1. No category

DEPARTMENT OF ECONOMICS WORKING PAPER SERIES
Politics and Entrepreneurial Activity in the U.S.
Louis-Philippe Beland
Louisiana State University
Ozkan Eren
Louisiana State University
Bulent Unel
Louisiana State University
Working Paper 2015-04
http://faculty.bus.lsu.edu/workingpapers/pap15_04.pdf
Department of Economics
Louisiana State University
Baton Rouge, LA 70803-6306
http://www.bus.lsu.edu/economics/
Politics and Entrepreneurial Activity in the U.S.∗
Louis-Philippe Beland, Ozkan Eren, and Bulent Unel
May 20, 2015
Abstract
There is a strong belief that Republicans are more pro-business than Democrats.
In this paper, we investigate the causal impact of partisan allegiance of governors
(Republican or Democratic) on business creation using entrepreneurship data
constructed from the Census Population Surveys and establishment creation
data from the Business Dynamics Statistics. By exploiting random variation
in close gubernatorial elections in 50 states over the last three decades in a
Regression Discontinuity design, we find, surprisingly, that Republicans are not
different than Democrats in business creation. Our findings are robust to a
number of different specifications, samples, and controls.
JEL Classification: L26, J24, D72, C23
Keywords: Entrepreneurship, Political Party, Regression Discontinuity.
∗
Louis-Philippe Beland, Ozkan Eren, and Bulent Unel: Department of Economics, Louisiana State University, Baton Rouge, LA 70803. E-mail: [email protected], [email protected] and [email protected]; Tel: (225)5783787 (Beland), (225)578-3900 (Eren), and (225)578-3790 (Unel).
1
Introduction
Many scholars have argued that entrepreneurship has been a pillar of economic growth by
advancing technology, fostering competition, and creating new jobs. For instance, Foster et
al. (2006) find that all of the labor productivity growth in the U.S. retail sector during the
1990s comes from replacing low productive establishments by relatively more productive
entering establishments. In another study, Haltiwanger et al. (2013) find that firm startups created about 3.5 million net new jobs in the U.S. private sector in 2005. Given its
essential role in the continued dynamism of market economies, policy makers around the
world are interested in implementing policies that foster entrepreneurship.
The importance of entrepreneurial activity, in particular small business entrepreneurship, has been a vastly debated subject in the U.S. politics. The common perception is
that Republicans are more pro-business. In their platform (2012, p.4), Republicans state
“[s]mall businesses are the leaders in the world’s advances in technology and innovation,
and we pledge to strengthen that role and foster small business entrepreneurship.”1 In
addition, they criticize the high corporate tax rate now faced by American business and
call for a reduction of the corporate rate to facilitate American business growth. All of the
current GOP presidential candidates call for tax cuts, especially for businesses.2 Naturally,
one wonders how substantial the impact of Republicans (relative to Democrats) has been
on entrepreneurial activity in the U.S.
In this paper, we investigate the causal impact of party affiliation of U.S. governors on
business creation in the U.S. More specifically, using more than 300 gubernatorial elections
in 50 states over the last three decades, we investigate the impact of Republican governors
on entrepreneurial activity by exploiting random variation associated with close elections in
a regression discontinuity (RD) design. To the extent that the variation in close elections is
1
“We Believe in America: Republican Platform” is a publicly available document at GOP website:
www.gopconvention2012.com.
2
In response to these criticisms, Democrats state “Democrats have continued to champion issues important to small businesses and have worked to expand access to resources that will enable greater success” on
their official website (www.democrats.org/people/small-business-community).
1
random, our estimation strategy yields causal effects of party affiliation on entrepreneurial
activity. As noted by Lee (2008), close elections may be used in a RD design under which
causal inferences can be as credible as those from a randomized experiment.
The identification of entrepreneurial activity at any level is challenging (Fairlie, 2014).
The main problem stems from the distinction between employer and non-employer firms;
popular measures of entrepreneurial activity generally focus on the former. New businesses
with employees, however, comprise only a small fraction of entrepreneurial activity, and
thus may not give us the big picture regarding business formation. To overcome these
challenges, we measure entrepreneurial activity in two different ways. First we measure
it at the individual level, which predominantly reflects small businesses and non-employer
firms. Using the matched data in the Census Population Survey (CPS) files over the period
1979–2012, we identify each individual who becomes self-employed by starting a business
as a new entrepreneur (Hipple, 2010 and Fairlie, 2014). Second, we use establishment
entry density as an alternative measure in an attempt to capture entrepreneurial activity
for businesses with employees (Klapper et al., 2010). Using Business Dynamics Statistics
(BDS) data from the U.S. Census Bureau over the same period, we define establishment
entry density in each state in each year as the number of new establishments created by
new firms normalized by the state’s working age population. The main advantage of the
BDS data is that it provides information about the size of establishments measured by their
number of employees. This allows us to investigate whether the impact of party affiliation
on creation of establishments differs with respect to establishment size.
To the best of our knowledge, this is the first paper analyzing the causal impact of party
affiliation on entrepreneurship. Contrary to common perception, we surprisingly find that
Republicans are not different than Democrats in business creation. Further examination
of entrepreneurial activity by individual characteristics (e.g., blacks vs whites, skilled vs
unskilled), industries, and firm size does not alter our conclusions. Several RD design
validity tests and robustness checks also support our findings.
