Abstract
This article analyzes patterns of geographic migration of black and white American families over four consecutive generations. The analysis is based on a unique set of questions in the Panel Study of Income Dynamics (PSID) asking respondents about the counties and states in which their parents and grandparents were raised. Using this information along with the extensive geographic information available in the PSID survey, the article tracks the geographic locations of four generations of family members and considers the ways in which families and places are linked together over the course of a family’s history. The patterns documented in the article are consistent with much of the demographic literature on the Great Migration of black Americans out of the South, but they reveal new insights into patterns of black migration after the Great Migration. In the most recent generation, black Americans have remained in place to a degree that is unique relative to the previous generation and relative to whites of the same generation. This new geographic immobility is the most pronounced change in black Americans’ migration patterns after the Great Migration, and it is a pattern that has implications for the demography of black migration as well as the literature on racial inequality.
Introduction
Research on the migration of black Americans is commonly studied from a temporal perspective, focusing on the scale and direction of migration at different historical periods or points in time. The most notable example is the extensive literature on the Great Migration of black Americans out of the rural South and into the cities of the Northeast, Midwest, and West (Tolnay 2003). This article takes a different approach, analyzing migration from a generational perspective. Drawing on a unique data set capturing geographic migration over four consecutive generations of black and white families, the analysis examines intergenerational migration during and after the Great Migration.
The multigenerational approach analyzes migration as a process in which the geographic locations of families are linked together across generations. This approach is consistent with a research tradition that views patterns of migration from a network perspective, where origin and destination places are linked by flows of individuals, families, or members of a residential, national, or ethnic community (Massey 1990). The generational approach also is consistent with the literature on black American migration based on oral histories and qualitative research, which emphasizes familial connections to places that extend over long periods of time and across generations of family members (Lemann 1991; Stack 1996; Wilkerson 2010). One implication of this research is that the migration patterns of each generation of family members cannot be understood as isolated from the migration patterns over the course of a family’s history.
Building on this idea, the analysis examines migration patterns over long periods of family histories. Whereas previous research has focused on flows of migration occurring in different historical periods, the generational perspective allows for comparisons of migration patterns from one generation of family members to the next. Patterns of continuity and change in families’ residential locations are observable over a longer period of families’ histories than is possible using other sources of data like the census, making it possible to consider new flows of migration in relation to a family’s geographic roots. This approach allows for a long-term analysis of “return migration” that extends beyond individual lifetimes and reaches across generations of family members.
Three research questions guide the analysis. First, what do flows of black and white migration look like in each of the past four generations, and to what extent do these patterns align with patterns analyzed from a temporal perspective? Second, how have patterns of migration changed for the most recent generation of black Americans in relation to previous generations and in relation to whites? Finally, how does demographic knowledge about return migration change when analyzed from a multigenerational perspective?
Migration of Blacks and Whites Over the Twentieth Century
Historian Ira Berlin described the black American experience as one that has been dominated, paradoxically, by both large-scale migration and by a unique attachment to places (Berlin 2010). Berlin argued that the scale, the coherent structure, and the historical context of black Americans’ migrations into and within the United States has led to a unique significance that is attached to both geographic movement itself and to the places in which black Americans have located. Whereas Berlin’s study extended all the way back to the Middle Passage to the Americas, the current study is similar to much of the demographic literature on the topic in focusing attention on migration of black Americans from a period beginning roughly at the start of the twentieth century and running through the latter half of the 2000s.
This literature is dominated by research on the Great Migration of black Americans out of the rural South and into the cities of the Northeast, Midwest, and West. Beginning in the second decade of the twentieth century and ending in the 1960s, the Great Migration resulted in more than 4 million black Americans who were born in the South living elsewhere by 1980 (Tolnay 2003). The migration of black Americans out of the South led different segments of the black population to different destinations at different times. Movement to cities in the Northeast and Midwest was prevalent throughout the Great Migration, whereas migration to cities on the West Coast began in earnest in the 1940s in response to growing demand for labor around World War II (Tolnay 2003; Tolnay et al. 2005).
Migration was driven by a complex combination of “push” and “pull” factors (Fligstein 1981; Lemann 1991; Marks 1989). In the South, obstructed opportunities for economic mobility combined with declining opportunity and government intervention into the agricultural economy of the South, the persistence of Jim Crow, and racial violence to drive black Americans to seek opportunity elsewhere (Fligstein 1981; Tolnay and Beck 1992). Growing demand for low-skilled labor, active recruitment by firms from the North, and the emergence of “ethnogenic” supports drew them to cities of the Northeast and Midwest (Marks 1989; Price-Spratlen 1999, 2008). The relative influence of the various push and pull factors driving the Great Migration remains a topic of central interest in the field, but in this article, I bracket the question of why black Americans left the South and instead focus on how they moved. This article considers the directional flows of migration with each passing generation and the degree to which members of individual families stay in place or return to the geographic locations of their families, generations earlier.
