Social Networks, Ethnicity, and Entrepreneurship

William R. Kerr; Martin Mandorff

doi:10.3368/jhr.58.3.0719-10306R2

Article Figures & Data

Figures

Tables
Additional Files

Download figure
Open in new tab
Download powerpoint
Figure 1
Model Depiction of Entrepreneurial Specialization
Notes: Left panel shows production possibilities with specialized occupational distributions. The ray v is the preference parameter over goods in a Leontief utility function. Along the curve with kink V(1, 0), all entrepreneurs belong to group A (belowkink), or all members of group A are entrepreneurs (above). Along the curve with kink V(0, 1), all entrepreneurs belong to group B (below), or all members of group B are entrepreneurs (above). The right panel shows the efficient occupational distribution for different values of v assuming convex productivity in interactions. The minority group A specializes as entrepreneurs if that sector is small enough.
Download figure
Open in new tab
Download powerpoint
Figure 2
Entrepreneurial Specialization by Ethnic Group
Notes: Figure shows the weighted average overage for the entrepreneurial concentration by ethnic immigrant group. The top 20 and bottom 20 values are shown, along with the national average.
Download figure
Open in new tab
Download powerpoint
Figure 3
Nonparametric Estimations of Specialization
Notes: See Table 1. Ethnic groups are divided into equal-sized bins based on group size and social isolation using terciles. Within each isolation triplet, groups are ordered smallest to largest as shown for the most isolated groups. Effects are measured relative to largest and least isolated ethnic groups. Coefficient estimates and 95 percent confidence bands are reported. Full results are provided in the Online Appendix.
Download figure
Open in new tab
Download powerpoint
Figure 4
Industry Distributions of Self-Employment
Notes: Figure shows the cumulative distribution of self-employment for groups moving from the smallest [#1] to largest [#126] industries for self-employment. Immigrant ethnic groups are divided equally into four groups based on being above or below median group size and social isolation.

View popup

Table 1

OLS Estimations of Weighted Average Overage across All Industries

	Baseline Estimation with OVER1 (1)	Using Total Worker Sample (2)	Excluding Natives from Denominator Shares (3)	Imposing Min. Counts on Ethnic Industry Presence (4)	Excluding New Arrivals over the Prior Five Years (5)	Excluding the Taxi Industry (6)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	0.582^***	0.440^***	0.615^***	0.588^***	0.475^***	0.472^***
	(0.076)	(0.074)	(0.078)	(0.081)	(0.074)	(0.088)
Isolation of ethnic group	0.325^***	0.557^***	0.307^***	0.326^***	0.529^***	0.483^***
	(0.076)	(0.090)	(0.085)	(0.086)	(0.085)	(0.095)
Adjusted R-squared value	0.373	0.373	0.385	0.363	0.378	0.337
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	0.465^***	0.325^***	0.460^***	0.432^***	0.370^***	0.416^***
	(0.079)	(0.072)	(0.082)	(0.082)	(0.071)	(0.087)
Isolation of ethnic group	0.447^***	0.674^***	0.428^***	0.491^***	0.672^***	0.561^***
	(0.094)	(0.103)	(0.113)	(0.111)	(0.104)	(0.125)
Adjusted R-squared value	0.455	0.479	0.501	0.494	0.484	0.403

Notes: Estimations describe the OLS relationship between industry concentration for ethnic entrepreneurship and ethnic group size and in-marriage isolation in 2000. The outcome variable is the weighted average overage ratio across industries for each ethnic group, where the weights are levels of self-employment in each industry per group. Variables are winsorized at their 1 percent / 99 percent levels and transformed to have unit standard deviation for interpretation. Regressions include 131 observations, are weighted by log ethnic group counts, and report robust standard errors. Column 2 considers the metric that uses all employed workers for the ethnic group. Column 3 compares industry-level overages only to rates of other immigrant groups. Column 4 drops ethnic–industry settings where fewer than three observation counts exist. Column 5 excludes new arrivals to America during the prior five years, and Column 6 excludes the taxi cab industry. Panel B controls for the traits of the ethnic group in 2000: share who are 36–55 years old, share who are 55–70 years old (reference group is aged 22–35), share who are married, share who speak English well, share who have some college education, and share who have a college degree or higher (reference group is high school or less). Significance:
* p < 0.10,
** p < 0.05,
↵*** p < 0.01.

