Supply of immigrant entrepreneurs and native entrepreneurship

Broadly, this paper is related to the vast literature, across disciplines, on socio-economic impacts of immigration. Most previous studies, both theoretical and empirical, on the topic examine the impact of immigrant workers on wages and employment prospects of native workers. Conclusions from theoretical models in this class of studies depend on the assumptions made on the relative skill structure of immigrant workers and native workers (which ultimately determine the elasticity of substitution between the two groups) and the elasticity of capital supply (Friedberg and Hunt, 1995; Dustmann et al., 2008; Manacorda et al., 2012). For example, if capital supply is perfectly elastic and immigrant workers are perfect substitutes for native workers, immigration does not affect the wages and employment prospects of natives. However, if immigrant workers have different skill composition than native workers, there will be winners and losers among natives even with perfectly elastic capital supply.

Empirical studies on the topic are abundant (for example Card (1990), Borjas et al. (1997), Winter-Ebmer and Zweimüller (1999), Hu (2000), Borjas et al. (2011), Ottaviano and Peri (2012), Dustmann et al. (2017), and Prantl and Spitz-Oener (2020) to state only a few) but report widely varying results. Nevertheless, a general consensus is that in the long run, immigrant workers, on average, neither lower wages of native workers (Peri, 2014) nor displace them from their jobs (Constant, 2014). Increased productivity, better skill mix, and high rate of entrepreneurship among immigrants are some of the benefits of immigration highlighted in these studies.

See Kerr and Kerr (2011) for a survey of recent empirical studies on the economic impacts of immigration.

More specifically, however, this paper aligns with a smaller group of studies within the migration literature that focuses on immigrant entrepreneurship. Studies within this space are scattered across disciplines including ethnic studies, sociology, urban studies, economics, entrepreneurship, economic geography, management studies, political science, and policy studies (Rath and Kloosterman, 2000). A majority of these studies describe immigrant entrepreneurship as a form of ethnic entrepreneurship and are often concerned with identifying distinct characteristics that make certain groups of immigrants more successful entrepreneurs than others (Evans, 1989; Yoon, 1995; Mesch and Czamanski, 1997; Min and Bozorgmehr, 2000; Kahn et al., 2017). Researchers have mainly examined entrepreneurship experiences of Hispanics, Chinese, Koreans, Indians, Iranians, and Filipinos in the North America and the experiences of Polish and Turkish immigrants in Europe (Aliaga-Isla and Rialp, 2013). Altogether, these studies highlight the various forms of human capital, social capital, and cultural attributes associated with immigrant entrepreneurs. Admittedly, the scope of this line of research is steadily expanding and scholars have explored areas as varied as the institutional environment in the context of immigrant entrepreneurship (Kloosterman, 2000), role of immigrant entrepreneurship in the high-tech sector (Hart and Acs, 2011), returns to immigrant entrepreneurship (Lofstrom, 2011), motivations and obstacles to immigrant entrepreneurship (Fatoki and Patswawairi, 2012), survival and growth of immigrant-founded businesses (Kerr & Kerr, 2016, 2020), and the role of global forces in immigrant entrepreneurship (You and Zhou, 2019). The other area of growing interest among researchers is the role of immigrants in transnational entrepreneurship (Kwak and Hiebert, 2010; Brzozowski et al., 2017). Dabić et al. (2020) provide an excellent up to date review of studies and evolving research agenda in this area. Despite large differences in terms of focus, more often than not, studies in this group discuss immigrant entrepreneurship as an important benefit of immigration, – for example, see Rath and Kloosterman (2000), Lofstrom (2002), and Peroni et al. (2016) – but credible research on how immigrant entrepreneurs affect native entrepreneurship is inadequate.