Research on black migration from the 1970s forward is much less extensive, but one central finding from descriptive analyses of census records is that there has been a reversal of the direction of net migration over this period. Since the 1970s, the South has drawn a growing stream of black American migrants from the rest of the country, as documented with census data, ethnographic evidence, and anecdotal evidence from black Americans who have decided to return to the South or move there for the first time (Adelman et al. 2000; Bilefsky 2011; Copeland 2011; Curtis 2011; Frey 2004; Hunt et al. 2008; Long and Hansen 1975; Robinson 1986; Stack 1996). Frey (2004) identified the movement South as the “New Great Migration” of black Americans, and argued that this new pattern is driven by a combination of improving racial climate and expanding economic opportunity in southern states, along with a lingering cultural attachment to the region (see also Falk et al. 2004). In a qualitative study of return migration to rural areas in the Carolinas, Stack (1996) emphasized the importance of familial ties to specific locations in the South, where family members may still be located, as a primary explanation for the new stream of migration southward.
The shift in the direction of migration is unmistakable, but Tolnay (2003) noted that demographic research on return migration to the South is based largely on responses to questions from the census asking respondents where they were born, where they currently live, and whether they have moved in the prior five years. Tolnay (2003:224) suggested further that longitudinal data may provide a more refined picture of return migration, and pointed to the Panel Study of Income Dynamics as a data set that “holds promise” for studying such patterns. This article responds to this call for longitudinal research on the migration patterns of American families and confirms the promise for such approaches to provide a more refined and nuanced description of migration patterns across generations.
Although the primary focus of this article is on the migration of black Americans, patterns of migration over four generations of white Americans are described to provide a point of comparison. The literature on white migration over the past century is less extensive, perhaps because patterns of long-range migration among whites are less coherent and less structured than those for black Americans. That said, patterns of white migration are related to patterns of black migration over time. Whereas the Great Migration out of the South is associated with black Americans, whites also moved out of the South in large numbers for much of the twentieth century, drawn by some of the same economic push and pull factors as black Americans (Tolnay et al. 2005). In the decades since 1970, whites have migrated to the South just as black Americans have, although more of the white migration is “primary” migration rather than return migration, meaning that whites are more likely to be moving to the South for the first time in their lives (Hunt et al. 2008).
Migration From a Network Perspective
This article follows in the tradition of research on international migration in adopting a network perspective on flows of migration. Perhaps the most succinct statement of this perspective is found in Massey’s theory of cumulative causation, which provides a theory to explain flows of migration from specified origins to specified destinations (Massey 1990; see also MacDonald and MacDonald 1964; Massey et al. 1993; Tilly 1968). Acknowledging the role of push and pull factors present in the origin and destination locales, this theory draws attention to the ways that geographically linked social or familial networks can facilitate flows of migration across places. According to the theory of cumulative causation, each migrant from a specified origin to a specified destination increases the likelihood of additional migrants by expanding knowledge of the destination place, establishing institutions and residential settlements in the destination place, and providing resources to facilitate housing and economic opportunities. The theory suggests that specific origin and destination places are linked together through flows of migrants in ways that are at least partially independent of any general push and pull factors that might lead to regional migration.
Much of the research on black American migration out of the South supports this perspective through written and oral histories about the networks of support and information that migrants utilized to help facilitate the transition to a new part of the country and a new social world (Ballard 1984; Lemann 1991; Marks 1989; Tilly 1968; Wilkerson 2010). The demographic literature on the Great Migration also supports the idea that flows of migration out of the South were highly structured, with specific origin and destination places tightly linked by flows of migrants. However, much of this research has considered only networks of migration at the state level, making it difficult to assess whether the mobility decisions of individuals and families from specific towns and cities in the South were influenced or facilitated by friends or neighbors who had previously moved to a new destination. The analysis in this article provides a more localized portrait of migration networks by describing the degree to which counties are linked by flows of migration.
This article considers networks of family members across generations in addition to networks of places linked by migration. Whereas much of the analysis examines patterns of migration from one generation to the next, the final analysis examines return migration that occurs across all four generations of family members. In this application, return migration to the county of origin occurs when a family member moves from County A in one generation to a different county in a different part of the country in a subsequent generation, and then returns to County A in a subsequent generation. In her ethnographic study of black Americans who have migrated to the South, Stack (1996) argued that family and social ties to places that persist for generations are a primary factor in explaining why blacks choose to “return” to the South, even if they have never personally lived there before (see also Cromartie and Stack 1989; Falk et al. 2004). Because prior demographic research on return migration has considered only where individuals were born and where they currently live, this literature has failed to capture the types of multigenerational ties that Stack identified as central to the migration patterns of black Americans. This study provides a multigenerational perspective on return migration that looks beyond the experiences of a single individual to consider family histories that extend across generations.
