman, 2000; Benz & Frey, 2008; Block & Koellinger, 2009; Hurst & Pugsley, 2011; and Astebro &
Thompson, 2011. A common limitation of these studies is that scant attention is paid to individual characteristics
and local economic activity in explaining self‐employed income. To this end and in contrast with these prior studies,
this study examines the determinants of the individual self‐employed income level within the context of: (i) entrepre-
neur personal attributes, (ii) resources available to the entrepreneur, (iii) characteristics of the industry in which the
entrepreneur operates and, (iv) characteristics of the region's economy. Importantly and uniquely as well, this study
introduces unconditional quantile analysis as a novel approach to examine determinants of entrepreneurial income
and how the characteristics of entrepreneurs vary by their income level.
An examination of possible differences in returns to self‐employment at different income level is important
because policy makers, practitioners, and academics argue that self‐employment is a way to provide opportunity
for lower income households and less affluent areas to raise their status (Bruton, Ketchen, & Ireland, 2013; Chliova,
Brinckmann, & Rosenbusch, 2015; Heemskerk, 2003; Karnani, 2007; London & Hart, 2011; Sutter, Bruton, & Chen,

1
Calculated as total county proprietor income divided by county proprietor employment using BEA‐REIS data.
WILLIS ET AL. 75

2019; Webb, Bruton, Tihanyi, & Ireland, 2013). Policy makers often institute programs aimed at easing poverty
through self‐employment education efforts (Laney, Bowles, & Hillard, 2013) or other means (Wolff & Nivorozhkin,
2012). Advocates of social entrepreneurship especially see it as a means of easing poverty (Barki, Comini, Cunliffe,
Hart, & Rai, 2015). However, self‐employment may affect household incomes in different ways with different impacts
of key determinants. A better understanding of how self‐employment influences returns to different levels of house-
hold income provides greater insight to the process of possible poverty alleviation and reduction in inequality of
income. Accordingly, results of this paper could be used to better target efforts aimed at using self‐employment to
reduce social problems.

1.1 | Study Setting

The three southeastern states of Georgia, North Carolina, and South Carolina serve as the study area. Over the past
decade, per capita income in each state has consistently been well below the national average. In 2016, South
Carolina ranked 44th, Georgia 40th, and North Carolina 39th among the 50 states and District of Columbia in per
capita income (U.S. Department of Commerce, 2018). The three‐state percentage of self‐employed income to total
earnings2 has remained relatively stable despite a significant increase in the share of entrepreneurial employment
to total employment. In 1998, the percentage of self‐employed income to total earnings was 9.9%; this value
declined to a low of 9.7% in 2011 before increasing to 10.9% in 2016. Self‐employment as a share of total employ-
ment, however, increased from 14.8% in 1998 to 22.4% in 2016 reflecting a 51.3% increase (U.S. Department of
Commerce, 2018). Thus, between 1998 and 2016 in these states, per job entrepreneurial earnings decreased relative
to per job wage and salary income. The decline in per entrepreneur earnings relative to wages is inconsistent with the
argument that growth in entrepreneurship increases average wealth and highlights the need to examine the drivers of
entrepreneurial income.3,4 The year 2008 is selected as the year of this analysis. This year was the first full year of the
Great Recession and, due to this, was thought to offer a useful case year for analysis.5

2 | LITERATURE REVIEW

Entrepreneurship has been defined as “the process of creating or seizing an opportunity and pursuing it regardless of
the resources currently controlled” (Fayolle, 2007, p.37). From this perspective, business owners can be viewed as
entrepreneurs as they are the business decision makers and risk takers.6 Despite the role of self‐employment as
an agent of economic growth, and the three‐decade increase in the share of national self‐employment relative to

2
Total earnings are defined as the sum of personal income‐wage, salary disbursements, and proprietors' income (Regional Economic Information System
2018).
3
It is worth noting that entrepreneurial activity in these states is generally representative of national entrepreneurship activity. A review of Kauffman Foun-
dation (2019) statistics for 2008 indicates that these three states are representative of the US. Specifically, the data indicates small differences in rate of
new entrepreneurs (0.33% for our three states versus 0.30% nationally), opportunity share of the new entrepreneurs (85.6% versus 81.4%), startup early
job creation (5.81 versus 5.51) and startup early survival rate (77.7% versus 77.6%).
4
This decline predated the national recession. Still, in 2016, U.S. per entrepreneurial real (2000 dollars) income was only $22,520. The same variable for the
three states in our analysis shows the same trend.
5
Intuitively we had thought that the economic circumstances of 2008 would offer a particularly interesting time period for our case study, it turns out that
this is not supported by literature and other evidence. The most pertinent literature in this regard is the relationship between self‐employment and unem-
ployment levels and provides mixed evidence (Block & Wagner, 2010; Halicioglu & Yolac, 2015; Schuetze, 2000). Quoting Blanchflower (2000), “there is
some disagreement in the literature on whether high unemployment acts to discourage self‐employment because of the lack of available opportunities
or encourage it because of the lack of viable alternatives (P. 477).”We have undertaken some additional analyses to examine this issue further. More rele-
vant, we have also analyzed US entrepreneurship data provided by the Kauffman Foundation (2019) which indicates that for the US the share of opportu-
nity entrepreneurs was 80.7% in 2008 versus an average of 80.1% for 1998 through 2017. For several measures, data indicates that 2008 is not an atypical
year. Specifically, across all states, the rate of new entrepreneurs (no difference), opportunity entrepreneur share (a 3.1% relative difference), startup job
creation (a decline of 11.2%), and startup survival rates (a 2% difference), in general showed little change from 2008 to 2017.
6
This is in contrast to the role of stockholders who must take investment risks but do not make day‐to‐day business decisions and hired managers, who
make business decisions but do not take investment risk.
76 WILLIS ET AL.

total employment, neoclassical economists continue to treat entrepreneurs as a black box when explaining regional
growth. As noted by Goetz (2003, p.4), “there have been only scant attempts to develop formal theories of entrepre-
neurship and even fewer efforts to formally study proprietorship formations.”

2.1 | Attributes of entrepreneurs

The entrepreneurial business literature has tended to focus on identifying individual and cultural attributes correlated
with entrepreneurial success, or on teaching entrepreneurial skills. Prior work experience in an industry closely
related to the entrepreneurial activity is frequently a key determinant of entrepreneurial success (Colombatto &
Melnik, 2007). Why, when, and how entrepreneurship opportunities are recognized by some individuals and not
others is influenced by idiosyncratic knowledge and skills acquired via various prior work experiences (Aldrich & Cliff,
2003; Shane & Venkataraman, 2000). This literature also suggests entrepreneur age is correlated with labor market
experience and improves entrepreneurial decision‐making (Holtz‐Eakin et al., 1994). As such, age is frequently used
as a proxy for individual work experience.
Other individual attributes such as education, gender, and ethnicity have been identified as important individual
characteristics that influence self‐employed income levels (Gurley‐Calvez & Hammond, 2010; Kusmin, 2010).
Additional education is believed to enhance one's ability to make profitable business decisions. Prior studies have
found males generally have higher incomes than females, and whites tend to have higher incomes than non‐whites,
which may be attributable to labor market discrimination (Kusmin, 2010). Evans (1989) found ethnic entrepreneurs in
concentrated ethnic markets often succeed because they understand ethnic preferences and their customers often
prefer to do business with individuals sharing similar cultural experiences.
In addition, a number of studies have found that the entrepreneur's lifestyle goals are a major determinant in the
decision to pursue self‐employment. Entrepreneurs are often willing to trade income for non‐pecuniary outcomes
such as greater independence, the ability to “be one's own boss” (Benz & Frey, 2008; Blanchflower, 2000;
Blanchflower & Oswald, 1998; Blanchflower, Oswald, & Stutzer, 2001), or a greater variety in work activities (Astebro
& Thompson, 2011). In a survey of 290 small Western Australia business owners, Walker and Brown (2004) found
non‐financial objectives were often the primary motive for many small business start‐ups. These authors report that
personal satisfaction, pride in work, being one's own boss, and a flexible lifestyle were often more highly valued than
earning additional income. Hurst and Pugsley (2011) found that factors such as age, education, individual wealth, and
characteristics of the business affect the magnitude of the work/lifestyle tradeoff. By ignoring the varied goals of
entrepreneurs, the literature may have overstated the importance of talent, luck, and access to capital in explaining
the size distribution of entrepreneurial firms.

2.2 | Assets of entrepreneurs

Access to financial institutions is also important to successful self‐employment. Evans and Jovanovic (1989) found
that access to capital is a key determinant in starting a business. Goetz and Freshwater (2001) examined the role
of regional financial institutions in explaining entrepreneurial activity. In examining all 50 states, these authors esti-
mated a positive relationship between the concentration of regional financial services and entrepreneurial income
that plateaued at high self‐employment income. Glaeser and Kohlhase (2004) provide evidence of the strong associ-
ation between the concentration of financial services and urban growth as an agglomerative effect.
Similarly, financial assets owned by the entrepreneur's family represent additional resources and increase the
probability of entrepreneurial success (Aldrich & Cliff, 2003; Shane & Venkataraman, 2000). Goetz and Freshwater
(2001) found that wealthier individuals are more likely to engage in entrepreneurial activities because their wealth
provides easier access to additional capital. Walzer (2007) and Todorovic (2004) report that home ownership is an
indicator of entrepreneur success because the skills needed to maintain a home, such as risk taking, being proactive,
WILLIS ET AL. 77

and having a desire to succeed, parallel those of creating and maintaining a business. At the regional level, Fairlies and
Krashinsky (2012) report that housing value appreciation is positively correlated with entrepreneurial growth, and
Parker and Belghitar (2006) found that nascent entrepreneurs are much more likely to become permanently
self‐employed if they owned a home. However, Hurst and Lusardi (2004) found no relationship between regional
entrepreneurial start rates and local housing appreciation.
Prior research has found that individuals lacking health insurance are less likely to be self‐employed (Fairlie, Kapur,
& Gates, 2011; Gumus & Regan, 2015; Holtz‐Eakin, Penrod, & Rosen, 1996; Wellington, 2001). This finding has given
support to the job lock hypothesis that posits the loss of employer provided health insurance inhibits potential entre-
preneurial employment. Holtz‐Eakin et al. (1996) found that wage and salary earners are less likely to become
self‐employed if they fear losing employer‐provided health care coverage. Thus, it is likely that individuals with
spouses who have health insurance are more likely to be self‐employed all else equal.

