Great Migration’s great return? An examination of second-generation return migration to the South

Abstract

Using novel panel data spanning 1940–2000, we examine the adult offspring of the Great Migration who returned to the South. We observe two types of return migrants: (1) southern-born, “lifetime” return migrants who were born in the South, resided outside of the South in 1940, and returned to the South by 2000, and (2) northern-born, “generational” return migrants whose parents were born in the South but who, themselves, were born in the North, resided in the North in 1940, and had returned to the South by 2000. These data also allow us to observe return migrants and their parents over a longer period of time than any previous data source, permitting us to consider the early-life predictors of return migration. We find that generational migrants comprise a sizeable segment of all second-generation return migrants to the South and that these migrants are positively selected on their own and their parents’ socioeconomic characteristics, relative to the second-generation migrants who remain in the North. Conversely, southern-born, lifetime, return migrants are negatively selected. Our investigation provides a broader and more representative view of who return migrants are and illustrates the underappreciated importance of generational ties to place for migration decisions.

1. Introduction

Through at least the first half of the twentieth century the demographic and economic landscape of the American South told a story of regional decline. The South’s economy relied on a declining agricultural sector and outdated modes of production compared to other regions in the United States (Mandle, 1992); its educational system was denigrated by critics for inadequate investment and poor results (e.g., Anderson, 1989; Margo 1990); and it experienced the exodus of millions of Black and White residents during the Great Migration (Berry, 2000; Fligstein, 1981; Gregory, 1989, 2005; Grossman, 1989; Lemann, 1991; Marks, 1989; Spring et al., 2016; Tolnay, 2003; U.S. Bureau of Education, 1916; Wilkerson, 2010).

Conditions changed dramatically in the latter half of the 20th century. By the 1950s, the South began to experience positive net in-migration of Whites, a change that accelerated in the 1960s (Falk et al., 2004; Frey, 2004; Fuguitt et al., 2001; Hunt et al., 2008). The Great Migration concluded somewhat earlier for Whites than for Blacks. However, by the 1970s, the South was recording net in-migration of Blacks as well. These transformations were due, in part, to increasing economic opportunities in the South and to improving racial relations as de jure Jim Crow era barriers were weakened and more political, social, and economic opportunities opened up for Blacks. Some individuals with no previous ties to the region migrated to the South to capitalize on new job opportunities, the milder climate, and other amenities (Frey, 2004; Hunt et al., 2008). Another group of migrants, who were born in the South or had relatives in the South, was also migrating South (Sharkey, 2015). For this group of migrants, family ties and cultural roots likely influenced the decision to move (Brown, 2017a; Stack, 1996). Together, these two groups of migrants—primary and return—have contributed to the South’s transition to a growing, economically dynamic region.

Previous research has examined who these primary and return migrants were and where they headed (Sharkey, 2015). However, because of data limitations, we have little knowledge of the extent to which return migration to the South is influenced by generational ties to the South. This is in part because previous research has often used census questions concerning respondents’ birthplaces, residences five years prior to the year of census enumeration, and residences as of census enumeration in order to identify return migrants (Adelman et al., 2000; Smith et al., 1992). Using this information, return migrants are defined as individuals who are born in the South, living outside of the South five years prior to the census, and living in the South as of census enumeration. However, these questions preclude any investigation of ties to the South that existed prior to an individual’s birth, such as the potential influence of parents’ birthplaces, and they limit analyses of return migrants to individuals who moved back to the South in the very recent past. Given that migration experiences, particularly recent migration experiences, are likely to influence the characteristics of migrants, such as their employment probabilities and incomes, such data may provide an incomplete or even biased picture of who return migrants are and how they affect southern society. Likewise, by failing to identify migrants who have generational ties to the South, we may be underestimating the contributions of return migrants to the South and its economic and demographic changes. The Great Migration and the subsequent return migration of Blacks and Whites to the South represent two of the most demographically important internal migration streams in the United States (Frey, 2004; Tolnay, 2003). By underestimating the size of return migration streams and failing to understand how generational ties to the South influence return migration, we miss the important opportunity to theorize about the formation of migration streams and the motives and characteristics of migrants.

Using novel, linked 1940-to-2000 Census data, we are able to identify the birthplaces of individuals and their parents, as well as their childhood residences in 1940 and their late-adulthood residences in 2000. These data allow us to distinguish between two different types of second-generation return migrants born between 1922 and 1940 who returned to the South¹ by the year 2000: (1) southern-born return migrants, that is, those who were born in the South, resided outside of the South in 1940, and returned to the South by 2000 and who we refer to throughout the paper as “lifetime” return migrants (because the return migration happens within their individual lifetimes), and (2) northern-born return migrants, whose parents were born in the South but who, themselves, were born in the North, resided in the North in 1940, and had returned to the South by 2000.² We refer to this latter group as “generational” return migrants because they are returning to an intergenerational homeland. This is an important distinction because traditional census-based definitions of “return migrants” would not include the “generational” group, thereby preventing us from examining whether they too are drawn to their ancestral homes. Given that the immigration literature has found that second-generation migrants frequently exhibit ties to ancestral homelands that motivate them to return migrate (King and Christou 2011), it is possible that this holds for second-generation Great Migration migrants as well, though we do not know if this is the case. In summary, the three migration status groups that comprise the basis for the comparisons made throughout the remainder of the manuscript are defined as follows:

Non-return migrants	Parent birthplace: South (either parent)
	Child birthplace: South/non-South
	Residence in 1940: Non-South
	Residence in 2000: Non-South
Lifetime Return/Southern-born	Parent birthplace: South (either parent)
	Child birthplace: South
	Residence in 1940: Non-South
	Residence in 2000: South
Generational Return/Northern-born	Parent birthplace: South (either parent)
	Child birthplace: Non-South
	Residence in 1940: Non-South
	Residence in 2000: South

Throughout our investigation we conduct separate analyses for Black and White return migrants, allowing us to examine whether race differentially structures these processes and relationships. Specifically, we seek answers to the following questions:

• How prevalent was the phenomenon of generational return migration among the northern-born offspring of Great Migration migrants?

• How do the characteristics and level of selectivity of lifetime return migrants and generational return migrants vary relative to each other and relative to second-generation Great Migration migrants who remained in the North?

• Do return migrants tend to settle in their or their parents’ birthplaces, and how does this phenomenon differ depending on whether one is a lifetime return migrant or a generational return migrant?

Our results indicate that roughly 9% of northern-born second-generation migrants (i.e., generational migrants) in 1940 returned to the South by 2000, compared to 15% of southern-born second-generation migrants (i.e., lifetime migrants). Further, we find that generational return migrants are more positively selected on their own and their parents’ socioeconomic characteristics than are lifetime return migrants and second-generation Great Migration migrants who remain in the North (i.e., second-generation non-return migrants). Lifetime return migrants are negatively selected on these characteristics relative to second-generation Great Migration migrants who remain in the North.

For both groups of return migrants, returning to one’s or one’s parents’ birth state is common, though it is particularly likely among lifetime return migrants. Moreover, those who return to their own or their parents’ birth state tend to be more negatively selected than return migrants who migrate to other areas within the South, which offers support for common theoretical assumptions about the characteristics of migrants who are motivated to move for family-related reasons compared to migrants who are motivated to move for other, often economic, reasons. These results hold generally for both Black and White return migrants.