Data
The analysis draws on data from the Panel Study of Income Dynamics (PSID), an ongoing longitudinal survey begun in 1968 with about 4,800 families (Hill and Morgan 1992). The PSID is composed of multiple samples. First, the Survey Research Center (SRC) sample is a nationally representative sample of American households as of 1967. Second, the Survey of Economic Opportunity (SEO) sample is an oversample of low-income households. A document describing the selection of the SEO sample notes several uncertainties about the procedures that were used to select the original sample (Brown 1996). However, using the Current Population as a reference point, several subsequent studies have shown that the SEO sample is representative of the low-income population (Becketti et al. 1988; Fitzgerald et al. 1998a, 1998b). The analysis files use both the SRC and SEO samples, which is necessary in order to obtain an adequate sample of blacks and whites. Because not enough members of other ethnic groups are available in the data to include in the analysis, the sample is limited to blacks and whites. Several decades after the PSID began, additional samples were included in order to make the survey representative of the changing population in the United States. These supplementary samples are excluded from the analysis because only a small number of the supplementary sample members have been in the survey for a sufficient number of years to appear in the data as children and again as adults. Interviews of sample families were conducted annually from 1968 through 1997, when the PSID switched to a biennial survey. The file used for this article includes survey years from 1968 through 2007.
The PSID follows family members of the original sample as they split off from the sample family and form their own households, allowing for the analysis of intergenerational relationships. The analysis sample is based on a focal cohort of black and white individuals in PSID households who were born between 1952 and 1982 and were observed as members of PSID sample households as children and as adults. Measures of childhood geographic locations for this cohort are based on geocoded data on the family’s residential address from the age of 0 through 17. To identify the single county in which the child was “raised,” I use the county at which the child’s family lived for the greatest number of survey years during childhood. If the child lived in more than one county for the same number of years, the earliest county is used. I use the same approach to identify the single county in which individuals in the focal cohort lived in adulthood, defined over all years in which the individual was at least 26 years old and was the head of household or the spouse of the head of household. Again, the county of residence is defined as the county in which the individual lived for the greatest number of years during adulthood.
The two measures of geographic location described to this point represent the locations in which the focal cohort lived in childhood and in adulthood, respectively. Throughout the rest of the article, these measures are identified as the locations of Generations 3 and 4. To be clear, Generation 4 does not represent a distinct set of individuals. Rather, Generation 4 represents the geographic locations of the same individuals from Generation 3, but observed in adulthood.
It is possible to measure the childhood geographic locations of the focal cohort’s parents and grandparents through a unique set of questions asked of PSID household heads about the county and state in which they were raised and the county and state in which their parents were raised.1 These questions are used to measure the geographic locations during childhood of the focal cohorts’ grandparents and parents, who are referred to as Generations 1 and 2, respectively. In most cases, the household heads are fathers; thus, data from Generations 1 and 2 usually refer to the geographic locations of the paternal side of the focal cohort’s family. Data are available for the household head’s mother as well, or the paternal grandmothers of the focal cohort. This information is used to create a more comprehensive measure of return migration that considers migration to the childhood location of the paternal grandfather or paternal grandmother.2 A summary of the definitions for each generation is included in Table 1.
The final sample includes 4,543 black and white families with nonmissing data on the county and state of four consecutive generations. Missing data arise from several sources. Although some data are missing on the locations of Generations 3 and 4, it is minimal. Most of the missing data are on the locations of Generations 1 and 2, which is not surprising given that these measures are based on retrospective accounts of where the household head and his or her parents were raised. Data are also missing if the individual was not raised in the United States, leaving new immigrants to the United States underrepresented in the sample. The sample can be thought of as representative of white and black families with children who have been in the United States over the four consecutive generations. It is possible to explore how the restrictions on inclusion in the sample affect results describing migration patterns in the most recent generations by comparing findings using the sample with nonmissing data in all four generations to a larger sample that includes individuals with nonmissing geographic information from only Generations 3 or 4 (regardless of whether data are missing for Generations 1 or 2). These sensitivity tests showed some slight differences in results, but none of the differences were substantively meaningful.
The analyses in Tables 2 and 3 are based on a smaller sample of individuals in the focal cohort who are the firstborn children in their households.3 I use this restriction to avoid “double-counting” the migration patterns of Generation 1 and Generation 2 individuals in families with more than one child. This restriction is not necessary for the remainder of the analyses, which focus on the migration patterns of Generations 3 and 4.
The geographic locations of the counties of residence for all four consecutive generations are geocoded based on the centroid of the county. Because data on the age of individuals in each generation are not available, it is not possible to know exactly when the individual lived within the given county. For this reason, there is error in the geocoded locations of the counties resulting from changes in county borders over time. The geocoded locations of counties for Generations 3 and 4 in the analysis are likely to be coded with less error than the locations of Generations 1 and 2. Also, it is important to acknowledge that the measures of geographic location are not perfectly consistent across all four generations. Data on the counties and states in which Generations 1 and 2 were raised are based on the recollections of household heads, whereas data on the counties and states in which Generations 3 and 4 were raised are based on geocoded address data supplied by staff from the PSID. Again, data from the first two generations of the analysis are likely to contain more measurement error, although it is difficult to know exactly how this error influences the patterns documented in the analysis. One might speculate that families who have moved long distances across generations would be more likely to have missing data on geographic locations in Generations 1 and 2. If this is the case, then the patterns describing change in geographic location over time are likely to be biased toward geographic immobility in Generations 1 and 2.
