Moving upward but not onward? Occupational Mobility and Migration in - - PDF document

Moving upward but not onward? Occupational Mobility and Migration in South Africa

Becca Wang Population Studies Training Center, Sociology Department, Brown University IPC 2017 Paper

October 9, 2017

1 Introduction

The vast body of migration literature over the past 3 decades has established a rich understanding

• f the determinants of and motivations for population mobility. That is, two major lines of inquiry

within migration research have focused on demonstrating who moves (versus who stays) and why. Across both developing and developed country contexts, ample evidence points to a strong selection effect: migrants tend to be male, healthier, and better educated (Feliciano 2005; Garip and Curran 2010; Lindstrom and Ramirez 2010; Tong and Piotrowski 2012). While migration is undertaken for a variety of reasons, including forced displacement (Rwamatwara 2005) and major life transitions such as marriage (Bernard et al. 2014), the migration literature in Africa has focused predominantly

• n labor migration. Motivated by spatial inequalities in employment opportunities (Adepoju 2003),

a striking 2.5 million individuals are estimated to engage in labor migration in South Africa alone (Collinson et al. 2006). However, while migrants tend to be positively selected on sociodemographic and economic characteristics, not all migrants are successful in obtaining a better job—or even any job—after migrating to a new locale (Dejong and Legazpi Blair 1994; Cushing and Poot 2004). Limited access to information about the labor market (Nogel 1994); lack of social networks to provide settlement assistance (Korinek et al. 2005); and discrimination against migrants (Zuberi and Sibanda 2004) are all significant challenges that may prevent migrants from securing desirable employment opportunities. Given this potential heterogeneity in migration outcomes, this paper considers whether the occupational outcome from a migration event shapes migrants’ decisions about whether to engage in further residential moves. Particularly in developing countries, it is well documented that most migration events are not

• ne-time residential relocations (Skeldon 2010). Instead, individuals engage in migration multiple

times throughout their life (Goldstein 1964; Morrison 1971). Despite this recognition of the re- peated nature of migration, few studies have explicitly considered the dynamic interplay between

• ccupational attainment and migration behavior. Sub-Saharan Africa exhibits some of the highest

levels of internal labor migration in the world (Lucas 2015), and repeat migration contributes sub- stantially to the overall totals. In this context, migrants’ decisions to settle in one place or engage in further migration becomes a salient issue for migration scholars and policymakers alike. In this 1

paper, I make use of occupational attainment data, measured both before and after each migra- tion event, to better understand how occupational outcomes might influence the repeat migration

• process. Using retrospective residence histories from a small but nationally representative survey
• f black South Africans, I focus specifically on a population for whom migration has represented a

critical livelihood strategy both historically and in contemporary periods (Reed 2013).

2 Background Literature

2.1 Migration and Occupational Mobility

Labor migration is commonly treated as a response to the uneven geographic distribution of eco- nomic resources and opportunities. The seminal work of Todaro and Harris (1969) and Todaro (1970) established the well-known framing of the rural migrant as a utility-maximizing agent who

• pts to relocate to an urban destination with the expectation of better employment opportunities

and higher wages. While migration theory has evolved a fair amount beyond the simplistic under- standing of a single migrant as a rational actor driven solely by economic factors, the fundamental insight that geographic mobility and social mobility are inextricably linked remains a central theme in the social demography literature. This treatment of labor migration as an avenue for social mobility has generated a large body

• f literature that investigates the returns to migration. However, the cumulative evidence on so-

cioeconomic gains from migration remain mixed (Cushing and Poot 2004). The positive effects of migration documented by classical studies (Blau and Duncan 1967; Wilson 1985) continue to be corroborated (Chattopadhyay 1998; Flippen 2014). At the same time, other studies also find migra- tion is not always socio-economically rewarding and can be associated with downward occupational mobility (DeJong and Legazpi Blair 1994; Ogena and DeJong 1999; Lindstrom 2013). Given the high prevalence of migration undertaken as a livelihood strategy, the occupational consequences of migration, and therefore social mobility of labor migrants, remains an important line of inquiry. Existing scholarship—including research on internal and international migration as well as developed and developing countries—tends to focus on how geographic mobility drives changes in occupational mobility. Relatively less attention has been paid to the reverse relationship. Consequently, it is unclear how the occupational status attained after a move may affect subsequent 2