This study is the first to definitively identify generational and lifetime return migrants and to capture migrants who returned to their homeland at almost any point throughout their adulthood. Our study is also unique in being able to account for both childhood and adulthood characteristics influencing the probability of return migration, thereby limiting the extent to which the characteristics of the migrants we observe are biased by recent migration experiences and allowing us to examine how early-life experiences are associated with later-life outcomes. This is also the first study we know of to observe return migrants from 1940 to 2000. This is a particularly informative span of time, including an initial period during which net out-migration from the South was substantial, a subsequent era during which net-migration into the South reversed from negative to positive and, finally, a time during which migration into the South accelerated in the latter part of the twentieth century. These unique characteristics of our data and analysis allow us to better understand whether and how migrants to the South are influenced by generational family ties to the South and whether, as is commonly hypothesized, return migrants are negatively selected. We therefore offer important contributions to theories of return migration and internal migration streams more broadly. We also offer unique insights into an important group of migrants who are participating in the transformation of the South from a declining, migrant-sending region, into a dynamic, migrantreceiving region.

2. The Great Migration and return migration to the U.S. South

The increase of return and primary migration to the South represents a dramatic turnaround from the situation that characterized the first half of the 20th century. Prior to 1960, millions of Black and White southerners migrated to the North, leading to a systematic decline in the southern population. This flight of Black and White southerners was a reaction to a variety of social and economic push factors, including agricultural transformations that produced a surplus rural labor force (Heinicke, 1994; Lemann, 1991; Mandle, 1992), an underdeveloped industrial sector, and, for Blacks, the persistence of oppressive racial conditions. Tenant farming, one of the dominant occupations among African Americans in the South, was little better than indentured servitude and provided few opportunities for economic security, let alone advancement (Mandle, 1992; Ransom and Sutch, 2001; Tolnay, 1999). Southern towns were hardly more promising, offering few nonagricultural employment options to unskilled and unschooled workers, particularly African Americans (Daniel 1972; Kirby 1987; Mandle, 1978, 1992; Wilkerson, 2010).

In comparison, the social and economic climate in the North appeared inviting. Industrial production and economic opportunity soared in response to World War I. The influx of international immigrants into the North was halted by the war’s hostilities in Europe and, subsequently, by restrictive federal immigration legislation.³ As a result, competition for jobs declined, opening up a multitude of employment opportunities to southerners willing to migrate (Boustan, 2016; Collins, 1997; Tolnay, 2003; Wilkerson, 2010). Indeed, scholars argue that the availability of white immigrant labor and discriminatory hiring practices delayed the black Great Migration by decades (Collins, 1997). Additionally, at the beginning of the Great Migration, African Americans were overwhelmingly concentrated in the South—over 90% of the population as late as 1910 (Tolnay, 2003). As a consequence, southern Blacks stood to benefit considerably less from an established network of family and friends in northern cities relative to Whites, who were more geographically dispersed. These weaker network ties retarded the flow of information about employment opportunities in the North and also reduced the support available to Black southern migrants upon their arrival in northern destinations in the initial years of the Great Migration. Labor agents helped to partially fill this gap by actively recruiting African American workers in the South, despite the steep licensing fees imposed by southern jurisdictions and the constant threat of arrest and intimidation to which they were exposed (Bodnar et al., 1982; Brown, 2017b). Nevertheless, these factors may help explain why White migration out of the South started earlier, with distinct southern White out-migration streams occurring as early as the post-Civil War period (Gregory, 2005).

A second surge in migration to the North occurred during World War II as factories geared-up for another war effort, creating renewed demand for workers in the North and West. In addition to pursuing economic opportunities in the North, southerners, particularly Black southerners, were fleeing the stultifying Jim Crow restrictions in the South, which included severe racial segregation, the disenfranchisement of black voters, mob violence and periodic lynchings (Tolnay and Beck, 1992). The southern racial state remained largely intact until the successes of the Civil Rights Movement in the 1950s and 1960s. It is therefore no surprise that the South struggled to maintain its African American population for decades.

Many southerners who migrated north maintained ties to the South. Black and White southerners frequently returned home for brief periods of time, sent their children south to be raised by grandparents and other relatives, and often dreamed of returning to the South once the social and/or economic conditions became more favorable (Alexander, 2005; Falk, 2004; Stack, 1996). In particular, White southerners were known for their tenuous ties to the North and the frequency with which they returned home on weekends and during the summer to care for their farms and families (Alexander, 2005; Gregory, 2005). This phenomenon, which Berry (2000) describes as White southern migrants having a “divided heart,” speaks to Lee (1966) and Ravenstein’s (1885) foundational theories of migration, which, among other precepts, hypothesize that for every migration stream, there is a counterstream. For many southern Black migrants, their southern roots held less appeal. The oppressive racial conditions and very real threat of physical violence that prevailed through the first half of the 20th Century lent some sense of a “refugee movement” to their exodus from the South (see, for example, Ballard, 1984; Brown, 2017b).

As the 20th century progressed, outmigration from the South slowed, and was increasingly balanced by a growing counterstream of in-migrants, eventually giving way to net in-migration of Whites in the 1960s and net in-migration of Blacks by the 1970s (Falk et al., 2004; Frey, 2004; Fuguitt et al., 2001; Gregory, 2005; Hunt et al., 2008; Tolnay, 2003). These migration streams consist of both primary migrants—migrants to the South with no previous residential ties to the area—and return migrants—migrants returning to a region where they or, in some cases, their close relatives previously lived.

This reversal from net out-migration to net in-migration has been credited to increasing income levels and job opportunities in the South as the North began to experience deindustrialization (Frey and Speare, 1988; Kasarda, 1980; Long, 1988), an improving southern racial climate (Gregory, 2005; Tolnay, 2003; Wilkerson, 2010), increasing political-representation of southern Blacks (Dreyfuss, 1977), and to disillusion with northern inner-cities and the dysfunction many people began to associate with them (DaVanzo and Morrison, 1981; Falk, 2004; Massey and Denton 1993; Stack, 1996). Economic opportunities in the South also became increasingly plentiful as companies migrated South to avoid the unionization of their employees and to take advantage of southern states’ “Right to Work” policies (Newman, 1983). Likewise, decreases in transportation costs (Frey, 1987; Wardwell, 1980), innovations in technology and employment that enhanced the mobility of firms and the spatial diffusion of workers (Frey, 1987), and the implementation of NAFTA all worked to make the South an increasingly economically viable option for companies and, thereby, a magnet for workers.

Additionally, ethnographic work has found that return migrants to the South often cite the importance of ties to place and cultural roots in their decision to return to the region (Brown, 2017a; Brown and Cromartie, 2006; Falk, 2004; Stack, 1996; Smith et al., 1992). Indeed, as mentioned above, Whites’ periodic returns to the South and their “divided hearts” likely prompted their earlier return migration to the South relative to southern Blacks. The earlier start to southern White out-migration mentioned above (Gregory, 2005) and the advances in opportunities for Blacks as a result of the Civil Rights Movement in the 1950s and 1960s may also be important explanatory factors behind the later start to return migration streams among African Americans.