2.3 | Strategies of entrepreneurs

Many new business start‐ups focus on providing services to existing local markets, as opposed to forging new
markets and, once established, often do not anticipate growing larger or innovating in any major way. These entre-
preneurial characteristics tend to be associated with small business owners that specialize in either service provision
(e.g., lawyers, physicians, restaurant owners) or retail sales.
On a related note, Moskowitz and Vissing‐Jorgensen (2002) and Hamilton (2000), found full‐time entrepreneurs
earn less than comparable full‐time wage and salary workers. Hamilton (2000) estimated that entrepreneurs in busi-
ness for 10 years had a median annual income 35% lower than their salaried counterparts. This earning differential
was found to be consistent across industries and he concluded the lower median entrepreneurial income level was
related to the possibility that some entrepreneurs are partially influenced by non‐pecuniary goals. Other researchers
disagree with Hamilton's conclusion and postulate the lower income level may be attributable to overconfidence and
excessive optimism by entrepreneurs regarding their potential entrepreneurial earnings relative to their skill level
(Camerer & Lovallo, 1999; Koellinger, Minniti, & Schade, 2007).

2.4 | Macroeconomic and local characteristics affecting entrepreneurship

Business cycle phases have been found to influence the decision to become an entrepreneur. Those whose entrepre-
neurial start is influenced by the business cycle are commonly classified as either “necessity” or “opportunity” entre-
preneurs. It has been observed that during periods of high unemployment, less‐skilled individuals may opt for self‐
employment due to lost or reduced income (Koellinger & Thurik, 2012; Low & Weiler, 2012). In contrast, “opportu-
nity” entrepreneurs are not forced‐in by economic circumstances, but rather see an opportunity to increase earnings
relative to their current employment situation. Somewhat counterintuitively, an economic downturn may increase
the number of opportunity entrepreneurs due to lower costs of capital, materials and/or labor (Parker, 2009). During
economic downturns, already established “opportunity” entrepreneurs focus on the adoption of new production
technologies, and take advantage of depressed capital costs to reduce per‐unit production costs (Bowen & De Clercq,
2008; Hessels, Gelderen, & Thurik, 2008).
In any particular year, regions with lagging economic growth are likely to have more necessity entrepreneurs than
more prosperous regions, as individuals living in less prosperous regions start businesses due to asset fixity (Hite,
1997), or social capital considerations (Sabatini, 2008). In an analysis of returns to self‐employment, Evans and
Leighton (1989) found that entrepreneurs starting a new business, who had less formal education and a long period
of unemployment prior to starting their business, were often less skilled than individuals doing the same job for either
wage or salary income. Supporting this argument, Low and Weiler (2012) found that limited employment prospects
increased the likelihood of individuals opting to become entrepreneurs within regional commuting zones.
78 WILLIS ET AL.

Local economic structure tends to influence entrepreneurial success (Evans & Jovanovic, 1989; Goetz & Freshwa-
ter, 2001; Low & Weiler, 2012; Walzer, 2007). In particular Goetz and Freshwater (2001), and Walzer (2007) found
regional self‐employment is positively correlated with capital access as measured by the per capita number of
regional financial institutions. Aggressive marketing of regional amenities designed to draw tourism can also attract
entrepreneurship development. Walzer (2007, p.67) states, “tourism opportunities differentiated counties with
respect to growth during the 1990s” and since then “there has been increased interest in amenities.”
There is growing support for the hypothesis that agglomeration economies play a pivotal role in the growth of
regional economies (Shaffer, Deller, & Marcouiller, 2004). Population density is a driver of Jacobs' type agglomerative
economies.7 Glaeser and Kohlhase (2004) found a strong connection between worker productivity and population
density; they argue that increases in population density accelerates the spread of knowledge, attracts skilled workers,
increases social capital, and improves entrepreneurial opportunities. Moreover, central place theory suggests that the
availability of financial resources and services increases with population density, and thus an entrepreneurs' chance
to obtain outside financial resources should increase with population density (Shaffer et al., 2004). Higher population
density also increases access to skilled employees with diverse skill sets, increase the size of the local market, and
access to specialized business services (Fujita & Krugman, 2004). Thus, entrepreneurs located in metropolitan areas
are more likely to benefit from agglomeration economies and have larger entrepreneurial income than those in
non‐metro areas.

2.5 | Approaches to the study of determinants of entrepreneurial income

In prior research ordinary least squares (OLS) regression has generally been used to estimate the relationship
between the level of self‐employment income and the determinants of self‐employment income. A well‐known
limitation of the OLS estimator is the assumption that the relationship between self‐employment income and its
determinants is constant across the income distribution. Observation and, to some degree theory, however, suggest
that these relationships may change over the income distribution.
Unconditional quantile regression (UQR; Firpo, Fortin, & Lemieux, 2009) provides a promising tool for relaxing
assumption of fixed relationships over the income distribution. UQR allows for the marginal effect of an independent
on the value of dependent variable to change over the distribution of the dependent variable. This research utilizes
UQR to examine the impact of various explanatory factors on income at quantiles of the self‐employed income
distribution. We are unaware of any prior study that has used this approach to analyze the determinants of self‐
employed income. A traditional conditional quantile regression (CQR) could be used to estimate the differing impacts
that sociodemographic factors have across income quantiles. CQR estimates, however, cannot be interpreted as
marginal impacts because parameter estimates explicitly depend on the values of the other conditioning variables
in each quantile (Firpo et al., 2009). In contrast, UQR parameter estimates can conveniently be interpreted in the
same manner as OLS parameters. That is, the estimated quantile parameters capture the unconditional marginal
effect of a change in each independent variable because the parameter estimates are independent of the values of
the conditioning variables.
Traditional CQR estimates at each quantile are conditional on the values of other explanatory covariates in the
quantile. Historically, the appeal of the CQR estimator over OLS estimator is that it provided a means to estimate
conditional marginal effects at different points (quantiles) in the conditional distribution of the dependent variable
(Koenker, 2005). However, policy makers are not typically interested in the conditional quantile marginal effect,

7
These are the interactions derived from the “the cramming of individuals, occupations, and industries into close quarters” (Glaeser, Kallal, Scheinkman, &
Shleifer, 1992) that generates ideas and innovations; the critical knowledge transfers coming from the variety and diversity of industries in the region as
opposed to the core industry (Glaeser et al., 1992). Additional agglomeration economy theories include the Marshall‐Arrow‐Romer theory which suggests
a core industry, similar to a local monopoly, drives knowledge spillovers within firms which, in turn, drives growth of the core industry and region (Glaeser
et al., 1992). Porter (2000) proposes that growth is driven by a core industry, however, local competition drives firms to innovate; if firms do not maintain
innovation parallel with other firms in the region, the firm will fail (Glaeser et al., 1992).
WILLIS ET AL. 79

but rather want to know the effect of a change in an explanatory variable on the unconditional quantile distribution.
The nonlinearity of the CQR estimator generates outcomes where the conditional quantile marginal effect is not
equal to the unconditional quantile marginal effect (Borah & Basu, 2013; Firpo et al., 2009) because the dependent
variable distribution is reshaped by the conditioning variables prior to estimating the conditional independent variable
coefficients. Using CQR, control variables can redefine quantiles such that the lowest earning entrepreneurs are not
necessarily in the lowest conditional quantiles. Instead, entrepreneurs in the lowest conditional quantiles are those
whose income is lower than expected given the explanatory covariates in the model (Maclean, Webber, & Marti,
2014). Thus, the marginal effect of a specific explanatory variable is defined by the values of other covariates for a
given quantile. Alternatively stated, the linearity of the OLS expectation operator that assures the conditional mean
E (YǀX) = Xβ equals the unconditional mean E(Y) = E(X)β does not apply to nonlinear operators such as quantile esti-
mators. Firpo et al. (2009) note that it is possible for CQR and UQR parameter estimates at a given quantile to be
statistically significant but have opposite signs using the same data set. Thus, CQR is inappropriate for estimating
the unconditional marginal effect of an explanatory variable on income at each quantile of the unconditional income
distribution.
In contrast, the UQR estimator combines the statistically attractive interpretation of the OLS estimator with the
ability to estimate unconditional marginal parameter effects at various quantiles of the dependent variable distribu-
tion. The UQR estimator allows the researcher to interpret parameter coefficients in a manner similar to OLS coef-
ficients because the UQR estimator respects the law of total expectations (also referred to as the law of iterated
expectation). Moreover, UQR defines quantiles pre‐estimation; including additional explanatory variables helps to
net out spurious associations between each independent variable and income, but does not affect which observa-
tions are assigned to a given income quantile.

3 | M E T H O D O L O G I C A L A P P R O A C H A N D DA T A

3.1 | Conceptual model

This research posits a functional relationship between the self‐employed income level and individual personal
attributes, economic and social/institutional assets available to the entrepreneur, and the economic environment
the entrepreneur operates within. The dependent variable is self‐employment income for entrepreneur i (Yi),and is
defined by the following broad categories of variables:


Y i ¼ f Pin ; Rik ; Sij; Air : (1)

Pin is a vector of n personal and family attributes for entrepreneur i, Rik is the k resources available to the ith entre-
preneur, Sij is a set of variables which characterize the industry j the entrepreneur operates in, and Air captures a set
of regional agglomerative effects in geographic region r. Regions are defined as the PUMA region which is used by
the U.S. Census Bureau in reporting economic and demographic data. The specific description of the variables
included in this study, plus the construction of additional variables used is presented in the data section.

3.2 | Data

The 2008 American Community Survey (ACS) data contained in the Integrated Public Use Microdata Series (IPUMS)
is used in this study. This data source was selected because it provides a large, unbiased sample of a selected popu-
lation of the states of interest.8 The ACS is conducted annually by the U.S. Census Bureau and provides data on

8
The annual ACS survey excludes areas (counties or cities of less than 65,000 in population (U.S. Census Bureau, 2018) and hence might be biased toward
urban areas. However, our dataset included a sizeable number of observations outside urban areas (27.3% non‐metropolitan residents) and we feel that it is
representative. We would like to thank the reviewers for clarification of this point.
80 WILLIS ET AL.

population and housing characteristics. In the IPUMS database, U.S. Census microdata is converted “into a single har-
monized database with uniform documentation, without losing any significant information contained in the existing
samples” (Ruggles et al., 2008, p.4). This database also includes constructed variables which use the original Census
data to allow family member records to be linked (Ruggles et al., 2008).
Several selection criteria were used to identify entrepreneurs from this dataset. Individuals outside the ages of 25
and 65 were excluded as they are not likely to be fully active in the labor market. In addition, individuals active in the
labor force were identified using IPUMS variables employment status and labor force status.9 The class of worker
variable10 was used to identify individuals who were identified as “self‐employed” as their primary labor market activ-
ity. Entrepreneurs who have a full‐time wage or salary job, or who operate a secondary business, or who farm only on
the weekends or as a hobby, are also excluded. The final data set contains 8,808 entrepreneurs in the three state
region during the year of analysis (2008).

