Migration and self-employment: the case of internal migrants

Abstract

In this paper, I compare the self-employment entry rate of internal migrants within the United States to that of individuals who have not moved. Internal migrants are defined as native born Americans who have migrated from one state to another. After controlling for a broad range of demographic and labor market characteristics, I find that Americans who have migrated to a state that does not border their initial state of residence are approximately 43% more likely to enter self-employment than Americans who have not moved. This higher rate of entry into self-employment among internal migrants remains robust to the inclusion of a wide range of controls. Since internal migrants are individuals who have made the decision to migrate, but are otherwise similar to native born non-migrants, this finding supports the idea that the higher self-employment entry rate among immigrants is driven in part by self-selection.

1 Introduction

It has become well established that immigrants to Western countries are more likely to be self-employed than native born individuals (Borjas 1986; Fairlie and Lofstrom 2015; Dheer 2018; Hammarstedt and Miao 2020; Azoulay et al. 2022). However, even though the related academic literature has consistently found this gap in self-employment rates, the evidence on why this gap exists has been inconsistent. Understanding the primary reasons behind this gap in self-employment rates is important, because it can shed light on whether immigrants tend to be more inclined towards entrepreneurship than natives, or whether their higher self-employment rates simply emerge as a symptom of difficulties with obtaining paid employment in their new country.

Explanations for the higher self-employment rates among immigrants generally fall into one of two different categories. The first category are theories which contend that immigrants tend to have post-migration experiences which increase their propensity to become self-employed, such as difficulty with finding a paid job, or high rates of contact with co-ethnic entrepreneurs. The second category of theories contend that immigrants tend to come from countries with stronger entrepreneurial traditions than most Western countries, leading immigrants to enter self-employment at higher rates than native-born individuals. A large literature, which is summarized in the literature review section of this paper, has emerged that examines in detail various theories that fall into one of these two categories.

However, it has been increasingly recognized that several of the characteristics that are predictive of entry into self-employment are also predictive of whether or not an individual will choose to migrate. In particular, individuals who are relatively tolerant of risk are both more likely to migrate (Jaeger et al. 2010; Gibson and McKenzie 2011; Akgüç et al. 2016; Dustmann et al. 2020) and are more likely to choose self-employment (Ekelund et al. 2005; Ahn 2010; Hvide and Panos 2014; Hsieh et al. 2017; Levine and Rubinstein 2017) than relatively risk averse individuals. Internal locus of control (Hansemark 2003; Caliendo et al. 2019) and personality characteristics including extraversion and openness to new experiences (Zhao et al. 2010; Crown et al. 2020) have also been linked to both the migration decision and the self-employment entry decision. This raises the question of whether the greater propensity for immigrants to enter self-employment can be explained, at least in part, by self-selection. That is, do individuals who make the choice to migrate tend to possess characteristics that lead to a higher probability of entering self-employment? If this is the case, then it would suggest that non-random selection into migrating is at least partially responsible for the immigrant-native gap in self-employment rates. The gap between the self-employment rates of immigrants and natives would therefore be expected to persist even if immigrants did not face post-migration experiences that differ from their native born counterparts, such as disproportionate difficulties with obtaining paid employment.

Examining whether self-selection is responsible for the immigrant-native self-employment rate gap is a difficult empirical exercise. In addition to being individuals who have self-selected into migrating, immigrants tend to go through other experiences that may also explain their greater propensity towards self-employment. Among these are; discrimination by prospective employers, a greater tendency to live in ethnic enclaves, and a lack of fluency in the host country language. In this paper, I contribute to the literature by examining the self-employment entry rates of a group of individuals who have self-selected into migrating, but are not expected to possess many of the aforementioned characteristics that might also explain a greater propensity towards becoming self-employed. These individuals are internal migrants within the United States.

Internal migrants are defined as native-born Americans who have migrated from one state to another state. If migrants are more likely to be self-employed than non-migrants exclusively because of reasons such as; cultural differences, strong networks of co-ethnics, discrimination by paid employers, or a lack of English language abilities, then we should not observe internal migrants being more likely to become self-employed. This is because these factors are unlikely to apply in the case of internal migrants, once a broad range of observable characteristics have been accounted for. However, if migrants are more likely to become self-employed in part due to self-selection on the basis of personality characteristics that are predictive of both migration and self-employment entry, then we should expect to see higher self-employment propensities among internal migrants than among individuals who have not moved. While it is worthwhile noting that a higher rate of self-employment entry among internal migrants does not preclude other factors from contributing to the higher self-employment propensity of immigrants, it would suggest that self-selection forms at least part of the explanation.