View popup

Table 2

Robustness Checks on OLS Estimations

	Baseline Estimation (1)	Without Sample Weights (2)	Without Winsorization (3)	Including Fixed Effects for Origin Continent (4)	Using Median Regression Format (5)	Using Boot-Strapped Standard Errors (6)	Including Expected Overage Control (7)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	0.582***	0.612***	0.571***	0.481***	0.322***	0.612***	0.384***
	(0.076)	(0.082)	(0.077)	(0.070)	(0.063)	(0.086)	(0.089)
Isolation of ethnic group	0.325***	0.331***	0.329***	0.279***	0.220***	0.331***	0.334***
	(0.076)	(0.081)	(0.077)	(0.074)	(0.073)	(0.088)	(0.074)
Adjusted R-squared value	0.373	0.368	0.364	0.507	0.198	0.368	0.435
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	0.465***	0.488***	0.460***	0.453***	0.353***	0.488***	0.286***
	(0.079)	(0.085)	(0.079)	(0.079)	(0.080)	(0.092)	(0.091)
Isolation of ethnic group	0.447***	0.441***	0.450***	0.367***	0.390***	0.441***	0.486***
	(0.094)	(0.100)	(0.095)	(0.107)	(0.088)	(0.107)	(0.084)
Adjusted R-squared value	0.455	0.452	0.454	0.529	0.262	0.452	0.517

Notes: See Table 1. Columns 2–6 provide robustness checks on the baseline specification. Regressions in Columns 5 and 6 are unweighted and should be referenced against Column 2. Column 5 reports pseudo R-squared values. Column 7 adds a control for the expected overage level for an ethnicity based on 100 Monte Carlo simulations with the number of observations in the sample.

View popup

Table 3

OLS Estimations of Overage Metric Designs

	Weighted Average Overage across All Industries [OVER1] (1)	Weighted Average Overage Using Three Largest Industries for Ethnic Group [OVER2] (2)	Average of Three Largest Overage Ratios for Ethnic Group [OVER3] (3)	Largest Overage Ratio for Ethnic Group [OVER4] (4)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	0.582***	0.458***	0.578***	0.526***
	(0.076)	(0.076)	(0.068)	(0.077)
Isolation of ethnic group	0.325***	0.423***	0.234***	0.160**
	(0.076)	(0.086)	(0.079)	(0.068)
Adjusted R-squared value	0.373	0.298	0.335	0.280
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	0.465***	0.335***	0.507***	0.466***
	(0.079)	(0.084)	(0.083)	(0.085)
Isolation of ethnic group	0.447***	0.531***	0.342***	0.258***
	(0.094)	(0.100)	(0.089)	(0.099)
Adjusted R-squared value	0.455	0.380	0.364	0.291

Notes: See Table 1 notes. Estimations consider variations in the overage metric design.

View popup

Table 4

OLS Estimations of Panel Changes, 1980–2018

	Baseline Panel Estimation [OVER1] (1)	Including Control for Expected Overage from Simulations (2)	Including Control for Linear Time Trend in 1980 Overage Level (3)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	0.970***	0.568***	0.207**
	(0.264)	(0.186)	(0.084)
Isolation of ethnic group	0.342*	0.242**	0.197**
	(0.190)	(0.119)	(0.093)
Ethnic group and year FE	Yes	Yes	Yes
Adjusted R-squared value	0.484	0.593	0.735
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	0.948***	0.501**	0.258*
	(0.285)	(0.208)	(0.151)
Isolation of ethnic group	0.300*	0.228*	0.165*
	(0.179)	(0.119)	(0.092)
Ethnic group and year FE	Yes	Yes	Yes
Adjusted R-squared value	0.500	0.602	0.742