geographic mobility. My study addresses this gap in the literature by investigating whether occu- pational transitions shape decisions to stay in one location or engage in further migration. I argue inverting this well-studied relationship can enhance our understanding of migration as a social mo- bility strategy. By treating migration as the outcome variable, I can interrogate how individuals dynamically formulate migration decisions in response to their experiences with the labor market. The lack of attention to the dynamic interplay between occupational attainment and migration events is surprising given the long-standing recognition that migration in sub-Saharan Africa often involves multiple moves over time (Goldstein 1964). Early documentation of the ubiquitous pattern

• f step-wise migration—where individuals make a series of sequential moves up the urban hierarchy,

for example, from a village to a small town to a metropolis (White and Lindstrom 2005)—can be traced back to Ravenstein (1889). Additionally, circular migration emerged as a prominent theme in the literature starting in the 1970s to describe the repetitive back-and-forth movement between

• rigin and destination that dwarfed other more permanent forms of migration (Chapman 1978;

Hugo 1985; Skeldon 1977). Studies that consider migration as a complex sequence of moves have also shown that those who have migrated at least once are likely to migrate again (Reed et al. 2010; Reed 2013). This is consistent with the longstanding consensus that migration is a highly selective process, and logically the traits (socioeconomic, health characteristics, etc) that would select an individual into migrating once would also affect subsequent migration (Reed et al. 2010). This research further highlights that each migration trip is not an independent event, and migration analyses should therefore distinguish between lower and higher order moves because the process leading up to each of these moves may be different. I hypothesize the occupational outcomes at each migration event may factor into the decision to engage in subsequent migration. More specifically, those who experience upward occupational mobility after a move are more likely to stay in a location while those who experience no or down- ward occupational mobility after a move are more likely to engage in subsequent migration. To date, studies have rarely taken advantage of data that combine migration histories and employ- ment histories. The few exceptions have focused on the domestic U.S. context (DaVanzo 1981)

• r on international migration to the U.S. (Lindstrom 2013). Relatively little is known about how
• ccupational mobility drives migration decisions in sub-Saharan Africa, where prevalence of repeat

migration is high and there is great interest in understanding the socioeconomic returns to migra- 3

• tion. This study contributes to this agenda by investigating whether a post-migration occupational

transition may shape propensities to remain in one place or engage in further migration.

2.2 Conceptual Framework

The conceptual framing of the following study combines Human Capital Theories of migration with a simple job search model. The set of Human Capital Theories is an influential theoretical per- spective in the internal migration literature which suggests individuals engage in labor migration as a long-term strategy for investing in their human capital (Dierx 1988). Individuals undertake residential moves, perhaps to multiple destinations, as long as each subsequent migration event

• ffers a positive expected return on the additional time and effort that the move requires. Suc-

cessful upward occupational transition will then depend on the migrant’s ability to successfully translate skills, knowledge and other forms of capital into a job (Lindstrom and Massey 1994). A key assumption underlying such theories is that integrating into a new labor market at the migrant destination is not an easy process (Zourkaleini and Pich 2007). First, skills are not perfectly trans- ferable across labor markets (Chiswick 1977). Second, migration decisions are often undertaken with extreme uncertainty. Knowledge about settlement assistance, employment opportunities, and alternative locations may be limited (Nogel 1994; Zuberi and Sibanda 2004). Thus, while migration is a prevalent strategy for trying to obtain upward social mobility, there is no guarantee the strategy will pan out as expected. DaVanzo (1983) proposes a dynamic framework of migration move-stay decisions that combines the concepts of human capital, location-specific capital, and information

• costs. Under this framing, migrants move to make an investment in their human capital, but due

to imperfect information about the destination labor market, the desired employment outcome is