In order to focus on significant moves, I focus on natives who have migrated to a state that does not border their initial state of residence. Furthermore, for the empirical analysis, I exclude individuals who report that they moved because of a new job or a job transfer. This restriction is imposed since immigrants are unlikely to have migrated primarily to work for a specific employer as is the case for these internal migrants.¹

After accounting for demographic characteristics and one’s recent labor market history,² I find that natives who have recently migrated from one state to another state that does not share a border with their initial state of residence are approximately 43% more likely to enter self-employment than individuals who have not moved. The self-employment entry rates are smaller for individuals who have moved to states that border their initial state of residence, and smaller still among those who have moved to another location within the same state. These findings are robust to the addition of a wide array of controls, including local labor market characteristics, the incidence of a post-migration period of unemployment, and wealth. Furthermore, consistent with past findings that individuals who migrate to culturally distant areas tend to be more tolerant of risk, I find that internal migrants who move to states with different political leanings from their initial state of residence are more likely to become self-employed.

By demonstrating that internal migrants are more likely to enter self-employment than non-movers, this paper provides evidence supporting the idea that the higher self-employment rate among immigrants is, at least in part, attributable to self-selection. This means that even if immigrants were to no longer experience discrimination or other difficulties in the paid labor market, we may still expect to observe higher self-employment rates among immigrants due to their typically unobservable characteristics which drive both the migration and the self-employment entry decisions.

The remainder of this paper is structured as follows. Section 2 reviews the related literature. Section 3 describes the data and methodology that is used. Section 4 reviews the results. Section 5 contains a series of robustness checks. Section 6 concludes.

2 Literature review

One class of theories for why immigrants are more likely to be self-employed than natives contends that immigrants have a higher propensity to enter self-employment due to characteristics of the source countries from which they arrive. Most frequently, it is asserted that immigrants from countries with high self-employment rates are more likely to become self-employed themselves (Yuengert 1995; Uwaifo and Belton 2012; Lassmann and Busch 2015). This could be due to cultural factors, or their own pre-migration experiences with self-employment. However, support for this theory has been far from universal, with Fairlie and Meyer (1996), Van Frank (2005) and Hammarstedt and Shukur (2009) failing to find evidence supporting this hypothesis. However, when using a direct measure of whether an individual was self-employed in their home country pre-migration, Akee et al. (2013) ,Tibajev (2019) and Giambra and McKenzie (2021) find evidence that individuals who were self-employed prior to migrating are subsequently more likely to be self-employed in the destination country.

Another set of explanations consists of theories which contend that immigrants enter self-employment at higher rates due to their post-migration experiences. This frequently takes the form of the assertion that, compared to natives, immigrants would receive low earnings if they work for an employer. These low prospective earnings in paid employment are predicted to drive immigrants towards the relatively more attractive option of self-employment (Fairlie and Meyer 1996; Clark and Drinkwater 2000; Hammarstedt 2006; Fisher and Lewin 2018; Wang and Lofstrom 2020). Similarly, Kone et al. (2021) find that refugees, who are the lowest earning class of immigrants to the United Kingdom, are also the most likely to enter self-employment. Another post-migration factor that has been put forward as an explanation for the higher self-employment rates among immigrants is the tendency for immigrants to reside in ethnic enclaves. Immigrants who reside in an ethnic enclave may have a higher propensity to enter self-employment due to their ability to learn from co-ethnic entrepreneurs (Martín-Montaner et al. 2018; Tavassoli and Trippl 2019; Andersson 2021; Andersson et al. 2021). However, support for the enclave hypothesis has not been uniform, with Klaesson and Öner (2021) finding mixed results.

Should the explanations that are listed above be the primary reasons for the higher self-employment rate among immigrants, we would expect an individual who is randomly selected to migrate from their country of origin to a host country to be more likely to enter self-employment than the average native. This could be because their background makes it more likely that they would choose self-employment, or because they will experience conditions in the host country that make self-employment relatively more attractive. However, individuals who migrate to a given host country are clearly not randomly selected subsamples of individuals from their country of origin. Increasingly, it has been recognized in the literature that the higher self-employment rates among immigrants may be, at least in part, an artefact of non-random selection into migrating. In other words, the characteristics that predict migration are in some cases also the characteristics that predict entry into self-employment.

Individuals who migrate are often found to more tolerant of risk than individuals who do not migrate (Jaeger et al. 2010; Gibson and McKenzie 2011; Akgüç et al. 2016; Dustmann et al. 2020). Using German data, and proxying for cultural distance using data on regional dialects, Bauernschuster et al. (2014) finds that this is strongest when the cultural distance between the home and destination region is larger. Umblijs (2012) finds that since migrating to a new region entails more risk when there are fewer co-ethnics present, earlier migrants to a destination region are more risk-tolerant than later migrants. Going beyond risk tolerance, Caliendo et al. (2019) finds that individuals who have an internal locus of control are more likely to migrate than individuals who have an external locus of control. Furthermore, Crown et al. (2020) documents that extraverted individuals, and those who are open to new experiences, are more likely to migrate between regions of Australia.