2
The plan of this paper is as follows. The next section reviews the previous studies and
emphasizes our contribution to the existing literature. Section 3 describes the data, explains
in detail how we measure entrepreneurial activity, and provides summary statistics related
to the main features of the data. Section 4 introduces the empirical strategy by describing
our RD design and presents the main results. Section 5 conducts sensitivity analysis and
Section 6 concludes the paper.
2
Related Literature
This paper is related to a large literature on entrepreneurship. One strand of the literature, dating back to Schumpeter (1950), examines the role of entrepreneurship on economic
growth and development.3 In endogenous growth models, for example, growth is driven
by profit seeking entrepreneurs who develop new products (Romer, 1990; Grossman and
Helpman, 1991; Aghion and Howitt, 1992). Several other studies have identified different
channels through which entrepreneurs affect economic development. For example, Murphy
et al. (1991) show how allocation of entrepreneurial ability determines firm size and economic growth (see also Ghatak et al., 2007), whereas Hausmann and Rodrik (2003) draw
attention to the positive externality created by entrepreneurs through sharing information
on the profitability of new activities.4
Another strand of the literature focuses on the determinants of becoming an entrepreneur.
These factors include financial constraints (Holtz-Eakin et al., 1994; Dunn and Holtz-Eakin,
2000 ; Fairlie and Krashinsky, 2012), taxes (Gordon and Cullen, 2002), family background
and race (Hout and Rosen, 2000; Fairlie, 2008), immigration status (Hunt, 2010), entry
3
Schumpeter (1950) was one of the first economists emphasizing the necessity of entrepreneurs in continued dynamism of market economies. In his classic work, Capitalism, Socialism and Democracy, he states
“the function of entrepreneurs is to reform or revolutionize the pattern of production by exploiting an invention or, more generally, an untried technological possibility for producing a new commodity or producing
an old one in a new way, by opening up a new source of supply of materials or a new outlet for products,
by reorganizing an industry and so on.” (Schumpeter (1950), p.132)
4
The literature on the impact of entrepreneurship on development and growth is vast, and its full rendering
is beyond the scope of this paper. Naude´e (2013) provides a review of the recent studies in this literature.
3
regulations (Dajkov et al., 2002; Branstetter et al., 2014), risk attitudes (Blanchflower,
1998; Skriabikova et al., 2014), and macroeconomic conditions (Fairlie, 2013; Klapper et
al., 2014). Our paper complements these studies by investigating whether politics has been
a significant determinant of entrepreneurial activity in the U.S.
Relatedly, a few other studies look at measures and identification of entrepreneurial
activity (Acs et al., 2008; Klapper et al., 2010; Fairlie, 2014). For example, the Kauffman
Index of Entrepreneurial Activity (KIEA) constructed by Fairlie (2014) is created using
matched data from the CPS to measure business creation at the individual level in the
U.S. covering 1996 to 2013, whereas the World Bank database constructed by Klapper et
al. (2010) presents the number of newly registered firms with limited liability in private
and formal sectors across about 140 countries over the period 2004–2012.5 In measuring
entrepreneurial activity at the individual level, we follow the procedure used by Fairlie
(2014); and at the establishment level, we follow Klapper et al. (2010).
Our paper is also related to a growing literature in political economy that investigates
the impact of elected officials’ policies on economic outcomes. In the U.S., governors have a
high degree of autonomy in exercising their power in policy choices. As such, they prepare
and administer state budget, implement industrial and fiscal policies (e.g., changing tax
rates, minimum wages), sign laws, appoint department heads, recommend legislation, etc,
and several studies have shown that the party affiliation has a significant impact on the
elected politicians’ actions.
Besley and Case (1995 and 2003), using data from the U.S. for the time period 1950-1998,
estimate the impact of the party affiliation of governors on state taxes and expenditures.
They find that Democratic governors are associated with higher state taxes, spending per
capita, and higher minimum wage (see also Reed, 2006 and Leigh, 2008). Employing an RD
design on gubernatorial elections in the U.S., Beland (forthcoming) shows that Democratic
5
The Global Entrepreneurship Monitor is another research program that has collected data on selfemployment (across more than 85 countries since 1999) to measure individuals’ perceptions of entrepreneurship, their involvement in entrepreneurial activity (Acs eta al., 2015).
4
governors increase the annual hours worked by blacks relative to whites; and Beland and
Unel (2015) find that immigrants are more likely to be employed, work longer hours and
more weeks, and have higher earnings under Democratic governors. In a recent paper, using
the network of university classmates among corporate directors and politicians in the U.S.
from 1999 to 2010, Do et al. (2013) find that firms connected to elected governors in the
U.S. saw an average increase in stock-market value by 1.4 percent surrounding the election
date.6
3
Data and Descriptive Statistics
The data used in this paper are drawn from several sources. As mentioned in the introduction, we use self-employed individuals and new establishments as our measures of
entrepreneurial activity. The data on self-employed workers come from the Current Population Survey Outgoing Rotation Group (CPS-ORG) files from Unicon Research Corporation (2015) over the period 1979–2013. The CPS is a monthly household survey where
each household is interviewed for four consecutive months in one year, followed by four
consecutive months one year later (and then leaving the sample permanently); we use the
CPS-ORG data which contain observations in the fourth or eight month in the observation
sample.