Notes: See Table 1. Estimations describe the OLS panel relationship between industry concentration for ethnic entrepreneurship and ethnic group size and in-marriage isolation from 1980, 1990, 2000, 2010, and 2018 combining censuses and the American Community Survey. The analysis considers 107 ethnic groups with full panel data, for 535 observations. Regressions are weighted by log ethnic group counts and report standard errors clustered by ethnic group. Column 2 adds a control for the expected overage level for an ethnicity and year based on 100 Monte Carlo simulations with the number of observations in the sample. Column 3 adds a linear time trend in the 1980 overage level.

View popup

Table 5

OLS Estimations with Variations on Ethnic Group Isolation Metric

	Weighted Average Overage across All Industries
	(1)	(2)	(3)	(4)	(5)	(6)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	0.582***	0.718***	0.466***	0.507***	0.488***	0.624***
	(0.076)	(0.103)	(0.070)	(0.086)	(0.084)	(0.118)
Isolation of ethnic group using in-marriage rate in United States	0.325***					0.331***
	(0.076)					(0.109)
Residential segregation in the United States		0.302***				0.030
		(0.066)				(0.112)
Genetic distance (country)			0.153**			0.189**
			(0.071)			(0.081)
Linguistic distance (country)				0.054		−0.081
				(0.047)		(0.075)
Religious distance (country)					0.151**	0.051
					(0.076)	(0.086)
Adjusted A’-squaied value	0.373	0.325	0.288	0.271	0.288	0.415
Observations	131	131	131	113	113	112
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	0.465***	0.597***	0.378***	0.302***	0.303***	0.440***
	(0.079)	(0.103)	(0.077)	(0.084)	(0.086)	(0.104)
Isolation of ethnic group using in-marriage rate in United States	0.447***					0.439***
	(0.094)					(0.125)
Residential segregation in the United States		0.335***				0.063
		(0.078)				(0.110)
Genetic distance (country)			0.103			0.115
			(0.082)			(0.094)
Linguistic distance (country)				0.110*		−0.019
				(0.056)		(0.067)
Religious distance (country)					0.149	0.060
					(0.102)	(0.113)
Adjusted R-squared value	0.455	0.394	0.349	0.392	0.391	0.509
Observations	131	131	131	113	113	112

Notes: See Table 1. Column 1 repeats the baseline estimation with social isolation measured through the 2000 in-marriage rate for the ethnic group. Column 2 uses average residential segregation of ethnic group across Public Use Micro Areas within Metropolitan Statistical Areas that have two or more PUMAs. Columns 3–6 use the average genetic, linguistic, and religious distances from home countries to the weighted ethnic composition of the United States as measured by Spolaore and Wacziarg (2016). Columns 3–6 cluster standard errors by 120 groups from the Spolaore and Wacziarg data.

View popup

Table 6

IV Estimations

	Instrumenting with Predicted Ethnic Group Size from Gravity Model and In-Marriage Rates in the United Kingdom in 1991			Instrumenting with Predicted Ethnic Group Size from Gravity Model and Average of In-Marriage Rates in the United Kingdom in 1991 and Spain in 2011
	First Stage for Size (1)	First Stage for Isolation (2)	Second Stage (3)	First Stage for Size (4)	First Stage for Isolation (5)	Second Stage (6)
Panel A: Baseline Estimation without Controls
Instrument for size	0.648***	−0.118		0.622***	−0.023
	(0.064)	(0.116)		(0.066)	(0.121)
Instrument for isolation	−0.135*	0.540***		−0.142*	0.490***
	(0.076)	(0.097)		(0.075)	(0.089)
F-statistic	52.2	17.4		48.2	18.9
Inverse of log ethnic group size			0.459***			0.469***
			(0.130)			(0.140)
Isolation of ethnic group			0.316**			0.419***
			(0.125)			(0.119)
Exogeneity test p-value			0.140			0.023
Panel B: Including Controls for Ethnic Group’s Traits
Instrument for size	0.503***	−0.081		0.496***	−0.054
	(0.077)	(0.067)		(0.079)	(0.064)
Instrument for isolation	−0.077	0.358***		−0.078	0.313***
	(0.069)	(0.072)		(0.052)	(0.063)
F-statistic	21.7	12.7		23.2	12.8
Inverse of log ethnic group size			0.294**			0.342**
			(0.143)			(0.174)
Isolation of ethnic group			0.484***			0.728***
			(0.153)			(0.209)
Exogeneity test p-value			0.237			0.006