The characteristics that are listed above are also characteristics that have been found to be related to the self-employment entry decision. Risk tolerant individuals are more likely to enter self-employment (Ekelund et al. 2005; Ahn 2010; Hvide and Panos 2014; Hsieh et al. 2017; Levine and Rubinstein 2017), with Batista and Umblijs (2014) having documented this specifically among immigrants to Ireland. Individuals with an internal locus of control are more likely to become self-employed (Hansemark 2003), as are individuals who are extraverted and open to new experiences (Zhao et al. 2010).

Despite this, empirical research linking self-selection into migrating to the self-employment decision has been scarce. Hormiga and Bolívar-Cruz (2014) finds that immigrants to Spain are less likely to perceive the formation of a new business as being a risky endeavour than non-immigrants, which may be the result of individuals who migrate having already undertaken a risky endeavour by moving, and perceive business formation to be less risky in comparison. Vandor (2021) uses survey evidence to document that Austrians who express a willingness to migrate are also more likely to have an intention to start a business in the future.

This paper contributes to the above literature by documening self-employment entry rates among internal migrants within the United States. Since these are migrants who are similar to non-migrants in most respects other than the decision to migrate, this approach is a cleaner way to establish whether self-selection drives higher self-employment entry rates among migrants than what is typically found in the literature.

3 Data

This project uses Current Population Survey (CPS) microdata files from 1995 to 2021 downloaded from IPUMS (Flood et al. 2020). The CPS is a monthly survey administered to approximately 60,000 American households every month.³ The CPS uses a well known 4-8-4 sampling scheme. Each household is initially interviewed every month for a total of 4 consecutive months. After an 8 month gap, these households are later interviewed for another period of 4 consecutive months. Since there is considerable attrition in the sample following the 8 month gap, I focus on observing changes in labor market status between the first month and fourth month of a 4 consecutive month period. For individuals who appear in the sample for 4 consecutive months in 2 different years,⁴ I include only observations from the first year.

The analysis is restricted to individuals who appear during the month of March, when the CPS administers a supplementary questionnaire to respondents that is commonly known as the March Supplement. Crucially, this supplementary questionnaire asks respondents whether they have moved within the previous year. While the March supplement contains a larger sample than the basic CPS monthly files, due to the need to follow individuals over a 3 month period, only respondents who can be matched to the basic monthly files are included in the sample.

If a March supplement respondent indicates that they had moved within the previous year, they are then asked whether they moved to a new location within the same county, to another county within the same state, to another state, or whether they moved to the United States from another country. They are also asked what their primary reason for moving was. Individuals who report having moved for a “new job or a job transfer” are eliminated from the sample since this reason implies a strong tie to a particular employer that is unlikely to exist for most recent immigrants.⁵

Individuals who have moved from one state to another are also asked which state they resided in 1 year earlier. Those who give inconsistent responses regarding their migration status are eliminated from the sample. These are individuals who report having moved states within the previous year, but also report having lived in the same state 1 year earlier, as well as individuals who report that they have not moved states, but report a different state of residence 1 year earlier.⁶ Immigrants are eliminated from the sample, as are individuals who report having been self-employed in their first month in the CPS, since by definition these individuals can not be observed entering self-employment. Furthermore, in order to be able to observe an individual’s starting occupation and industry, the sample is restricted to consistent exclusively of individuals who were engaged in paid employment during their first month in the CPS. This leaves a sample of 697,822 individuals, approximately 0.77% of whom entered self-employment between their first month and fourth month in the CPS.

A set of dummy variables are created which reflect a respondent’s recent migration history. $Moved {0.1cm} in {0.1cm} County_{i}$ indicates whether the respondent reports moving to another location within the same county. $Moved {0.1cm} Counties_{i}$ indicates whether the respondent reports moving to another county within the same state. $Moved {0.1cm} Border {0.1cm} State_{i}$ indicates whether the respondent reports moving to another state in the U.S. that shares a land border with their initial state of residence. $Moved {0.1cm} NonBorder {0.1cm} State_{i}$ indicates whether the respondent reports moving to another state in the U.S. that does not share a land border with their initial state of residence.

Table 1 reports the percentage of the sample that fall into each movement category. As can be seen, the overwhelming majority of the sample did not move within the year prior to their first appearance in the March Supplement. Just over 11% of individuals in the sample moved but remained within the same county. This is the largest group of movers, followed by those who moved counties but remain within the same state, at 3.3% of the sample. Moving states is a relatively rare event, with only 2% of the sample having moved from one state to another, split between 0.8% who moved to a bordering state and 1.2% who moved to a non-bordering state.