3. Data and methodology

We use newly constructed data linking individuals from the complete-count 1940 Census to the 2000 Census Long Form to investigate return migration to the South by the adult progeny of parents who left the region before 1940. Recently fully-digitized and released, the full-count 1940 Census is the first fully-released historical census containing individuals who are young enough at the time of the historical census to be linked to the 2000 Census and the 2010 Census, which are the only two linkable contemporary censuses currently available.⁴ Linkage of the 1940 Census to the 2000 Census enables detailed, careful analyses of internal migration in the U.S. as it identifies place of residence at birth and in 1935, 1940, 1995, and 2000.

By linking the 1940 Census to the 2000 Census, we are able to establish parent-child relationships for observations in the 2000 Census by utilizing the 1940 Census’s “Relationship to the household head” question, which determines whether respondents were the household head, the spouse of the household head, or child of the household head. Therefore, through use of the 1940 Census, we observe intergenerational migration information and are able to observe both characteristics of children’s parents and their childhood household. Since we focus on return migrants who could be linked to the 2000 Census, the average individual in our analysis is between 65 and 70 years old in 2000. Therefore, our analyses and the evidence they yield pertain to the older adult population in 2000, and appropriate caution must be exercised when extrapolating our findings beyond this age group.

The linkage of the 1940 and 2000 Censuses relied on the record linkage infrastructure developed at the U.S. Census Bureau’s Center for Administrative Records Research and Applications (CARRA). CARRA supports a linkage infrastructure to facilitate the agency’s use of administrative records. The 1940, 2000 Censuses were both processed using a set of programs and tools called the Person Identification Validation System (PVS). The PVS appends unique Protected Identification Keys (PIKs) to census records and other files using a probabilistic matching algorithm (Fellegi and Sunter, 1969). This algorithm compares the characteristics of records in a given census to the characteristics of records in administrative data including the Social Security Administration’s (SSA) Numident file and Internal Revenue Service (IRS) 1040 tax returns. These characteristics include full name, age, state of birth, and parents’ names for children in the 1940 Census (see Massey et al., forthcoming for a detailed description of the matching process) and full name, full date of birth, and street address for the 2000 Census. If the characteristics of a census record and an administrative record sufficiently agree, the PVS assigns the PIK from the administrative data to the census record. Because the PIKs correspond one-to-one with Social Security Numbers in the administrative data, and therefore uniquely identify a particular person, they can be used to link records between any files that have been processed using PVS.⁵

We limit our sample to individuals 18 years or younger (as of the 1940 Census) who lived with their parents in 1940 and who could be linked to the 2000 Census.⁶ Among those cases, we further limit the sample to individuals who are Black or White and born in the U.S. They must also have native-born parents, reported income and education in 2000, at least one parent who was born in the South, and a consistent race response in 1940 and 2000.⁷ They must also have moved to the North by 1940 in order to identify those who had the potential to return to the South. These sample restrictions reduced the number of linked cases used in our analysis to 17,212 Black respondents and 125,550 White respondents.

Respondents in our sample were grouped into one of three categories for the bivariate and multivariate analyses: (1) second-generation, non-return migrants (southern- or northern-born) whose parents were born in the South but had migrated to the North by 1940, and who remained in the North in 2000, (2) “lifetime return migrants” who were born in the South, lived in the North in 1940, and were located in the South in 2000, and (3) “generational return migrants” whose parents were born in the South but who themselves were born in the North, who were located in the North in 1940, and who had moved to the South by 2000. Without the benefit of the linked data, generational return migrants would be considered “primary migrants” using traditional definitions applied to cross-sectional census data and many of the lifetime migrants who returned prior to 1995 would be missed entirely.

As noted, by the time of the 2000 census, the second-generation Great Migration participants that form the basis for our study were in their 60s and 70s. This places many of them at a point in the life course where they are no longer economically active, and some likely had experienced retirement-related migration. In addition, by these ages, most of the individuals in our analytic sample have adult children as well as grandchildren. While we have access to some information that captures earlier stages in their lives (e.g., a variety of their parents’ and own characteristics in 1940 and their completed level of educational attainment), the linked 1940–2000 Census data do not allow us to capture other elements of individuals’ childhood, adolescence, and intervening adult years that may be valuable for studying their migration experiences. Eventually, a successful linkage for intervening census years for these individuals will help remedy this issue and yield further insights into the earlier life course stages of migrants (see footnote #4).

To assess the relationships between second-generation return migrant status and socioeconomic characteristics, we conduct bivariate and multivariate analyses separately for Black and White offspring of Great Migration participants.⁸ Table 1 reports the number of observations within each migrant and racial group. We observe 15,450 Black second-generation non-return migrants, comprising approximately 89.7 percent of our African American sample. We also observe 389 Black lifetime return migrants and 1373 Black generational return migrants. Among Whites, we observe 101,669 second-generation non-return migrants, 6620 lifetime return migrants, and 17,261 generational return migrants. Consequently, the likelihood of returning to the South by 2000 was considerable for second-generation migrants born in the South (15.0% and 30.9% for Blacks and Whites, respectively) as well as those born in the North (9.4% and 16.5% for Blacks and Whites, respectively). In our sample of second-generation migrants who resided in the North in 1940, the number of generational return migrants (1373 and 17,212 for Blacks and Whites, respectively) actually exceeds substantially the number of lifetime return migrants (389 and 6622 for Blacks and Whites, respectively). We therefore would miss a large portion of return migrants by neglecting generational ties to the South. However, any conclusions about the relative dominance of either migration stream (i.e., lifetime or generational) must be drawn with caution, given the different processes responsible for determining the population of second-generation migrants at risk of returning to the South for the two groups.⁹

Table 1.

Count and proportions of return and non-return migrants by race and birthplace.

Migration Status	Black		White		Total
	N	%	N	%	N	%
Non-return migrants	15,450	89.7	101,669	81.0	117,119	82.0
Lifetime return migrants	389	2.3	6620	5.3	7009	4.9
Generational return migrants	1373	8.0	17,261	13.7	18,634	13.1
Total	17,212	100.0	125,550	100.0	142,762	100.0

Source: 1940 Census data linked to 2000 Census data.

Notes: At least one parent was born in the South for each of the three migrant groups. Non-return migrants include those born in the South who resided outside the South in 1940 and 2000 (southern-born, non-return migrants) and those whose parents were born in the South but who themselves were born in the North and lived in the North in 1940 and 2000 (northern-born, non-return migrants). Lifetime return migrants include those who were born in the South, living outside of the South in 1940, and resided in the South in 2000. Generational return migrants had parents who were born in the South but were themselves born outside the South and living outside of the South in 1940, but resided in the South in 2000.

To observe the characteristics that are associated with each group and whether lifetime and generational return migrants are positively or negatively selected relative to non-return migrants, we use multinomial logistic regression, which allows us to examine the odds of belonging to either group of return migrants (lifetime or generational) relative to the odds of belonging to the group of second-generation non-return migrants (reference group). We also observe the frequency with which both groups of return migrants return either to their own birth state or their parents’ birth state and, using binary logistic regression, examine what factors are associated with returning to one’s or one’s parents’ state of birth conditional on return migrating. It should be emphasized that our analytic strategy is meant to examine the characteristics associated with return and non-return migrants and to describe these return migration streams. Our analysis is not meant to offer strong causal evidence or to be used to draw causal claims, however, we include pre-migration controls of the parents and childhood household in an attempt to control for selection into migration.