3.2.1 | Description of variables

The hypothesized relationships between entrepreneurial income and the variables in this analysis are summarized in
Table 1 and are discussed below.

Dependent Variable (Yi): The dependent variable is entrepreneur i's income earned through self‐employed activ-
ities. Data for this variable is drawn from the self‐reported INCBUS00 variable in the annual ASC survey which
reflects pre‐tax self‐employment income derived from sampled businesses or farms (Ruggles et al., 2008).

Personal and Family Characteristics (Pin): Factors which affect individual and family productivity are hypothe-
sized to affect the self‐employed income level. These variables include demographic characteristics such the entre-
preneur's age, gender, race, educational level, and household type. Binary variables were used to identify gender and
race (black, non‐black), and education (beyond high school = 1). Consistent with literature, we hypothesize that edu-
cation has a positive effect on the self‐employed income, while being black and/or female will have a negative effect
(Table 1). In particular, women may be more likely to combine household work (Hundley, 2001) and possibly other
part‐time salary based‐employment with self‐employment; hence, we would expect the negative value for female.11
Prior research has often used age as a proxy for experience, and hypothesized age to be positively related to self‐
employment income. However, more recent research by Zwan, Thurik, Verheul, and Hessels (2016) found that
opportunity entrepreneurs tend to be younger and more technically skilled and earn higher income than older entre-
preneurs, thus we hypothesize a negative relationship between age and self‐employment income.

Additional variables are used to characterize entrepreneurial and retirement households. An entrepreneurial
household is defined as one in which at least fifty percent of total household income is derived from entrepreneurial
activities. Self‐employed income can be greater than total family income if the entrepreneur's spouse had a net neg-
ative income. In these few instances, self‐employed income is set equal to household income (100%). A retirement
household is one where at least fifty percent of total household income is from retirement funds. We hypothesize
that entrepreneurial households have a stronger commitment to entrepreneurial activities and expect a strong pos-
itive relationship between self‐employed income and entrepreneurial households. Conversely, a strong negative rela-
tionship is expected between self‐employment income and retirement households all else being equal.
In the unconditional quantile analysis, the impact of being black is hypothesized to be negative in each income
quantile, but that the marginal effect decreases at higher income quantiles. Thus, while race based income discrimi-
nation may still exist, its affects are expected to be less impactful for higher self‐employed income earners. Zwan

9
The IPUMS EMPSTATD and IPUMS LABFORCE variables, respectively.
10
The IPUMS CLASSWKRD variable.
11
We would like to thank an anonymous reviewer for this important insight.
WILLIS ET AL. 81

TABLE 1 Hypothesized relationship between log of self‐employed income and independent variables for OLS and
unconditional Quantile regressions

Category of variable Variable OLS sign Quantile Sign1

Personal and family characteristics (Pin) Age − ‐ and decreasing

Gender (male = 1) + + and increasing
Race (black =1, non‐black =0) + + and decreasing
Education (greater than H. S = 1) + + and increasing
Entrepreneurial HH (yes = 1) + + and increasing
Retirement HH (yes = 1) − −
Resource availability (Rik) Mortgage (yes = 1) + + and increasing
Health insurance (yes =1) + + and increasing
Corporation (yes =1) + + and increasing
Industry structure (Sij) Retail (yes = 1) − −
FIRE industry (yes = 1) + +
Farm (yes = 1) − −
Regional economic structure (Aij) Financial services % of PUMA income + +
Per capita income in 2005 + + and increasing
Metro PUMA (yes = 1) + + and increasing
PUMA workforce utilization ? ?
1
Increasing and decreasing refer to the change in the appropriate parameter value as the quantile increases.

et al. (2016) state that opportunity self‐employed attend to be male, younger, and wealthier and as indicated by Block
and Wagner (2010), opportunity self‐employed tend to receive higher remunerations than necessity self‐employed.
(As indicated earlier, women are also more likely to have salaried employment and hold a part‐time self‐employed
job.) Accordingly, we hypothesize that the marginal effect of being male increases at higher income quantiles We fur-
ther hypothesize the relationship between education and self‐employed income is stronger for higher self‐employed
income earners (i.e., the marginal effect is more positive). Our hypothesis is based on the observation by Block and
Wagner (2010) that opportunity entrepreneurs exploit opportunities due to better information flows based on life
experience, social networks, and search processes. Absorptive capacity (Cohen and Levinthal (1990)), intelligence,
and cognitive abilities (Sarasvathy, Simon, and Lave (1998)) are personality traits that facilitate use of information.
We see education as positively correlated with these elements. On the other hand, based on Zwan et al. (2016) that
opportunity entrepreneurs are younger, we expect the relationship between age and self‐employed income to be
negative and stronger (i.e., the marginal effect is more negative) at higher income quantiles. Based on observation
by Zwan et al. (2016) that opportunity motivated self‐employed individuals have a better understanding of available
financial resources and potential markets we expect the entrepreneurial household coefficient to be positive and
increasing across income quantiles. For retirement households, we have no a priori expectation regarding the change
in the value of the coefficient estimate across quantiles except that the coefficient sign remains negative over the
unconditional quantiles.

Resource Availability (Rik): Entrepreneurs with greater access to resources or assets can use these resources as
collateral to gain access to additional resources; as such, these entrepreneurs are expected to earn higher profits
and have a lower rate of business failure. Capital is especially important for entrepreneurial starts‐ups, as investment
by owners is the major source of capital for these firms followed by personal debt (Renski & Wallace, 2012).

Direct measures of resource availability (Rik) would ideally be used to provide information on an entrepreneur's
resources and success in obtaining outside sources of financing. Unfortunately, this data is not available for the
82 WILLIS ET AL.

entrepreneurs in this study so proxy variables are instead developed to control for resource availability. Specifically,
entrepreneurs with access to health insurance, and those who have a mortgage or who own their own home, are
hypothesized to have greater access to resources; these outcomes are, in turn, hypothesized to be positively corre-
lated with self‐employed income. In addition, as owners of incorporated businesses may have enhanced business
skills and access to a greater resource base relative to those who opt to operate under alterative business forms, a
dummy variable is used to identify these entrepreneurs. We hypothesize that entrepreneurs of incorporated busi-
nesses are likely to have higher income than those of non‐incorporated businesses. This hypothesis is consistent with
findings by Shane (2003) who found incorporated businesses start‐ups are generally more successful than sole
proprietorships.
For the unconditional quantile analysis, we hypothesize a stronger positive relationship for the resource availabil-
ity variables (health insurance, mortgage/own‐home, incorporation), at lower levels of self‐employed income distribu-
tion because these resources may be more beneficial for individuals with constrained (as opposed to voluntary) lower
levels of self‐employed income. Parker (2009) suggests that such individuals are likely to be necessity entrepreneurs.
As a corollary, Zwan et al. (2016) state that opportunity self‐employed tend to be wealthier.

Industry Sector (Sij): Entrepreneurs operating in the finance, insurance, and real estate (FIRE) industry sectors
tend to have higher self‐employed income than do entrepreneurs in other sectors of the economy, and those with
retail businesses tend to earn below average self‐employment income (Hurst & Pugsley, 2011). Consistent with this
prior research, we hypothesize that self‐employment in the FIRE sectors will have a positive effect on self‐employed
income and self‐employment in the retail sector will generate below average self‐employment income. Entrepre-
neurs in the farming sector are expected to have below average self‐employment income at all income quantiles
because most farm operations are small in the three‐state region.12
Regional Economic Structure (Aij): Characteristics of a region can also affect entrepreneurial income. Goetz and
Freshwater (2001) and Walzer (2007), found the level of regional self‐employment income is positively correlated
with capital access. Further research by Glaeser and Kohlhase (2004), find that easy capital access is one character-
istic of an agglomerative economy. Because disaggregated within‐state information concerning capital access is not
available, the percentage of wage and salary income earned in the banking and the insurance sector (NAICS 52) rel-
ative to total wage and salary income earned in each PUMA region is used to proxy for capital access.

Agglomeration economies are believed to play a pivotal role in the growth of regional economies (Fujita &
Krugman, 2004; Glaeser & Kohlhase, 2004; Low & Weiler, 2012; Shaffer et al., 2004). An entrepreneur located in
a metropolitan area should benefit from agglomeration economies and have higher income than one in a non‐
metropolitan area. Moreover, as the economic geography literature suggests, agglomeration plays an increasingly
prominent role at higher self‐employment income levels (Fujita & Krugman, 2004), thus we expect the agglomeration
effect will be stronger at higher‐income quantiles (Table 1). In the same vein, higher income levels imply fewer neces-
sity entrepreneurs (Parker, 2009). Thus we expect fewer necessity entrepreneurs in higher income and agglomerative
affected areas.
The relationship between self‐employed income and the concentration of financial services in a PUMA is less
clear over the entire income distribution. One could be tempted to hypothesize that a stronger positive relationship
between the two variables exists at lower quantiles, because increased access to financial services may be more crit-
ical to the ability of low‐income entrepreneurs to increase their income. However, financial services are typically
associated with urbanization (Glaeser & Kohlhase, 2004), and often primarily serve higher income businesses. Thus,
we advance no a priori expectation regarding changes in the magnitude of the expected positive relationship between
the concentration of financial services and self‐employed income at alternative quantiles (Table 1).