Table 1 Percentage of sample engaging in moves of each type

Basic demographic characteristics of this sample, split by an individual’s migration history, can be found in Table 2. As can be seen, individuals who have migrated are considerably younger than individuals who have not migrated. However, there is little difference between the average age of different migrant groups. Similarly, migrants are also less likely to be married than non-migrants. With regards to education, those who have moved smaller distances (within their county or state) tend to have less formal education that non-movers, while those who have moved larger distances tend to have the highest levels of education among all groups.

Table 2 Sample means of demographic characteristics

4 Results

The probit model in Eq. (1), found below, estimates the probability that an individual enters self-employment during the 3 months between their first and fourth months in the CPS, split by their migration history. The estimated probabilities from this analysis are shown in Table 3. This table demonstrates that 0.74% of individuals who had not recently moved enter self-employment within a given 3 month period. Among individuals who have moved but remain within the same county, this figure is only slightly higher at 0.77%. However, individuals who have undertaken more significant moves have a higher probability of entering self-employment. Individuals who have left their county, but remain within the same state have a 0.83% chance of entering self-employment. Individuals who have moved to a bordering state have a 1.03% chance of entering self-employment, and individuals who have moved to a non-bordering state have a 1.19% chance of entering self-employment. Thus it is clear that natives who have undertaken significant internal moves within the United States have a higher probability of becoming self-employed than natives who have not done so. The marginal effects from this analysis are reported in Column 1 of Table 4, which demonstrates that the gap in entry rates into self-employment between individuals who have moved to non-bordering states, and those who have not moved, is statistically significant at the 1% level.

$$\begin{aligned}{} & {} Pr(Enter \; Self_{i}=1)=\Phi (\beta _{0}+\beta _{1}Moved \; in \; County_{i}+\beta _{2}Moved \; Counties_{i}\\ & + \beta _{3}Moved \; Border \; State_{i}+\beta _{4}Moved \; NonBorder \; State_{i}) \end{aligned}$$

(1)

$$ \begin{aligned}{} & {} Pr(Enter \; Self_{i}=1)= \; \Phi (\beta _{0}+\beta _{1}Moved \; in \; County_{i}+\beta _{2}Moved \; Counties_{i}+\beta _{3}Moved \; Border \; State_{i}+\\& \beta _{4}Moved \; NonBorder \; State_{i}+\varvec{\beta _{5}X_{i}}+\varvec{\beta _{2}State_{i}}+\varvec{\beta _{6}Time_{i}})\end {aligned} $$

(2)

In order to determine whether these results hold while controlling for demographic characteristics, Eq. (2) is estimated, which controls for a wide set of demographic characteristics in $\varvec{X_{i}}$ consisting of; gender, marital status, age, race, level of education, and region of birth of one’s parents.⁷ This model also controls for one’s current state of residence (for internal migrants this is their state of destination), as well as a series of dummies for the time period (year and month) in which an individual is observed. Column 2 of Table 4 demonstrates the marginal effects after controlling for these covariates. In this case, it is clear that the size of the effects have increased relative to the raw gaps that are demonstrated in Column 1. This is primarily attributable to the inclusion of controls for age, as younger individuals are more likely to migrate, but are less likely to enter self-employment. After accounting for demographic characteristics, individuals who have moved within the same county are found to be statistically more likely to enter self-employment than those who have not moved at all, albeit the magnitude of this effect is small. The size of the effect is increasing as bigger moves are considered, with individuals who have moved to a non-bordering state being more than half a percentage point more likely to enter self-employment than individuals who have not.

Equation 3 adds additional controls for one’s recent labor market history. For migrants, these controls reflect their labor market experiences post-migration. I control for the two-digit NAICS industry code that is recorded during an individual’s first month in the CPS, as well as their two-digit SOC occupation code during that month. Furthermore, I control for whether an individual was observed as being either unemployed or not in the labor force during their second or third month in the CPS, as well as an interaction between these two terms. The results from this estimation can be found in Column 3 of Table 4. Adding this rich set of controls attenuates, but does not eliminate, the effect of internal migration on the probability of self-employment entry. Individuals who have moved to states that do not border their initial state of residence remain considerably more likely to enter self-employment than non-movers.

$$\begin{aligned}{} & {} Pr(Enter {0.1cm} Self_{i}=1)=\Phi (\beta _{0}+\beta _{1}Moved {0.1cm} in {0.1cm} County_{i}+\beta _{2}Moved {0.1cm} Counties_{i}+\beta _{3}Moved {0.1cm} Border {0.1cm} State_{i} \\{} & {} \quad + \beta _{4}Moved {0.1cm} NonBorder {0.1cm} State_{i}+\varvec{\beta _{5}X_{i}}+\varvec{\beta _{2}State_{i}}+\varvec{\beta _{6}Year_{i}} +\varvec{\beta _{7}Month_{i}}+\varvec{\beta _{8}Initial Occ_{i}}+\varvec{\beta _{9}Initial Ind_{i}} \\{} & {} \quad +\beta _{10}Was {0.1cm} Unemployed_{i}+\beta _{10}Was {0.1cm} NotLabor_{i}+\beta _{10}Was {0.1cm} Unemployed_{i}*Was {0.1cm} NotLabor) \end{aligned}$$

(3)

Table 3 Probability of self-employment entry by migration group

This table shows the estimated probability that an individual enters self-employment during the 3 months between their 1st month and 4th month in the CPS. This sample includes non-immigrants who were not self-employed during the first month in which they appear during the CPS. Individuals who moved but reported that they moved for a new job or a job transfer are also excluded from the sample. These probabilities are calculated following estimation of Eq. (1).