The surveys provide detail employment information on individuals, and we record each
individual’s employment status, worker class, industry, weekly hours as well as worker’s
gender, age, race, marital status, and education level. Our sample includes all people
between 20 and 64 years old, but excludes individuals with imputed employment status,
worker class, and weekly hours as in Fairlie (2014).7 Individuals with a missing class of
6
Lee et al. (2004), using a RD design, find that the party affiliation has a large impact on a legislator’s
voting behavior. However, Ferreira and Gyourko (2009) investigate whether cities are as politically polarized
as states. Using close U.S. municipal elections between 1950 and 2000, they find that whether the mayor is
a Democrat or Republican has no impact on the size of local government, the composition of local public
expenditure, or crime rate. Our paper, methodologically, is closely related to their work.
7
To ensure consistency over time, we aggregate 3-digit industries under the following 11 broad indus-
5
worker information are dropped as well. We identify self-employed individuals working at
least 15 hours per week as entrepreneurs, who constitute about 10 percent of our sample
which has more than 7 million observations. There are two groups of self-employed workers:
incorporated (those who work for themselves in corporate entities) and unincorporated
(those who work for themselves in other entities).
We identify new entrepreneurs in year t as those individuals who change their worker
class to “self-employed” from time t to t + 1. It is important to emphasize that the CPS is
a household survey, and does not have individual identifiers. However, uniquely matched
pairs were identified with identical household ID, household number, record lines, survey
month, sex, and race (Card, 1996; Fairlie, 2014). We only consider individuals with age and
schooling difference in two successive years less than two. All unmatched individuals are
dropped from the sample.8 Subsequent to the matching process, our sample includes about
1.7 million observations, about 2.4 percent of them are new entrepreneurs. We choose to
match individuals across years to ensure that those with strong entrepreneurial ability and
interest are considered. In this way, we also eliminate individuals who hold temporary jobs
and identify themselves as self-employed.9
Table 1 reports key statistics about the main characteristics of entrepreneurs (selfemployed workers). For the sake of comparison, we also report the corresponding statistics
for non-entrepreneurs, overwhelming majority of whom are wage and salary workers.10 Column II in Table 1 presents the statistics related to all existing entrepreneurs, whereas the last
tries: Agriculture & Mining, Construction, Manufacturing, Transportation/Utility/Information, Wholesale,
Retail, Finance & Insurance, Professional, Repair, Personal & Entertainment, and Public Administration.
8
The matching rate in the above process across two consecutive years is usually around 60 percent
(consistent with Ziliak et al., 2011). Household IDs assigned in 1985 and 1995 are problematic, and thus
matching rates between 1984 and 1985, 1985 and 1986, 1994 and 1995, and 1995 and 1996 were around 25
percent. As a robustness check, we run regressions excluding these years and the results qualitatively remain
the same.
9
Our construction is slightly different than Fairlie (2014) who matches the data across months in each
year. We choose to match individuals across years for two reasons. First, as mentioned above, we would like
to focus on individuals with strong entrepreneurial ability and interest. Second, using annual data is also
consistent with the data on elections which are reported on a yearly basis.
10
Our sample includes unemployed individuals as well as discouraged workers.
6
Table 1. Summary Statistics on Entrepreneurs and Wage & Salary Workers, 1979–2012
NonEntrepreneurs
I
Incorporated
Existing
Entrepreneurs
II
New
Entrepreneurs
III
0.295
(0.456)
0.298
(0.457)
Female
0.477
(0.499)
0.304
(0.460)
0.364
(0.481)
Age
38.469
(11.878)
43.839
(10.845)
41.123
(10.851)
0.832
(0.373)
0.118
(0.323)
0.905
(0.293)
0.047
(0.212)
0.888
(0.314)
0.058
(0.235)
0.479
(0.499)
0.263
(0.440)
0.169
(0.375)
0.439
(0.496)
0.240
(0.427)
0.189
(0.391)
0.431
(0.495)
0.235
(0.424)
0.204
(0.403)
0.061
(0.240)
0.174
(0.379)
0.726
(0.444)
0.158
(0.365)
0.048
(0.214)
0.687
(0.463)
0.130
(0.336)
0.098
(0.297)
0.720
(0.448)
Race
White
Black
Education
High School or Less
Some College
College
Industry
Construction
Manufacturing
Service
Notes: The data draw on the CPS-MORG Files from Unicon Corporation. Incorporated refers to entrepreneurs who work for themselves in corporate entities. Numbers in parentheses are standard deviations.
7
column reports the statistics for new entrepreneurs that we identify through our matching
process. A comparison of the first two columns reveals that the majority of entrepreneurs
are white, slightly older, and male. Although education compositions of both groups are
similar to each other, they differ in the sectors where they work. For example, about 17
percent of non-entrepreneurs work in the manufacturing sector, whereas the corresponding
statistics for entrepreneurs is less than 5 percent. The last two columns indicate that the
majority of entrepreneurs are unincorporated. Furthermore, these columns show that the
composition of new entrepreneurs in terms of gender, race, education, and industry is very
similar to that of the existing entrepreneurs.
Entrepreneurial activity can also be measured by the new establishment density, defined
as the number of newly registered establishments per 1000 working-age population (Klapper
et al., 2010).11 To this end, we use the Business Dynamic Statistics (BDS) from the U.S.