In our models we include a host of individual- and family-level covariates measured in 1940 and 2000, permitting us to examine the individual and familial characteristics associated with belonging to either return migrant group and to the group of migrants who return to their own or their parents’ birth state. Specifically, we control for a number of the respondent’s characteristics in 2000, when we observe the respondent in late adulthood, including the respondent’s age, residence in a metropolitan statistical area (MSA),¹⁰ gender, and education.¹¹ We also include logged income in 1999 in the binary logistic regressions evaluating the odds of returning to one’s or one’s parents’ birth state. Additionally, we control for a number of characteristics of the respondent’s parents in 1940, when the respondent was a co-resident child. These 1940 covariates include residence in a metro area, parental homeownership, the highest grade achieved by either parent, and the highest occupational status exhibited by either parent.¹² We present the descriptive statistics for the 1940 and 2000 covariates in Table 2.

Table 2.

Demographic Characteristics by Return or non-Return Migration Category in 1940 and 2000.

	(1)	(2)	(3)	(4)	(5)	(6)
	Black, lifetime return migrant	White, lifetime return migrant	Black, generational return migrant	White, generational return migrant	Black, Non-return migrant	White, Non-return migrant
	mean/sd	mean/sd	mean/sd	mean/sd	mean/sd	mean/sd
Age, 1940	11.02 (4.95)	8.97 (4.88)	6.70 (4.80)	7.11 (5.11)	8.01 (5.10)	7.99 (5.18)
Age, 2000	70.83 (4.96)	68.91 (4.88)	66.63 (4.80)	67.06 (5.12)	71.36 (4.84)	70.00 (4.82)
Male	0.42 (0.49)	0.48 (0.50)	0.48 (0.50)	0.50 (0.50)	0.41 (0.49)	0.47 (0.50)
Metro area, 1940	0.86 (0.34)	0.45 (0.50)	0.88 (0.32)	0.56 (0.50)	0.88 (0.33)	0.47 (0.50)
Owned home, 1940	0.09 (0.28)	0.21 (0.41)	0.17 (0.38)	0.36 (0.48)	0.17 (0.38)	0.38 (0.49)
Highest parent grade, 1940	7.84 (3.34)	9.84 (3.14)	9.04 (3.05)	10.56 (3.01)	8.56 (2.96)	10.02 (2.88)
Father’s earnings, 1940	651.41 (486.04)	1128.92 (1089.59)	834.66 (593.79)	1280.48 (1130.82)	780.38 (553.85)	1049.09 (1015.14)
Mother’s earnings, 1940	132.08 (264.39)	47.21 (202.36)	63.09 (197.01)	50.70 (274.36)	56.25 (191.99)	43.96 (212.12)
Parent occupational score, 1940	2.86 (0.47)	3.06 (0.47)	3.00 (0.33)	3.20 (0.42)	2.95 (0.38)	3.13 (0.44)
Married, 2000	0.48 (0.50)	0.69 (0.46)	0.55 (0.50)	0.71 (0.45)	0.41 (0.49)	0.67 (0.47)
Years of Education (2000	11.54 (3.71)	12.55 (2.99)	13.30 (2.96)	13.31 (2.69)	12.05 (2.85)	12.24 (2.70)
Total personal income, 2000	24420.89 (47574.26)	31114.63 (55148.44)	32658.42 (42544.13)	35872.12 (63696.61)	24220.95 (39298.42)	28121.03 (51670.76)
Not MSA, 2000	0.24 (0.43)	0.38 (0.49)	0.15 (0.35)	0.28 (0.45)	0.03 (0.16)	0.30 (0.46)
Migrated in past 5 years, 2000	0.22 (0.42)	0.25 (0.43)	0.30 (0.46)	0.30 (0.46)	0.19 (0.39)	0.21 (0.41)
Observations	389	6620	1373	17261	15450	101669

Source: 1940 Census data linked to 2000 Census data.

Notes: At least one parent was born in the South for each of the three migrant groups. Non-return migrants include those born in the South who resided outside the South in 1940 and 2000 (southern-born, non-return migrants) and those whose parents were born in the South but who themselves were born in the North and lived in the North in 1940 and 2000 (northern-born, non-return migrants). Lifetime return migrants include those who were born in the South, living outside of the South in 1940, and residing in the South in 2000. Generational return migrants had parents who were born in the South but were themselves born outside the South and living outside of the South in 1940, but residing in the South in 2000.

The descriptive statistics indicate that both Black and White generational return migrants tend to be more advantaged on average than non-return migrants. Specifically, generational (northern-born) return migrants have higher levels of education and higher incomes than either non-return or lifetime (southern-born) return migrants. Their parents also tended to have higher incomes and higher levels of educational attainment in 1940. Lifetime return migrants, in contrast, tend to be somewhat disadvantaged relative to non-return migrants. They are less likely to have parents who owned their own home in 1940 and have lower levels of education than either generational return migrants or non-return migrants. The fathers of Black lifetime return migrants also tended to earn less than the fathers of generational return migrants and non-return migrants.

These bivariate comparisons provide suggestive evidence that focusing solely on lifetime return migrants produces a misleading picture of return migrants. Distinguishing between generational and lifetime return migrants reveals contrasting sociodemographic profiles that are obscured when second-generation familial ties to the South are not considered.

4. Multivariate evidence

4.1. Assessing the selectivity of return migration

While the descriptive evidence reported in Table 2 is suggestive, multivariate analyses allow for a more definitive assessment of the factors that are most closely associated with belonging to either return migrant group and, by extension, the potential selectivity associated with each group of return migrants. Table 3 reports the results from the multinomial logistic regression analyses determining the odds of belonging to the group of second-generation Black southern-born lifetime return migrants (Column 1) and the group of second-generation Black northern-born, generational return migrants (Column 2), relative to belonging to the group of second-generation, non-return Great Migration migrants (reference group). A value greater than 1.0 for a coefficient indicates that a given characteristic is more common, or has a higher value, for the given return migrant group than for the reference group. A value less than 1.0 suggests the opposite relative magnitude for the variable.¹³

Table 3.

Results from multinomial logistic regression analysis (odds ratios) of membership in migration history groups for Blacks. Non-return migrants serve as the reference group.

	Lifetime return migrants	Generational return migrants
	(odds ratio)	(odds ratio)
Male	0.877 (0.0979)	1.014 (0.0596)
Age	1.115** (0.0118)	0.957**a (0.00555)
Metro Area, 1940	0.879 (0.136)	1.016 (0.0912)
Owned Home, 1940	0.432** (0.0793)	0.972a (0.0751)
Highest Parent Grade	0.968 (0.0199)	1.023*a (0.0105)
Highest parent occupation score	0.691** (0.0824)	1.013a (0.0824)
Education	0.929** (0.0196)	1.102**a (0.0141)
Married	1.409** (0.156)	1.492** (0.0880)
Constant	0.000** (0.000)	0.301* (0.151)
Observations	17,212	17,212

^**p < 0.01

^*p < 0.05.

Source: 1940 Census data linked to 2000 Census data.

Notes: Standard errors are in parentheses.

^aDifferences between the coefficients for northern- and southern-born return migrants are significant (p < 0.05).