Table 4 Marginal effects of migration status on the probability of self-employment entry

This table shows the marginal effects of having recently migrated on the probability of entering self-employment during a 3 month period. This sample includes non-immigrants who were not self-employed during the first month in which they appear during the CPS. Individuals who moved but reported that they moved for a new job or a job transfer are also excluded from the sample. The marginal effects in Column (1) are calculated following estimation of Eq. (1), while the marginal effects in Columns 2 and 3 are calculated following the estimation of Eqs. (2) and (3), respectively.

These results clearly demonstrate that natives who have migrated across larger distances within the United States exhibit greater tendencies to become self-employed than natives who have not migrated. Factors that are commonly thought of as being related to the high self-employment rates of immigrants, such as poor paid employment opportunities or ethnic enclaves, are less likely to impact these natives. This is consistent with the idea that individuals who self-select into migrating possess unobserved characteristics that are associated with a higher propensity towards self-employment.

5 Robustness checks

The results in the previous section are consistent with the idea that individuals who have self-selected into migrating are more likely to enter self-employment than non-movers. In order to strengthen the claim that the higher rates of self-employment entry among internal migrants is due to self-selection, I address a series of alternative explanations. For ease of comparison, Column 1 in all of the following tables contains the primary results from estimation of Eq. (3), which are also shown in Column 3 of Table 4.

5.1 Political differences across states

Bauernschuster et al. (2014) has shown that individuals who migrate to regions of Germany that are more culturally distant from their initial region of residence are more tolerant of risk than those who migrate to regions that are culturally similar to their initial region of residence. This is true even after accounting for geographic distance between the two regions. Differences in linguistic dialects are used to proxy for cultural differences in their paper. If self-selection on the basis of risk tolerance, or other unobservable characteristics that are also predictive of migrating to more culturally distant areas, is driving the higher self-employment entry rates among internal migrants, then we should expect to see that those who migrate to more culturally distant regions of the U.S. will have a greater tendency to enter self-employment than those who migrate to culturally similar regions.

In the absence of large linguistic differences within the United States, I proxy for cultural differences across states using their voting tendencies in the 2016 and 2020 Presidential Elections. While the majority of the data is taken from before this time period, these voting patterns are expected to reflect longstanding cultural and political differences that pre-date these elections. I create a dummy variable $To {0.1cm} Different_{i}$, which takes on a value of 1 if an individual either migrates from a Republican-leaning state to a Democratic-leaning state, or from a Democratic-leaning state to a Republican-leaning state,⁸ and 0 otherwise. I also create a variable $To {0.1cm} Similar_{i}$, which takes on a value of 1 if an individual moves across states, but to a state with similar politics (either Democratic or Republican) as their initial state of residence.

These variables are added to the model in Eq. (3). It is worthwhile noting that this model includes controls for an individual’s current state of residence, which means that the effects that are captured by $To {0.1cm} Different_{i}$ and $To {0.1cm} Similar_{i}$ are being driven by migration, and not by characteristics of states of a given political leaning that impact the self-employment entry rate for all residents of that state. The results are presented in Column 2 of Table 5. It is clear that individuals who migrate to states with different political leanings from their initial state of residence are significantly more likely to become self-employed after migrating than those who migrate to states with similar political leanings. Furthermore, in Column 3 of Table 5, this is further broken down based on whether an individual migrated to a state that borders their initial state of residence. The results show that an individual who migrated to a state with a different political leaning from their initial state of residence is more likely to enter self-employment than an individual who migrated to a state with a similar political affiliation, regardless of whether their new state of residence borders their initial state of residence. These results are consistent with the idea that individuals who self-select into migrating to more culturally distant areas possess characteristics that are predictive of their likelihood of becoming self-employed.

5.2 Difficulty obtaining quality paid employment

Relative to the comparison between immigrants and natives, internal migrants and non-migrants are expected to possess similar characteristics. However, it is still possible that a typical internal migrant may face barriers to obtaining paid employment in their new state of residence. Such barriers may include regional discrimination, or employer uncertainty regarding the quality of one’s previous employers or education institutions. If such reasons lead to internal migrants being disproportionately likely to be “pushed” into self-employment relative to non-migrants in their new state of residence, then the higher entry rate into self-employment among internal migrants should be observed primarily among individuals who had been unemployed or employed in low quality jobs following migration.