Census Bureau that provides annual measures of establishment births by initial size, firm
age, and state over the period 1979–2012.12 We only consider the number of establishments
created by new firms, and in each year in each state we record the number of new establishments according to their sizes measured by the number of employees. The advantage
of using these data is that we can investigate whether the impact of party affiliation differs
across different establishment sizes. We categorized establishments into four groups: those
with less than 20 employees, those that have between 20 and 99 employees, those that have
between 100 and 499 employees, and establishments with more than 500 employees. Table
2 presents summary statistics related to the establishment data. It is interesting to note
that over the years the number of big establishments increases, whereas the number of small
ones decreases.
The final data that we use in our analysis are elections results recorded between 1979
and 2012. The data come from two main sources: for elections prior to 1990, we use
11
Working-age population in each state in each year is the total number of people who are in the labor
force and between 20 and 64 years old. The CPS data are used to construct these statistics.
12
The data are publicly available at http://www.census.gov/ces/dataproducts/bds/data firm.html, and
we use the data from row “Firm Age by Initial Firm Size by State.”
8
Table 2. New Establishment Density in the U.S.
Firm Size
Year
1980
1990
2000
2010
1–19
20–99
100–499
500 +
3.361
(0.936)
2.972
(0.520)
2.670
(0.544)
1.814
(0.412)
0.159
(0.048)
0.146
(0.033)
0.142
(0.027)
0.087
(0.016)
0.016
(0.011)
0.017
(0.007)
0.020
(0.006)
0.032
(0.015)
0.002
(0.003)
0.003
(0.002)
0.004
(0.003)
0.005
(0.005)
Notes: The data draw on the Business Dynamic Statistics (BDS) from U.S. Census Bureau. New establishment density is calculated as the number of newly registered establishments per 1000 working-age population.
Numbers in parentheses are standard deviations.
ICPSR 7757 (1995) files; and for election outcomes since 1990, we use the “Atlas of U.S.
Presidential Elections” (Leip, 2015). We drop all elections where a third-party candidate
won. From 1979 to 2012, there are more than 1800 state×year observations, and over this
period Republicans were in power 53 percent of the time. The margin of victory (MRV) is
defined as the proportion of votes cast for the winner minus the proportion of votes cast for
the candidate who finished second. In our analysis, a positive (negative) MRV indicates a
Republican (Democratic) governor won.
As a part of our sensitivity analysis, we also investigate whether the impact of party
affiliation on entrepreneurial activity is different when legislatures and governors are from
the same party. We use data on state legislatures from Klarner (2013).13
4
Empirical Strategy and Main Results
4.1
Econometric Specifications
The identification strategy in this paper relies on the exogenous variation generated by
close gubernatorial elections. As mentioned in the previous section, one way to measure
new entrepreneurial activity is to examine whether an individual becomes a self-employed
13
Data are available at www.indstate.edu/polisci/klarnerpolitics.htm
9
worker. More specifically, let Eist be an indicator variable such that Eist = 1 if individual
i in state s at time t is a self-employed worker, and is 0 otherwise. We then define
+
Eist
=
0 if Eist = 0, Eist+1 = 0
1 if Eist = 0, Eist+1 = 1
(1)
and thus E + is an indicator variable representing a new entrepreneur. With the above
specification, two points are worth emphasizing. First, an individual first interviewed in
year t will be interviewed in year t + 1, but after that the individual will be dropped from
the sample.14 Second, according to (1), all self-employed individuals at time t are excluded
from the sample.
Using (1), we then estimate the following equation
+
0
Eist
= β0 + βR Repst + f (M RVst ) + Xist
δ + αs + τt + ist ,
(2)
where Repst is an indicator variable that takes on the value one if a Republican governor is
in power in state s at time t, zero otherwise. M RVst denotes the margin of victory in the
most recent gubernatorial election prior to time t in state s and is given by the proportion of
votes cast for the winner minus the proportion of votes cast for the candidate who finished
second.15 The functional form between M RVst and entrepreneurial activity is described
by the polynomial f (·). The variable Xist is a vector of observed covariates including each
individual’s gender, race, age, marital status, education, and sector where the individual
works.16 The variables αs and τt capture state and year fixed effects, respectively, and ist
is the error term.
The key identifying assumption underlying equation (2) is that the functional form f (·)
is continuous through the election victory cutoff, i.e., unobserved state characteristics are
14
For example, if an individual in year 2000 is not self-employed, but becomes a self-employed worker in
2001, the individual will be considered a new entrepreneur in 2000.
15
For example, in Louisiana for 2008, 2009, 2010, and 2011 years we used the party of the winner of the
2007 gubernatorial election and the corresponding margin of victory.
16
Our regressions include dummies for sex, marital status, three race dummies (white, black, other), five
education dummies (less than high school, high school, some college, college, and advanced), and age.
10
smooth around the winning cutoff. Under this assumption, the coefficient estimate βR can
be interpreted as the effect of the Republican Party on entrepreneurial activity.