Notes: See Table 1. Estimations describe the IV relationship between industry concentration for ethnic entrepreneurship and ethnic group size and in-marriage isolation. Instruments are the predicted ethnic group size from gravity model and in-marriage rates in U.K. 1991 or an average of the in-marriage rates in U.K. 1991 and Spain 2011. The null hypothesis in Wu-Hausman exogeneity tests is that the instrumented regressors are exogenous. Regressions have 130 and 129 observations, respectively, as U.K. and Spain are excluded when used in the instrument. Regressions cluster standard errors by U.K. 1991 data set ethnic groups.

View popup

Table 7

OLS Estimations of Individual Incomes and Group Concentration

	Log Yearly Income in 2000
	(1)	(2)	(3)	(4)	(5)
Panel A: Self-Employed Individuals
Percentage of individual’s group who are self-employed in the industry	0.069***	0.030***	0.029***	0.024***	0.015*
	(0.009)	(0.009)	(0.010)	(0.008)	(0.009)
Percentage of individual’s group who are self-employed		0.100***	0.101***
		(0.008)	(0.008)
Percentage of individual’s group who are working in the industry			0.002		0.035**
			(0.017)		(0.014)
Person-level traits FE	Yes	Yes	Yes	Yes	Yes
MSA–industry FE	Yes	Yes	Yes	Yes	Yes
Ethnicity FE				Yes	Yes
Adjusted R-squared value	0.244	0.250	0.250	0.260	0.260
Observations	49,026	49,026	49,026	49,026	49,026
Panel B: Wage Workers
Percentage of individual’s group who are self-employed in the industry	0.045***	0.018	0.012	0.002	−0.004
	(0.012)	(0.012)	(0.013)	(0.007)	(0.007)
Percentage of individual’s group who are self-employed		0.059***	0.062***
		(0.007)	(0.007)
Percentage of individual’s group who are working in the industry			0.020		0.021**
			(0.013)		(0.009)
Person-level traits FE	Yes	Yes	Yes	Yes	Yes
MSA–industry FE	Yes	Yes	Yes	Yes	Yes
Ethnicity FE				Yes	Yes
Adjusted R-squared value	0.252	0.254	0.254	0.266	0.266
Observations	355,441	355,441	355,441	355,441	355,441

Notes: Estimations describe the OLS relationship between log yearly income of individuals and entrepreneurial activity of their ethnic group. Sample is taken from 2000 Census IPUMS. Sample includes immigrant males who arrived into the United States at age 16 or later and who are aged 22–70 in 2000. Estimations include fixed effects for the following person-level traits (category counts in parentheses): age (5), age at immigration for migrants (5), education (4), and English language fluency (2). Regressions use person weights and cluster standard errors by ethnic group.