Most of the analysis consists of descriptive comparisons of migration patterns for blacks and whites, but two additional analyses of geographic immobility and return migration compare blacks and whites while controlling for a set of family demographic and social and economic characteristics. These analyses include several dimensions of family background that are measured over the duration of the focal cohort’s childhood years (age 0–17). Similar information on family background is not available for Generations 1 and 2 because they were not in the sample during their childhood years. Information on family characteristics in Generation 4 is not included in any analyses because these characteristics are endogenous to migration patterns in the previous generation and thus are not appropriate as control variables.
The control variables in these additional analyses include the following: total family income, which is the log of summed income from all members of the sample family, inflated using the CPI-U-RS; parental years of schooling, which measures the total number of years schooling for the parent who has the most education in the household; the total annual hours worked, which represents the estimated hours worked on all jobs in the year prior to the interview grouped into three categories (less than 250 hours (the reference category), 250–1999, and 2000 hours or more); marital status, which is measured with an indicator for whether the child’s parents were married over the full period of childhood (relative to families in which the household head is always not married or sometimes married); the number of children in the household, again averaged over all years in which the individual is observed; the age and squared age of the household head and of the child; the gender of the child (male = 1); homeownership, which represents the proportion of years observed in which the family owned their home; and welfare receipt, measured with a dichotomous indicator for whether the individual or his/her spouse ever reported receiving any income from programs typically referred to as “welfare,” including ADC, AFDC, or TANF. An additional specification (not shown in the tables) includes a measure of public housing, an indicator for whether the child’s family lived in a public housing project or otherwise lived in an apartment or home owned by a public agency. This measure is excluded from reported analyses because it contains extensive missing data, but I include a note to describe how model results change when this measure is included.
Results
Migration Flows of Black and White Families Over Four Generations
The direction of migration flows for blacks and whites is displayed in Table 2, which examines the most common direction of movement in each generation. In this table, directed migration flows are defined by movement in any direction that leads the family out of its origin state and at least 100 miles in any direction (that is, either 100 miles to the east or west, or 100 miles to the north or south). The table shows the percentage of all flows among blacks and whites in each generation, respectively, that led the family northward or southward, westward or eastward. For instance, in the top panel of Table 2, the cell in the top right corner of the matrix for black Americans indicates that 6 % of all families left their origin state and moved at least 100 miles northward and at least 100 miles eastward from Generation 1 to Generation 2. Five percent of black American families left their origin state and moved at least 100 miles northward but did not move eastward or westward; another 2 % moved northward and westward. The cell in the middle indicates that 84 % of black Americans either did not leave their origin state or did not move at least 100 miles in any direction.
Focusing on migration from Generation 1 to 2, the dominant direction of migration for black Americans was northward (5 %) and northward and eastward (6 %). For whites, the most common direction of movement was northward and westward, a path taken by 5 % of all white families. Note that most of the black American population was located in the South at this point, which affects the interpretation of racial differences in the direction of mobility. However, when the sample is limited to white and black families living in the South in Generation 1, there are no substantive differences in the direction of migration for blacks or whites.
From Generation 2 to 3, black Americans’ migration northward intensified: 27 % of black American families moved out of their origin state and at least 100 miles north. The largest share of this group, 16 %, moved northward and eastward, reflecting migration to the cities of the Midwest and Northeast. Whites did not follow the same coherent pattern of migration. Whites were more likely to move westward than eastward, but their migration was fairly evenly distributed across all directions. When the sample is limited to families originating in the South, the direction of migration for whites and blacks looks similar to what is shown in the table, although a larger percentage of black families moved northward. Analyses not shown in the table indicate that 30 % of black families originating in the South moved northward from Generation 2 to Generation 3, compared with 14 % of white families originating in the South.
From Generation 3 to 4, migration among whites is close to evenly distributed in all directions, with the largest migration moving southward and westward. For black Americans, the steady flow of movement northward ended in the latest generation. The largest share of blacks moved southward and westward, although only 4 % of black American families moved in this direction. Whereas previous research has focused attention on the new migration of blacks to the South, the figures in Table 2 suggest that this is little more than a trickle southward: overall, 8 % of black American families moved southward in this latest generation, and 6 % moved northward. It is certainly true that the strong flow migration northward has come to a halt in the latest generation of black Americans, but results provide little support for the conclusion that there is a new, large-scale migration southward. The more appropriate conclusion from the figures in Table 2 is that there is no dominant direction of migration among the latest generation of black Americans. Most notable in this latest generation is the proportion of black Americans who have stayed in place, as represented by the 85 % of families that did not leave their origin state and move at least 100 miles in any direction. This finding is explored in more depth in the following section.
The results from black Americans shown in Table 2 can be summarized succinctly for each generation. From Generation 1 to Generation 2, migration across states was not extensive, but migration followed a dominant directional pattern northward as well as northward and eastward. From Generation 2 to Generation 3, growth in the proportion of families moving out of their home states was pronounced, and the dominant direction of migration was, again, northward. From Generation 3 to Generation 4, movement out of families’ origin states was much less prevalent, and the movement northward ended. A new flow of migration emerged moving southward, which was virtually nonexistent in previous generations; however, the direction of migration was fairly balanced overall.