12
Half of all U.S. crop farms have less than 45 acres in production (MacDonald, Korb, & Hoppe, 2013), and in our three‐state region, our analysis of Census
of Agriculture data (US Department of Agriculture, 2014) indicates that 53.7% of all farms have under $5,000 in annual sales.
WILLIS ET AL. 83

The PUMA workforce utilization variable is constructed as the percentage of 25 to 65 year‐old individuals
employed relative to the total population in that age range in a PUMA. Because the literature offers competing hypoth-
esis, we offer no hypothesis concerning the relationship between labor market tightness and self‐employed income.
For example, as the workforce utilization rate increases, one might expect the number of necessity entrepreneurs to
decline due to increased opportunity for wage and salary work. Given the well‐established relationship between
increases in necessity entrepreneurs and decreases in average earnings of self‐employed individuals (Koellinger &
Thurik, 2012; Low & Weiler, 2012), one could argue the relationship between workforce utilization and entrepreneurial
income is positive because low‐income entrepreneurs may shift into higher paying salaried opportunities causing aver-
age entrepreneurial income to increase. Alternatively, as high‐income entrepreneurs often hire employees, an increase
in the workforce utilization rate might require entrepreneurs to pay higher wages to compete for workers. In this case, a
negative relationship between workforce utilization and self‐employed income might be expected.
A region's income level may be related to the extent of opportunity and income‐earning potential of entrepre-
neurs in a PUMA region. Given this, per capita PUMA income in 2005 is included as a control variable. We hypoth-
esize a positive relationship exists between self‐employment income and per capita income in the OLS analysis. We
further hypothesize that the marginal impact of PUMA per capita income becomes larger at higher income quantiles
in the UQR model based on research that opportunity entrepreneurs have higher incomes and that opportunities
should be greater with higher local incomes.

3.3 | Estimation approach

This analysis assesses the effects of personal attributes (age, gender, race, education), and household type (primarily
an entrepreneurial household or retirement household) on an entrepreneur's self‐reported income. We, also test the
role of assets (mortgage/own‐home, health insurance, incorporation), type of business (retail, financial, or farm), and
agglomerative effects of the regional economy (concentration of financial services in a PUMA, 2005 PUMA per capita
income, and the importance of a PUMA being located in a metropolitan area). Stata statistical software was used to
estimate all results. Firpo et al. (2009) RIF‐REG UQR program (a Stata add‐in) was used to estimate the UQR results.
The natural log of self‐employed income is the dependent variable in all regressions. Thus, all reported regression
parameters are semi‐elasticities. Briefly, a regression model where the dependent variable is individual self‐employed
income (Yi) consisting of n explanatory variables, estimated in semi‐log form has the functional form. For the semi‐log
form, the marginal impact on the dependent variable for a change in explanatory x variable (say x1), is calculated as
Δln(Y) = β1Δx1. Following Thornton and Innes (1989), the percentage change in predicted self‐employment income
for a one unit change in each explanatory variable is calculated as %ΔY = exp(β1)‐1.
The UQR estimator is implemented using Recentered Influence Functions (RIF) to derive unconditional quantile
estimates (see Firpo et al. (2009) for details). In the RIF‐OLS estimation approach each unconditional quantile is eval-
uated marginally over the distribution of the vector of explanatory variables and is defined independently of the
values of the covariates in the model. Thus, the RIF‐OLS UQR procedure accurately measures the unconditional mar-
ginal effect of each explanatory variable on dependent variable at different quantiles of the unconditional dependent
variable distribution (Park, 2015).13 Residuals of initial OLS estimates exhibited a significantly skewed heteroscedastic
error distribution. Due to this, the natural log of self‐employed income was used to create a homoscedastic error
structure.
In addition to standard testing of the statistical significance of coefficient estimates, useful insights can also be
gained by testing for differences in variable significance between quantiles. A significant difference in a specific
parameter estimate between quantiles would provide evidence of a changing relationship between the dependent
variable (natural log income) and the independent variable across the distribution of the dependent variable. To

13
In quantile analysis, all observations are to estimate coefficient values at each quantile level. Thus in this analysis, all quantiles are estimated with 8,808
observations.
84 WILLIS ET AL.

assess this differences between two quantiles, a confidence interval was constructed for each quantile at a given
probability level. If the two confidence intervals do not overlap the null hypothesis that the parameter value is equal
in both quantiles is rejected at the specified probability level.

4 | RESULTS AND DISCUSSION

Descriptive statistics for the explanatory variables used in the OLS and UQR models are presented in Table 2. The top
and bottom portions of this table respectively summarize the data for the continuous and categorical variables.
Self‐employed income is highly varying, with a mean of $58,015, and ranges from a low of $50 to a high of
$673,000 with a standard deviation of $79,755 for the 8,808 observations. The average age of the sampled self‐
employed is 47.0 with a standard deviation of 9.84. Other continuous variables reflect the nature of the economic
region (the PUMA region) in which the self‐employed individual operates. Average PUMA per capita income is
$24,312 but varies widely across the 148 PUMA regions in the three states. Although not reported in Table 2, pop-
ulation density also varies widely by PUMA region. Average population density is 354 individuals per square mile with
a standard deviation of 408 individuals and reflect the heterogeneity of the environment in which entrepreneurs
operate. The share of financial services in PUMA income ranges from 2.45% to 21.51% with a mean of 6.98% across
all PUMA units.
Self‐employment income varied considerably with the entrepreneur's gender; average self‐employed male income
($69,343) is almost twice the average self‐employed female income ($35,394). Similarly, average self‐employed
income for individuals with a mortgage or who own their own home ($61,379) is nearly twice those who are not
home owners ($35,178). Having health insurance, an incorporated business, more than a high school education,
and being of a race other than black are all associated with higher self‐employed income. Self‐employed individuals
working in FIRE industries have higher average income than those in retail, farming and other industries. These out-
comes are highly consistent with descriptive results found elsewhere in the literature.

4.1 | Regression results

The ordinary least squares (OLS) regression and unconditional quantile regression (UQR) results are presented in
Table 3. The OLS results are consistent with the literature; the impact of all personal and family attributes on self‐
employed income have the expected sign. Being older, male, non‐black, and better‐educated all led to greater levels
of self‐employed income and are statistically significant (p < 0.01). UQR estimates are reported for the 0.10, 0.25,
0.50, 0.75, and 0.90 quantiles. The UQR coefficients for gender, race, and education are generally significant and con-
sistent with the OLS results in that the coefficient for each quantile has the expected positive sign and is statistically
significant. UQR results for age, however, reveal a more varied picture. The age coefficient is either insignificant, or
negative and significant, for all quantiles equal to or below the 0.50 quantile, before turning positive and highly sig-
nificant at all quantiles above the 0.50 quantile. Thus the UQR results do not support our premise that the impact of
age is consistently negative over the entire income distribution.
We hypothesized that the advantages of being non‐black would be positive throughout the income distribution
but monotonically decrease at each subsequent higher quantile level. However, our hypothesis only holds through
the 0.50 quantile. Above the 0.50 quantile, the estimated race coefficient unexpectedly increases at each successive
higher quantile. As subsequently discussed, these results along with others speak to the disadvantage that some
demographic groups may still have at higher income levels.
Table 4 translates our empirical estimates into percentage impacts for all OLS and reported quantile parameters.
The UQR results support our contention that education has an increasingly positive effect on self‐employed income
at higher income quantiles. The marginal effect for the education parameter monotonically increases across the five‐
reported quantile groups. The impact of additional education on entrepreneurial income at the 0.10 quantile is 12.8%
 WILLIS

TABLE 2 Summary statistics for continuous and categorical variables

ET AL.

Self‐employed income

Mean Range % of
Variable Description (Std. Dev.) (min.‐ max.) observations
Self‐employed Incomea Dependent variable $58,015 ($79,755) $50–$673,000
Agea 46.96 (9.84) 25–65
a
Per capita Income Per capita income by PUMA, 2005 $24,312 (7,238) $13,243–$59,919
Financial Servicesa PUMA income derived from financial services (NAICS 52; %) 6.98% (3.82%) 2.45%–21.51%
a
Workforce Utilization Workforce utilization by PUMA (%) 73.04% (5.32%) 55.20%–82.57%
b
Race Black $40,393 10.7%
Non‐black $60,122 89.3%
Genderb Female $35,394 33.4%
Male $69,343 66.6%
Educationb Greater than high school $71,218 59.4%
High school or less $38,696 40.6%
Ent. HHb Self‐employed share of household income >0.5 $78,749 54.6%
b
Retire HH Retirement share of household income >0.5 $15,430 1.2%
b
Industry‐FIRE Employed in FIRE $74,050 9.8%
Industry‐Retailb Employed in retail $46,395 8.0%
b
Industry‐Farm Employed in farming $64,337 2.2%
b
Industry‐Other Employed in other Ind. $57,033 80.0%
Mortgageb Has mortgage/own $61,379 87.2%

Note:
a
N = 148 for PUMA variables (SC = 27, NC = 58, GA = 63).
b
N = 8,808 for individual variables.
85
 86

TABLE 3 OLS and select Quantile estimates and standard errors (0.10, 0.25, 0.50, 0.75, and 0.90)

Quantile regression
Parameter OLS 0.10 0.25 0.50 0.75 0.90

Constant 8.655*** (0.184) 7.284*** (0.545) 8.162*** (0.318) 8.810*** (0.239) 9.296*** (0.223) 9.341*** (0.427)
Age 0.00218** (0.00107) −0.00237 (0.00292) −0.00499** (0.002) 0.000754 (0.00113) 0.00411*** (0.00136) 0.0114*** (0.00257)
Gender (male = 1) 0.372*** (0.0229) 0.554*** (0.0702) 0.486*** (0.0619) 0.299*** (0.0338) 0.221*** (0.0286) 0.351*** (0.0464)
Race (black = 1) −0.146*** (0.0333) −0.145 (0.0918) −0.122** (0.0613) −0.0576* (0.0309) −0.0976*** (0.0373) −0.213*** (0.0602)
Education (greater than H.S. = 1) 0.257*** (0.0217) 0.120* (0.0616) 0.173*** (0.0435) 0.249*** (0.0338) 0.346*** (0.0301) 0.462*** (0.0549)
Entrepreneurial HH (yes = 1) 1.099*** (0.0277) 1.489*** (0.182) 1.266*** (0.114) 0.916*** (0.0819) 0.943*** (0.0487) 1.069*** (0.0701)
Retirement HH (yes = 1) −0.862*** (0.172) −2.474*** (0.748) −1.408*** (0.315) −0.525*** (0.128) −0.277** (0.109) −0.355*** (0.0956)
Mortgage/own (yes = 1) 0.340*** (0.0322) 0.359*** (0.102) 0.399*** (0.0686) 0.243*** (0.0392) 0.278*** (0.0366) 0.391*** (0.0566)
Health insurance (yes = 1) 0.383*** (0.0250) 0.287*** (0.055) 0.412*** (0.0543) 0.383*** (0.0388) 0.431*** (0.0329) 0.449*** (0.0413)
Corporation (yes =1) 0.485*** (0.0204) 0.675*** (0.073) 0.565*** (0.0606) 0.420*** (0.0393) 0.400*** (0.0329) 0.451*** (0.0611)
Ind. ‐ retail (yes = 1) −0.106*** (0.375) −0.111 (0.0907) −0.0168 (0.0583) 0.00029 (0.0386) −0.109** (0.0425) −0.279*** (0.0780)
Ind. ‐ FIRE (yes = 1) 0.123*** (00340) 0.138 (0.0899) 0.110** (0.0533) 0.116*** (0.0316) 0.114*** (0.0441) 0.0345 (0.0826)
Ind. ‐ farm (yes = 1) 0.180** (0.089) 0.358** (0.176) 0.318** (0.131) 0.168* (0.0919) 0.0458 (0.103) 0.141 (0.200)
Financial services % of PUMA Total 0.325 (0.228) 1.072* (0.614) 0.747* (0.399) 0.445* (0.237) −0.331 (0.330) −0.187 (0.581)
PUMA per capita income 2005 (1,000 s) 0.0118*** (0.00213) 0.00325 (0.00486) 0.00503 (0.00344) 0.00443** (0.00213) 0.0142*** (0.00284) 0.0256*** (0.00646)
Metro PUMA (yes = 1) 0.0948*** (0.0277) 0.0839 (0.0825) 0.118** (0.0473) 0.0465* (0.0282) 0.0653* (0.0364) 0.151** (0.0590)
PUMA workforce utilization (%) −0.906*** (0.282) −0.998 (0.703) −0.862* (0.465) −0.521** (0.261) −0.682** (0.342) −1.058* (0.627)

