http://www.smallbusinessinstitute.biz A B S T R A C T Keywords: Journal of Small Business Strategy 2021, Vol. 31, No. 03, 72-87 ISSN: 1081-8510 (Print) 2380-1751 (Online) ©Copyright 2021 Small Business Institute® w w w. j s b s . o rg Introduction 1Purdue University, Department of Agricultural Economics, 403 W. State Street, KRAN 606, West Lafayette, IN 47907, USA, torres2@purdue. edu 2Purdue University, Department of Agricultural Economics, 403 W. State Street, KRAN 617, West Lafayette, IN 47907, USA, delgado2@pur- due.edu 3Purdue University, Department of Agricultural Economics, 403 W. State Street, KRAN 778, West Lafayette, IN 47907, USA, mimarsha@ purdue.edu The economic implications of clustering on Hispanic entrepreneurship in the US Hispanic entrepreneurship, Endogeneity, Generalized propensity score, Intergenerational mobility APA Citation Information: Torres, A., Delgado, M. S., & Marshall, M. I.. (2021). The economic implications of clustering on Hispanic entrepreneurship in the US. Journal of Small Business Strategy, 31(3), 72-87. The Hispanic labor force is on track to increase by 77% over the 2000-2020 period (Suro & Passel, 2003), main- taining its status as the largest US minority (Liu, 2012). A growing Hispanic population is likely to contribute to an increase in the number of Hispanic-owned businesses, which are already outpacing the growth of non-Hispanic businesses, according to data from Census Bureau’s Sur- vey of Business Owners. By 2015, over 4 million Hispanic businesses reported $661 billion in sales (USHCC, 2015); most of these businesses are located in Hispanic clusters, where the access to ethnic clientele, low-cost labor, and key suppliers is crucial for their success. It is unclear if the benefits of living in clusters mo- tivates or deter Hispanics to start a business. On the one hand, there is evidence that Hispanic self-employment is highly influenced by the community (Fairlie, 2004; Fisher & Lewin, 2018). In close proximity or in clusters, Hispan- ics are more likely to know each other and develop socio- economic ties (Kim & Aldrich, 2005). Other researchers have found the relationship between Hispanic clustering is either not correlated or negatively correlated with Hispan- ic self-employment (Liu, 2012; Yuengert, 1995). Here, we address this open question by investigating the impact of Hispanic clustering on Hispanics’ choice to become entre- preneurs. Our secondary goal is to address the role of heteroge- neity on Hispanic entrepreneurship. This article advances the literature by incorporating Hispanic heterogeneity and derives insight on the labor choices for different generations of Hispanics. While Hispanic heterogeneity is, to a large extent, related to the place of birth (Dávila & Mora, 2013), we show that generational differences are also important factors contributing to Hispanic diversity. Generations are measured via categorical variables that combine place of birth and the length of time the individual has been in the US. We expect differences in economic behavior across for- eign- and US-born Hispanics, and across different genera- tions (e.g. first, 1.5, second, and third generation). We develop an identification strategy that allows us to address several potential sources of endogeneity. Specifi- cally, we use a series of robust econometric techniques to control for macroeconomic, individual, and unobserved peer influences that may affect Hispanic self-employment; these are issues that have not yet been addressed in the cur- This article assesses the effect of Hispanic clustering on Hispanic self-employment in the US and the extent to which endogenous social factors within a cluster may encourage Hispanics to start a business. We address key identification issues in the clustering literature by applying a series of robust econometric techniques to US census data. The study provides empirical evidence on the role of Hispanic clustering on Hispanic entrepreneurship. This article also tackles the constructs of Hispanic entrepreneurial heterogeneity and suggests the clustering of second-generation Hispanics as a potential indicator of the Hispanic entrepreneurial environment. The study derives insight on the economic implications of Hispanic clustering and its benefits and suggests policy recommendations to promote success among Hispanic entrepreneurs. We propose that generational differences across Hispanics is not merely an ethnic control variable, but rather an important factor for the design of strategies and incentives at the federal, state, and local level. Ariana Torres1, Michael S. Delgado2, Maria I. Marshall3 http://www.smallbusinessinstitute.biz http://www.jsbs.org 73 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 rent Hispanic entrepreneurship literature. We used an exten- sive list of covariates at the individual and community lev- el control for numerous factors. We then use instrumental variables to control for unobserved macroeconomic factors leading Hispanics to sort into different regions (i.e. Public Use Microdata Areas (PUMAs)) that may also be correlated with the decision to be an entrepreneur. Lastly, we use a generalized propensity score (GPS) approach to adjust for non-randomness in the effect of Hispanic clustering on His- panic self-employment. We contribute to the literature in three ways. First, we develop robust econometric models on the effect of Hispan- ic clustering on Hispanic entrepreneurship, allowing us to push past empirical hurdles such as self-selection into clus- ters and the presence of unobservable factors. Second, we provide empirical evidence that the probability of Hispan- ic entrepreneurship responds to specific levels of Hispan- ic clustering. Lastly, we tackle the constructs of Hispanic heterogeneity and find potential indicators for the Hispanic entrepreneurial environment. Policymakers, scholars, and stakeholders can use our findings to fuel future generations of Hispanic entrepreneurs. If clustering is relevant for His- panic entrepreneurship, public and private programs are more likely to effectively increase the success of Hispan- ic entrepreneurs if targeted through community linkages. Lastly, one-size-fits-all policies disregarding the heteroge- neity among Hispanics will likely have contrasting implica- tions for certain groups. Literature Review Hispanic Self-Employment The US Census Bureau defines self-employed indi- viduals to be those who “operate their own business, pro- fessional practice, farm, or who in any other way regularly work independently to earn a living.” Historically, self-em- ployment is an important means through which immigrants obtain economic mobility in the US (Fairlie & Meyer, 1996), especially for those facing labor market barriers (Shinnar & Young, 2008). Major drivers of self-employment are re- gional factors, age, marital status, human capital, and other individual, family, business, and community characteris- tics (Liu, 2012; Robinson & Sexton, 1994). The literature has reported Hispanics are more likely to start a business than non-Hispanics, especially in retail, services, and con- struction industries (Liu, 2012). However, Hispanic-owned businesses tend to have lower returns than non-Hispanic en- trepreneurs, enter industries with lower barriers, and report additional sources of income other than self-employment (Evans & Leighton, 1989). There is evidence that immigrant self-employment is highly influenced by the socioeconomic environment. Fac- tors influencing the likelihood of