In order to examine whether this is the case, I estimate the average recorded earnings within each occupation among all CPS respondents, and categorize respondents based on whether average earnings in the occupation that they worked in during their first month in the CPS were below or above the median across all respondents.⁹ The marginal effect of migration on self-employment entry is subsequently permitted to vary based upon the earnings category of their occupation in their first month in the CPS, which for migrants occurred post-migration.

The results from this estimation can be found in Columns 2 and 3 of Table 6. Internal migrants are more likely to enter self-employment than their non-migrant counterparts regardless of whether they worked in lower paying or higher paying occupations following migration. Furthermore, this effect is strongest among those internal migrants who found relatively high paying work post-migration, suggesting that difficulty obtaining quality paid work following migration is unlikely to be a primary driver of the higher self-employment entry rate among internal migrants.

Although all respondents in the sample are employed in their first month in the CPS, some respondents went through a period of unemployment in their second and/or third month in the CPS. These individuals may have been more likely to subsequently enter self-employment out of necessity, as per Fairlie and Fossen (2020). Therefore, I further allow the marginal effect to vary based upon whether an individual had such a recorded period of unemployment. The results from this can be found in Columns 4 and 5 of Table 6. Although the marginal effects of migration for those who were observed as unemployed are imprecise because of the relatively small sample size, the marginal effect among those who were never unemployed during the CPS period remains significant and only slightly attenuated in magnitude, suggesting that the higher rate of self-employment entry among internal migrants is unlikely to be driven by necessity self-employment that follows a period of unemployment.

5.3 Migrating for family/relationship reasons

Individuals who possess some form of family/relationship tie in their new area of residence are expected to be selected differently than migrants without such ties. Earlier work in the literature by Umblijs (2012) has shown that migrants to areas with a large number of co-ethnics are more risk averse than migrants to areas with a significant co-ethnic population. Although it is impractical to perform such an analysis in this paper due to the overwhelmingly white sample, since migrants are asked what their primary reason for moving was, I am able to determine whether an individual migrated for family and/or relationship reasons. Given that such individuals possess a strong tie in their new region of residence, selection based on characteristics that are related to self-employment entry (such as risk tolerance) is expected to be weaker for these individuals.

When allowing the marginal effect of migration to differ between individuals who migrated for family/relationship related reasons,¹⁰ and those who did not, the results are consistent with expectations. Column 2 of Table 7 shows that internal migrants with ties to their new state of residence are no more likely to enter self-employment than non-migrants. However, among those individuals who do possess such ties, the marginal effect is large and significant, as is shown in Column 3. This supports the idea that the results are primarily driven by self-selection¹¹.

5.4 Controlling for state of origin

It has been hypothesized that immigrants may be more likely to be self-employed than natives in part because they tend to disproportionately originate from countries with strong self-employment traditions. In order to address concerns that internal migrants tend to originate from states with high entry rates into self-employment, I examine whether the primary results change when controlling for an individual’s initial state of residence, instead of their current state of residence as is the case for the primary results that are estimated following Eq. (3). The marginal effects from this estimation can be found in Column 2 of Table 8. I fail to document any notable change in the results, indicating that internal migrants are not disproportionately drawn from states with high self-employment entry rates.

5.5 Lack of attachment to a particular employer

Individuals who have migrated are expected to have little to no ties with pre-migration employers following their move. This is especially the case for individuals who have migrated between non-bordering states. If individuals who lack strong ties with an employer are more likely to enter self-employment, then this could be driving the higher rates of self-employment entry among internal migrants. In order to address this, I take advantage of the fact that respondents to the March Supplement indicate the number of employers for whom they worked in the previous year. I add a series of dummy variables to the model in Eq. (3) which indicate whether a respondent worked for one employer, two employers, or three of more employers in the previous year. The base group consists of individuals who worked for no employer in the previous year. Note that the CPS information on how many employers an individual worked for in the previous year confusingly treats a period of self-employment as working for an employer. Therefore, the base group consists of individuals who did not work at all in the previous year, and does not include individuals who were self-employed.

If individuals who have migrated are unlikely to have continued working for a pre-migration employer after migrating, and if these internal migrants have a similar propensity to enter self-employment as non-migrants who have changed employers, controlling for the number of employers in the previous year should account for the difference in self-employment entry rates. The results, shown in Column 3 of Table 8, show that individuals who worked for a single employer in the previous year are indeed the least likely to enter self-employment, consistent with the idea that strong attachment to a particular employer reduces the probability of becoming self-employed. Individuals who worked for no employer or two employers have similar rates of entry into self-employment, while those who worked for three or more employers have the highest rate of entry into self-employment. After accounting for the number of employers for whom an individual worked in the previous year, I find that the higher entry rates into self-employment among migrants are attenuated, but remain statistically significant for those who migrated to non-bordering states.