Note. N = 8,808 in all unconditional quantile regressions and the OLS regression.
*( **, ***) denotes significance at the 10 (5, 1) % level.
WILLIS
ET AL.
WILLIS ET AL. 87

TABLE 4 Change (%) and tests of differences in Quantile estimates of self‐employment income

Quantile regression Comparison of Quantile Results1

0.10 Vs. 0.50 Vs. 0.10 Vs.

Variable OLS 0.10 0.25 0.50 0.75 0.90 0.50 0.90 0.90

Age 0.2% −0.2% −0.5% 0.1% 0.4% 1.2% * *** **

Gender (male = 1) 45.1% 74.0% 62.6% 34.8% 24.7% 42.0% *** NS *
Race (black = 1) −13.6% −13.5% −11.5% −5.6% −9.3% −19.2% NS * NS
Education (> than H.S. = 1) 29.3% 12.8% 18.9% 28.3% 41.3% 58.7% NS *** ***
Entrepreneurial HH (yes = 1) 200.1% 343.3% 254.7% 149.9% 156.8% 191.2% *** * *
Retirement HH (yes = 1) −57.8% −91.6% −75.5% −40.8% −24.2% −29.9% *** * ***
Mortgage/own (yes = 1) 40.5% 43.2% 49.0% 27.5% 32.0% 47.8% NS *** NS
Health insurance (yes = 1) 46.7% 33.2% 51.0% 46.7% 53.9% 56.7% NS * *
Corporation (yes =1) 62.4% 96.4% 75.9% 52.2% 49.2% 57.0% *** NS *
Industry ‐ retail (yes = 1) −10.1% −10.5% −1.7% 0.0% −10.3% −24.4% NS * NS
Industry ‐ FIRE (yes = 1) 13.1% 14.8% 11.6% 12.3% 12.1% 3.5% NS NS NS
Industry ‐ farm (yes = 1) 19.7% 43.0% 37.4% 18.3% 4.7% 15.1% NS NS NS
PUMA‐financial service (%) 0.4% 1.9% 1.1% 0.6% −0.3% −0.2% NS NS NS
PUMA PCI 2005 ($1,000 s) 1.2% 0.3% 0.5% 0.4% 1.4% 2.6% NS *** **
Metro PUMA (yes = 1) 9.9% 8.8% 12.5% 4.7% 6.8% 16.3% NS NS *
Workforce utilization (%) −0.6% −0.6% −0.6% −0.4% −0.5% −0.6% NS NS NS

Note. Bolded percentages are not statistically significant different from 0 at a 10% or higher level of significance.
1
***, **, and *denote statistical significance at the 0.001, 0.05, and 0.10 level respectively for a difference in the quantile
parameter estimate at the tested quantile levels. NS = not statistically different between quantiles.

and monotonically increases to 58.7% at the 0.90 quantile. This reflects an increasingly strong interaction between
education and income at higher income quantiles.
In the OLS analysis, the effect of the household type control variables for residing in either an entrepreneurial or a
retirement household had the hypothesized signs (positive and negative, respectively). These results are not surpris-
ing and are consistent with the hypothesized relationships. From the UQR results we find the marginal entrepreneur-
ial household impact on additional entrepreneurial income exhibited a “U‐shaped” pattern over income quantiles. At
quantiles less than or equal to 0.50, the marginal effect monotonically decreases with increases in the quantile before
consistently increasing successively higher quantiles. However, the percentage effect on self‐employment income is
much larger for quantiles below 0.50 than above this level (Table 4). Conversely, as expected, the marginal effect of
residing in a retirement household significantly decreased self‐employment income across all quantiles. Not unex-
pectedly, however, in percentage terms, being in a “retirement household” has a smaller negative impact on self‐
employment earnings at higher than at lower income levels.
Variables that serve as proxies for resource availability (mortgage, health insurance, and corporation status) had
the expected positive sign and were statistically significant (P < 0.01) in both the OLS and UQR analysis (Table 3).
Moreover, the UQR results generally support our hypothesis that resource access increases in importance at higher
quantiles. As reported in Table 4, the impact of having access to health insurance increased almost monotonically
over the five reported quantiles.14 Incorporation status was statistically significant (P < 0.01) for all quantiles. The
impact of business incorporation on entrepreneurial income followed a pattern somewhat similar to gender and race;

14
At the 0.50 quantile, health insurance had a slightly lower percentage impact on self‐employment income relative to the 0.25 quantile coefficient.
88 WILLIS ET AL.

the effects of incorporation decline up to the 0.50 quantile (as expected) before increasing at higher quantiles. The
greatest percentage impact of incorporation was found at quantiles less than or equal to 0.25, suggesting low income
entrepreneurs that incorporate may have superior business management skills. The mortgage marginal effect pattern
over quantiles also exhibits a general U‐shaped pattern. The percentage impact is lowest at the 0.50 quantile
(27.51%) before increasing to 47.85% at the 0.90 quantile. As discussed later, these results along with others speak
to the advantage of resource availability at higher income quantiles.
The OLS results for the industry variables were somewhat consistent with our maintained hypotheses. FIRE was
associated with higher, and retail was associated with lower self‐employed income relative to the aggregated group
of “other” industries. However, counter to our expectation, in the OLS analysis farming is associated with relatively
higher self‐employed income.15 The UQR results, however, offer a much more nuanced look at these results.
Working in the retail sector had a significant negative impact on income only for income quantiles in excess of 0.50
(Table 3). Moreover, the positive impact of working in the FIRE sector of the economy only had a statistically signif-
icant positive income impact on quantiles between 0.25 and 0.75. Finally, the quantile analysis revealed the statisti-
cally significant positive income impact of working in the farming sector is limited to quantiles less than or equal to
the 0.50 quantile. Moreover, the marginal effect of the farming sector on entrepreneurial income monotonically
decreases at sequentially higher quantiles in this quantile range (10% to 50% quantiles).
The results for the regional economic structure parameter estimates (2005 per capita income, metropolitan status,
relative size of the financial sector, and workforce utilization rate) are somewhat mixed in their ability to explain the
variance in self‐employed income. In the OLS analysis, the concentration of financial services on self‐employed
income was insignificant (Table 3). However, the UQR analysis tells a different story. The concentration of financial
services has a significant positive relationship on all quantiles less than or equal to 0.50 (Table 3), but the percentage
impact monotonically decreases over this quantile range (Table 4). This implies greater access to a robust financial
services sector benefits lower‐income entrepreneurs more than higher‐income entrepreneurs and is consistent with
Conroy, Low, and Weiler (2017).
In the OLS analysis, the workforce utilization rate parameter is negative and statistically significant (P < 0.01).
Thus, tighter labor markets seem to reduce self‐employment income. The UQR analysis refines this result; the
estimated UQR parameters are statistically significant only for quantiles at or above the 0.25 quantile. One possible
explanation may be that as labor markets tighten, individuals lacking the skills to compete for higher wage jobs
become lower earning necessity entrepreneurs. As reported in Table 4, for each one percent increase in labor force
utilization, entrepreneurial earnings decrease by about 0.50 percent in the upper income quantiles. Alternatively, the
decrease in entrepreneurial earning may result from high‐income entrepreneurs, who likely have hired employees,
paying their employees higher wages in tight labor markets. The greatest percentage reduction in entrepreneurial
income occurs at the 0.90 quantile.
This analysis provides some support for agglomerative type impacts on self‐employed income. In the OLS results,
both 2005 PUMA per capita income and metropolitan status are significant and positively related to 2008 self‐
employment income (P < .01). More detail is offered through the UQR results which support our contention that
agglomerative type effects are stronger at higher income quantiles. The impact of lagged 2005 per capita income
is significant only at the 0.50 quantile and higher quantiles (P < .01) and the impact of working in a metropolitan
PUMA is significant at the 0.25 and higher quantiles (Table 3). Moreover, as reported in Table 4, the percentage
impacts of these two variables is greatest in the 0.90 quantile.