Entrepreneurial activity can also be measured by using the establishment density in
each state. Let EDjst denote the new establishment density for establishments of size j in
state s at time t. We then estimate the following slightly revised version of equation (2)
EDjst = β0 + βR Repst + f (M RVst ) + Sj + αs + τt + uist ,
(3)
where Sj is a categorical variable that controls for different sizes as indicated in Table
2. All other variables are as previously defined, and the coefficient of interest is βR as in
specification (2).17
Before proceeding further, several remarks are worth emphasizing. First, we exclude
state-by-election observations where neither a Democrat nor a Republican won. Second, in
estimating specifications (2) and (3), we use a cubic spline as the functional form between
the outcomes of interest and the margin of victory. Third, we limit our estimation sample
to within 50 percent of the margin of victory. This roughly corresponds to more than 98
percent of each data set, although we provide several sensitivity checks using different orders
of polynomials and varying bandwidths. Finally, to account for possible serial correlation,
standard errors are clustered at the state level.
4.2
Main Results
We begin our analysis with a graphical representation of the effect of the Republican Party
on entrepreneurial activity (business creation). Figure 1.a displays the impact of a Republican governor on entrepreneurial activity at the individual level in close gubernatorial
elections. We plot the unconditional means over a window of 50 percent of margin of victory. Fitted values from a cubic spline are superimposed over these averages. Looking at
17
We also run regressions with Rep with longer time-lags (e.g., Rist−1 or Rist−2 ) since construction of
new establishments may take time. Our results (available upon request) remain qualitatively the same. We
would also like to include additional controls such as sector fixed effects, but such data are not publicly
available.
11
a. Self-employed
b. Establishments
Figure 1: The Impact of Republican Governors on Entrepreneurial Activities
Figure 1.a, we do not observe any visible discontinuity at the cutoff, suggesting no effect
of the Republican Party on entrepreneurial activity (small business creation). When we
switch our attention to new establishment entry density in Figure 1.b, we do not observe
any discontinuity either.
12
Table 3.A. Impact of Party Affiliation on Self-employment
Variable
Rep
All
I
−0.0006
(0.0008)
Construction
II
0.0009
(0.0056)
Manufacturing
III
−0.0000
(0.0011)
Service
IV
−0.0006
(0.0008)
Table 3.B. Impact of Party Affiliation on Self-employment in Different Sub-Groups
Variable
Rep
Whites
I
Blacks
II
−0.0003
(0.0009)
0.0006
(0.0016)
Unskilled
III
−0.0013
(0.0010)
Skilled
IV
0.0000
(0.0011)
Notes: The data draw on the CPS MORG files from Unicon Corporation for 1979–2012. All regressions
include state fixed and time effects, and all other control variables specified in equation (2). Numbers in
parentheses are standard errors based on clustering data at the state level.
We now turn to our regression analysis. Table 3.A presents the RD design estimation
of the Republican Party on entrepreneurial activity at the individual level. Consistent with
Figure 1.a, we do not find any impact of Republican governors: the coefficient estimate
on Rep is insignificant and very small in magnitude (see Column I). We also investigate
potential heterogeneity in the effects of a Republican governor on entrepreneurial activity
at different industries. The results from this exercise for selected industries are reported in
columns II–IV in Table 3.A. Note that they are qualitatively the same as those reported in
Column I.
In Table 3.B, we report the results from RD designs using different subgroups. Columns I
and II present results by race. Investigating the impact of party affiliation on entrepreneurial
activity for different racial groups is important for two reasons. First, recent studies have
shown that the level of entrepreneurial activity differs substantially across racial groups
(Fairlie, 2008). Second, employing a RD design on gubernatorial elections between 1977
and 2008, Beland (forthcoming) shows that Democratic governors increase the annual hours
13
Table 4. Impact of Party Affiliation on New Establishments
Variable
Rep
All
I
[1–19]
II
[20–99]
III
−0.0239
(0.0269)
−0.0650
(0.0989)
−0.0013
(0.0060)
[100–499]
IV
−0.0010
(0.0012)
[500+]
V
−0.0001
(0.0010)
Notes: The data draw on the BDS from U.S. Census for 1977–2012. All regressions include state fixed and
time effects, and all other control variables specified in equation (3). Numbers in parentheses are standard
errors based on clustering data at the state level.
worked by blacks relative to whites. It is therefore possible that Democratic governors might
have a positive impact on the likelihood of being an entrepreneur among blacks. However,
the results presented in Columns I and II show that the impact of Republican governors on
entrepreneurial activity among whites and blacks is insignificant for both races.
Columns III and IV present results based on different skill levels. Individuals with high
school or less education are considered as unskilled, and those with at least some college
education are considered as skilled. According to these columns, there is no skill-based
heterogeneity in the effect of party affiliation.
We have so far concentrated on the entrepreneurial activity at the individual level, which
is likely to reflect small business formation. It is conceivable to argue that the pro-business
effects of the Republican Party are more pronounced at the level of mid- to large-size
business establishment formation. To examine this, we take our analysis one step further
and look at the Republican Party effect on new establishment density. Table 4 report the
results based on equation (3) using different samples. Column I presents the results using
all establishments by controlling differences in establishment size. Columns II–V in Table
4 reports the results under different establishment sizes. The coefficient estimates are not
statistically different from zero irrespective of establishment size.