A complementary visual depiction of black migration flows in each generation is provided in Fig. 1. The maps in the figure provide perhaps the clearest description of the dominant direction of the network structure of black migration flows over successive generations.4 The maps confirm a pattern in which the flow of migration out of the South is clearly visible in the map showing migration from Generation 1 to 2, then intensifies from Generation 2 to Generation 3, and finally dissipates and partially reverses in the most recent generation. Consistent with prior research, Fig. 1 reveals how the geography of the Great Migration changed over time, with greater movement westward occurring in the latter stages. The figure also confirms that the structured migration out of the South ended in the most recent generation.
Geographic Immobility
One of the interesting findings from Table 2 is the growth in the proportion of black families who did not migrate from Generation 3 to Generation 4. Table 3 explores this finding further by showing the prevalence of geographic immobility, defined as the proportion of individuals who remained in the same county, state, and region/division in each successive generation of black and white families. The table reveals differences by race in the prevalence of geographic immobility in each passing generation. From Generation 1 to 2, only minimal racial differences exist in the degree to which blacks and whites remained in the same county, state, and region/division as their parents. Black Americans were slightly more likely to remain in the same county from Generation 1 to 2, but this difference is substantively small: 56 % of black Americans in Generation 2 remained in the same county in adulthood, compared with 51 % of whites, a nonsignificant difference.
Racial differences in the prevalence of geographic immobility are more pronounced in the subsequent generation. From Generation 2 to 3, black Americans were less likely than whites to remain in the same state or the same region/division. Fifty percent of blacks remained in the same county as their parents, 65 % remained in the same state, and 74 % remained in the same region/division. For whites, 46 % remained in the same county from Generation 2 to 3, 74 % remained in the same state, and 82 % remained in the same region/division. Racial differences in the prevalence of geographic immobility at the state and region/division level are statistically significant. Compared with the previous generation and with whites of the same generation, black Americans were more likely to make long-distance moves that brought them across state and regional lines from Generation 2 to Generation 3. The high rates of cross-state and cross-region/division mobility among black Americans in this generation is consistent with previous research on the scale of migration flows during the Great Migration. It is in this generation that black Americans were highly likely to cross the borders of states and regions in moving to the Northeast, the Midwest, and the West.
This pattern is reversed for the most recent generation. Displaying immobility between Generation 3 and 4, the bottom row of Table 3 reveals that black Americans in Generation 4 were much more likely than whites to remain in the same county, state, and region/division upon reaching adulthood. Among black Americans in Generation 4, 69 % remained in the same county from childhood to adulthood, compared with 45 % of whites; 82 % remained in the same state, compared with 74 % of whites; and 90 % remained in the same region/division, compared with 81 % of whites. All racial differences are highly significant. The degree of intergenerational geographic immobility among black Americans not only is much greater than for whites but also represents a marked shift from the prior generation. Whereas flows of black Americans from Generation 2 to 3 reflect the prevalence of long-range migration, flows from Generation 3 to 4 reflect a unique geographic stability, a “transmission” of places from one generation to the next. The most recent generation of black Americans is thus unique, relative to the same generation of whites and relative to the previous generation of black Americans, in the degree to which it has remained tied to places.5
I further examine the racial difference in geographic immobility from Generation 3 to 4 in an additional analysis that estimates the probability of remaining in place as a linear function of race and a set of family-level demographic and economic characteristics measured over the course of the focal cohort’s childhood.6 I use two specifications for analyses of geographic immobility at the county, state, and region/division levels, respectively. The first specification documents the unadjusted difference in the probability of black Americans staying in place from Generation 3 to 4. The second specification shows the same difference after adjusting for observable measures of family characteristics. In both specifications, the outcome is dichotomous, and the table displays results from a linear probability model that is weighted by the population weights used in all other analyses.
The first column from Table 4 shows the unadjusted racial gap in geographic immobility and indicates that black Americans are 22 percentage points more likely than whites to have remained in the same county from Generation 3 to 4. When family demographic and economic characteristics are controlled, blacks are 15 percentage points more likely to have remained in the same county from one generation to the next. The unadjusted racial gap in remaining in the same state is much smaller. Black Americans in Generation 3 are 8 percentage points more likely than whites to have remained in the same state in Generation 4, and the adjusted gap is 6 percentage points after family characteristics are controlled. The fifth column of results in Table 4 shows that blacks are 9 percentage points more likely to have remained in the same region/division from Generation 3 to 4, and the racial gap is 8 percentage points and remains statistically significant after family characteristics are controlled.