4.2 | Discussion of results

Overall, the UQR estimates for the marginal effects of personal characteristics (male, black, education, and house-
hold type), access to resources (mortgage/own home and health insurance), and the nature of the regional economy

15
As one reviewer noted, given the nature of the data used in this analysis, perhaps the result of farming should have been expected.
WILLIS ET AL. 89

(workforce utilization, metropolitan status, 2005 per capita income) vary with the quantile level. Focusing on per-
sonal characteristics, the marginal benefit of being older, male, non‐black, and better educated significantly increases
with the income quantile. For example, an additional year of age increases self‐employment income by 1.15% at the
0.90 quantile, versus 0.41% at the 0.75 quantile. Entrepreneurial income of black individuals is 19.18% less than
those of other races at the 0.90 quantile but (only) 9.30% less at the 0.75 quantile. Males earn 42.05% more than
females at the 0.90 quantile, compared to 24.73% more at the 0.75 quantile. Having greater than a high school edu-
cation increases income by 58.72% at the 0.90 quantile, versus 41.34% at the 0.75 quantile. While the results for
education and female are not surprising, the results concerning black entrepreneurs are somewhat contrary to
our expectations; we expected lower income levels for black entrepreneurs but hypothesized the impact would
decrease at higher income quantiles. Likewise, based on Zwan et al. (2016) our hypothesis that age would lead to
relatively lower incomes in the higher quantile levels did not hold. Instead, this analysis offers support for the more
traditional approach that age is a proxy for experience and is positively and increasingly important at higher income
quantiles.
Resource access generally has the greatest impact at the highest income quantile. For example, entrepreneurs
with a mortgage or who own a home had 47.85% more income at the 0.90 quantile, and 27.51% more at the 0.50
quantile. Having health insurance had a nearly monotonically increasing positive impact across quantiles, with the
largest at the 0.90 quantile of 56.67%. These findings are consistent with our hypothesis that having access to these
resources would have a greater percentage impact on high income entrepreneurs.
Several of the regional agglomeration variables also have their highest percentage impact at the 0.90 quantile. A
$1,000 increase in 2005 PUMA per capita income increases 2008 entrepreneurial income by 2.59% at the 0.90
quantile, but by no more than 0.50% at any quantile less than 0.75. Similarly, working in a metro region increases
entrepreneurial income by 16.30% at the 0.90 quantile, by 4.76% at the 0.50 quantile and is insignificant at the
0.10 quantile. Collectively these results imply that regional agglomeration offers relatively more benefits to higher
income entrepreneurs. As an exception to this general rule, the concentration of financial services within a PUMA
is positive and statistically significant only at quantiles less than or equal to the 0.50 quantile, indicating that an
increased availability of financial services primarily benefits lower income entrepreneurs as noted in recent research
by Conroy et al. (2017).

4.3 | Comparing results across quantiles

The three rightmost columns of Table 3 contain significance tests for differences between quantile coefficient esti-
mates at three quantile levels. Due to space limitations, the statistical tests are limited to testing for differences
between the 0.10, 0.50, and 0.90 quantiles.16 Across these quantiles, three coefficients were found to be statistically
different at the 0.10 significance level or higher. The coefficients for age, entrepreneurial household, and retirement
household are significantly different from each other at all three quantiles at the 0.10 or higher level. The coefficient
for age has the lowest statistical significance of a difference between the 0.10 versus the 0.50 quantiles. This outcome
is driven by the 0.10 quantile age coefficient outcome which is not significantly different from zero and has a very high
standard error. However, due to the generally increasing age coefficient over quantiles the age coefficients are statis-
tically different at the 0.95 significance level when compared across the 0.10 versus 0.90 quantile level.17
In addition to the entrepreneurial household and retirement household coefficients being highly significant in
each quantile regression (P < 0.01) with the appropriate sign, there are significant differences in the magnitude of
their respective marginal effect between quantiles. The marginal effect for an entrepreneurial household is signifi-
cantly greater, and for a retirement household is significantly lower on entrepreneurial income at the 0.90 quantile
than at the 0.10 or 0.50 quantiles.

16
These three quantiles were selected because they offer the starkest test of possible differences in coefficient estimates between quantiles.
17
This is also the result of the smaller standard deviation of the age coefficient in the 0.90 than the 0.50 quantile.
90 WILLIS ET AL.

Although all resource availability coefficients (mortgage/own home, health insurance, and corporation status) are
positive and highly significant (P < 0.01) in each quantile regression, these coefficient estimates do not always signif-
icantly differ between quantiles. For example, the marginal effect of having health insurance on entrepreneurial
income is significantly greater at the 0.90 quantile than at either the 0.50 or 0.10 quantiles, but there is no significant
difference between the 0.10 and 0.50 quantiles. Conversely, corporation status has a much greater marginal impact
at low income quantiles (0.10) than higher income (0.50, 0.90) quantiles.
There are also differences in the marginal effect of the agglomeration variables, lagged per capita income and
metropolitan location, over quantiles. The coefficients for these agglomeration variables are significantly more posi-
tive at the 0.90 than the 0.10 quantile (P < 0.05). In summary, the marginal income effect of explanatory variables can
significantly vary between income quantiles and this initial case analysis supports the use of unconditional quantile
analysis in understanding entrepreneurial earnings.

5 | CO NCLUSIO N S

It has been claimed that entrepreneurship growth is one path to building and sustaining urban and rural regional
economies (Edmiston, 2007; Henderson, 2002). Proponents argue that economic development practices that
enhance and support entrepreneurship are essential because they cultivate innovation that, in turn, creates new jobs,
new wealth, and a better quality of life. However, the income of self‐employed workers, as opposed to just the num-
ber of self‐employed, is a critical policy concern because economic development occurs through increases in income
and not just employment. This distinction is important. Identifying and quantifying the personal, cultural, and regional
economic factors that influence self‐employed income provides policy makers with another tool to examine and bet-
ter understand the mechanisms through which entrepreneurship can enhance regional economic development. This
study represents an initial step in the analysis of a comprehensive set of determinants of individual self‐employed
income. Uniquely as well, this study employs an unconditional quantile regression approach to disentangle and sep-
arately examine the relationships between individual socio‐economic factors and regional economic characteristics
across different strata of the self‐employed income distribution.
The OLS results generally met our a priori expectations. Being male, non‐black, better educated, and having access
to resources (mortgage/own‐home, health insurance and incorporation) all led to higher income levels, and entrepre-
neurs living in retirement households had much lower self‐employed income. However, in contrast to our expecta-
tion, age had a highly significant positive impact (P < 0.01), and not negative impact on income. As expected,
entrepreneurs working in the retail sector had lower, and those working in FIRE sectors had higher self‐employment
income relative to those who derived entrepreneurial income from other industries. At the regional level, residing in a
metropolitan area, or a location with higher than average lagged PUMA per capita income both increased self‐
employment income.
The UQR analysis provides insight into how the marginal effect of the explanatory variables vary over self‐
employed income distribution. The UQR results were generally consistent with our hypothesis for education and gen-
der but not for age or race. We correctly hypothesized that the negative marginal effect of being a woman would
increase with the quantile level. However, we incorrectly hypothesized that the positive marginal effect of being
non‐black would decrease at higher quantile levels. Our hypothesis that age would be negative and increasingly more
negative across quantiles is rejected. Instead, age is significant and increasingly more positive across quantiles begin-
ning at the 0.50 quantile. The most profound and, upon reflection, a not surprising outcome of this analysis is the
clear advantage that more education, access to resources and, to a lesser extent, a more prosperous regional econ-
omy contributes to self‐employment income across all income quantiles.
Using the UQR analysis offers additional insights into the impact of explanatory variables not evident through
OLS regression. In the OLS analysis, the importance of the financial service sector in a PUMA was insignificant in
explaining entrepreneurial income. However, in the UQR, the relative presence of the financial service sector was
WILLIS ET AL. 91

found to be significant at or below the 0.50 quantile where it had a significant positive impact on self‐employment
earnings, suggesting that lower income earners benefit from increased access to financial services. Another interest-
ing contrast between the OLS and UQR results is the impact of age on earnings. In the OLS results age is highly sig-
nificant and positively related to income (P < 0.01), but the UQR results suggest a different story. At low quantiles,
the estimated coefficient for age is either insignificant or negative before turning positive and highly significant at
quantiles above 0.50 (P < 0.01). Several circumstances may account for this observation. This might, for example,
reflect that when establishing their businesses, younger (new) entrepreneurs are drawn into a different set of
(more profitable) industries than are older entrepreneurs, or that younger and older entrepreneurs differ in their strat-
egies concerning income withdrawal from their business.18 Also, it may be that older entrepreneurs own more
established businesses from which they can afford to draw a larger income. The agglomeration effect of working
in a metropolitan area also varied across quantiles, and the percentage effect monotonically increased beginning with
the 0.50 quantile.

5.1 | Opportunities for future research

The analytical approach used herein offers unique and nuanced insights into how the determinants of self‐employed
income vary across self‐employed income strata. Implementing this approach to other settings and time periods is
likely to reveal additional important information which could be useful in explaining both past entrepreneur success
(or lack thereof), and offer guidance as to how entrepreneurs of different types can be better supported. Micro‐level
survey data would be particularly useful to more completely quantify the relationship between entrepreneurial access
to resources and self‐employment income. Using such data in analysis that examines the determinants of self‐
employed income, in general or for those in a particular industry, in a longitudinal manner could further reveal how
the relative importance of determinants of entrepreneurial income change with an entrepreneur's experience or
industry environment. Alternatively, an expended cross‐section analysis which considers other time periods would
be useful in gaining insights into the determinants of success for opportunity entrepreneurs who may be more prev-
alent during different phases of the business cycle. In a similar vein, the move to “gig economy” jobs such as driving
rideshares and freelancing (Katz & Krueger, 2017) offers another potentially useful application of this analytical
approach as this form of employment would potentially have a different set of factors affecting entrepreneur income
and success. Future research could also include a better accounting for the motivation for self‐employment, by for
example better determining whether individuals are drawn into self‐employment through opportunities or necessity.
In this regard, Bunten et al. (2015) used business deaths and establishments to assess the impact of entrepreneurs on
regional economies.

5.2 | Policy implications

An array of public policies are designed to support economic growth through generating new jobs and/or increasing
the income of those who are already employed. Among these are programs geared toward current entrepreneurs, or
those considering a switch to this form of employment. General entrepreneur support programs, however, are
unlikely to close the equity gaps that, while previously known, have been further revealed though this analysis.
Likewise, being non‐white is a barrier to income growth in higher income quantiles. Designing programs that promote
minorities as opportunity entrepreneurs could be an appropriate policy response. Results for the UQR demonstrate
the important extent to which an entrepreneur's income is affected in statistically and economically significant ways
by their personal attributes and resources, and characteristics of their industry and region. Attributes and circum-
stances faced by self‐employed workers in high income versus low income quantiles appear to be bundled, allowing
a multi‐dimensional composite of these entrepreneurs to be developed. This information can be used to better tailor

18
This suggestion is consistent with findings by Zwan et al. (2016) and Block and Wagner (2010).
92 WILLIS ET AL.

and target entrepreneur training and support programs, investment opportunity funds, and other advisory services, to
specific demographic groups and/or geographic areas.