• ften not realized and the individual therefore makes a ‘corrective’ move back to the origin or on-

ward to a new location. This framework highlights that not all moves are planned and sometimes a move is made in response to a prior move that did not work out as expected. My paper draws upon this dynamic decision-making framework as well as a simple job search model proposed in Lindstrom (2013). This job search model begins with the assumption that workers compare potential wage offers to their current wages, and therefore, without a better wage offer, employed workers prefer to remain in their current job (Parsons 1973). Those that are unemployed compare potential wage offers 4

to a reservation wage, the lowest wage a worker is willing to accept (Lippman and McCall 1976). Importantly, there is both an opportunity cost (the lost wages during a period of unemployment) and diminishing returns in the job search process (Matilla 1974; Mortensen 1986). Thus, the reservation wage decreases the longer the unemployment spell and the job search is expanded to consider less desirable jobs in other labor markets (Kapser 1976). While the conceptual job search model is premised on wage comparisons, this research measures occupational attainment rather than wages directly. This is consistent with prior studies investigating the socioeconomic returns to migration (Toussaint-Comeau 2006; Wilson 1985; Flippen 2014). Operationalizing occupational changes rather than wage differences is considered more theoretically relevant, as occupation levels convey a socioeconomic status that incorporates not only wage compensation but also the skill and educational endowments required for a specific occupation (Seibert et al. 1970).

3 Data and Methods

3.1 Study Context

Migration patterns in South Africa are inextricably linked to the legacy of apartheid. Historically, a combination of coercive legislation and restrictions on access to land prompted labor migration into mining and industrial centers (Coovadia et al. 2009). Lacking employment opportunity in rural areas, it is estimated between 60-80% of working-age males were forced to seek work away from home in any given period during the apartheid era (Coovadia et al. 2009). The gradual dissolution of the apartheid system led many to anticipate dramatic changes in migration trends and a massive increase in permanent rural-to-urban migration (Posel 2004). However, recent em- pirical studies show that temporary (repeat) migration remains very much entrenched in South Africa (Collinson 2006; Reed 2013). Thus, both in the pre-democracy and contemporary periods, temporary migration represents a critical livelihood strategy for the rural population. Despite the unique history of apartheid, the migratory context in South Africa is, in many ways, indicative of the broader migration trends across sub-Saharan Africa. Migration affects a high proportion of households across South Africa (Kok et al. 2003; Posel and Casale 2003) and often migration is undertaken by adolescence or young adulthood (Collinson et al. 2007). While individuals and families are motivated by a variety of reasons to migrate, labor migration 5

still dominates migration trends. Furthermore, migration flows are diverse, and in addition to urbanward migration flows, a substantial proportion of migration occurs between rural-to-rural areas and urban-to-rural areas as well (Reed 2013).

3.2 Data Source

In this paper, I make use of the South African Migration and Health Survey (SAMHS), a nationally representative survey of black South African adults (age 18+) that was fielded in 2000 (Reed 2013). Prior to 1996, the census did not capture any explicit migration information (SAMHS report). The SAMHS was thus undertaken to understand mobility trends for the black population. The survey includes retrospective lifetime residence histories for individuals starting at age 12, where any cross district move lasting one month or longer is considered a migration event. Additionally, the survey captures occupational status (formal or informal) both in the month prior to each move and in the month immediately following each move. Thus, this data set is particularly relevant for my research question as I can determine occupational transitions associated with every migration event. As I am interested in modeling the likelihood of repeat labor migration, I restrict my sample to individuals whose first migration is motivated by employment-related reasons and have migrated at least once. Accounting for differences in the challenges of settlement for international and domestic migration is beyond the scope of this article, so I also restrict my sample to native-born South Africans. After further excluding individuals with missing values on model covariates, my final analysis sample contains 1,159 individuals with complete residential histories. These data are restructured to use person-year observations as the unit of analysis, where each record in the data represents one calendar year in which a person is at risk of migrating. Risk of migration begins the year after the individual’s first migration trip and censoring occurs either at calendar-year 2000 or when the individual is over age 64. The person-year level analysis file used in the following event history analysis consists of 15,133 records.