5.6 Wealth

It is well established in the literature that wealthier individuals are more likely to become self-employed (Disney and Gathergood 2009; Simoes et al. 2016). If wealthier individuals are also more likely to migrate than less wealthy individuals, this could account for the result that internal migrants are more likely to enter self-employment than non-migrants. The CPS March Supplement asks respondents to report their previous year’s income from a wide variety of sources, including rental income, interest income, and dividend income. In order to account for wealth effects, I add controls for the natural log of each of these 3 different types of income (adjusted for inflation) into the model. The results in Column 4 of Table 8 suggest that dividend income and especially rental income are positive predictors of self-employment entry. Curiously, interest income negatively predicts self-employment entry. Although determining why this is the case is beyond the scope of this paper, it is plausible that higher levels of interest income are predictive of higher levels of risk aversion.

When these 3 types of income are controlled for, the results continue to suggest that migrants to non-bordering states are still more likely to become self-employed than non-migrants. This indicates that higher levels of wealth are unlikely to be simultaneously driving the migration and self-employment entry decision in a way that accounts for the higher self-employment propensity of internal migrants.

5.7 Native born parents

If internal migrants are predominantly 2nd generation Americans (those with at least one foreign born parent), and if 2nd generation Americans are more likely to become self-employed, then the results that are documented in this paper could plausibly be explained by cultural differences between 2nd generation Americans and 3rd generation (or more) Americans. In order to account for this, I include a control for whether an individual reports that both their mother and father were born in the United States. A variable called $Native {0.1cm} Parents_{i}$ is added to the model in Eq. (3), with the results found in Column 5 of Table 8. When this variable is added to the model, there is no meaningful change to the results. Furthermore, there is no evidence that individuals with two native born parents are any more or less likely to enter self-employment than individuals with at least one foreign-born parent.

5.8 Local business opportunities

If individuals who plan to enter self-employment have a tendency to migrate to locations where self-employment opportunities in their industry of interest are strongest, this could explain the relationship between migration and self-employment that is documented in this paper. In order to address this possibility, I estimate a model where the variable $Log {0.1cm} Business_{i}$ is included as a control. This variable consists of the inflation-adjusted average log non-farm business income in the previous year, among individuals who work in the same industry as individual i, and live in the same state and metropolitan area that individual i resides in¹².

The results from this estimation can be found in Column 2 of Table 9. It is clear that individuals are considerably more likely to enter self-employment when, in the previous year, average business income in one’s industry and location was high. This is consistent with individuals taking advantages of opportunities that are present within one’s local area. However, this only modestly attenuates the finding of higher self-employment entry rates among internal migrants. This is because internal migrants are only somewhat more likely to migrate to areas with strong self-employment opportunities in their eventual industry.

5.9 Local labor market conditions

Since the decision to become self-employed depends upon local labor market conditions, non-random migration of internal migrants to areas with favorable economic conditions for self-employment could account for the results. In order to address this, I include a variable that consists of the unemployment rate in one’s metropolitan area of current residence.¹³ From Column 3 of Table 9, it is clear that the local unemployment rate has only a weak negative influence on the probability that an individual enters self-employment, once their own unemployment history is accounted for. Furthermore, accounting for the local unemployment rate fails to explain the higher rates of self-employment entry among internal migrants.

5.10 Previous year’s wage/salary income

In order to account for the possibility that individuals with a history of low earnings in paid employment are both more likely to become self-employed out of necessity, and more likely to migrate, I add a control for an individual’s wage/salary income in the previous calendar year. This controls consists of the log of inflation adjusted income from wage and salary sources in the calendar year prior to their first appearance in the CPS. The results from this estimation can be found in Column 4 of Table 9. While it is clear that, as anticipated, individuals with higher wage/salary income in the previous year are less likely to enter self-employment, this only accounts for a modest portion of the higher self-employment entry rates among internal migrants.

5.11 Self-employment entry rates over a 15 month period

The primary results in this paper show that individuals are more likely to enter self-employment within 3 months of their first appearance in the CPS if had moved to their current area of residence from a non-bordering state within the preceding year. However, since CPS respondents are contacted as late as 15 months after their first appearance in the CPS, it is in theory possible to examine self-employment entry rates over a longer time horizon than 3 months. However, many CPS respondents only respond to the March supplementary questionnaire in their second 4 month period in the CPS, making it impossible to observe these individuals 15 months later. Furthermore, even among respondents who do respond to this supplementary questionnaire in their first 4 month period in the CPS, many do not continue to respond to the CPS after the 8 month gap. Due to these sources of attrition in the sample, it is only possible to follow 57% of the original sample over a 15 month period.