ORCID

David B. Willis https://orcid.org/0000-0002-8592-7942

David W. Hughes https://orcid.org/0000-0001-6238-2145

RE FE R ENC ES
Acs, Z. (2007). How is Entrepreneurship Good for Economic Growth? Paper Presented at the 25th Economic Conference of
the Progress Foundation, 2007, October 31. Great Barrington, MA
Aldrich, H. E., & Cliff, J. E. (2003). The pervasive effects of family on entrepreneurship: Toward a family Embeddedness per-
spective. Journal of Business Venturing, 18(5), 573–598. https://doi.org/10.1016/S0883‐9026(03)00011‐9
Astebro, T., & Thompson, P. (2011). Entrepreneurs: Jacks of all trades or hobos? Research Policy, 40(5), 637–649. https://doi.
org/10.1016/j.respol.2011.01.010
Barki, E., Comini, G., Cunliffe, A., Hart, S., & Rai, S. (2015). Social entrepreneurship and social business: Retrospective and
prospective research. Rae‐Revista De Administracao De Empresas, 55(4), 380–385.
Benz, M., & Frey, B. (2008). Being independent is a great thing: Subjective evaluations of self‐employment and hierarchy.
Economica, 75(298), 362–383. https://doi.org/10.1111/j.l468‐0335.2007.00594.X
Blanchflower, D. (2000). Self‐employment in OECD countries. Labour Economics, 7(5), 471–505. https://doi.org/10.3386/
w7486
Blanchflower, D., & Oswald, A. (1998). What makes an entrepreneur? Journal of Labor Economics, 16(1), 26–60.
Blanchflower, D., Oswald, A., & Stutzer, A. (2001). Latent entrepreneurship across nations. European Economic Review,
45(4–6), 680–691. https://doi.org/10.1016/S0014‐2921(01)00137‐4
Block, J., & Koellinger, P. (2009). I Can't get no satisfaction: Necessity entrepreneurship and procedural utility. Kyklos, 62(2),
191–209. https://doi.org/10.1111/j.1467‐6435.2009.00431.x
Block, J. H., & Wagner, M. (2010). Necessity and opportunity entrepreneurs in Germany: Characteristics and earnings differ-
entials. Schmalenbach Business Review, 62(1), 154–174.
Borah, B. J., & Basu, A. (2013). Highlighting differences between conditional and unconditional Quantile regression
approaches through an application to assess medication adherence. Health Economics, 22(9), 1052–1070. https://doi.
org/10.1002/hec.2927
Bowen, H. P., & De Clercq, D. (2008). Institutional context and the allocation of entrepreneurial effort. Journal of International
Business Studies, 39(4), 747–767.
Bruton, G. D., Ketchen, D. J., & Ireland, R. D. (2013). Entrepreneurship as a solution to poverty. Journal of Business Venturing,
28(6), 683–689.
Bunten, D., Weiler, S., Thompson, E., & Zahran, S. (2015). Entrepreneurship, information, and growth. Journal of Regional Sci-
ence, 55(4), 560–584. https://doi.org/10.1111/jors.12157
Camerer, C. F., & Lovallo, D. (1999). Overconfidence and excess entry: An experimental approach. American Economic
Review, 89(1), 306–318.
Chliova, M., Brinckmann, J., & Rosenbusch, N. (2015). Is microcredit a blessing for the poor? A meta‐analysis examining
development outcomes and contextual considerations. Journal of Business Venturing, 30(3), 467–487.
Cohen, W. M., & Levinthal, D. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Sci-
ence Quarterly, 35(1), 128–152.
Colombatto, E., & Melnik, A. (2007). Early work experience and the transition into entrepreneurship. The Journal of Entrepre-
neurship Finance and Business Ventures, 12(1), 9–25.
Conroy, T., Low, S. A., & Weiler, S. (2017). Fueling job engines: Impacts of small business loans on establishment births in
metropolitan and nonmetro counties. Contemporary Economic Policy, 35(3), 578–595.
Edmiston, K. (2007). The role of small and large businesses in economic development. Economic Review, 2nd quarter, Federal
Reserve Bank of Kansas City.
Evans, D. S., & Jovanovic, B. (1989). An estimated model of entrepreneurial choice under liquidity constraints. Journal of
Political Economy, 97(4), 808–827.
WILLIS ET AL. 93

Evans, D. S., & Leighton, L. S. (1989). Some empirical aspects of entrepreneurship. The American Economic Review, 79(3),
519–535.
Evans, M. D. R. (1989). Immigrant entrepreneurship: Effects of ethnic market size and isolated labor Pool. American Sociolog-
ical Review, 54(6), 950–963.
Fairlie, R. W., Kapur, K., & Gates, S. (2011). Is employer‐based health insurances a barrier to entrepreneurship? Journal of
Health Economics, 30(1), 146–162. https://doi.org/10.1016/j.jhealeco.2010.09.003
Fairlies, R. W., & Krashinsky, H. A. (2012). Liquidity constraints, household wealth, and entrepreneurship revisited. The
Review of Income and Wealth, 58(2), 279–306. https://doi.org/10.1111/j.1475‐4991.2011.00491.x
Fayolle, A. (2007). Entrepreneurship and new value creation: The dynamic of the entrepreneurship process. Cambridge: Cam-
bridge University Press. https://doi.org/10.1017/CBO9780511488689
Firpo, S., Fortin, N. M., & Lemieux, T. (2009). Unconditional Quantile Regressions. Econometrica, 77(3), 953–973. https://doi.
org/10.3386/t0339
Fleming, D. A., & Goetz, S. J. (2011). Does local firm ownership matter? Economic Development Quarterly, 25(3), 277–281.
https://doi.org/10.1177/0891242411407312
Formaini, R. L. (2001). The engine of capitalist process: Entrepreneurs in economic theory. Economic and Financial Review 4th
quarter. Federal Reserve Bank of Dallas
Fujita, M., & Krugman, P. (2004). The new economic geography: Past, present and the future. Papers in Regional Science,
83(1), 139–164. https://doi.org/10.1007/s10110‐003‐0180‐0
Gallardo, F., & Scammahorn, R. (2011). Determinants of innovative versus non‐innovative entrepreneurs in three southern
states. The Review of Regional Studies, 41(1), 103–117.
Glaeser, E., Kallal, H., Scheinkman, J., & Shleifer, A. (1992). Growth in cities. Journal of Political Economy, 100(6), 1126–1152.
Glaeser, E. L., & Kerr, W. R. (2010). The secret to job growth: Think small. Harvard Business Review, 88(7‐8), 26.
Glaeser, E. L., & Kohlhase, J. (2004). Cities, regions and the decline of transport costs. Papers in Regional Science, 83(1),
197–228. https://doi.org/10.3386/w9886
Goetz, S. J. (2003). Proprietor employment‐dependent counties. Rural development paper RDP‐17. The northeast regional
Center for Rural Development.
Goetz, S. J., & Freshwater, D. (2001). State‐level determinants of entrepreneurship and a preliminary measure of entrepre-
neurial climate. Economic Development Quarterly, 15(1), 58–70.
Gumus, G., & Regan, T. L. (2015). Self‐employment and the role of health Insurance in the U.S. Journal of Business Venturing,
30(2), 357–374.
Gurley‐Calvez, T., & Hammond, G. W. (2010). Determinants of local self‐employment growth by gender, age, and selected
industry: 2000‐2006. Paper presented at the 49th southern regional science association meeting, 2010, march 26
https://doi.org/10.1080/00343404.2011.653335
Halicioglu, F., & Yolac, S. (2015). Testing the impact of unemployment on self‐employment: Evidence from OECD countries.
Procedia ‐ Social and Behavioral Sciences, 195(1), 10–17.
Hamilton, B. H. (2000). Does entrepreneurship pay? An empirical analysis of the returns to self‐employment. Journal of Polit-
ical Economy, 108(3), 604–631. https://doi.org/10.1086/262131
Heemskerk, M. (2003). Self‐employment and poverty alleviation: Women's work in artisanal gold mines. Human Organization,
62(1), 62–73.
Henderson, J. (2002). Building the rural economy with high‐growth entrepreneurs. Economic Review Third Quarter: 45–70
Hessels, J., Gelderen, M., & Thurik, R. (2008). Entrepreneurial aspiration, motivation and their drivers. Small Business Econom-
ics, 31(3), 323–339. https://doi.org/10.1007/s11187‐008‐9134‐x
Hite, J. C. (1997). The Thunen model and the new economic geography as a paradigm for rural development policy. Applied
Economic Perspectives and Policies, 19(2), 230–240. https://doi.org/10.2307/1349738
Holtz‐Eakin, D., Joulfaian, D., & Rosen, H. S. (1994). Entrepreneurial decisions and liquidity constraints. RAND Journal of Eco-
nomics, 25(2), 334–347. https://doi.org/10.3386/w4526
Holtz‐Eakin, D., Penrod, J. R., & Rosen, H. S. (1996). Health insurance and the supply of entrepreneurs. Journal of Public Eco-
nomics, 62(1), 209–235. https://doi.org/10.3386/w4880
Hundley, G. (2001). Why women earn less than men in self‐employment. Journal of Labor Research, 22(4), 817–829.
Hurst, E., & Lusardi, A. (2004). Liquidity constraints, household wealth, and entrepreneurship. Journal of Political Economy,
112(2), 319–347. https://doi.org/10.1086/381478
94 WILLIS ET AL.