3.3 Estimation Strategy

I use discrete-time logistic models to estimate the impact of occupational mobility on subsequent migration while controlling for both time-varying and time-invariant background characteristics of the individual. Formally, the hazard function, defined as 6

hi(t) = Pr(Yi(t) = 1|yi(t − 1) = 0) (1) represents the probability that individual i experiences migration event y during interval t, given that no event has occurred prior to t. I then fit a discrete-time logistic regression model with the form logit[hi(t)] = log( hi(t) 1 − hi(t)) = α(t) + β′Xi(t) (2) Where hi represents the hazard of migration at discrete time t for respondents i; α(t) is the logit of the baseline hazard function; Xi is the vector of covariates for individual i; and β′ is the vector of regression parameters. Due to the nature of the data collection, time to migration is measured in discrete-time intervals

• f calendar years. As my goal is to estimate the risk of higher-order migration, the outcome is

treated as a repeatable event and the model allows for multiple observations from each individual.

3.4 Variables

The outcome variable is a binary variable indicating whether a migration event occurred during a given person-year. Migration is defined as a residential move that crosses a district boundary and lasts for at least one month or longer. This analysis does not distinguish between return migration to the origin location or migration onward to a new location. It is possible there are differences between return moves or moves to a new location. However, the SAMHS does not contain a large enough sample of return movements to distinguish between the two types of moves. My analysis is thus focused on the general decision to migrate again and the outcome variable measures any migration event. The independent variables in this analysis are the occupational transitions experienced at key moments along the migration trajectory. These are created from the occupational categories that are recorded the month before and the month after each migration event (see figure 1). The five categories are: professional–1; professional–2; semi-skilled; unskilled; and unemployed. In the absence of occupational prestige measures, I treat these five categories as an ordered occupational hierarchy, with professional–1 at the top and unemployed at the bottom. In terms of occupational 7

transitions, there are three possible outcomes after a migration event:

• upward transition (a move into an occupation that is higher on the occupation hierarchy than

previous occupation held)

• lateral transition (a move into a job at the same level on the hierarchy)
• downward transition (a move into a job that is lower on the hierarchy)

As depicted in figure 1, the primary independent variable of interest measures occupational transitions by comparing the occupational statuses immediately before and immediately after each migration event. The goal is then to examine whether the occupational transitions at one migration event is related to the likelihood of subsequent migration. Figure 1: Occupational categories recorded before and after migration events for sample individual with two migration events The remaining covariates control for individual-level demographic, human capital, regional, and period effects. The number of moves is captured in the model by dummy variables representing the residence spell number, and a duration variable is entered in the model to represent the years of exposure time in each spell. Gender, age, educational attainment and marital status are included to control for respondent’s background demographic characteristics. To capture the declining risk

• f migration as an individual gets older, the age variable here measures age at the start of each
• spell. Unfortunately, only education at time of survey was collected, therefore its inclusion in the

model is used as a proxy for highest level of educational attainment. I include an indicator for 8

marriage as an event, coded 1 in the year of marriage and 0 otherwise as well as an indicator for marital status. This is set equal to 1 in the years after a marriage took place. I also include a dummy variable to indicate residence in an urban location as well as the province at start of exposure to control for any regional effects and more broadly, urban-rural differences. As labor migration is very unevenly distributed in the sample, I collapse some of the origin provinces to account for small sample sizes. Following Reed (2013), KwaZulu-Natal is the only province accounted for independently in the model. I combine the Northern, Eastern and Western provinces into the category of Cape Provinces because of geographic proximity and because few respondents reside in the Northern Cape. Gauteng and Free State are similarly collapsed to one category because of geographic proximity and the inter-connected labor markets I use Northern/Limpopo, North-west and Mpumalanga as the reference category of rural Northern provinces. Lastly, period effects are accounted for with four dummy variables. Again, I employ the same approach as Reed (2013) and the four time periods are: pre-1976 to account for the time prior to Soweto uprising; 1976-1985 to account for time-period up to repeal of the Pass Laws, 1986-1994 to represent the time period leading up to the first democratic election in 1994; and post-1994. All time-varying covariates are lagged by one year.