Examining the results in Column 1 of Table 3 for Black lifetime return migrants, we see that, consistent with the descriptive evidence, this group is somewhat socioeconomically disadvantaged relative to non-return migrants. Parental homeownership in 1940 is negatively associated with the probability of belonging to the group of lifetime return migrants, as is parental education and parental occupational status, though parental education is not statistically significant. Own years of education also decreases the probability of belonging to this group. We also find that lifetime return migrants tend to be older and are more likely to be married than are non-return migrants.

Turning to the results for Black generational return migrants in Column 2, we see substantially different patterns, with generational return migrants more likely to be socioeconomically advantaged relative to non-return migrants, as was found in the descriptive results. Specifically, parental and own education are positively associated with belonging to the group of generational return migrants, as is parental occupation score, though the latter advantage is not statistically significant. Furthermore, generational return migrants exhibit higher levels of education (parental and own) and higher parental occupation scores than lifetime return migrants (as indicated by the symbol †). This group is also much more likely to be married and tends to be younger than non-return migrants.

The corresponding results for Whites are presented in Table 4. The multivariate results are largely consistent with both the descriptive evidence for Whites described earlier and the multivariate results presented for Blacks.

Table 4.

Results from multinomial logistic regression analysis (odds ratios) of membership in migration history groups for Whites. Non-return migrants serve as the reference group.

	Lifetime return migrants	Generational return migrants
	(odds ratio)	(odds ratio)
Male	0.973 (0.0256)	1.004 (0.0172)
Age	1.049** (0.00245)	0.972**a (0.00164)
Metro Area, 1940	0.916** (0.0253)	1.267**a (0.0221)
Owned Home, 1940	0.393** (0.0124)	0.913**a (0.0160)
Highest Parent Grade	1.003 (0.00544)	1.011** (0.00337)
Highest parent occupation score	0.886** (0.0291)	1.339**a (0.0288)
Education	1.009 (0.00602)	1.069**a (0.00393)
Married	1.075* (0.0307)	1.088** (0.0206)
Constant	0.004** (0.001)	0.152** (0.021)
Observations	125,550	125,550

^**p < 0.01

^*p < 0.05.

Source: 1940 Census data linked to 2000 Census data.

Notes: Standard errors are in parentheses.

^aDifferences between the coefficients for northern- and southern-born return migrants are significant (p < 0.05).

White lifetime return migrants report lower parental occupation scores and a lower likelihood of home ownership in 1940 than non-return migrants (column 1), paralleling our findings of lower early-life socioeconomic statuses among Black lifetime return migrants. In contrast, white generational return migrants evince higher levels of education (parental and own) and higher parental occupation scores than do non-return migrants (column 2), though they were less likely than non-return migrants to have lived in an owned home in 1940. Additionally, White generational return migrants are characterized by own education levels and parental occupation scores that are superior to those for lifetime return migrants, and they were more likely to reside in an owned home in 1940 relative to lifetime return migrants (as indicated by the symbol †).

In light of the fact that we are observing these second-generation Great Migration participants relatively late in the life course, we also examined the possibility that differential proportions of recent return migrants (i.e., those who moved within the last five years) across the migration history groups are responsible for the generally negative socioeconomic profile of lifetime return migrants and the more positive profile for generational return migrants. Supplementary analyses that included an additional covariate for recency of migration (migrated within the last 5 years) and its multiplicative interaction with parents’ education and occupation score, as well as the respondent’s own education, had little consequence for our major substantive conclusions. While Black recent return migrants appear to be slightly positively selected relative to more distant return migrants, the relationships reported earlier remain significant for more distant return migrants and there are no other significant differences between recent and more distant return migrants, with the exceptions that recent Black generational return migrants exhibit slightly higher parental occupational scores and that recent Black lifetime return migrants exhibit higher levels of parental education than more distant return migrants. For Whites, the predictors of migration status are virtually identical for those who migrated within the last five years and those who migrated earlier.¹⁴

These findings have important implications for how we understand return migration to the South as well as the relative levels of human capital that different migration groups brought with them upon their return. We find that, overall, return migrants with generational ties to the South are positively selected. Therefore, focusing only on lifetime (southern-born) return migrants, as is typically done, yields a misleading picture of return migrants. Doing so paints an overly negative picture of the potential human capital contributions these migrants are bringing with them and misses a significant segment of individuals who are returning to the South.

4.2. Assessing the likelihood of returning to a southern “homeland.”

An important question regarding the motivations of these return migrants is whether they are returning to the South because of perceived economic opportunities or because they are being drawn back to a childhood or generational “homeland.” These motivations are not mutually exclusive and could easily operate simultaneously. To be sure, in the absence of detailed information concerning the motivations of migrants, we are unable to definitively determine what factors are predominantly drawing migrants to the South. Nevertheless, by examining the rates at which return migrants to the South reside in their own or their parents’ state of birth we can obtain an idea, albeit an indirect and imprecise one, of what proportion of migrants might be particularly likely to be drawn “home” and what proportion are perhaps more likely to be returning South for other economic or social reasons. Additionally, by examining the factors associated with the likelihood of returning home, we can determine whether, as previous authors have contended, return migrants who return to their state of birth are more negatively selected relative to other return migrants and migrants to the South more generally. Further, our data allow us to examine whether the potential negative selection of migrants who return to their birth state also applies to migrants who return to their parents’ state of birth.

Table 5 shows the proportion of lifetime migrants who return to their own or their parents’ birth state and the proportion of generational return migrants who return to their parents’ birth state. Recall that the generational return migrants, themselves, were born in the North. Because the vast majority (roughly 90 percent) of respondents who were born in the South and whose parents were born in the South share the same birth state, we attain very little additional information by separately examining those who return to their own birth state and those who return to their parents’ birth state.

Table 5.

Count and proportion of return migrants who return to own or parents’ birth state by race and return migration group.

Migration Status	Black		White		Total
	Lifetime return migrant	Generational return migrant	Lifetime return migrant	Generational return migrant	Lifetime return migrant	Generational return migrant
Return to birth state or parents’ birth state	189 (48.6%)	289 (21.0%)	3556 (53.7%)	3763 (21.8%)	3745 (53.4%)	4052 (21.8%)
Return to other southern state (not birth state)	200 (51.4%)	1084 (79.0%)	3064 (46.3%)	13,489 (78.2%)	3264 (46.6%)	14,573 (78.2%)
Total	389	1373	6620	17,252	7009	18,625

Source: 1940 Census data linked to 2000 Census data.