A handful of studies examine the closely related issue of how immigration affects entrepreneurship among natives. Light and Sanchez (1987), probably the earliest among them, utilize data collected from the 1980 public-use samples of U.S. Census to find that a higher percentage of foreign-born in labor force results in a higher aggregate self-employment rate. Similarly, Fairlie and Meyer (1997) use census data from 1980 and 1990 to estimate the impact of immigration on African American self-employment and find no significant effect. However, in a subsequent related study, the authors look at the impact of immigration on the self-employment propensities of the non-African American population in the US and find a small negative impact (Fairlie and Meyer, 2003). Hunt and Gauthier-Loiselle (2010) examine the impact of skilled immigration on innovation, measured by patenting behavior, in the US and find that a one percentage point increase in the share of immigrant college graduates in the population increases patents per capita by 6%. Similarly, Kerr and Lincoln (2010) examine the impact of high-skilled immigrants, entering on H-1B visas, on US technology formation. The authors conclude that total innovation, which is measured by patenting and employment in science and engineering, increases with higher H-1B admission levels. Olney (2013) examines the related issue of how firms respond to immigration and find that greater influx of immigrants translates into more firms. More recently, Unel (2018) examines the effect of immigration on entry and exit of US born entrepreneurs and finds negative impact on entry and exit of female entrepreneurs but no impact on male entrepreneurs. Specifically, the author finds that a 10% increase in the share of immigrants in the population lowers the entry and exit rates of female entrepreneurs by 10 and 5% respectively. Taking a theoretical approach, Unel (2019) sets up a two country model to study the effect of immigration policies and concludes that policies that accelerate the integration of immigrants to host economy reduce offshoring of labor and increase the local supply of entrepreneurs.

Although important in their own right, existing studies closely related to this study investigate the impact of immigration on post-immigration aggregate levels of entrepreneurship or the entrepreneurship rate of natives. They fall short of disentangling the impact of immigrant entrepreneurs from that of immigrants in general. Since a wave of immigration adds both self-employed and employed to the economy, the impact of immigration is a result of the economic activities of the immigrant entrepreneurs and the non-entrepreneurs. Isolating the impact of immigrant entrepreneurs from immigration not only informs the policies related to immigration and entrepreneurship development, but also improves our understanding of the determinants of entrepreneurship among natives.

To empirically estimate the net impact of immigrant entrepreneurs on native entrepreneurship, I build up on a conceptual framework of self-employment outlined by Borjas (1986). In this econometric setup, an individual i chooses self-employment based on the difference between market wage w_i, and the expected net income from self-employment, y_i. If we define an index function I such that (1) Ii=yi-wi=Ciπ+vi {I_i} = {y_i} - {w_i} = {C_i}\pi + {v_i} where C_i is the vector of socio-economic characteristics which affect y_i and w_i, v_i is the disturbance term, and π is a parameter vector, then the individual chooses self-employment when I_i > 0. The probability of an individual being self-employed is then given by (2) Pi=Pr[Ii>0|Ci]=Pr[vi>-Ciπ|Ci] {P_i} = Pr [{I_i} > 0|{C_i}] = Pr [{v_i} > - {C_i}\pi |{C_i}]

If v_i is distributed symmetrically, then (3) Pi=Pr[vi<Ciπ|Ci]=F(Ciπ) {P_i} = Pr [{v_i} < {C_i}\pi |{C_i}] = F({C_i}\pi) where F is the cumulative distribution function of the random variable v_i. The parameter vector π can now be estimated in a standard regression framework by specifying the functional form of F and the elements of vector C.

It is hard to specify, with certainty, the exact variables that go into the vector C (Borjas, 1986). However the literature, discussed in the foregoing subsection, suggests that in addition to normally considered micro and macro level determinants of labor market outcomes, the supply of immigrant entrepreneurs is likely to play an important role in a native individual's choice of self-employment. Therefore, for the purpose of estimating the net impact of immigrant entrepreneurs on native entrepreneurship, I utilize a reduced form model that specifies the self-employment propensity of natives as a function of supply of immigrant entrepreneurs and other relevant controls. I discuss this model in detail in Section 4.