4 Results

4.1 Descriptive Results

Table 1 below presents selected descriptive statistics for respondent-level characteristics. Roughly half of the sample are male, and the average age at the time of survey in 2000 is 44. However, there is substantial spread in the sample’s age distribution. Educational attainment is quite high in this sample with 31 percent completing some primary education, 48 percent completing some high school, and 9 percent completing some education beyond high school. This is evidence of the well-documented selective nature of migration where migration is positively selected on human capital endowments (Massey et al. 1993). On average, respondents moved a total of 1.87 times and average age at start of each residence spell is 26 years. The descriptive statistics suggest the majority of individuals experience no change (neither upward nor downward movement) in occupational category after a migration experience. Eighty 9

Table 1: Descriptive Statistics %(mean) sd Male .49 Age (at time of survey) 44.00 15 Marital Status (by time of survey) .41 Educational Attainment No education .11 Some primary .31 Some high school .48 Some education beyond HS .09 Origin Province Cape Province .20 Kwazulu-Natal .32 Gauteng / Free State .23 Rural Northern Provinces .24 Urban Place .69 Age (start of each spell) 25.25 9.25 Total Migration Events 1.87 Occupational Transitions (all moves) Upward .13 Neutral .80 Donward .07 10

Table 2: Tabulation of Occupational Transitions Last Observed Occupation Prof-1 Prof-2 Semi-skilled Unskilled Unemployed Total First Occupation Prof-1 5.4% 0.0% 0.0% 0.0% 0.3% 6% Prof-2 0.0% 6.1% 0.1% 0.1% 0.3% 7% Semi-skilled 0.1% 0.3% 15.6% 0.5% 0.5% 17% Unskilled 0.3% 0.3% 0.8% 16.0% 3.4% 21% Unemployed 1.4% 1.8% 2.2% 3.7% 40.8% 50% Total 7% 8% 19% 20% 45% 100% percent of all migration events are a lateral move where the individual moves into the same occu- pational category. To provide further intuition on the prevalence of occupation transitions, table 2 provides occupational mobility between the first observed occupation status and the last observed

• ccupation status. What is striking is the large proportion of individuals that begin and end in

the unemployed and unskilled occupation category, roughly 45 percent and 20 percent respectively. Despite the high educational attainment in this sample, many of these individuals end up staying either unemployed or in unskilled employment positions. Thus, the high proportion of neutral transitions suggest upward occupational attainment does not materialize for many migrants. To examine the bivariate relationship between occupational attainment and propensity to mi- grate, I calculate Kaplan-Meier Survival Estimates. The survival curves in figure 2. are restricted to the second residence spell and figure 2a (left) uses the five occupational categories as the strata while figure 2b (right) uses the transition type at the first migration event. In both cases, the results are consistent with my hypothesis. Figure 2a. shows those in the lowest two occupational categories—unemployed or unskilled worker—after their first migration event are quicker to under- take a subsequent migration trip. Additionally, figure 2 b. shows those whose first migration event resulted in a downward occupational transition also engage in migration much quicker relative to those who experience no transition or an upward transition. For downward movers, over a quarter have engaged in a second migration 2 years after the first move. In contrast, a quarter of lateral movers engage in a second migration by 5 years. Duration of residence is much longer for upward

• movers. In the following section I estimate the strength and nature of the relationship between
• ccupational change and migration decisions in a multivariate regression framework.

11

Figure 2: Kaplan-Meier Survival Estimates

4.2 Regression Results

Table 3 presents parameter estimates from the hazard models predicting the relationship between

• ccupational mobility and subsequent migration. These models measure occupational transition by

comparing occupations in the month before and after each migration event. Model 1 demonstrates that the type of occupational transition is predictive of migration events. All else equal, respondents who experience a downward move into a lower occupation category are 124 percent more likely to make a subsequent migration trip relative to those that experience an upward move into a higher

• ccupation category.

Similarly, those who experience a lateral move into the same occupation category are 60 percent more likely to migrate again relative to the reference category. The direction

• f the coefficients on each of the occupation categories is consistent with this pattern. Individuals

who were in the unemployed, unskilled, semi-skilled or skilled-manual labor categories after their previous move, are all more likely to undertake another migration trip relative to those who were in the highest occupation category. Coefficients on background demographic characteristics mostly appear reasonable and intuitive. On average, migration is highly selective on age and education, with age having a curvilinear

• relationship. Higher educational attainment, here serving as a proxy for human capital, is associated

with a greater likelihood of repeat migration. Getting married in a given year greatly elevates the likelihood of migration in that year, however, this effect is only temporary, and on average, married individuals experience lower migration rates than single individuals overall. Reflecting the historical 12