Approximately one-half of all lifetime, Black and White southern-born return migrants reside in their own or their parents’ birth state. Moreover, roughly one-fifth (21–22 percent) of Black and White generational return migrants reside in their parents’ birth state.¹⁵ While this does not constitute a majority of return migrants overall, it is a high percentage given that these respondents are ostensibly separated from their parents’ state of birth by a generation. Racial differences in the proportion of return migrants residing in their parents’ southern state of birth in 2000 are quite modest. Using the marginal distributions for parental birth state and respondent’s state of residence in 2000, we determined that the observed probability of returning to one’s parents’ birth state is far higher than the expected probability of returning to one’s parents’ birth state based on chi-square tests (p < 0.0001 for both Black and White generational return migrants).¹⁶

It is possible that many of these generational return migrants spent summers, holidays, and/or family reunions in their parents’ home state and therefore maintained ties to those states before moving themselves. These high proportions suggest that the “call to home” documented by Stack (1996) may be a prevalent motivation for many migrants, Black and White. It is important, however, to also acknowledge powerful forces that operated to tarnish the memories and impressions that second-generation Great Migration migrants had of the South. For example, Thompson-Miller et al. (2015) leverage oral histories gathered from elderly African Americans residing in the Southeast and South Central regions of the country to document a strong and resilient “legacy of Jim Crow.” According to the authors, the horrors, insults, and discrimination inflicted by Jim Crow contributes to what they refer to as “segregation stress syndrome,” a condition that transcends generations. Without information about the factors that motivated return migration among the members of our sample, we are unable to offer an assessment of the relative influence of these countervailing forces.

We next turn to the characteristics associated with migrants who return to their own or their parents’ birth state. Table 6 illustrates the individual and familial characteristics associated with the odds of returning to one’s or one’s parents’ birth state for Black return migrants (Column 1) and White return migrants (Column 3). In these models, the indicator variable “Born in South” distinguishes lifetime return migrants from generational return migrants. Once again, we report odds ratios that reflect the relationship of the predictor variables to the likelihood that migrants returned to their own or parents’ state of birth.

Table 6.

Results from logistic regression analysis (odds ratios) of returning to own or parents’ birth state, by race.

	(1) Black respondents	(2) Black respondents	(3) White respondents	(4) White respondents
	(odds ratio)	(odds ratio)	(odds ratio)	(odds ratio)
Born in South	4.226 (0.776)**	14.37 (21.83)	5.162 (0.186)**	4.105 (1.445)**
Same birth state as parent	0.691* (0.114)	0.690* (0.115)	0.621** (0.0207)	0.619** (0.0207)
Male	0.926 (0.118)	0.922 (0.119)	0.983 (0.0338)	0.981 (0.0338)
Age	0.978 (0.0117)	0.978 (0.0118)	0.945** (0.00304)	0.945** (0.00304)
Metro Area, 1940	1.369 (0.250)	1.357 (0.249)	0.798** (0.0250)	0.798** (0.0251)
Owned Home, 1940	0.920 (0.162)	0.899 (0.159)	0.914** (0.0312)	0.914** (0.0312)
Highest Parent Grade	1.015 (0.0197)	1.004 (0.0242)	0.984** (0.00569)	0.988 (0.00707)
Highest Parental Occupation Score	0.699** (0.0964)	0.687* (0.116)	0.877** (0.0333)	0.857** (0.0418)
Education	0.923** (0.0175)	0.952* (0.0228)	0.934** (0.00600)	0.947** (0.00772)
Logged Personal Income	0.837** (0.0497)	0.846* (0.0558)	0.940** (0.0155)	0.923** (0.0175)
Married	0.917 (0.111)	0.911 (0.111)	0.938 (0.0325)	0.938 (0.0325)
Parent Grade*Born in South		1.034 (0.0418)		0.989 (0.0118)
Parent Occ*Born in South		1.027 (0.297)		1.069 (0.0804)
Own Grade*Born in South		0.919* (0.0391)		0.967* (0.0129)
Own Income*Born in South		0.944 (0.136)		1.060 (0.0344)
Constant	41.69** (44.68)	29.99** (34.19)	146.6** (41.24)	157.8** (48.65)
Observations	1762	1762	23,881	23,881

^**p < 0.01

^*p < 0.05.

Source: 1940 Census data linked to 2000 Census.

Notes: The sample is limited to return migrants for this analysis. Standard errors are in parentheses.

One of the strongest and seemingly most important predictors for returning to one’s or one’s parents’ birth state is whether one was born in the South and therefore is a lifetime return migrant. This is understandable, given that this group likely has stronger personal ties to the South and to specific sub-regions within the South. It may also indicate that some of these return migrants returned with their parents early on and are therefore reflective of the “disappointed migrant” experience (DaVanzo and Morrison, 1981). Surprisingly, we also find that for both Black and White return migrants, being born in the same birth state as one’s parents has a negative association with returning to one’s state of birth.¹⁷ Arguably, sharing a birth state with one’s parents would provide even stronger personal ties to the South and, specifically, one’s state of birth. Nevertheless, as mentioned above, only a small proportion of Black and White southern-born respondents do not share a birth state with their parents. Therefore, our results may be biased by this small sample of individuals or by the potential selectivity associated with this particularly mobile group.

For both Blacks and Whites, there is strong evidence that those with lower socioeconomic statuses were more likely to return to a familial birth state. Regardless of race, returning to a southern birth state (own or parent’s) was significantly more likely for those with less education, lower incomes, and with parents who had lower occupational scores. In addition, for Whites, returning to a familial birth state was also significantly more likely for those who had parents with less education and who resided in a rented home in 1940. Other findings suggest that younger Whites were more likely to return to a familial birth state.

Using interaction terms, we also examined whether the indicators of socioeconomic status have different relationships with the probability of returning to one’s or one’s parents’ birth state depending on whether the respondent is a generational or lifetime return migrant. These results are presented in Column 2 for Black return migrants and in Column 4 for White return migrants. With one exception, we found that the relationships did not differ significantly between these groups. For both Black and White return migrants the negative relationship between one’s own level of education and the likelihood of returning to a familial birth state was significantly stronger among lifetime return migrants (i.e., those who were born in the South).

Overall, return migrants residing in their own or their parents’ birth state in 2000 tend to be negatively selected relative to other return migrants. These results are consistent with the findings of previous literature that has found that return migrants who return to their birth state are negatively selected, though we add to the literature by showing the consistency of these relationships for both Black and White migrants. Our study is also the first to examine the propensity of, and factors associated with, returning to one’s parents’ state of birth. Our findings that those who return to their parents’ state of birth are also negatively selected provide further evidence for the strength and consistency of these relationships.

5. Discussion

Our study provides unique insights into who return migrants are and how their characteristics differ depending on whether they were born in the South or the North and whether they are returning to their home state or elsewhere in the South. This work is the first we know of to definitively identify today’s aging population of generational return migrants to the U.S. South and the characteristics associated with this group, outside of small-scale, ethnographic studies or small longitudinal surveys. It is also the first to compare generational return migrants to traditionally defined, lifetime return migrants. We find that the propensity for the northern-born offspring of Great Migration participants to return to the South by 2000 was substantial (nearly 1-in-10 for Blacks and nearly 1-in-3 for Whites) and that generational return migrants tend to be more positively selected on their own and their parents’ socioeconomic characteristics than either lifetime return migrants or second-generation Great Migration migrants who remain in the North. This holds for both Blacks and Whites and has important implications for how we view return migrants and their selectivity. Indeed, these findings demonstrate that by solely examining lifetime return migrants, we may seriously misjudge the selectivity of return migrants as well as the magnitudes of return migration streams. Family ties to place, even those that are removed by a generation, may therefore be more important motivators for migration location decisions than previously assumed.