14
a. Histograms
b. McCrary Density Plots
Figure 1: Distribution of the Margin of Victory
5
Sensitivity Analysis
In this section, we undertake several sensitivity checks to examine the validity of our RD
design estimates. To do so, we closely follow the checklist suggested by Lee and Lemieux
15
Table 5. RD Validity Test: Similarity of States in Close Elections
Variable
Rep
Proportion
of Females
Proportion
of Blacks
Proportion
of Col Grad
Employment
Rate
−0.0036
(0.0032)
−0.0000
(0.0021)
−0.0020
(0.0054)
−0.0039
(0.0060)
Notes: The data draw on the CPS MORG files from Unicon Corporation for 1979–2012. All regressions
include state fixed and time effects. Numbers in parentheses are standard errors based on clustering data at
the state level.
(2014). First, we investigate the validity of the key assumption that candidates should not
have any influence on election outcomes. Figure 2.a plots the distribution of the MRV (in
the form of histograms), and as can be seen from the figure that there are no unusual jumps
around the zero cutoff point. A more formal way to test the validity of the aforementioned
assumption is to use McCrary’s (2008) density test. Figure 2.b plots the density of the MV
based on the McCrary test, and again there are no unusual jumps around the cutoff point.
These approaches suggest that candidates from either party have no influence on election
outcomes.
Second, for our RD designs to be valid, the states where Republicans barely won should
be similar to the states where they barely lost elections. To examine this crucial assumption,
we regress state characteristics (such as proportion of females, blacks, college graduates, and
the employment rate in the year preceding the election) on the dummy variable Repst . If
states are similar, the estimated coefficients on Rep should be statistically insignificant, and
this is indeed what we find in our analysis (see Table 5).18
Third, we run placebo RD designs, using outcomes in the previous term to minimize
concerns on the persistence of election results. Specifically, one particular concern regarding
the identification strategy is that Republican governors might be more likely to be elected,
even in close elections, in state-years with relatively lower business creation. Entrepreneurial
18
Using data on campaign spending by Republicans and Democrats from Jensen and Beyle (2003), we
find that campaign spending by these parties across states with close elections are not statistically different.
Finally, if close elections won by Republican candidates are more likely to win under a Republican House
or Senate, they cannot be considered random. However, the analysis using the data on House and Senate
representatives shows that this is not the case either.
16
Variable
Table 6. RD Validation Test: Pre-Election Outcomes
Self-employed
Establishment
Rep
0.0022
(0.0024)
−0.0317
(0.0422)
Notes: The data draw on the CPS MORG samples from Unicon Corporation (1979–2012) and the BDS
database from U.S. Census (1977–2012). Numbers in parentheses are standard errors based on clustering
data at the state level.
activity could be state-specific, and thus RD designs may yield biased results even in the
presence of state-fixed effects. To further explore this potential state-specific confounding
effect, we use individual-level entrepreneurial activity and new establishment density from
the year prior to gubernatorial elections and run a placebo RD using equations (2) and (3).
The results presented in Table 6 indicate that there are no discontinuities in the outcomes
in the year prior to the election (T − 1).
Fourth, we investigate how sensitive our results are to different specifications for RD
designs. Following Lee and Lemieux’s (2014) suggestion, we present the results from locallinear regression and polynomials with different order and bandwidths. Table 7 reports
the results using individual- and establishment-level data from the CPS-MORG and BDS,
respectively. Column I presents the results from local-linear regressions based on Calonico
et al. (2014).19 Note that none of the estimated coefficients are statistically significant.
Columns II–IV in Table 7 present results using polynomials of different orders. Under
quadratic and cubic spline regressions, we limit our estimation sample to within 20 percent
of the margin of victory, whereas under quartic spline regressions we use the full sample.
The results presented in these columns indicate that party affiliation has no significant
impact on the entrepreneurial activities. We also run separate regressions as in Table 7 by
individual characteristics, industries, and firm size, and the results (available upon request)
qualitatively remain the same.
Another important point is to investigate the impact of party affiliation when legisla19
Using the procedure proposed by Imbens and Kalyanaraman (2012) yields qualitatively similar results.
17
Table 7. RD Designs with Different Polynomial Orders and Bandwiths
Local-Linear
Regression
I
Quadratic
Spline
II
Cubic
Spline
III
Quartic
Spline
IV
A. Self-employed
Rep
−0.0004
(0.0009)
−0.0002
(0.0009)
−0.0009
(0.0011)
−0.0005
(0.0009)
B. Establishment
Rep
−0.0264
(0.0024)
0.0011
(0.0325)
0.0309
(0.0382)
−0.0183
(0.0302)
Variable
Notes: The data draw on the CPS MORG samples from Unicon Corporation (1979–2012) and the BDS
database from U.S. Census (1977–2012). In Columns II and III, we limit our estimation sample to within
50 percent of the margin of victory, whereas in Column IV we use full sample. All regressions include state
fixed effects, time effects, and all other control variables specified in equations (2) and (3). Numbers in
parentheses are standard errors based on clustering data at state level.
tures and governors are from the same party. This is important, because one may argue
that neither Democrat nor Republican governors are likely to make much of a difference
unless they are matched with legislatures that are of the same party. For example, Wisconsin passed right-to-work laws only after the recent election of Republican legislatures
and governors. Consequently, pro-business policies may be more likely to be implemented
when governors and legislatures are of the same party. Table 8 present the results based on
the sample of states where governors and legislatures are of the same party. Note that the
estimated coefficients are still very small and statistically insignificant.