Beyond race, there are few consistent predictors of geographic immobility among the measures of family background included in the specifications. Children from families in which parents had more schooling and higher income are less likely to have remained in the same county, state, and region/division in adulthood (the estimated association between income and remaining in the same county is not statistically significant). Children from families in which the parents were married throughout childhood are more likely to have remained in the same region/division in adulthood, and children from larger families are more likely to have remained in the same state and region/division, respectively. Overall, these results suggest that social and economic status and family structure are predictive of remaining in place from one generation to the next. Children from families with higher social and economic status are more likely to have left, and children from larger families in which the parents were stably married are more likely to have remained in place.7 However, these family characteristics play only a limited role in explaining why black Americans have been more likely to remain in place in the most recent generation.
Intergenerational Return Migration
The last analysis considers the new flow of black migration toward the South that emerged in the most recent generation. Ethnographic studies and journalistic accounts of this new stream of migration have argued that migration to the South reflects a desire among black Americans to return to the locations of their ancestors in the period prior to the Great Migration (Bilefsky 2011; Copeland 2011; Curtis 2011; Stack 1996). The implication is that much of the southward flow of migration in the most recent generation is “return” migration undertaken by families that have lived in the South, either earlier in their own lives or in prior generations. The multigenerational perspective that is provided by the PSID allows for an empirical assessment of the prevalence of return migration, to the South and elsewhere, among blacks and whites over four generations.
Table 5 displays the prevalence of return migration for blacks and whites, defined in four different ways. In all cases, the family is identified as eligible for return migration if the household members in Generation 2 or Generation 3 lived outside the region/division in which either the mother or father in Generation 1 of the household was raised. The family is defined as a return migrant family if the Generation 4 household member returned to the original county (or state or region/division, respectively) in which either the mother or father from Generation 1 was raised.
In the first row of results, the prevalence of return migration is measured as a proportion of all families, regardless of whether the family ever left the origin region/division. When the full sample of black families is used as the denominator, about 1 % of black American families are defined as return migrants to the same county, 3 % are return migrants to the same state, and 3 % are return migrants to the same region/division. Among whites, less than 0.5 % of all families are return migrants to the same county of origin, 1 % are return migrants to the same state, and 3 % are return migrants to the same region/division. These figures indicate that the prevalence of return migration is low for all families.
The second set of rows displays the prevalence of return migration among families who left their origin region/division in any generation. With this subset of the sample used as the denominator, the prevalence of return migration is greater. Four percent of black American families are defined as return migrants to the same county, 8 % are return migrants to the same state, and 9 % are return migrants to the same region/division. For whites, 1 % of eligible families are return migrants to the same county of origin, 6 % are return migrants to the same state, and 11 % are return migrants to the same region/division. Only the difference between blacks and whites in return migration to the county of origin is significantly different from zero.
The third and fourth rows of the table limit the focus to families originating in the South, the group to whom the narrative of return migration is most commonly applied. With all families originating in the South used as the denominator, the third row of results shows that 1 % of eligible black families are return migrants to the same county, 3 % are return migrants to the same state, and 3 % are return migrants to the same region/division. Among whites originating in the South, less than half of 1 % of eligible families are return migrants to the same county of origin, 2 % are return migrants to the same state, and 4 % are return migrants to the same region/division.
The bottom row of the table shows the same figures with only families originating in the South and exiting their region/division used as the denominator. Among black families in this group, 4 % are return migrants to the same county, 8 % are return migrants to the same state, and 9 % are return migrants to the same region/division. For whites, 1 % of those who left their origin region/division in the South are return migrants to their county of origin, 6 % are return migrants to their state of origin, and 14 % are return migrants to their region/division of origin.
Over the four generations of data available in the PSID, the results indicate that blacks who left the South are less likely to have returned to the South than whites, although they are slightly more likely than whites to have returned to their families’ specific county of origin. This latter finding could help to explain the narrative of blacks’ returning “home” to the specific counties where their families lived in prior generations.
Table 6 shows results from analyses of return migration that control for demographic and economic characteristics measured over the course of the focal cohort’s childhood years. The first three specifications in the table include a measure of whether the family originated in the South as well as an interaction term (Originated in South × Black) that allows for a comparison of the prevalence of return migration for black families that originated in the South compared with white families that originated in the South. The second set of three specifications includes a different interaction term that allows for an alternative comparison. These specifications include a dichotomous indicator for race and the same interaction term (Originated in South × Black), but here the interaction allows for a comparison of the prevalence of return migration for black families that originated in the South with black families that originated outside the South.
Focusing first on family characteristics other than race, a few characteristics are strongly associated with return migration to the origin county, state, or region/division. Children from families in which the household head is more stably employed are more likely to have undertaken return migration, but this finding is countered by the additional finding that children whose family members received welfare also are more likely to have undertaken return migration in several of the specifications. The strong association between welfare receipt and return migration to the county of origin suggests that one reason for migration back to a family’s origin county may be to receive familial support.