self-employment are the characteristics of Hispanic communities; for instance, the level of educational attainment in the community, type of predominant industries, housing prices, population diversi- ty and density, and urban or metro status (Parker, 2004). According to Fairlie (2004b), the decision to be an entre- preneur is positively correlated with the agglomeration of Hispanics. While Wang (2010) suggests that Hispanics are more likely to be self-employed in areas with a high pro- portion of Hispanics because of community resources and opportunities, Liu (2012) does not find that Hispanic-con- centrated areas are correlated with higher rates of Hispanic self-employment. Yuengert (1995) reports no correlation. Hispanic Clustering Hispanics are likely to cluster in areas where other His- panics live and work (Stark, 1991). In clusters, successful Hispanics are observed and copied by others in their pursuit of achieving economic mobility (Danes et al., 2008). Indi- viduals sharing similar characteristics, such as ethnicity, are more likely to know each other and link their socioeconom- ic activities (Kim & Aldrich, 2005). The fact that clusters tend to remain stable over time encourage the creation of entrepreneurial ecosystems (Amit & Muller, 1995; Fisher & Lewin, 2018). The push and pull theory of entrepreneurship (Amit & Muller, 1995) helps us understand that the choice to start a business is a function of the individual’s motivational fac- tors. The framework is useful to accommodate the assump- tion that not one all-encompassing model can explain labor choices of immigrants (Clark & Drinkwater, 2000). Factors such as the demographic context, personal characteristics, and living and working environment shape labor choices (Shapero & Sokol, 1982). This overarching framework in- corporates the opposing pushing and pulling mechanisms driving the choice of self-employment among Hispanics (Cromie, 1987; Fisher & Lewin, 2018). Hispanic clusters can act as a pull factor by motivat- ing Hispanics to start a business. Clusters can provide en- trepreneurs with access to ethnic clientele, low-cost labor, and key suppliers. Hispanic clustering can also be a push factor by forcing Hispanics into self-employment (Reimers, 1983). Economically depressed communities with high un- employment can push Hispanics to start a business as the only way to achieve economic mobility (Kramer Mills et al., 2018). This is especially true among those facing labor discrimination and a lack of educational credentials. Few scholars have studied the mechanisms affecting 74 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 the correlation between Hispanic clustering and the creation of Hispanic-owned businesses. We posit that inconsisten- cies in the literature on the relationship between clustering and entrepreneurship may be due to the lumping of Hispan- ics into a homogenous group, which fails to account for the heterogeneity among Hispanics from different generations (Bradley, 2004). Generational heterogeneity is a key trait likely to affect the economic behavior of Hispanics (Portes & Rumbaut, 2003). While most of the current entrepreneur- ship studies have focused on foreign-born Hispanic entre- preneurs (Yuengert, 1995), the one-size-fits-all findings tend to overlook Hispanic heterogeneity and may lack gen- erality in the results. Method Data and Sample Data comes from the 2010 US Census, which is the latest count of every person living in the US This study takes advantage of the large sample size documented by the census to provide a representative sample of Hispanics who are self-employed. We use the person weights in the census database to make the sample representative of the national population. The sample contains 307,698 Hispanics living in 2,043 Public Use Microdata Areas (PUMAs) across the US. The choice of PUMA as the spatial scale is motivated by data availability, as PUMAs are the smallest geographic unit available in the census. Although PUMAs can include large areas with a low population, such as in rural areas, urban areas may contain one or more PUMAs. Thus, PUMAs are large geographic areas, but they are also small enough to capture the environment in which individuals interact. The sample contains Hispanics between 18 and 70 years old, which were categorized by generation such as first, 1.5, second, and third generation. First generation His- panics are individuals born in Latin America that arrived in the US at any age greater than 16 years. Generation 1.5 consists of Hispanics born in Latin America that arrived in the US at a maximum age of 16 years0F. The countries of origin for 1st and 1.5 generation Hispanics include Puer- to Rico, Cuba, Mexico, Caribe (the Dominican Republic, Haiti, Jamaica and the West Indies), Central America (Be- lize, Costa Rica, El Salvador, Guatemala, Honduras, Nic- aragua and Panama), and South America (Argentina, Bo- livia, Brazil, Chile, Colombia, Ecuador, French Guiana, Guyana, Paraguay, Peru, Suriname, Uruguay and Venezu- ela). Second-generation Hispanics include individuals born in the US with either parent born in Latin America. Lastly, third-generation Hispanics include individuals born in the US, but report Hispanic ethnicity or ancestry. In our sam- ple, 38% are first generation, and the 1.5, second, and third generations make up 14, 7, and 41% of the sample, respec- tively. A Probit Model of the Decision to be an Entrepreneur To understand how living in Hispanic clusters influ- ences the decision for a Hispanic individual to become an entrepreneur, we use a probit regression model Pr(Yi = 1│Xi ) = Φ(β0 + X1i β1 + X2i β2 ), i = 1,2,…,N (1) where the dependent variable if individual self-reported as being employed in his/her own. The Hispanic clustering variables we are interested in are: the share of the popula- tion living in individual ’s PUMA that is Hispanic, repre- senting the degree of Hispanic clustering within the PUMA; the share of the population (of any race or ethnicity) in the PUMA that is self-employed; and the share of the Hispanic population in the PUMA that is self-employed. The vector is a set of individual-level and PUMA-level control vari- ables. The parameter vector is to be estimated, and is the standard normal probability distribution function. This em- pirical model fits into the scope of models in the neighbor- hood effects literature (e.g., Durlauf, 2004; Graham, 2016), of which the central question pertains to the extent to which estimation of the regression model can be plausibly inter- preted as causal. The first Hispanic clustering variable, the share of His- panics living in the PUMA, captures the clustering of all Hispanics. The last two clustering variables, the percent of self-employed individuals in the PUMA and the percent of all Hispanics in the PUMA that are self-employed, measure the extent to which there are endogenous clustering fac- tors within the PUMA that encourage Hispanic individu- als to be self-employed. We anticipate that any increase in self-employment within the PUMA will lead to a significant increase in the probability that a Hispanic individual be- comes self-employed, with a particularly strong effect com- ing from the share of other Hispanics in the PUMA that are self-employed. In additional specifications of the model, we interact these clustering variables with generational indica- tors, to understand the extent to which these clustering ef- fects might be heterogeneous across Hispanic generations. Similarly, Lazear (1999) calculates immigrant clustering by the proportion of individuals in a geographic area. Identification of the social effects comes from several sources. First, we control for an extensive set of individu- al-specific and PUMA-specific factors that affect both the decision to become an entrepreneur and the shares of the population within each PUMA that are Hispanic and/or are 75 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 entrepreneurs (i.e., the variables in ). Our data include an extensive list of covariates measured at the individual and PUMA levels, which allows us to control for numerous fac- tors that might otherwise confound identification. At the in- dividual level, we observe many individual, household, and family background demographic variables, which include gender, age, household income, marital status, the number of children, English and Spanish language proficiency, ed- ucation, access to mortgage, industry, metropolitan status, the education level and employment status of the individu- al’s parents and spouse, and the ethnicity of the individual’s spouse. At higher levels of spatial aggregation, we control for broad geographic differences via eight US territory di- vision indicators from the Bureau of Economic Analysis (BEA): New England, Mideast, Great Lakes, Plains, South- east, Southwest, Rocky Mountain, and Far West.1F Alaska and Hawaii are excluded as the subsample does not report Hispanics within the age group living in these states. See the Supplemental Appendix for more details. At the PUMA lev- el, we control for the share of the population in the PUMA that is white or black (i.e., US-born, non-Hispanic), as well as the violent crime rate and the proportion of the popu- lation that is college-educated, unemployed, and self-em- ployed. In total, these variables control for numerous in- dividual-level and PUMA-level attributes that might drive self-selection into becoming an entrepreneur or the sorting of individuals into different neighborhoods (i.e., PUMAs). Table 1 presents a complete list of control variables included in our analysis. These variables control for the contextual socioeconomic status of each PUMA. Includ- ing other ethnic and racial groups permits us to control for potential labor segregation (Charles, 2003) since it is like- ly that other race and ethnic clusters affect the creation of Hispanic-owned businesses. Further, a high negative cor- relation between percent Hispanic and percent white (-0.73) validates our decision to include the share of other races and ethnicities in the list of explanatory variables. The second source of identification comes from the nonlinearity inherent in the probit specification (Brock & Durlauf, 2001; 2007). According to Manski (1993), there are three types of variables that drive economic activity or, in our case, the decision by a Hispanic individual to be- come an entrepreneur. One set of variables is the endoge- nous social effects – the probability of self-employment in response to Hispanic clustering, self-employment, and His- panic self-employment in the group. The other two sets of variables are the exogenous group effects – the probability of self-employment varies with the socioeconomic compo- sition of the PUMA and the exogenous individual effects – the probability of self-employment varies according to individual attributes, such as language proficiency or edu- cation. Brock and Durlauf (2001; 2007) show that nonlin- earities – which are inherent to the probit specification – are a critical source of identification of endogenous social ef- fects from contextual group characteristics. In other words, the non-identification result in Manski (1993) depends crit- ically on the assumed linearity of the regression function (among other assumptions; see for instance Durlauf, 2004). Hence, while the interplay between these types of variables are complex, the probit structural form aids us in separating these effects from one another. To the extent that our set of controls is not sufficient for eliminating bias caused by omitted variables, we deploy an instrumental variables probit and generalized propensity score approach as auxiliary models. While our set of control variables is extensive, it is possible that unobservable mac- roeconomic factors lead to systemic sorting of Hispanics into different PUMAs that would render the share of His- panics in the PUMA to be endogenous. To guard against this possibility, we deploy the two-stage instrumental variable approach of Card (2009). The advantage of the generalized propensity score model is that the flexible functional form may be more robust if the single-index form of the probit is overly restrictive; in other words, the generalized propensi- ty score approach is a flexible way of adjusting for observ- able differences that might be a source of self-selection of Hispanics into entrepreneurship. We use two instrumental variables designed to ex- tract the exogenous variation in Hispanic clustering (Card 2009). The first instrument is the supply-push component of immigration inflows, capturing the tendency of immi- grants to move to pre-existing clusters and controlling for macroeconomic shocks that increase the attractiveness of a city and leading to immigrant inflows. Specifically, the in- strument measures the expected number of Hispanics going to a PUMA, which is the multiple fraction of all arriving Hispanics who choose to live in a PUMA (e.g. the share of immigrants in a PUMA in an initial period 2000) (λs =Ms/ MUS) and the total number of new Hispanics to the US in 2010 relative to 2000 (ΔMUS). The instrument is multi- plied by the fixed multiple of the fraction of immigrants in the PUMA. In other words, the instrument captures how current Hispanic clustering is a product of historical settle- ment patterns of Hispanics and newly arriving Hispanics in a PUMA. Additional instruments for Hispanic clustering are based on weather (Coates & Gindling, 2010; Miguel et al., 2004). We use heating (and cooling) degree days as mea- sures of climatic comfort; these data were collected from the National Ocean and Atmospheric Administration (NOAA) for each county. Heating degree-days express the frequency in which the temperature falls low enough that heating is 76 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 Table 1 List of dependent and explanatory variables Variable Description selfempa 1= if individual is self-employed, 0 otherwise. Reported that is employed (empstat=1) and works for own enterprise (classwkr=1) platino average percentage share of Hispanic at the PUMA level lincome log of average household income age age in years agetwo square of age in years crime violent crime rate per 100,000 at the state level nchild average number of children in Hispanic household marrieda 1= if individual is married with either present or absent spouse speakenglisha 1= if individual speaks English well, very well, or only English spanisha 1= if individual speaks Spanish in household, 0 otherwise femalea 1= if individual is female collegea 1= if individual has 1 year of college or more metroareaa 1= if individual lives in a metro area accessmorta 1= if individual reports to have a mortgage or contract to purchase collparenta 1= if either mother or father have some college education or graduate studies selfemppara 1= if either mother or father is self-employed, 0 otherwise employspa 1= if spouse is employed sphispanica 1= if spouse is Hispanic collspousea 1= if spouse has some college education