Theoretical foundation for this analysis is the conceptual framework of self-employment outlined in Borjas (1986).

See Section 2.2 for details.

Under the assumptions that regressors in Eq. (4) are exogenous, parameter β₁ estimates the impact of the supply of specific type of immigrant entrepreneurs on the specific type of entrepreneurial propensity of natives. For instance, when SelfEmployed_ist is the indicator for whether a native individual is self-employed in an incorporated business, a positive β₁ implies that incorporated immigrant entrepreneurs crowd in natives into incorporated businesses. Similarly when SelfEmployed_ist represents self-employment in unincorporated business, a positive β₁ implies that unincorporated immigrant entrepreneurs, on average, crowd in natives into unincorporated businesses.

The estimates of β₁ obtained from Eq. (4) are unbiased and consistent only if the supply of immigrant entrepreneurs is uncorrelated with the error term conditional on the full set of controls. However, this assumption can be invalid for several reasons. First, if native individuals avoid or select states with large supply of immigrant entrepreneurs (endogenous sorting), the estimates from Eq. (4) can be biased and inconsistent. Second, it is unlikely that all micro and macro level determinants of native self-employment are accounted for in the model, so a potential omitted variable bias cannot be ruled out. Finally, endogeneity can result from reverse causality that runs from the outcome variable to the regressor of interest. To address these empirical concerns, I estimate Two Stage Least Squares (TSLS) models utilizing an instrumental variable described below.

In TSLS models, a suitable instrumental variable enables causal identification by acting as an exogenous source of variation in the endogenous regressor. Assuming monotonicity in its effect, an instrument must satisfy two conditions to ensure consistency of the estimates of β₁. First, the instrument must have sufficient correlation with the supply of immigrant entrepreneurs after partialling out the effect of other covariates (the relevance requirement) and second, it should not be correlated with the error term ɛ in Eq. (4) after conditioning on other covariates (the exclusion restriction). Intuitively, we require an exogenous source of variation in the supply of immigrant entrepreneurs which does not affect native self-employment except through the channel of supply of immigrant entrepreneurs. Fortunately, the quasi experiment provided by the implementation of the State Children's Health Insurance Program (SCHIP) across states in the late 90s provides such variation.

To understand how SCHIP induces a plausibly exogenous variation in the supply of immigrant entrepreneurs, one needs to go back to the 1996 welfare reform. As the welfare program known as Aid to Families with Dependent Children (AFDC) received increasing criticism in the 80s and 90s for “disincentivizing work” and “promoting welfare dependence” among the beneficiaries, the US passed the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) in 1996. Two most important goals of the Act were to encourage work among welfare recipients and achieve a significant decline in welfare caseloads (Clinton, 1996). To that end, PRWORA replaced AFDC program with the Temporary Assistance for Needy Families (TANF) and imposed stricter work requirements for welfare eligibility. Additionally, the Act banned immigrants who have been in the US for less than 5 years, henceforth newly arrived immigrants, from receiving federally funded means tested benefits such as TANF and Medicaid. In 1997, the Congress passed SCHIP to cover uninsured children living in low-income families that make too much to qualify for Medicaid but too little to afford private coverage. The program was rapidly implemented across states. By 2000, every state including the District of Columbia (DC) had some form of the SCHIP. The SCHIP promised a partnership between the federal and the state governments and gave states flexibility in designing their own eligibility requirements. Taking advantage of this autonomy, 15 states including the DC, henceforth “generous states,” chose to include newly arrived immigrants in their SCHIP.

Information on dates of SCHIP implementation and on whether states covered newly arrived immigrants is presented in Table A1 in Appendix. The generous states also provided Medicaid coverage to a select group of newly arrived adult immigrants who met eligibility criteria based on disability, age, etc. (Fortuny and Chaudry, 2011). In 2009, an amendment to the Children's Health Insurance Program Reauthorization Act (CHIPRA) allowed all states to provide medical coverage to lawfully present pregnant women and children without a waiting period.