More relevant to the focus of the article are the terms capturing differences between blacks and whites originating inside and outside the South. The first column of results indicates that families originating in the South are no more likely than those outside the South to have returned to the specific county in which their families originated, and the interaction term indicates that blacks originating in the South are 2 percentage points more likely than whites originating in the South to have returned to their families’ county of origin. Considering return migration to the families’ origin state, neither the main effect of originating in the South nor the interaction term is significantly different from zero. The third column of results indicates that families originating in the South are 7 percentage points more likely than families originating outside the South to have undertaken return migration to the region/origin in which the family began; however, the interaction term indicates that blacks originating in the South are 6 percentage points less likely than whites to have undertaken this form of return migration.
The second set of specifications allows for a comparison of return migration among black families that originate within the South versus outside the South. None of the interaction terms are statistically significant in these specifications because the number of black families that lived outside of the South in Generation 1 is minimal. Still, the findings are revealing. Results show that black families originating outside the south are more likely than white families to have undertaken return migration, but black families originating inside the South are less likely to have undertaken return migration than black families originating outside the South.
Although none of these findings are statistically significant, the finding that black families originating outside of the South actually have higher levels of return migration than black families originating within the South suggests that the narrative of black return migration to the South is overstated. Black families originating in the South are slightly more likely than white families to have returned to the specific county in which their families originated, but they are less likely to have returned to their origin region/division. In addition, black families originating in the South are less likely than black families originating outside the South to have undertaken return migration to the county, state, or region/division in which the family originated.
Discussion
This article provides new data and a new perspective on the way that families and places are linked together through four generations of geographic migration. The analysis is based on a cohort of black and white Americans born from the early 1950s through the early 1980s. Beginning with the childhood geographic locations of individuals in this cohort, the analysis looks backward in time to document the migration patterns of the two prior generations, and forward in time to track this cohort from childhood into adulthood. Putting all of this information together allows for a look at the geographic locations of four consecutive generations of black and white American families.
The unique nature of the data comes with important limitations. Measures of geographic location are not consistent across all four generations, and measurement error is more likely in the locations of Generations 1 and 2. Because the analysis is organized around generations, and not periods, the patterns cannot be compared directly to prior historical research on the demography of black and white migration over the past century. Despite the differences in perspective, there is strong convergence between the patterns of migration documented in this analysis and the patterns documented in the large empirical literature on the Great Migration. Research on the Great Migration has split the demographic phenomenon into two periods. The first period occurred in the first three decades of the twentieth century and was characterized by black migration out of the South and into the cities of the Northeast and Midwest. The second period, which ran from the 1940s through the 1960s, was characterized by a continuation of migration into the Northeast and Midwest and a new pathway of migration to the West and Northwest (Tolnay 2003). The generational flows of migration described in this article correspond closely. From Generation 1 to 2, the results in Table 2 and the map of black migration flows in Fig. 1a show a clear pattern of migration northward and eastward. From Generation 2 to 3, this migration flow intensified, and a new flow emerged to the West. The structured migration of black Americans northward and westward is no longer present from Generation 3 to 4, a finding that is consistent with the literature on the timing of the end of the Great Migration.
It is reassuring that the analysis of migration flows from an intergenerational network perspective produces patterns that are consistent with the literature on migration in different periods. The generational perspective, however, provides additional insights that cannot be gleaned from studies that focus on shifts in migration over time. Most notably, the generational perspective allows for evidence on the degree to which family members left and returned to places in which their parents or grandparents were raised. Despite the attention given to the new stream of black migration to the South, figures showing migration over four generations indicate that return migration has been rare among both blacks and whites and that only return migration to the specific county of origin is slightly more common among blacks than whites. Overall, there is no evidence indicating that black Americans whose families originated in the South have been disproportionately likely to migrate back to their families’ region of origin. Black families from the South are less likely than white families from the South to have returned to the region/division in which the family originated, and black families from the South are less likely than black families outside the South to have undertaken any form of return migration (although the differences are not statistically significant).
These results do not undermine or contradict the demographic and qualitative literature on the new stream of black migration southward, but they do provide a new perspective on the scale of this migration stream. The findings in this article confirm other research showing that after multiple generations in which black Americans were steadily moving out of the South, this stream of migration ended and a new stream of migration into the South emerged. These developments are important for understanding the demography of black migration (Frey 2004), and they are important for understanding the connections between black family life, cultural identity, and geography (Stack 1996). For instance, one could argue that the greater prevalence of black Americans’ return migration to the specific southern counties in which their ancestors lived is consistent with Stack’s (1996) findings about the importance of familial connections in understanding this new migration stream.
However, the analyses in this article suggest that the scale of the new stream of black migration into the South is not particularly large and may not represent the most important recent change in the demography of black migration. A more pronounced change in black migration has been the new geographic immobility of black Americans in the most recent generation. Black Americans in the most recent generation are more likely than their parents and more likely than whites in the same generation to remain in place from one generation to the next. Perhaps the most notable finding in the analysis is that the destinations of previous generations of black Americans have been transmitted to the most recent generation.