Finally, the Democratic Party has some conservative members whose political views are
similar to their Republican counterparts, and they are generally from the Southern states.
Consequently, we investigate the impact of party affiliation on entrepreneurial activity when
the Southern states are excluded from the sample.20 However, the analysis using this
subsample yields qualitatively the same results (which are available upon request).
In sum, our extensive set of sensitivity analysis using different specifications, samples,
and control variables indicate that the impact of party affiliation of U.S. governors on
20
We exclude Alabama, Arkansas, Kentucky, Louisiana, Mississippi, Oklahoma, Tennessee, and Texas.
18
Table 8. RD Validation Test: Governors and Legislatures from the Same Party
Variable
Self-employed
Rep
0.0007
(0.0018)
Establishment
−0.0154
(0.0371)
Notes: The data draw on the CPS MORG samples from Unicon Corporation (1979–2012) and the BDS
database from U.S. Census (1977–2012). Numbers in parentheses are standard errors based on clustering
data at the state level.
entrepreneurial activity is not significant.
6
Conclusion
Fostering entrepreneurship and supporting business is a contentious issue in American politics. Both parties claim that they have policies that are more conducive for business
creation. Republicans complain about high corporate tax rates imposed on American
firms, whereas Democrats claim that they have fought to remove barriers that stand in
the way of businesses, especially helping small businesses. The common public perception
is that Republicans are more pro-business. Are Republicans really more pro-business than
Democrats?
In this paper, we addressed the above question by investigating the causal impact of
Republican governors on business creation both at the individual and establishment levels.
More specifically, using data on gubernatorial elections across states over the last three
decades, we investigate the impact of Republican governors on business creation by exploiting the random variation in close election in a Regression Discontinuity (RD) design. We
find that creation of new businesses under Republican governors is not different than that
under Democratic governors. Our results are robust to a number of econometric specifications, samples, and control variables.
19
References
Acs, Zoltan J, Sameeksha Desai, and Jolanda Hessels, “Entrepreneurship, Economic
Development and Institutions,” Small Business Economics, 2008, 31, 219–234.
Acs, Zoltan, Szerb Laslo, and Autio Erkko, Global Entrepreneurship and Development
Index, Washington, D.C.: Global Entrepreneurship and Development Institute, 2015.
Aghion, Philippe and Peter Howitt, “A Model of Growth Through Creative Destruction,” Econometrica, 1992, 60, 323–351.
Beland, Louis-Philippe, “Political Parties and Labor Market Outcomes: Evidence from
U.S. States,” American Economic Journal: Applied, forthcoming.
and Bulent Unel, “The Impact of Party Affiliation of U.S. Governors on Immigrants’
Labor-Market Outcomes,” LSU Working Paper, 2015.
Besley, Timothy and Anne Case, “Does Electoral Accountability Affect Economic
Policy Choices? Evidence from Gubernatorial Term Limits,” The Quarterly Journal
of Economics, 1995, 110, 769–798.
and
, “Political Institutions and Policy Choices: Evidence from the United
States,” Journal of Economic Literature, 2003, 41, 7–73.
Blanchflower, David G and Andrew J Oswald, “What Makes an Entrepreneur?,”
Journal of Labor Economics, 1998, 16 (1), 26–60.
Branstetter, Lee, Francisco Lima, Lowell J Taylor, and Ana Venˆ
ancio, “Do Entry
Regulations Deter Entrepreneurship and Job Creation? Evidence from Recent Reforms
in Portugal,” The Economic Journal, 2014, 124 (577), 805–832.
Djankov, Simeon, Rafael La Porta, Florencio Lopez de Silanes, and Andrei
Shleifer, “The Regulation of Entry,” The Quarterly Journal of Economics, 2002, 117,
1–37.
Do, Quoc-Anh, Yen-Teik Lee, and Bang Dang Nguyen, “Political Connections and
Firm Value: Evidence from the Regression Discontinuity Design of Close Gubernatorial
Elections,” LIEPP Working Paper, 2013.
Dunn, Thomas and Douglas Holtz-Eakin, “Financial Capital, Human Capital, and
the Transition to Self-Employment: Evidence from Intergenerational Links,” Journal
of Labor Economics, 2000, 18, 282–305.
Fairlie, Robert, “Entrepreneurship, Economic Conditions, and the Great Recession,”
Journal of Economics and Management Strategy, 2013, 22, 365–89.
and Alicia Robb, Race and Entrepreneurial Success: Black-, Asian-, and WhiteOwned Businesses in the United States, Cambridge, MA: MIT Press, 2008.
and Harry Krashinsky, “Liquidity Constraints, Household Wealth, and Entrepreneurship Revisited,” Review of Income and Wealth, 2012, 58, 279–306.
Fairlie, Robert W, “Kauffman Index of Entrepreneurial Activity 1996-2013,” Technical
Report, Ewing Marion Kauffman Foundation 2014.
Ferreira, Fernando and Joseph Gyourko, “Do Political Parties Matter? Evidence from
U.S. Cities,” The Quarterly Journal of Economics, 2009, 124, 399–422.
20
Foster, Lucia, John Haltwanger, and C. J. Krizan, “Market Selection, Reallocation,
and Restructuring in the U.S. Retail Trade Sector in the 1990s,” Review of Economics
and Statistics, 2006, 88 (4), 748–58.