or graduate studies newenglanda percentage of Hispanics living in PUMAS located in Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont mideasta percentage of Hispanics living in PUMAS located in Delaware, District of Columbia, Maryland, New Jersey, New York, Pennsylvania greatlakesa percentage of Hispanics living in PUMAS located in Illinois, Indiana, Michigan, Ohio, Wisconsin plainsa percentage of Hispanics living in PUMAS located in Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota southeasta percentage of Hispanics living in PUMAS located in Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Virginia, West Virginia southwesta percentage of Hispanics living in PUMAS located in Arizona, New Mexico, Oklahoma, Texas. Reference group rockyma percentage of Hispanics living in PUMAS located in Colorado, Idaho, Montana, Utah, Wyoming farwesta percentage of Hispanics living in PUMAS located in California, Nevada, Oregon, Washington agmina percentage of Hispanics working in agriculture or mining 77 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 Variable Description construca percentage of Hispanics working in construction manufa percentage of Hispanics working in manufacturing tradea percentage of Hispanics working in trade transporta percentage of Hispanics working in transportation informa percentage of Hispanics working in information financea percentage of Hispanics working in finance profserva percentage of Hispanics working in professional services otherserva percentage of Hispanics working in other services. Reference group pwhite share of US born whites at the PUMA level pblack share of US born African-Americans at the PUMA level pminorit share of other US-born minorities at the PUMA level pimmigrant share of other immigrants at the PUMA level pselfemp share of self-employed individuals at the PUMA level punemployed share of unemployed individuals at the PUMA level pcollege share of individuals with college or higher education at the PUMA level aThe mean value for dummy variables represents the percentage of individuals showing that characteristic. 78 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 required in buildings, and cooling degree-days measure the frequency in which the temperature is high enough that air conditioning is needed in the buildings. In addition to being a flexible way of adjusting for non-randomness in the effect of Hispanic clustering on His- panic self-employment, the generalized propensity score approach also provides us with a continuously varying treatment response: a dose-response function. This allows us to assess the extent to which the impact of Hispanic clus- tering on Hispanic self-employment is nonlinear. Results Descriptive Statistics Approximately 8.5% of Hispanics in our sample are self-employed, compared to 11.1% of white US-born non-Hispanics, 12.6% of non-Hispanic immigrants, and 4.6% of black US-born non-Hispanics. Among Hispanics, first-generation Hispanics are the most entrepreneurial with 11.6% owning a business, followed by the 1.5 generation (8.3%), the third generation (6.7%), and the second-genera- tion (2.8%). The share of Hispanics living in a PUMA rang- es from 0.2% to 96.8% with a mean of 13.5%. The average age for Hispanics in the 2010 census is 39 years old. As expected, first generation Hispanics are the oldest group with an average age of 43 years, followed by the third generation (39 years old), the 1.5 generation (36 years old), and the second-generation (25 years old). On average, 52% of Hispanics are married and have 1 child. While 13% of Hispanics have a college education or high- er, educational attainment varies across Hispanic gener- ations. The third generation has the highest proportion of college-educated Hispanics (17%), which is significantly higher than first-generation Hispanics . On the other hand, only a minority of the respondents’ parents have a college education (2%) or have been self-employed (2%). Approx- imately, 32% of their spouses are employed, 17% have at least a college education, and 39% are Hispanic. Over 80% of Hispanics are English-proficient and a similar proportion speak Spanish at home. When looking at the group of entrepreneurs, our sample suggests that the proportion of Spanish-speakers (84%) is higher than En- glish-proficient (69%) Hispanics. As expected, first-gener- ation Hispanics are significantly more Spanish-fluent than other generations . Conversely, first-generation Hispanics are significantly less English-proficient than other genera- tions . Similar to Parker (2004), over 91% of Hispanics live in metro areas, and this is true for all generations. The pro- portion of Hispanics with mortgages is similar across gen- erations, but the second, third, and 1.5 generation are sig- nificantly different than first-generation Hispanics . There is a higher proportion of Hispanics that live in the Southwest (38%), Far West (28%), and Southeast regions (13%). The vast presence of Hispanics in the south may be explained by the variety of policies and characteristics of these states that motivate Hispanics to live in well-established Hispanic communities (Kochhar et al., 2005). Regression Results This section answers two main questions: how does the probability of self-employment of any Hispanic change as Hispanic clustering increases, and how does Hispanic het- erogeneity influence the probability of Hispanic self-em- ployment. The first question is answered by looking at the parameter platino in Table 2 and the dose-response function in Figure 1. The relationship between Hispanic heteroge- neity and Hispanic entrepreneurship is answered through Tables 3, 4, and 5 and in Figure 2 and 3. The probability to start a business for each generation of Hispanics as His- panic clustering increases is shown by the coefficient of the interaction terms in Table 2 and the graph of the marginal effects in Figure 2. Table 3 shows the probability of self-em- ployment for each generation, regardless of platino, and Ta- ble 4 shows how the clustering of each generation affects Hispanic entrepreneurship. How Does the Probability of Self-employment of any Hispanic Change as Hispanic Clustering Increases? Table 2 reports the coefficients and marginal effects from the standard probit and the instrumental variables probit. The left panel of Figure 1 shows the dose-response function from the GPS model, displaying how the average probability of self-employment varies depending on the (continuous) level of Hispanic clustering. The right panel of Figure 1 shows the treatment effect function, which is the derivative of the dose-response function with respect to the level of Hispanic clustering. We find that usual standard errors and robust standard errors are similar, which suggests that heteroskedasticity is not an issue. The results togeth- er provide robust empirical evidence that, on average, the probability of self-employment for Hispanics decreases as the share of Hispanics increases at the PUMA level, but this relationship is not necessarily linear. The standard probit suggests that the average Hispanic is significantly less likely to start a business as the share of all Hispanics clustering in a PUMA increases. Contrary to Borjas (1986) and Wang (2010), this article provides empir- ical evidence that the clustering of Hispanics discourages 79 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 the entrepreneurial motivation of Hispanics. One explana- tion is that Hispanic-dominated neighborhoods may suffer from limited