For the most part, our findings are generally consistent with the previous literature on southern-born return migrants that classified return migrants as those who were born in the South, living in the North five years prior to the census in question, and living in the South during the year the census in question was conducted. Indeed, we find that lifetime return migrants tend to be more negatively selected than both second-generation Great Migration migrants who are at risk of migrating but remain in the North and generational return migrants. However, our innovative data source allows us to examine a larger and more representative group of lifetime return migrants than has been possible with cross-sectional data and other available data sources. And, because we are able to follow migrants over the majority of their lifetimes, we are better able to examine how both early- and later-life characteristics are associated with the probability of return migration. Our results are also less likely to be influenced by “disappointed” migrants who leave and return to the South relatively rapidly. We are therefore able to provide a more robust and comprehensive view of return migrants to the South and their characteristics than has been possible with other large-scale data sources.

Moreover, a substantial proportion of return migrants return to either their own or their parents’ birth state, offering suggestive evidence that many of these migrants could be experiencing a “call to home” rather than merely a desire to return to an economically growing region of the United States. In general, migrants who return to a familial birth state tend to be more negatively selected than return migrants who return to other states in the South. Nevertheless, these migrants may still provide important resources for rural areas in the South, areas that may have lower average levels of human capital and population growth and that may therefore experience benefits from return migration. These results are largely consistent for both Black and White respondents. This, in itself, is an interesting finding that suggests that both Black and White return migrants may share similar motivations for returning to the South.

Notwithstanding the availability of a novel data source linking individuals between the 1940 and 2000 censuses, there are limitations associated with our study that need to be acknowledged. For example, we were unable to assign PIKs to approximately 30 percent of all children in 1940. Of those assigned PIKs in 1940, we were unable to locate 30 percent of those in the 2000 Census. It is possible that the inability to link all individuals to their 2000 Census records biases our results. For example, survival to 2000 is likely related to early-life conditions such as one’s own and one’s parents’ educational attainment and income. The inability to observe those who have passed away prior to the 2000 census may therefore potentially lead us to observe more advantaged individuals than average. However, it is unlikely that these factors substantially affect our results, because to do so would require that the PIK linkages operated differently for return migrant and non-return households, which we do not believe is the case. The substantial similarities between the respondents we are able to successfully link and those who we are not also suggests to us that this bias is important to consider, though minimal.

Additionally, our sample is limited to older individuals who were, on average, between 60 and 70-years-old in 2000. Our results are therefore not generalizable to younger groups of return migrants who would have moved at the height of their workforce productivity and might have exhibited different sociodemographic profiles from those included in our sample. And, as noted earlier, our data provide only very limited information about the stages of the life course that post-dated the 1940 census and pre-dated the 2000 census.

We are also unable to determine why individuals return migrated because we do not have information about the motivations of migrants. Such information would be helpful to determine whether generational return migrants are more likely to migrate for economic reasons than lifetime return migrants and whether desires to return “home” make important contributions to generational return migrants’ motivations to return to the South. Likewise, such information would help us to more definitively determine who is a “disappointed” migrant or who is migrating for family-related reasons. Given the vastly different contexts Blacks and Whites faced in the South, it is possible that understanding these motivations would reveal important differences between Black and White return migrants. For example, prior to the Great Migration, 90 percent of Blacks in the U.S. resided in the South, compared with a considerably smaller percentage of Whites. This density may cause the South to be perceived as an “ancestral homeland” to a greater extent for southern Blacks than for southern Whites whose kith and kin networks may have been more dispersed and who may therefore identify less with the “generational” migrant typology. That being said, the “divided hearts” of many White southern migrants and the frequent tendency for White southern migrants to engage in circular migration between the South and the North suggests that returning to an ancestral homeland may be an important motivation behind Whites’ return migration decisions. Differences in property arrangements between Blacks and Whites prior to the Great Migration may also influence return migration propensities. Specifically, while Penningroth (2003) has found that Black property ownership was a prevalent and underappreciated phenomenon, much of this property involved livestock and other items that were more mobile and transitory and land ownership tended to be less common among Blacks than among Whites, with many Black southerners becoming tenants to White land owners (Tolnay, 1999). Whites may therefore be more likely to return migrate in order to claim family land. For these individuals, the idea of being a “generational” return migrant would likely be particularly relevant. However, without information on the motivations behind return migrants’ decisions, we are only able to speculate on the potential role of these contextual factors in influencing return migrants’ decisions and in influencing the extent to which “generational” return migrants would identify themselves as such.

Lastly, our study considers only the adult progeny of Great Migrant participants who left the South prior to 1940. The exodus from the South was substantial during this early phase of the Great Migration (roughly 1910 to 1940) and it was during this period that the highest percentages of Black birth cohorts migrated (Black et al., 2015). However, we are unable to capture the post-World War II stream of Great Migrants and their children, which exceeded the pre-World War II stream of migrants in terms of its absolute magnitude and was associated with a different historical context. This limitation is imposed by the availability of the data used for our study. Whether similar findings would be obtained from a parallel study of the later phase of the Great Migration is an empirical question that must be left for future research when the required data become available.

Despite these limitations, our study harnesses a novel, newly-available, and uniquely expansive dataset to offer important contributions to the literature on return migration and internal migration more broadly, showing that return migrants are a diverse group that are not easily captured with typical cross-sectional data concerning birth place, residence in the recent past, and current residence. Moreover, the magnitude and selectivity of return migration streams may be severely underestimated if generational return migrants are not accounted for. Indeed, we find that generational return migrants comprise a substantial part of the return migration stream and that generational family ties therefore appear to be important motivators for the location decisions of migrants. This finding demonstrates the importance of considering the myriad ways that personal and family ties to place may play important roles in migration decisions and the destination choices of migrants.

An important part of the story we have told concerns the selectivity of return migration and how it might vary across meaningful subsets of return migrants. As such, our findings offer a cautionary tale for theoretical perspectives, or their application, that fail to consider such heterogeneity while assuming a “one size fits all” explanatory framework. We submit that the potential for such heterogeneity should be considered, more generally, in the study of migration streams whether they are composed of primary migrants or return migrants, internal migrants or international migrants.

Our study leaves ample room for future research into the larger phenomenon of return migration to the South, or the more restricted focus on the return of second-generation return migrants. First, as noted, our analytic sample is restricted to a relatively elderly group of return migrants from the early phase of the Great Migration. Important questions remain regarding the sociodemographic characteristics of more youthful returnees, as well as for the children of Great Migration participants who left the South after 1940. Second, our study is bereft of direct measures of the forces that motivated the return migrants, therefore we are constrained to drawing indirect inferences about those forces. Alternative data sources will need to be mined for further insights into the migrants’ specific motivations for their mobility, including the relative importance of a southern “call to home.” Third, greater attention should be devoted to the influence of geography on return migration, including the consideration of specific origin-destination migration streams and the role of distance in shaping the likelihood of return migration as well as the choice of southern destinations. The American South is a large and diverse region. It is very possible that patterns of return migration, including its selectivity and the allure of an ancestral “homeland,” vary across southern sub-regions. Finally, if the definition of “return migrants” is to be expanded, as we have done in this study, to include those who were not born in, or possibly never lived in, the South, then where do we draw the generational line? For example, should the grandchildren of original Great Migration participants be considered return migrants if they move to the South? These and other remaining questions comprise an interesting and challenging agenda for future research.