However, when the marginal effects of migration on self-employment entry are estimated for this sample, the results are qualitatively similar to the primary results in Table 4. Individuals who moved to a non-bordering state within the year preceding their first appearance in the CPS are more likely to enter self-employment within the following 15 months than non-movers are. The estimated difference in entry rates between these two groups is approximately 0.75 percentage points, after controlling for demographic and labor market characteristics. This indicates the higher self-employment entry rate among internal migrants is not restricted to the months immediately following migration but persists for more than year afterwards.

Table 5 Marginal effects of migration status on probability of self-employment entry (split by political leaning of state)

This table examines the primary results of interest when splitting internal migrants into those who have migrated to states with similar political leanings as their initial state of residence, and those who have migrated to states with different political leanings than their initial state of residence. Column 2 contains the results with variables including all individuals who have moved states, while Column 3 contains the results when the split by political leaning is also split based on whether one migrated to a bordering or non-bordering state.

Table 6 Marginal effects of migration status on probability of self-employment entry (based upon occupation earnings and unemployment status)

This table examines how the results in Table 4 change when allowing the marginal effect of migration on self-employment entry to vary based upon whether a respondent initially worked in an occupation with average earnings below the median (Column 1) or above the median (Column 2). The marginal effects are also permitted to vary on the basis of whether a respondent was never observed as unemployed during the CPS period (Column 3) or was observed as having been unemployed (Column 4).

Table 7 Marginal effects of migration status on probability of self-employment entry (split based on presence of family/relationship ties)

This table examines how the results in Table 4 change when allowing the marginal effect of migration on self-employment entry to vary based upon whether a respondent reports that they moved for non-Family/Relationship related reasons (Column 1) or for Family/Relationship related reasons (Column 2).

Table 8 Marginal effects of migration status on probability of self-employment entry (additional controls)

This table examines how the results in Table 4 change when accounting for a variety of possible confounding factors. In Column 2, an individual’s initial state of residence is controlled for, rather than their current state of residence. In Column 3, a control is added for the number of employers that an individual worked for in the previous year. In Column 4, controls are added for the natural log of one’s income from dividends, interest, and rent. In Column 5, a control is added indicating whether both of the respondent’s parents were born in the United States.

Table 9 Marginal effects of migration status on probability of self-employment entry (additional controls)

This table examines how the results in Table 4 change when accounting for a variety of possible confounding factors. In Column 2, a control is added for the mean log business income in one’s industry and location in the previous year. In Column 3, a control is added for the unemployment rate in one’s area of current residence. In Column 4, a control is added for an individual’s own log wage/salary income in the previous year (Table 10).

Table 10 Marginal effects of migration status on the probability of self-employment entry (results over 15 months)

This table examines shows the marginal effects estimated after Eq. (3) when individuals are recorded as having entered self-employment if they became self-employed between their first month in the CPS and the 8 month in the CPS, which occurs 15 months after their first appearance. This sample contains only respondents who can be followed up 15 months after their first appearance.

6 Conclusion

Migrants have long been documented as being more likely to be self-employed than non-migrants. In this paper, I provide evidence that this also holds true among internal migrants within the United States. Controlling for a wide array of characteristics suggests that this finding is attributable to self-selection. That is, natives who self-select into migrating from one state to another possess characteristics that positively predict their entry into self-employment.

Although this paper makes the case that the higher self-employment entry rate of migrants is at least in part attributable to self-selection, due to the absence of appropriate variables, I am not able to determine conclusively which unobserved characteristics are driving both the migration decision and the self-employment entry decision. Although previous literature suggests that risk tolerance is a likely contender, empirical research into which factors drive both migration and self-employment is a valuable avenue of future research.

Furthermore, given that the internal migrants in this paper have resided in their new state of residence for at most 15 months, I am unable to say whether their higher rate of entry into self-employment persists over longer periods of time. Furthermore, given the limited window of time, it is impossible to observe whether internal migrants who enter self-employment tend to remain self-employed for a long period of time, or whether self-employment tends to be a transitory state for these individuals. Among immigrants, exit rates from self-employment are higher than is the case among natives (Dheer 2018). Given the limited time frame of the CPS data, it is impractical to test whether this is the case for internal migrants.

Despite the above shortcomings, the primary contribution of this paper to the literature on immigration and self-employment is the fact that this paper examines two groups of individuals who are very similar to one another, with the exception of the decision to migrate. Internal migrants are more likely to enter self-employment than non-migrants, with this effect being stronger when they move larger distances. The importance of these findings comes primarily from the fact that this suggests that even if immigrants did not face discrimination in the paid employment sector, imperfect recognition of their credentials by prospective employers, or if they did not live in ethnic enclaves, we may still continue to observe higher self-employment rates among immigrants than among natives due to non-random selection into migrating.

Appendix

See Table 11.

Table 11 Marginal effects of migration status on probability of self-employment entry (split by reason for migration and mean occupation earnings)