The transmission of places to the current generation of black Americans carries a unique set of implications for black families and for scholarly understanding of racial inequality more broadly. To understand the implications of this new geographic immobility, it is important to go beyond the finding that black children have remained tied to places over time and to consider the trajectories of the places that they have inherited. A large portion of black children in the cohort born from the mid-1950s through the mid-1980s have remained tied to urban areas in the Northeast and Midwest that have fared poorly in the post–Civil Rights era. As documented in an extensive literature from urban sociology, the northern cities that were passed on to the current generation of black Americans experienced a set of changes since the 1970s (and earlier in some cases), including economic transformation, demographic change, and shifts in public investment (Sharkey 2013; Sugrue 1996; Wilson 1987). As a consequence, much of the urban history of the post–Civil Rights period in the Northeast and Midwest can be characterized by the erosion of the economic opportunity that drew families from the South, the erosion of the institutions that sustained urban communities prior to the 1970s, and the erosion of the federal investment necessary to allow urban communities to withstand the economic and social transformations of the post–Civil Rights period.
The urban history of northern cities since the 1970s adds important context to the patterns documented in this article. A common argument in the literature on racial inequality is that the economic advances made by black Americans in the first three quarters of the twentieth century are at least partly attributable to blacks’ willingness to make long-distance moves to seek out new economic opportunities outside the South (Smith and Welch 1989; but see Eichenlaub et al. 2010). The corollary to this argument is that low levels of long-range geographic migration might limit a group’s capacity to exit areas where opportunity is declining and to enter regions or cities that offer greater economic opportunities. From this perspective, the relatively low level of geographic mobility in the most recent generation of black Americans may be viewed as a factor contributing to persistent racial inequality (Sharkey 2013).
If black Americans in the most recent generation have been less able or less eager to make long-distance moves that lead them to new places, the patterns described in this article provoke a consideration of why this is the case. Analyses that control for a range of family social and economic characteristics indicate that these dimensions of family background do not account for racial differences in the prevalence of remaining in place from one generation to the next. The multigenerational perspective presented here also makes clear that any valid explanation cannot focus on fixed or stable features of black families, black culture, or black residential preferences to explain migration patterns; such an explanation is inconsistent with the empirical finding showing that previous generations of black Americans were more likely than whites to undertake long-range geographic migration. An explanation of the patterns documented in this article requires an explanation of change across generations.
This article stops short of providing such an explanation, as the empirical focus of this article is on description—description of migration patterns over long periods of time, racial differences in migration patterns in each generation, and the strength of families’ ties to places across four generations. The hope is that this descriptive analysis of migration patterns will provide a launching point for a second stage of inquiry, one that utilizes additional methods and data and one that is focused not only on description but also on explanation.
Acknowledgments
This project was supported with a grant from the UC-Davis Poverty Research Center. I received helpful comments from participants in the UC-Davis Poverty Research Center Small Grants Conference, and in particular from Abigail Wozniak. I would also like to thank Manish Nag, who offered extensive assistance in the construction of maps found in the text, which were developed using the Sonoma Network Mapping Software that he developed.
Notes
Household heads were asked a very simple set of questions about the county and state in which they grew up. The same questions were asked of the household head about the household head’s parents. In later years of the survey, the questions also were asked about the parents of the spouse of the household head, but this information is not included in the analysis because it is not available for most of the sample.
Children who are living in a household in which the household head is not the child’s mother or father are excluded from the analysis because some children are raised by grandparents or other relatives who are substantially older than their own parents. In this scenario, the analysis would skip a generation in the family: the analysis would span four generations for most families and five generations for a small portion of the sample, leading to a more muddled picture overall. This restriction excludes 565 individuals from the final sample. I conducted analyses of staying in place and return migration with this sample included in order to assess the sensitivity of the results. I found no differences in results.
Technically, the children who are observed over the longest period of adulthood are selected. This is usually the firstborn child in the household.
The maps were constructed using software developed by sociologist Manish Nag called “Sonoma Network Mapping Software” (Nag 2009), which allows for the visual display of network data with a spatial component. The software is available online (http://princeton.edu/~mnag/sonoma).
The racial gap in geographic immobility is reflected in national data from the census (Ruggles et al. 2010). An analysis of non-Hispanic black and white adults’ current state of residence compared with their state of birth shows that in census years from 1930 to 1980, black Americans age 26–45 were equally or less likely than whites to live in their state of birth. In census years 1990, 2000, and 2010, black American adults were more likely than whites to live in their state of birth. For instance, in 2010, 69 % of black American adults lived in the same state in which they were born, compared with 62 % of whites.
This analysis is not possible for earlier generations because survey data were not collected for Generations 1 and 2 during childhood. Because the analysis is restricted to Generations 3 and 4, the sample for the analysis is not restricted to the firstborn children in Generation 3 households but includes all individuals in families with nonmissing data on state and county of residence in all four generations.
Long (1988) argued that receipt of public assistance may tie families to specific places that offer limited economic opportunity. In the results reported in Table 4, receipt of cash welfare is not associated with remaining in place. However, in separate results (not shown here), a measure of residence in public housing is significantly associated with remaining in the same state and the same region/division from Generation 3 to 4. Including the measure of public housing does not reduce the size of the racial gaps in geographic immobility shown in Table 4. This measure is not included in the main results because it has extensive missing data.