View popup

Table 8

OLS Estimations of Industry Size and Group Size

	Self-Employed Members of Ethnic Group				Wage Workers of Ethnic Group
	Log Avg. Industry Size in Terms of Self-Employed (1)	Log Avg. Industry Size in Terms of Total Workers (2)	Log Size of Largest Industry Measured by Self-Employed Count (3)	Log Size of Largest Industry Measured by Total Worker Count (4)	Log Avg. Industry Size in Terms of Self-Employed (5)	Log Avg. Industry Size in Terms of Total Workers (6)	Log Size of Largest Industry Measured by Self-Employed Count (7)	Log Size of Industry Measured by Total Worker Count (8)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	−0.172***	−0.148***	−0.427***	−0.411***	−0.062	−0.036	−0.311	−0.221**
	(0.058)	(0.050)	(0.129)	(0.134)	(0.048)	(0.027)	(0.212)	(0.092)
Isolation of ethnic group	−0.189***	−0.148***	−0.700***	−0.695***	−0.055	−0.031	0.094	−0.150*
	(0.051)	(0.042)	(0.117)	(0.120)	(0.039)	(0.022)	(0.194)	(0.080)
Adjusted R-squared value	0.115	0.101	0.208	0.199	0.007	0.007	0.011	0.046
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	−0.107**	−0.083**	−0.189*	−0.125	0.026	0.022	−0.171	−0.119
	(0.047)	(0.040)	(0.108)	(0.115)	(0.034)	(0.020)	(0.207)	(0.080)
Isolation of ethnic group	−0.347***	−0.298***	−0.886***	−0.939***	−0.179***	−0.101***	−0.461**	−0.437***
	(0.055)	(0.048)	(0.144)	(0.142)	(0.036)	(0.021)	(0.201)	(0.097)
Adjusted R-squared value	0.578	0.582	0.451	0.443	0.600	0.582	0.437	0.459

Notes: See Table 1 notes.

View popup

Table 9

Sources of Advice in 2016 Annual Survey of Entrepreneurs

Source of advice	Accommodation and Food Services (NAICS 72) (1)	Construction (NAICS 23) (2)	Average for Other NAICS 2-Digit Sectors (3)	Ratio of NAICS 72 to Avg. in Other Sectors (4)
Customers	12.5	8.3	6.5	1.92
Family	15.9	12.9	9.0	1.76
Friends	12.0	10.8	7.1	1.70
Other	8.4	5.3	5.2	1.61
Suppliers	12.4	12.5	10.8	1.15
Government	1.8	2.4	1.6	1.14
Colleagues	42.7	49.8	39.0	1.10
Employees	13.7	12.6	13.6	1.01
Advisors	75.9	79.6	85.2	0.89

Notes: Tabulation of employment-weighted share of businesses reporting source of advice.

View popup

Table 10

OLS Estimations at Metropolitan Level for Size

	Baseline MSA-Level Estimation with Local Size Measure [OVER1] (1)	Including Control for Expected Overage from Simulations (2)	Including Control for Expected Overage from Simulations and Ethnicity Fixed Effects (3)
Panel A: Baseline Estimation without Controls
Inverse of log ethnic group size (small groups have larger values)	0.420***	0.209***	0.115***
	(0.023)	(0.034)	(0.026)
Isolation of ethnic group (national measure)	0.063**	0.066**
	(0.026)	(0.027)
MSA FE	Yes	Yes	Yes
Ethnic group FE			Yes
Adjusted R-squared value	0.175	0.209	0.247
Observations	6,649	6,649	6,649
Panel B: Including Controls for Ethnic Group’s Traits
Inverse of log ethnic group size (small groups have larger values)	0.384***	0.165***	0.115***
	(0.020)	(0.026)	(0.026)
Isolation of ethnic group (national measure)	0.087***	0.095***
	(0.023)	(0.023)
MSA FE	Yes	Yes	Yes
Ethnic group FE			Yes
Adjusted R-squared value	0.193	0.228	0.247
Observations	6,649	6,649	6,649

Notes: See Table 1. Estimations describe the OLS panel relationship between industry concentration for ethnic entrepreneurship and ethnic group size at a metropolitan level. The sample includes metropolitan areas for an ethnicity where self-employment activity is observed. Regressions are weighted by log ethnic group counts in the metropolitan area and report standard errors clustered by ethnic group. Variables are winsorized at their 10 percent / 90 percent values to guard against outliers. Column 2 adds a control for the expected overage level for an ethnicity and metropolitan area based on 100 Monte Carlo simulations with the number of observations in the sample. Column 3 adds a fixed effect for ethnicities.