There are several reasons for why the provision of SCHIP to newly arrived immigrants can increase the supply of immigrant entrepreneurs. First, in the US, adults are often compelled to work for large employers to ensure continued access to affordable insurance for themselves and their kids (Fairlie et al., 2011). Some of these adults may switch into self-employment when there is a guaranteed access to publicly funded children's health insurance. Second, public health insurance helps families to save on premiums that would otherwise be paid for private coverage. Savings accumulated over time because of public coverage may ease liquidity constraint, a commonly cited obstacle to business creation (Evans and Jovanovic, 1989; Fairlie and Krashinsky, 2012). Third, the supply of immigrant entrepreneurs may increase when adults in families without any insurance coverage, who had previously chosen to remain out of labor force for child care, decide to enter the labor force due to reduced risk of leaving a child at day care or with a family member. Nevertheless, the suitability of generous SCHIP (GenSCHIP) as an instrument depends on whether GenSCHIP meets both the relevance and the excludability criteria (Angrist and Pischke, 2008). I conduct several tests to address potential concerns related to these criteria in the following section.

I begin with a close examination of the summary statistics presented in Table 1. In the pooled sample that includes both the natives and the immigrants sampled during 1994–2016 period, the overall self-employment rate, defined as the proportion of self-employed individuals in non-agricultural labor force, is approximately 9.61%. Of the 9.61%, 6.06% are self-employed in unincorporated businesses and only 3.42% are self-employed in incorporated businesses. That a significantly smaller number of entrepreneurs chose to incorporate their businesses, in spite of a number of legal benefits including the advantage of limited liability, indicates that incorporated self-employed individuals may indeed be different from unincorporated self-employed in certain aspects. The third and fourth columns in Table 1 show that the ratio of incorporated to unincorporated self-employed individuals is similar across native and immigrant groups. In terms of demographic characteristics, the natives in the sample are slightly older, slightly more educated, significantly more white, and have fewer children in the family.

Table 2 presents summary characteristics of the self-employed natives and self-employed immigrants separated by self-employment types. The statistics demonstrate that incorporated entrepreneurs differ from the unincorporated in various ways. For example, the incorporated self-employed individuals are, on average, older – therefore, with potentially longer labor market experience – and more educated than the unincorporated self-employed. Similarly, incorporated self-employed are more likely to be married, have more children, and are more likely to live in metropolitan areas than their unincorporated counterparts. Women and minorities (Blacks and Hispanics) have a better representation in unincorporated than incorporated businesses. These differences are reflected within both native and immigrant groups and further support the unique treatment of the two types of self-employment.

Selected results generated in the baseline model are presented in Table 3. These estimates are obtained from estimating Eq. (4) by Ordinary Least Squares. Only coefficients on immigrant entrepreneurs (ImmiEnt) are shown for brevity (see Table A2 in Appendix for coefficients on full set of controls). All models control for individual characteristics, state level characteristics, state fixed effects, and year fixed effects. All coefficients in Table 3 are positive and statistically significant at conventional levels. Specifically, estimates in second column suggest that an increase of 100,000 self-employed immigrants (of any type) increases the entrepreneurial propensity of natives, on average, by 0.0035 percentage points. Given that only approximately 9.5% of natives are self-employed, it amounts to a 3.6% increase in the supply of native entrepreneurs. Similarly the third column shows that an increase of 100,000 incorporated immigrants increases the native propensity of incorporated self-employment by 0.0064 percentage points. Finally, the fourth column in Table 3 shows that an increase of 100,000 unincorporated immigrants increases the native propensity of unincorporated self-employment by 0.0025 percentage points. Overall, Table 3 indicates positive effect of immigrant entrepreneurship on native entrepreneurial propensities. However, as discussed earlier in the manuscript, the LPM estimates are likely to be biased due to endogeneity. Therefore, I report results from the preferred instrumental variable model in the following subsection.