resources possibly due to residential segre- gation. It is likely that Hispanic-dominated PUMAs may concentrate limited consumer demand, which can constrain the creation of Hispanic businesses. Further analysis indi- cates that employment, household income, and educational attainment proportionally decreases as the share of Hispan- ics increases. By concentrating Hispanics geographically, the decrease of Hispanic household income appears to de- teriorate the entrepreneurial environment. It is likely that in PUMAs in which household income is below-average, there will be limited demand and an unfavorable business atmosphere. This finding supports the theory that Hispan- ic clustering can act as a mobility trap and harm Hispanic entrepreneurship (Borjas, 1983; Fischer & Massey, 2000; Reimers, 1983). The standard probit regression further shows that oth- er factors are detrimental to Hispanic entrepreneurship, including financial factors, household characteristics, in- dustry, and gender. Study results illustrate that Hispanics are less likely to be an entrepreneur as household income increases . In our sample, higher household income may be associated with broader labor market prospects among Hispanics. Thus, Hispanics with higher household income are less likely to be self-employed. Further factors deterring the probability that Hispanics enter self-employment are ac- cess to a mortgage and Hispanic ethnicity of the spouse . Hispanics working in agriculture, manufacturing, trade, in- formation, or service industries are less likely to be self-em- ployed. PUMAs with a high concentration of white US- born and non-Hispanic immigrants are environments that are detrimental to Hispanic entrepreneurial endeavors. Re- sults provide evidence of a potential residential segregation and its negative effect on Hispanic entrepreneurial activity. Lastly, Hispanic women are less likely to be self-employed than men . This finding is supported by the literature that describes women’s lower entrepreneurial activity due to lower human capital accumulation, motherhood penalty, and lower work-force participation rates (Fairchild, 2010; Marshall & Flaig, 2014). The more human capital at the entrepreneur’s disposal, the greater the odds of self-em- ployment among Hispanics. Hispanics are more likely to be self-employed as the number of children increases, his/ her parents are self-employed, and the spouse is employed or has attended college. This may be due to the fact that age is associated with higher levels of human capital and access to financial capital that can improve the odds to start a busi- ness (Fairchild, 2010). Hispanics living in metro areas, the Southeast, or in PUMAs with a high concentration of self-employed indi- viduals are more likely to start their own business. Well-es- tablished Hispanic communities in the south of the US may bring adequate resources, such as clientele, that motivate Hispanic entrepreneurship. Consistent with the literature, areas with vibrant entrepreneurial ecosystems can offer the clientele and resources that motivate Hispanics into self-employment (Liu, 2012; Wang, 2010). The instrumental variable probit uses the Card (2009) and Coates and Gindling (2010) instruments to control for endogeneity from macroeconomic shocks that may induce Hispanic agglomeration. A key finding is that in the instru- mental variable probit, the parameter is not statistically sig- nificant . The parameter represents the correlation between the errors in the standard probit and the reduced-form equa- tion for the endogenous regressor. A that is not statistically significant is equivalent to saying that platino is unlikely to be endogenous. In other words, endogeneity is unlikely, and the results from the standard probit can be used to disentan- gle the effects of Hispanic clustering on the probability of self-employment. A reason why endogeneity is not an issue may be the extensive list of covariates included in the right- hand side in the standard probit. The results from the in- strumental variables probit are consistent with the standard probit, though the parameter estimates in the instrumental variables probit are somewhat larger in magnitude. The GPS estimates of the dose-response function are shown in Figure 1; generally, we find that Hispanic entre- preneurial activity decreases as the share of Hispanics in a PUMA increases. Thus, the GPS method confirms the find- ings from the probit model. An important contribution from the dose-response and treatment effect functions is how Hispanic self-employment responds to a specific level of Hispanic clustering. Figure 1 points to three regions where the direction of the response of Hispanic entrepreneurship changes with respect to Hispanic clustering. In region 1 (< 20% share of Hispanics) and region 3 (>80% share of His- panics), Hispanic clustering has a positive or zero effect on the entrepreneurial activity of any Hispanic. A low and high concentration of Hispanics in the PUMA increases the prob- ability of self-employment. Alternatively, Hispanics are less likely to start a business if living in PUMAs where the share of Hispanics is between 20% and 80% (region 2). Further analysis shows that Hispanic-dominated PU- MAs are mainly composed of foreign-born Hispanics with lower household income and educational credentials. For instance, Hispanic households living in low-concentrated PUMAs made on average $69,283 while medium-concen- trated and Hispanic-dominated PUMAs made $54,346 and $54,754 in 2010, respectively. Moreover, the higher the Hispanic clustering, the bigger the income gap between foreign-born and US-born generations. Hispanic-dominat- 80 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 Table 2 Standard probit and IV probit results on the probability of Hispanic self-employment Probit IV probit Coeff. Marginal Eff.a Coeff. Marginal Eff. platino -0.006 ** -0.074 ** -0.014 * -1.353 * lincome -0.185 *** -2.450 *** -0.185 *** -18.497 *** age 0.063 *** 0.837 *** 0.063 *** 6.340 *** agetwo -0.001 *** - b -0.001 *** - b crime -0.001 -0.001 0.000 -0.009 nchild 0.046 *** 0.609 *** 0.046 *** 4.593 *** married 0.053 ** 0.699 ** 0.052 ** 5.241 ** speakenglish -0.057 ** -0.754 ** -0.057 ** -5.736 ** spanish 0.054 *** 0.717 *** 0.056 *** 5.631 *** female -0.206 *** -2.726 *** -0.207 *** -20.657 *** college 0.033 0.442 0.034 3.378 metroarea 0.131 *** 1.733 *** 0.143 *** 14.295 *** accessmort -0.037 ** -0.493 ** -0.038 ** -3.774 ** collparent -0.011 -0.146 -0.011 -1.132 selfemppar 0.487 *** 6.430 *** 0.488 *** 48.783 *** employsp 0.060 *** 0.796 ** 0.060 ** 5.997 ** sphispanic -0.068 *** -0.900 *** -0.067 *** -6.686 *** collspouse 0.085 *** 1.125 *** 0.085 *** 8.490 *** newengland -0.023 -0.298 -0.009 -0.869 mideast 0.015 0.204 0.032 3.205 greatlakes -0.054 -0.717 -0.042 -4.233 plains -0.038 -0.507 -0.033 -3.260 southeast 0.103 *** 1.355 *** 0.112 *** 11.170 *** rockym -0.064 -0.847 -0.059 -5.940 farwest 0.101 *** 1.334 *** 0.087 ** 8.711 *** agmin -0.419 *** -5.535 *** -0.420 *** -42.033 *** construc 0.383 *** 5.058 *** 0.382 *** 38.241 *** manuf -0.811 *** -10.707 *** -0.811 *** -81.051 *** trade -0.170 *** -2.246 *** -0.170 *** -16.977 *** transport 0.001 0.019 0.002 0.184 inform -0.435 *** -5.740 *** -0.434 *** -43.399 *** finance -0.018 -0.239 -0.017 -1.746 profserv -0.046 ** -0.605 ** -0.045 ** -4.546 ** pamerican -0.006 ** -0.084 ** -0.014 ** -1.441 ** pblack -0.002 -0.027 -0.010 * -0.990 * pminority -0.004 -0.049 -0.011 -1.124 pimmigrant -0.007 ** -0.099 ** -0.017 ** -1.742 ** pselfemp 0.039 *** 0.521 *** 0.041 *** 4.107 *** 81 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 punemployed -0.006 -0.081 -0.006 -0.614 pcollege 0.003 ** 0.043 ** 0.003 ** 0.281 ** constant -0.898 *** - -0.127 - Prob>F 0.00 0.000 athrho - 0.022 lnsigma - 0.860 *** N. Obs. 111,132 111,132 aMarginal effects are expressed in percentage points b The marginal effects on variables that are not linear, that is interaction or quadratic variables, are omitted due to the lack of a parameter expressing the flexibility in the relationship between the variables. Note: Asterisk (*), double asterisk (**), and triple asterisk (***) denote variables significant at 10%, 5%, and 1% re- spectively. Figure 1. Dose-Response Function on the Effects of Hispanic Clustering on the Entrepreneurial Activity of an Average Hispanic ed PUMAs have below-average levels of self-employment and individuals with college education. How Does Heterogeneity Among Hispanics Influence the Probability of Hispanic Self-Employment? One of the main goals of the article is to understand the role of Hispanic heterogeneity on entrepreneurship. First, we investigate whether the probability of self-em- ployment varies across generations. Second, we investigate how the share of different generations of Hispanics within the PUMA influences the probability that Hispanics become self-employed. Table 2 displays the coefficients for interac- tion between platino and first, second, and third-generation Hispanics. The interaction coefficients in Table 2 measure the extent to which the probability to start a business varies across generations as the clustering of Hispanics increas- es. The coefficient estimates indicate that first-generation Hispanics are significantly more likely to start a business if they live in Hispanic-dominated PUMAs relative to 1.5 generation Hispanics. An explanation is that Hispanic-dom- inated PUMAs may encourage first-generation Hispanics to start a business by providing access to Hispanic clustering resources. On the other hand, second and third generation Hispanics are less likely to start a business as the cluster- ing of Hispanics increases. Hispanic-dominated areas may be pushing US-born Hispanics out of self-employment and 82 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 into wage-salary sectors due to the prospect of lower-reve- nue and lower-growth Hispanic-owned businesses. Figure 2 illustrates the marginal effects of the inter- actions terms in Table 2 over the entire range of Hispanic clustering. Using Equation (1), Figure 2 a, b, and c show how the probability of self-employment for first, second, and third-generation Hispanics, respectively, changes as the value of platino increases. In other words, this figure shows the relationship between self-employment and His- panic clustering for first (a), second (b), and third (c) gener- ation Hispanics over the range of Hispanic clustering. The figure also illustrates the marginal effects with 95% confi- dence intervals. Similar to the coefficients in Table 2, Fig- ure 2 illustrates that the probability of self-employment for first-generation Hispanics is positive, and it increases as the clustering of Hispanics increases . While the probability to start a business increases also for second and third-genera- tion Hispanics , the probability of self-employment is nega- tive at all levels of Hispanic clustering. In other words, the marginal effects indicate the true effect of Hispanic cluster- ing on the probability of self-employment for first, second, and third-generation Hispanics. Figure 2. Marginal Effects on the Probability of Self-Employment for First (a), Second (b), and Third (c) Generation Hispanics as Hispanic Clustering Increases. Table 3 demonstrates the probability of entrepreneur- ship for each generation, regardless of the clustering effect. Using Equation (1), Table 3 replaces the interaction terms with dummy variables if the respondent is first, second, and third-generation relative to the 1.5 generation. Table 3 shows the estimates and marginal effects of the probability of self-employment across generations of Hispanics. Simi- lar to Model 1, the instrumental variables probit shows that endogeneity is not an issue, and the standard probit regres- sion is interpreted below. A key finding is that first-generation Hispanics are more likely to start a business relative to the 1.5 genera- tion . One explanation is that first-generation Hispanics may face larger labor market constraints to enter the wage-sala- ry sector than the 1.5 generation due to lower human capi- tal. Georgarakos and Tatsiramos (2009) explain that many first-generation Hispanics enter self-employment from un- employment or underemployment. It is likely that first-gen- eration Hispanics start their own business pushed by labor barriers and make use of clustering resources as a way out of poverty. On the other hand, second and third-generation Hispanics are less likely to be self-employed relative to the 1.5 generation . We also investigate if there is a specific generation that drives the decrease in the probability to start a business for all Hispanics. Table 4 displays the probability that any His- 83 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 Table 3 Standard probit results on the probability of self-employment across generations of Hispanics Probit Coeff. Marginal Eff.a platino -0.006 ** -0.073 ** first 0.075 *** 0.979 *** second -0.359 *** -4.672 *** third -0.136 *** -1.777 *** lincome -0.173 *** -2.259 *** age 0.055 *** 0.712 *** agetwo 0.000 *** - b crime 0.000 -0.001 nchild 0.042 *** 0.543 *** married 0.036 0.471 speakenglish 0.027 0.357 spanish -0.012 -0.157 female -0.203 *** -2.644 *** college 0.035 0.450 metroarea 0.121 *** 1.580 *** accessmort -0.047 *** -0.615 *** collparent 0.006 0.084 selfemppar 0.549 *** 7.159 *** employsp 0.068 *** 0.881 *** sphispanic -0.087 *** -1.139 *** collspouse 0.094 *** 1.220 *** newengland -0.052 -0.679 mideast -0.010 -0.133 greatlakes -0.069 -0.894 plains -0.070 -0.909 southeast 0.064 ** 0.831 ** rockym -0.080 -1.039 farwest 0.088 *** 1.145 *** agmin -0.421 *** -5.482 *** construc 0.374 *** 4.875 *** manuf -0.815 *** -10.614 *** trade -0.165 *** -2.152 *** transport 0.010 0.134 inform -0.413 *** -5.384 *** finance -0.006 -0.076 profserv -0.037 * -0.482 * pwhite -0.007 ** -0.087 ** pblack -0.003 -0.035 84 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 pminority -0.003 -0.033 pimmigrant -0.008 ** -0.108 ** pselfemp 0.040 *** 0.518 *** punemployed -0.007 -0.091 pcollege 0.003 ** 0.036 ** constant -0.718 ** - Prob>F 0.00 N. Obs. 111,132 Data source: 2010 census. a The marginal effects are expressed in percentage points. b Marginal effects on quadratic variables are omitted due to the lack of a parameter expressing the flexibility in the rela- tionship between the variables. Note: Asterisk (*), double asterisk (**), and triple asterisk (***) denote variables significant at 10%, 5%, and 1% respec- tively. panic becomes self-employed as the share of first, 1.5, sec- ond, or third-generation Hispanics increases. For the sake of simplicity, Table 4 only illustrates the key explanatory variables and excludes the set of covariates . Table 4 illustrates the key variables of four probit re- gressions. Each regression replaces platino with the share of each generation pfirst, ponehalf, psecond, and pthird, respectively. The instrumental variables probit shows that endogeneity is not an issue, and we use the results from the standard