5.1. Characteristics of migrants: the question of selectivity

Research on migration to the South has helped to illuminate who these return migrants are, where they come from, and where they are going. Specifically, it has been found that Black and White female return migrants to the South (conventionally defined) tend to be younger and more highly educated than both southern-born female migrants remaining in the North and stationary southerners (Adelman et al., 2000). Likewise, the general population of White and Black migrants to the South tends to be younger, more highly educated, and claim higher status occupations than northern and southern populations in general (Falk et al., 2004; Hunt et al., 2008; Robinson, 1990). Indeed, some have claimed that the South is experiencing a “brain gain” as relatively highly educated individuals enter the region (Frey, 2004; Nord and Cromartie, 2000).

However, this “brain gain” largely seems to be accruing in the urban South, while southern nonmetropolitan areas have actually experienced a “brain drain” to southern cities (Fuguitt et al., 2001). Some evidence also suggests that migrants to southern rural areas are negatively selected. For example, the majority of migrants to the Yazoo Mississippi Delta, a persistently poor and predominantly Black region, end up in poverty, with 53% of them having less than a high school education (Brown and Cromartie, 2006, p. 197). Other research has shown that return migrants who settle in nonmetropolitan areas tend to be less educated, less likely to be employed, and have lower earnings than the general northern and southern populations (Falk et al., 2004; Li and Randolph, 1982; Wilson et al., 2009). Thus, migrants to southern cities and, particularly, primary migrants to the South tend to be positively selected on socioeconomic characteristics and likely are more influenced by economic motivations, while migrants to the rural South, who are often return migrants, may be negatively selected and, therefore, more likely to be motivated by familial interests and obligations or “push” factors in the North.

These conclusions, however, often depend on the population to which migrants are compared and how return migrants are defined. Because studies frequently use cross-sectional data from the census, which only provides information on individuals’ birth places, residences five years prior to the census, and residences as of the census enumeration, return migrants have classically been defined as individuals who were born in the South, lived outside of the South five years prior to the census, and live in the South as of the census enumeration. This definition misses many potential return migrants, such as those who returned to the South earlier in their lives or who returned to the South after the census was conducted. Further, this definition may tend to capture a higher proportion of disappointed (and therefore more negatively selected) migrants who find their experiences in the North unfulfilling economically and/or socially, compelling them to quickly return home (DaVanzo and Morrison, 1981). Some studies also include children in their definitions of return migrants, which influences the size and characteristics of various migration streams and may bias downward estimates of age and socioeconomic status.

Additionally, Cromartie and Stack (1989) demonstrate that many individuals who are classified as “primary migrants”—migrants who move to the South and did not previously reside in the South—are actually children of return migrants or have parents or grandparents who lived in the South at some point. They attempt to identify these generational return migrants by finding primary migrants who share households with other individuals who can be classified as return migrants. Presumably, sharing a household with a return migrant indicates that the primary migrant was influenced by the motivations of the “return migrant” and can therefore also be considered a return migrant. Cromartie and Stack (1989) demonstrate that, by conventional definitions, approximately 42 percent of Black migrants to the South between 1975 and 1980 were return migrants (1989:307). However, using a more expansive definition in which return migrants are classified as individuals with recent generational ties to the South, 69 percent of migrants are classified as return migrants (1989:307).

While Cromartie and Stack examine generational return migration with a compelling and innovative analytical strategy, they are only able to identify generational return migrants who live with family members who are return migrants. It is therefore difficult to assess the prevalence of generational return migrants or how their characteristics differ from those of other migrant groups.

Recent work by Sharkey (2015) begins to fill the gap in our understanding of generational return migration by examining the location choices of Black and White grandchildren relative to their parents and paternal grandfathers using the Panel Study of Income Dynamics (PSID). He finds that the descendants of Great Migration migrants were more likely to remain immobile relative to their parents and grandparents and that few characteristics were associated with returning to ancestral regions, with the exception of the employment status of the household head and familial receipt of welfare payments. Although the first systematic study of return migration among the children of participants in the Great Migration, and a significant contribution, Sharkey’s reliance on PSID data required important compromises. For example, his analysis is restricted to children born between 1952 and 1982, relies on limited recollected information about parents and grandparents, and does not include enough observations to estimate the relationship between demographic and economic characteristics and the probability of returning to the South.¹⁸ Therefore, while Cromartie’s and Stack’s (1989) and Sharkey’s (2015) work provides important insights into generational return migration, important questions remain concerning who return migrants are and whether it is primarily those who were born in the South who return home, or whether having generational ties to a place is also an important motivator for return migration.

The question of who return migrants are has important implications for our understanding and theorization of migrants’ motivations, migration decisions, and outcomes. Primary migrants’ characteristics, motivations, and destinations tend to be systematically different from those of return migrants (Cromartie and Stack, 1989; Hunt et al., 2008; Newbold, 1997). Specifically, return migrants are more likely than primary migrants to head to nonmetropolitan destinations, while primary migrants are far more likely to migrate to metropolitan destinations or to areas with high levels of natural amenities (Ambinakudige et al., 2012; Cromartie and Stack, 1989; Falk, 2004; Frey, 2004; Fuguitt et al., 2001; Nord and Cromartie, 2000; Stack, 1996). Moreover, as many as 50–60 percent of return migrants move to their birth state (Adelman et al., 2000; Long and Hansen, 1975). Primary migrants also tend to be younger and to have higher levels of education than conventionally-defined return migrants, regardless of race (Falk et al., 2004; Hunt et al., 2008; Li and Randolph, 1982; Wilson et al., 2009). Thus, differentiating between return and primary migrants has important implications for understanding migration streams into the South.

Furthermore, our understanding of return migrants has been hampered by inconsistencies across studies in the definitions used to identify return migrants, as described above. It is unclear how lifetime return migrants—those who were born in the South, moved out of the South, and then returned to the South—compare to generational return migrants—those who return to their parents’ southern homeland. These two groups of return migrants have vastly different life experiences and have spent varying amounts of time in the North and South, likely corresponding to differences in socioeconomic status, family characteristics, and the resulting selectivity of these migration streams. For example, among Black migrants to the North and children of Black migrants to the North, educational attainment and income tend to increase as time spent in the North increases (Alexander et al., 2017; Restifo et al., 2013; Tolnay and Bailey, 2006). However, health outcomes and mortality rates tend to worsen with time spent in the North (Black et al., 2015; Erikkson and Niemesh, 2016). Moreover, it is likely that lifetime return migrants, who returned to the South within 5 years of the census enumeration (the traditional definition), are more prone to belonging to the group of disappointed migrants, because many seem to return to the South relatively quickly, while “generational” return migrants may be more likely to move in order to maximize their utility and economic prospects in the South. Southern-born return migrants who returned to the South sometime between childhood and late adulthood may be more positively selected than return migrants who returned to the South within five years of the census enumeration because they may be less likely to belong to the group of “disappointed migrants” (DaVanzo and Morrison, 1981). However, they may be negatively selected relative to generational return migrants because they spent less time in the North. While these are compelling possibilities, we know very little about these groups of second-generation return migrants, hampering our ability to hypothesize about these relationships. We also know very little about whether migrants with generational ties to the South are likely to return to their parents’ home states, or whether they are returning to other southern destinations. If the former case is particularly prevalent, it would provide evidence that migrants are experiencing a “call to home” (Stack, 1996), while the latter would provide more evidence for a generalized response to improving economic and social conditions in the South.