Hurst, E., & Pugsley, B. W. (2011). What do Small Businesses Do? National Bureau of Economic Research, NBER Working
Paper Series, Working Paper 17041. October. Cambridge, MA. https://doi.org/10.3386/w17041
Karnani, A. (2007). The mirage of marketing to the bottom of the pyramid: How the private sector can help alleviate poverty.
California Management Review, 49(4), 90–111.
Katz, L. F., & Krueger, A. B. (2017). The Rise and Nature of Alternative Work Arrangements in the United States, 1995‐2015.
June 18. Access at https://scholar.harvard.edu/files/lkatz/files/katz_krueger_cws_resubmit_clean.pdf
Kauffman Foundation (2019). Kauffman Indicators of Entrepreneurship. https://indicators.kauffman.org/
Koellinger, P., Minniti, M., & Schade, C. (2007). ‘I think I can, I think, I can’: Overconfidence and entrepreneurial behavior.
Journal of Economic Psychology, 28(4), 502–527. https://doi.org/10.1016/j.joep.2006.11.002
Koellinger, P. D., & Thurik, A. R. (2012). Entrepreneurship and the business cycle. The Review of Economics and Statistics,
94(4), 1143–1156.
Koenker, R. (2005). Quantile Regression. Cambridge: Cambridge University Press. ISBN: 0‐521‐60827‐9.
Kusmin, L. D. (2010). Returns to education by metro status, gender, and race/ethnicity: Evidence from the American commu-
nity survey 2006‐2008. Paper presented at the 49th southern regional science association meeting, 2010, march 26.
Laney, K., Bowles, J., & Hillard, T. (2013). Launching Low‐Income Entrepreneurs. Center for an Urban Future. April.
Lazear, E. P. (2004). Balanced skills and entrepreneurship. The American Economic Review, 94(2), 208–211.
Leistritz, L., & Sell, R. S. (2001). Socioeconomic impacts of agricultural processing plants. Community Development: Journal of
the Community Development Society, 32(1), 130–159.
London, T., & Hart, S. L. (2011). Creating a fortune with the base of the pyramid. In T. London, & S. L. Hart (Eds.), Next Gen-
eration Business Strategies for the Base of the Pyramid (pp. 1–18). upper Saddle River, New Jersey: Pearson education Inc.
Low, S. A., Henderson, J., & Weiler, S. (2005). Gauging a Region's Entrepreneurship. Economic Review Federal Reserve Bank
of Kansas City, Kansas City
Low, S. A., & Weiler, S. (2012). Employment risk, returns, and entrepreneurship. Economic Development Quarterly, 26(3),
238–251. https://doi.org/10.1177/0891242412452445
Low, S. H. (2009). Defining and measuring entrepreneurship for regional research: A new approach. Unpublished Ph.D. dis-
sertation, University of Illinois at Urbana‐Champaign. Retrieved from http://hdl.handle.net/2142/14269
MacDonald, J. M., Korb, P., & Hoppe, R. A. (2013). Farm size and the Organization of U.S. crop farming, ERR‐152. U.S.
Department of Agriculture, Economic Research Service, August
Maclean, J. C., Webber, D. A., & Marti, J. (2014). An application of unconditional Quantile regression to cigarette taxes. Jour-
nal of Policy Analysis and Management, 33(1), 188–210. https://doi.org/10.1002/pam.21724
McGranahan, D. A., Wojan, T. R., & Lambert, D. M. (2010). The rural growth trifecta: Outdoor amenities, creative class and
entrepreneurial context. Journal of Economic Geography, 11(3), 529–557. https://doi.org/10.1093/jeg/lbq007
Moskowitz, T. J., & Vissing‐Jorgensen, A. (2002). The returns to entrepreneurial investment: A private equity premium puz-
zle? NBER working paper no. 8876, April https://doi.org/10.3386/w8876
Park, T. (2015). Direct marketing and the structure of farm sales: An unconditional Quantile regression approach. Journal of
Agricultural and Resource Economics, 40(2), 266–284.
Parker, S. C. (2009). The economics of entrepreneurship. Cambridge: Cambridge University Press. https://doi.org/10.1017/
CBO9780511817441
Parker, S. C., & Belghitar, Y. (2006). What happens to nascent entrepreneurs? An econometric analysis of the PSED. Small
Business Economics, 27(1), 81–101. https://doi.org/10.1007/s11187‐006‐9003‐4
Porter, M. E. (2000). Location, competition, and economic development: Local clusters in a global economy. Economic Devel-
opment Quarterly, 14(1), 15–34. https://doi.org/10.1177/089124240001400105
Renski, H., & Wallace, R. (2012). Entrepreneurship in rural America. In S. White, & Z. Kotval (Eds.), Financing Economic Devel-
opment in the 21st Century. New York: Routledge.
Romer, P. M. (1994). The origins of endogenous growth. The Journal of Economic Perspectives, 8(1), 3–22.
Ruggles, S., Sobek, M., Catherine, T. A., Fitch, A., Goeken, R., Hall, P. K., …, Ronnander, C. (2008). Integrated public use
microdata series: Version 4.0. Minneapolis, MN: Minnesota population center https://doi.org/10.18128/D030.V4.0
Sabatini, F. (2008). Social capital and the quality of economic development. Kyklos, 61(3), 466–499. https://doi.org/10.1111/
j.1467‐6435.2008.00413.x
Sarasvathy, D. K., Simon, H., & Lave, L. (1998). Perceiving and managing business risks: Differences between entrepreneurs
and bankers. Journal of Economic Behaviour and Organization, 33(2), 207–225.
WILLIS ET AL. 95

Schuetze, H. J. (2000). Taxes, economic conditions and recent trends in male self‐employment: A Canada‐US comparison.
Labour Economics, 7(5), 507–544.
Shaffer, R., Deller, S. C., & Marcouiller, D. W. (2004). Community Economics: Linking Theory and Practice. Ames, IA: Blackwell.
Shane, S. (2003). A General Theory of Entrepreneurship: The Individual–Opportunity Nexus. Edward Elgar, Cheltenham: New
Horizons in Entrepreneurship.
Shane, S., & Venkataraman, S. (2000). The promise of entrepreneurship as a field of research. Academic Management Review,
25(1), 217–226. https://doi.org/10.2307/259271
Shrestha, S., Goetz, S. J., & Rupasingha, A. (2007). Proprietorship formations and U.S. job growth. The Review of Regional
Studies, 37(2), 146–168.
Spindler, C. J. (1994). Winners and losers in industrial recruitment: Mercedes Benz and Alabama. State and Local Government
Review, 26(3), 192–204.
Stephens, H. M., & Partridge, M. D. (2013). Do entrepreneurs enhance economic growth in lagging regions? Growth and
Change, 42(4), 4310–4465. https://doi.org/10.1111/j.1468‐2257.2011.00563.x
Sutter, C., Bruton, G. D., & Chen, J. (2019). Entrepreneurship as a solution to extreme poverty: A review and future research
directions. Journal of Business Venturing, 34(1), 197–214.
Thornton, R. J., & Innes, J. T. (1989). Interpreting Semilogarithmic regression coefficients in labor research. Journal of Labor
Research, 10(3), 443–447. https://doi.org/10.1007/BF02685335
Todorovic, Z. W. (2004). The framework of static and dynamic components: An examination of entrepreneurship orientation
and university ability to teach entrepreneurship. Journal of Small Business and Entrepreneurship, 17(4), 301–316. https://
doi.org/10.1080/08276331.2004.10593326
U.S. Census Bureau (2018). American Community Survey. “When to Use 1‐year, 3‐year, or 5‐year Estimates.” https://www.
census.gov/programs‐surveys/acs/guidance/estimates.html
US Department of Agriculture (2014). 2012 Census of Agriculture: United States Summary and State Data (Vol. 1). Geographical
Area Series Part 51, AC‐12‐A‐51. Washington, DC: Government Printing Office. May.
US Department of Commerce (2018). Regional Economic Information System, Bureau of Economic Analysis, January.
Van Praag, C. M., & Versloot, P. H. (2007). What is the value of entrepreneurship? A review of recent research. Small Business
Economics, 29(4), 351–382.
Walker, E., & Brown, A. (2004). What success factors are important to small business owners. International Small Business
Journal, 22(6), 577–594. https://doi.org/10.1177/0266242604047411
Walzer, N. (Ed.) (2007). Entrepreneurship and local economic development. Lanham: Lexington Books. https://doi.org/
10.1111/j.1468‐2257.2008.00467.x
Webb, J. W., Bruton, G. D., Tihanyi, L., & Ireland, R. D. (2013). Research on entrepreneurship in the informal economy: Fram-
ing a research agenda. Journal of Business Venturing, 28(5), 598–614.
Wellington, A. J. (2001). Health insurance coverage and entrepreneurship. Contemporary Economic Policy, 19(4), 465–478.
https://doi.org/10.1093/cep/19.4.465
Wolff, J., & Nivorozhkin, A. (2012). Start me up: The effectiveness of a self‐employment Programme for needy unemployed
people in Germany. Journal of Small Business & Entrepreneurship, 25(4), 499–518.
Yenerall, J. (2008). What entrepreneurs mean for your community and how entrepreneurship can be fostered. Rural devel-
opment paper no. 40. The northeast regional Center for Rural Development, Pennsylvania State University.
Zwan, P. V. D., Thurik, R., Verheul, I., & Hessels, J. (2016). Factors influencing the entrepreneurial engagement of opportunity
and necessity entrepreneurs. Eurasian Business Review, 6(2), 273–295.

How to cite this article: Willis DB, Hughes DW, Boys KA, Swindall DC. Economic growth through entrepre-
neurship: Determinants of self‐employed income across regional economies. Pap Reg Sci. 2019;99:73–95.
https://doi.org/10.1111/pirs.12482
DOI: 10.1111/pirs.12482

Resumen. El conocimiento de los determinantes de los ingresos del trabajo por cuenta propia es fundamental para las
estrategias de desarrollo empresarial cuando el objetivo de desarrollo es aumentar los ingresos y no sólo el empleo.
Se utiliza la regresión de cuantiles incondicionales con datos de la Encuesta de la Comunidad Estadounidense para
investigar las diferencias en la relación entre los ingresos empresariales y una serie de características individuales,
industriales y regionales de la distribución de los ingresos del empleo por cuenta propia. Los atributos personales,
como la educación, la raza, la edad y el género, explican las diferencias en los ingresos del trabajo por cuenta propia
y su importancia varía en toda la distribución de los ingresos. Los efectos de aglomeración regionales son
significativamente positivos y más fuertes en el extremo superior de la distribución de los ingresos de los trabajadores
por cuenta propia.

抄録: 成長目標が、雇用だけでなく所得も増加させることである場合、アントレプレナーの成長戦略では、個人
事業主の所得の決定要因に関する知識が非常に重要となる。アメリカン・コミュニティ・サーベイ (American
Community Survey)のデータを用いて、無条件分位点回帰法により、アントレプレナーの所得と、個人事業主
の所得分布全体における個人、産業、地域の特性との関連性における相違を検討する。学歴や人種、年齢などの
個人特性および性別のどちらによっても個人事業主の所得の差を説明することができ、所得分布全体ではこれら
の重要性は様々である。地域における集積効果は、有意にプラスであり、個人事業主の所得分布の上位において
強力である。

-------------------------------------------------------------------------------------------------------
© 2019 The Author(s). Papers in Regional Science © 2019 RSAI