probit regression. This suggests an intergeneration- al Hispanic clustering effect for Hispanic self-employment. an indicator of the Hispanic entrepreneurial environment. An explanation for the negative effect of second-generation clustering is that their socioeconomic behavior is probably the most divergent due to assimilation and economic mobil- ity, compared to first-generation Hispanics. Thus, increas- ing the proportion of second-generation Hispanics may be causing co-ethnic segregation. Intergenerational segrega- tion will likely decrease the social and economic interaction between foreign-born and US-born Hispanics. Reduced in- tergenerational interaction in US-born-dominated PUMAs is likely to shrink the entrepreneurial atmosphere for for- eign-born Hispanics. This assumption is echoed in Fairchild (2010): the clustering environment and intergenerational interaction significantly influence ethnic entrepreneurship. These results suggest that clustering resources, market con- ditions, and social norms that can encourage Hispanic en- trepreneurship are minimal in neighborhoods dominated by second-generation Hispanics. Conclusions and Implications The US Census Bureau projects that by 2060, one out of three Americans will be Hispanic. If Hispanics are twice as likely to start a business as native-born Americans, it is likely that Hispanics will help define the US entrepreneurial landscape in the coming years. However, different studies define Hispanics differently: some articles define Hispanics as those that migrate from Latin America, and others con- sider Hispanics to be any foreign- or US-born individuals of Hispanic descent. Before designing policies, decision-mak- ers should first recognize the heterogeneity among Hispan- ics and how policies may affect different groups of Hispan- ics differently. It is also unclear what motivates Hispanics to become entrepreneurs. Table 4 Standard probit results The probability that the share of each generation of His- panics influences the decision to self-employment. Probit Coeff. Marginal Eff.a pfirst 0.001 0.011 ponehalf -0.003 -0.036 psecond -0.005 ** -0.071 ** pthird -0.001 -0.012 a Marginal effects are expressed in percentage points. Note: Asterisk (*), double asterisk (**), and triple aste- risk (***) denote variables significant at 10%, 5%, and 1% respectively. Our results, consistent with a growing body of litera- ture (Fairchild, 2010), show that generational clustering has an effect on Hispanic self-employment. The results provide evidence that the clustering of second-generation Hispanics drives the decline of entrepreneurial activity among all His- panics. Thus, second-generation clustering can be used as 85 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 One major contribution of this article is the empiri- cal evidence related to Hispanic heterogeneity. While most studies lump Hispanics together in a homogeneous group, we advance the literature by focusing on Hispanic hetero- geneity. Using census data, we derive insight on the labor choices for different generations of Hispanics. This article proposes that generational differences across immigrants is not merely an ethnic control factor, but rather an important factor for the design of strategies and incentives at the fed- eral, state, and local level. Effective policies should focus on these generational differences to accurately promote suc- cess among Hispanic entrepreneurs. The American Dream is the story of immigrants arriv- ing to the US to find better opportunities. Many Hispanics are able to achieve economic mobility, but we find that dif- ferent generations of Hispanics make different labor choic- es. While first-generation Hispanics are the most entrepre- neurial group of Hispanics, they have on average lower income, education, and English proficiency than US-born Hispanics. Kramer Mills et al. (2018) reports Hispanic en- trepreneurs to be the second-most dominant group among low-income entrepreneurs. We expect that policies that sup- port low-income entrepreneurs will likely assist first-gen- eration Hispanics more profoundly than other generations. Most foreign-born Hispanic businesses are related to manual low-skilled sectors, especially construction and manufacturing. It seems that entrepreneurship is the way out of poverty for many Hispanics but can also be a source of employment for many low-wage workers. Hispanic-owned businesses, which tend to be located in low-income in- ner-city communities, may be helping reduce unemploy- ment and poverty at the community level. Our results are consistent with Kramer Mills et al. (2018), who reported that many new Hispanic business ventures tend to be con- centrated in economically depressed areas. While entrepreneurship can be the ladder out of pov- erty for foreign-born Hispanics and their communities, they are likely to experience different obstacles to success than US-born entrepreneurs. Lack of access to capital and train- ing, work regulations, and a high failure rate are among the most common barriers to Hispanic entrepreneurs (Kramer Mills et al., 2018). Targeted policies and incentives that provide training, education, and information to low-income immigrant entrepreneurs may be one way of helping immi- grant entrepreneurs to succeed while shrinking poverty and expanding prosperity in economically-depressed commu- nities. Policies that assist immigrant entrepreneurs through the strengthening of ethnic communities, access to resourc- es, and bilingual information could improve the socioeco- nomic status of Hispanics and their communities. The communities where immigrants live matter. Re- sults show that Hispanic entrepreneurship significantly de- pends on the clustering of Hispanics. A low and high degree of clustering in Hispanic communities encourage Hispanics to start their own businesses. These communities may be bringing social resources or offering niche markets that en- courage Hispanics to create businesses. Our results suggest the existence of a potential indi- cator for the Hispanic entrepreneurial environment. The agglomeration of second-generation Hispanics seems to be driving the decline of Hispanic entrepreneurship at the PUMA level. A major policy implication is that policy- makers should consider policies that encourage social and economic interaction between foreign- and US-born im- migrants. Increasing intergenerational bonds and social re- sources is likely to improve the entrepreneurial atmosphere for Hispanic entrepreneurs. We also directly address identification issues in the ethnic clustering literature by applying a series of robust econometric techniques to census data. We develop econo- metric procedures to address the potential endogeneity be- tween social interactions and individual economic behav- ior. Our identification strategy addresses several sources of potential endogeneity such as macro, individual, and peer unobserved characteristics that may affect the probability of self-employment and Hispanic clustering. This econo- metric procedure yields consistent results and allows us to draw causal conclusions on the main drivers of Hispanic entrepreneurship. Researchers and policymakers can use our findings to increase the availability and efficiency of community-based programs to encourage immigrant entre- preneurship. Further research should investigate other sources of heterogeneity such as nationality and its interaction with generational differences. We expect that Hispanics from Mexico and Central America are likely to make different economic decisions than those from South America and the Caribbean. Future research should also focus on the role of government-sponsored and community-based organizations to offset the barriers to entrepreneurship and support cred- it access for Hispanic entrepreneurs. Many strategies can emerge from the interaction of local organizations and im- migrants to provide the resources needed to start a business. While we do not focus on the legal barriers to immigrant en- trepreneurship, this article can be helpful for understanding Hispanic entrepreneurship. Researchers and policymakers can use this article to advocate for ideas at the local, state, and federal level that aim to capture the economic gains from immigrant entrepreneurship. 