Ghatak, Maitreesh, Massimo Morelli, and Tomas Sjostrom, “Entrepreneurial Talent, Occupational Choice, and Trickle up Policies,” Journal of Economic Theory, 2007,
137, 27–48.
Gordon, Roger H and Julie Berry Cullen, “Taxes and Entrepreneurial Activity: Theory and Evidence for the US,” Technical Report, National Bureau of Economic Research 2002.
Grossman, Gene M. and Elhanan Helpman, “Quality Ladders in the Theory of
Growth,” The Review of Economic Studies, 1991, 58, 43–61.
Haltiwanger, John, Ron S Jarmin, and Javier Miranda, “Who Creates Jobs? Small
versus Large versus Young,” Review of Economics and Statistics, 2013, 95 (2), 347–361.
Hausmann, Ricardo and Dani Rodrik, “Economic Development as Self-Discovery,”
Journal of Development Economics, 2003, 72, 603–33.
Hipple, Steven F., “Self-Employment in the United States,” Monthly Labor Review, 2010,
133 (9), 17–32.
Holtz-Eakin, Douglas, David Joulfaian, and Harvey S. Rosen, “Sticking it Out:
Entrepreneurial Survival and Liquidity Constraints,” Journal of Political Economy,
1994, 102, 53–75.
Hout, Michael and Harvey Rosen, “Self-Employment, Family Background, and Race,”
Journal of Human Resources, 2000, 35 (4).
Hunt, Jennifer and Marjolaine Gauthier-Loiselle, “How Much Does Immigration
Boost Innovation?,” American Economic Journal: Macroeconomics, 2010, 2, 31–56.
Imbens, Guido and Karthik Kalyanaraman, “Optimal Bandwidth Choice for the
Regression Discontinuity Estimator,” Review of Economic Studies, 2012, 79 (3), 933–
959.
Jensen, Jennifer M and Thad Beyle, “Of footnotes, Missing Data, and Lessons for
50-State Data Collection: The Gubernatorial Campaign Finance Data Project, 1977–
2001,” State Politics & Policy Quarterly, 2003, 3 (2), 203–214.
Klapper, Leora, Inessa Love, and Douglas Randall, “New Firm Registration and the
Business Cycle,” The World Bank Policy Research Working Paper: 6775, 2014.
, Raphael Amit, and Mauro F Guill´
en, “Entrepreneurship and Firm Formation
Across Countries,” in “International Differences in Entrepreneurship,” University of
Chicago Press, 2010, pp. 129–158.
Klarner, Carl, William Berry, Thomas Carsey, Malcolm Jewell, Richard Niemi,
Lynda Powell, and James Snyder, “State Legislative Election Returns (19672010),” ICPSR34297-v1. Ann Arbor, MI: Inter-university Consortium for Political
and Social Research, 2013.
Lee, David S., “Randomized Experiments From Non-Random Selection in U.S. House
elections,” Journal of Econometrics, 2008, 142, 675–697.
21
and Thomas Lemieux, “Regression Discontinuity Designs in Social Sciences,” in
Henning Best and Christof Wolf, eds., The SAGE Handbook of Regression Analysis
and Causal Inference, SAGE Publications, 2014, pp. 301–27.
, Enrico Moretti, and Matthew J. Butler, “Do Voters Affect or Elect Policies?
Evidence from the U. S. House,” The Quarterly Journal of Economics, 2004, 119,
807–859.
Leigh, Andrew, “Estimating the Impact of Gubernatorial Partisanship on Policy Settings
and Economic Outcomes: A Regression Discontinuity Approach,” European Journal
of Political Economy, 2008, 24, 256–268.
Leip, David, Dave Leip’s Atlas of U.S. Presidential Elections, http://uselectionatlas.org,
2015.
McCrary, Justin, “Manipulation of the Running Variable in the Regression Discontinuity
Design: A Density Test,” Journal of Econometrics, February 2008, 142, 698–714.
Murphy, Kevin M, Andrei Shleifer, and Robert W Vishny, “The Allocation of
Talent: Implications for Growth,” Quarterly Journal of Economics, 1991, 106, 503–30.
Naud´
e, Wim, “Entrepreneurship and Economic Development: Theory, Evidence, and
Policy,” IZA Discussion Paper Series, No: 7507, 2013.
Reed, W. Robert, “Democrats, Republicans, and Taxes: Evidence that Political Parties
Matter,” Journal of Public Economics, 2006, 90, 725–50.
Romer, Paul M., “Endogenous Technological Change,” Journal of Political Economy,
1990, 98, 71–102.
Schumpeter, Joseph, Capitalism, Socialism and Democracy, third ed., New York, NY:
Harper and Row, 1950.
Skriabikova, Olga J, Thomas Dohmen, and Ben Kriechel, “New Evidence on the
Relationship between Risk Attitudes and Self-Employment,” Labour Economics, 2014,
30, 176–184.
Unicon Research Corporation, “Outgoing Rotations Monthly Earner Study Data Files
(1979-2014),” Santa Monica, CA, 2015.
Ziliak, James P., Bradley Hardy, and Christopher Bollinger, “Earnings Volatility
in America: Evidence from Matched CPS,” Labour Economics, 2011, 18, 742–754.
22