86 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 References Amit, R., & Muller, E. (1995). “Push” and “pull” entrepreneurship. Journal of Small Busi- ness & Entrepreneurship, 12(4), 64-80. Borjas, G. J. (1983). The substitutability of black, Hispan- ic, and white labor. Economic Inquiry, 21(1), 93-106. Borjas, G. J. (1986). The self-employ- ment experience of immigrants. The Jour- nal of Human Resources, 21(4), 485-506. Bradley, D. E. (2004). A second look at self‐em- ployment and the earnings of immigrants. In- ternational Migration Review, 38(2), 547-583. Brock, W. A., & Durlauf, S. N. (2001). Dis- crete choice with social interactions. The Re- view of Economic Studies, 68(2), 235-260. Brock, W. A., & Durlauf, S. N. (2007). Identifica- tion of binary choice models with social interac- tions. Journal of Econometrics, 140(1), 52-75. Card, D. (2009). How immigration affects US cit- ies. In R. P. Inman, (Ed.), Making cities work: Prospects and policies for urban Ameri- ca (pp. 158-200). Princeton University Press. Charles, C. Z. (2003). The dynamics of racial residential seg- regation. Annual Review of Sociology, 29(1), 167-207. Clark, K., & Drinkwater, S. (2000).Pushed out or pulled in? Self-employment among ethnic minorities in En- gland and Wales. Labour Economics 7(5), 603-628. Coates, D. & Gindling, T. H. (2010). Are Hispanic im- migrant families reviving the economies of ameri- ca’s small towns? (IZA Discussion Paper No. 4682). Cromie, S. (1987). Similarities and differences between women and men business proprietorship. Internation- al Small Business Journal 5(3), 43-60. http://papers. ssrn.com/sol3/papers.cfm?abstract_id=1545109## Danes, S. M., Lee, J., Stafford, K. & Heck, R. K. Z. (2008). The effects of ethnicity, families and cul- ture on entrepreneurial experience: An extension of sustainable family business theory. Journal of De- velopmental Entrepreneurship, 13(03), 229-268. Dávila, A., & Mora, M. (2013). Hispanic Entrepre- neurs in the 2000s: An economic profile and pol- icy implications. Stanford University Press. Durlauf, S.N. (2004). Neighborhood effects. In J. V. Hen- derson & J. F Thisse, (Eds.), Handbook of region- al and urban economics (pp. 2173-2242). Elsevier. Evans, D. S., & Leighton, L. S. (1989). Some em- pirical aspects of entrepreneurship. The Amer- ican Economic Review, 79(3), 519-535. Fairchild, G. B. (2010). Intergenerational ethnic enclave influences on the likelihood of being self-employed. Journal of Business Venturing, 25(3), 290-304. Fairlie, R. W. (2004). Race and the digital divide. Contri- butions in Economic Analysis & Policy, 3(15), 1–38. Fischer, M. J., & Massey D. S. (2000). Residential segregation and ethnic enterprise in U.S. metro- politan areas. Social Problems, 47(3), 408-424. Fisher, M., & Lewin, P. A. (2018). Push and pull factors and Hispanic self-employment in the USA. Small Business Economics, 51(4), 1055-1070. Georgarakos, D., & Tatsiramos, K. (2009). Entrepreneurship and survival dynamics of immigrants to the U.S. and their descendants. Labour Economics, 16(2), 161-170. Kim, P., & Aldrich, H. (2005). Social capital and entrepreneurship. Now Publishers Inc. Kochhar, R., Suro, R., & Tafoya, S. (2005). The New Lati- no South: The Context and Consequences of Rap- id Population Growth. Pew Hispanic Center Report. Kramer Mills, C., Battisto, J., Lieberman, S., Oroz- co, M., & Lee, N. S. (2018). Latino-owned busi- nesses: Shining a light on national trends. Stan- ford Business. https://www.gsb.stanford.edu/sites/ gsb/files/publication-pdf/slei-report-2018-latino- owned-businesses-shinging-light-national-trends.pdf Lazear, E. (1999). Culture and language. Jour- nal of Political Economy, 107(6), S95-S126. Liu, C. Y. (2012). Intrametropolitan opportunity structure and the self-employment of Asian and Latino immigrants. Economic Development Quarterly, 26(2), 178-192. Manski, C. F. (1993). Identification of endogenous social effects: The reflection problem. The Re- view of Economic Studies, 60(3), 531-542. Marshall, M. I., & Flaig, A. (2014). Marriage, chil- dren, and self-employment earnings: An analy- sis of self-employed women in the US. Journal of Family and Economic Issues, 35(3), 313-322. Miguel, E., Satyanath, S., & Sergenti, E. (2004). Economic shocks and civil conflict: An instrumental variables ap- proach. Journal of Political Economy, 112(4), 725-753. Parker, S. C. (2004). The economics of self-employment and entrepreneurship. Cambridge University Press. Portes, A., & Rumbaut, R. G. (2003). Children of im- migrants longitudinal study. [Conference ses- sion]. The Immigrant Second Generation in North America and Europe. Bellagio, Italy. Reimers, C. W. (1983). Labor market discrimina- tion against Hispanic and black men. The Re- view of Economics and Statistics, 65(4), 570-579. Robinson, P. B., & Sexton, E. A. (1994). The effect of education and experience on self-employment suc- cess. Journal of Business Venturing, 9(2), 141-156. http://papers.ssrn.com/sol3/papers.cfm?abstract_id%20=1545109 http://papers.ssrn.com/sol3/papers.cfm?abstract_id%20=1545109 https://www.gsb.stanford.edu/sites/gsb/files/publ%20ication-pdf/slei-report-2018-latino-owned-businesses-shinging-light-national-trends.pdf https://www.gsb.stanford.edu/sites/gsb/files/publ%20ication-pdf/slei-report-2018-latino-owned-businesses-shinging-light-national-trends.pdf https://www.gsb.stanford.edu/sites/gsb/files/publ%20ication-pdf/slei-report-2018-latino-owned-businesses-shinging-light-national-trends.pdf 87 A. Torres, M. S. Delgado, & M. I. Marshall Journal of Small Business Strategy / Vol. 31, No. 3 (2021) / 72-87 Shapero, A., & Sokol, L. (1982). The social dimensions of entrepreneurship. University of Illinois at Ur- bana-Champaign’s Academy for Entrepreneurial Leadership Historical Research Reference in En- trepreneurship. https://ssrn.com/abstract=1497759 Shinnar, R. S., & Young, C. A. (2008). Hispanic immigrant entrepreneurs in the Las Vegas metropolitan area: Moti- vations for entry into and outcomes of self‐employment. Journal of Small Business Management, 46(2), 242-262. Suro, R., & Passel, J. S. (2003). The rise of the second gen- eration: Changing patterns in Hispanic population growth. Pew Hispanic Center Report. Washington, DC. USHCC (2015). Hispanic business & entrepre- neurs drive growth in the new economy. h t t p : / / r e g i o n a l h i s p a n i c c c . o rg / w p - c o n t e n t / u p - loads/2016/02/HispanicBusinessReport2015.pdf Wang, Q. (2010). Immigration and ethnic entre- preneurship: A comparative study in the Unit- ed States. Growth and Change, 41(3), 430-458. Yuengert, A. M. (1995). Testing hypothe- ses of immigrant self-employment. Jour- nal of Human Resources, 30(1), 194-204. https://ssrn.com/abstract=1497759