Table 3: Discrete-Time Logistic Model Predicting Repeat Migration (Odds Ratios) Model 1 Model 2 Upward transition (Ref) Downward transition 2.24∗∗∗ [1.46; 3.40] 1.99∗∗ [1.29; 3.05] Neutral transition 1.60∗∗∗ [1.23; 2.10] 1.57∗∗∗ [1.21; 2.06] Professional-1 (Ref) Professional-2 1.44 [0.93; 2.26] 1.52∗ [0.97; 2.38] Semi-skilled 1.50∗ [0.99; 2.31] 1.53∗ [1.01; 2.37] Unskilled 1.47∗ [0.99; 2.22] 1.50∗ [1.00; 2.27] Unemployed 1.50∗∗ [1.03; 2.23] 1.60∗ [1.09; 2.38] Age 1.17∗∗∗ [1.12; 1.22] 1.18∗∗∗ [1.13; 1.23] Age-squared 1.00∗∗∗ [1.00; 1.00] 1.00∗∗∗ [1.00; 1.00] Male 1.30∗∗ [1.10; 1.54] 1.28∗∗ [1.08; 1.52] Marriage-event 4.71∗∗∗ [3.00; 7.13] 3.75∗∗∗ [2.38; 5.70] Marital Status 0.44∗∗∗ [0.37; 0.52] 0.52∗∗∗ [0.43; 0.62] No edu Primary 1.32∗∗∗ [1.02; 1.74] 1.32∗ [1.00; 1.74] HS 1.94∗∗∗ [1.49; 2.54] 1.89∗∗∗ [1.43; 2.50] Degree 3.31∗∗∗ [2.30; 4.77] 3.19∗∗∗ [2.17; 4.67] Urbanplace 0.67 [0.56; 0.81] 0.69 [0.58; 0.83] Cape province 0.99 [0.78; 1.25] 1.00 [0.79; 1.27] Kwazulu-Natal 0.46∗∗∗ [0.35; 0.60] 0.46∗∗∗ [0.35; 0.60] Gauteng or Free State 1.22 [0.98; 1.51] 1.13 [0.91; 1.41] Pre-1976 2.11∗∗∗ [1.64; 2.69] 1976-1984 1.11 [0.88; 1.38] 1985-1993 1.21∗ [1.00; 1.46] Post-1994 (Ref) Duration 0.94∗∗∗ [0.92; 0.95] Spell2 0.87 [0.57; 1.37] Spell3 0.84 [0.55; 1.35] Spell4 0.89 [0.55; 1.47] Spell5 0.97 [0.56; 1.71] AIC 5151.41 5032.09 BIC 5296.28 5237.95 Log Likelihood −2556.71 −2489.04 N 15133 15133

Exponentiated coefficients; 95% confidence intervals in brackets Data Source: South Africa Health and Migration Study

∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

13

male dominance of labor migration, males are more likely than females to engage in subsequent migration trips. Finally, there may be some regional differences as captured by the provincial dummy variables. I do not, however, find a statistically significant effect for residing in an urban location. Model 2 adjusts the baseline model specification by conditioning on the number of moves and duration of time in a spell. The relationship between occupational attainment and migration remains robust. Those who experience downward occupational transitions are more likely than upward movers to move on. Similarly, those who experience a neutral transition have a higher propensity of engaging in further migration. Parameter estimates demonstrate the longer an in- dividual stays in a given location, the lower the likelihood of subsequent migration becomes. The interactive terms on duration and occupational transitions were not found to be significant. This is somewhat contrary to my expectations, as I anticipated downward movers and neutral movers would exhibit positive duration dependence—that is, the risk of migration increasing over time— while upward movers would exhibit negative duration dependence. The signs of the coefficients on the interaction terms at least suggestive of this, but I am unable to disentangle whether there is a significant difference in duration dependence with the present data.

4.3 Additional Analyses

As was evident in the descriptive results, movement into and out of the unemployment category accounted for a substantial portion of migration events. Unemployment is less straightforward of an occupational status because it potentially captures different types of individuals depending on whether a prior occupation was held or the duration of the unemployment spell. For instance, a white-collar worker who becomes temporarily unemployed is quite different from a worker who re- mains chronically unemployed. To further test the robustness of the main findings, I run additional analyses that exclude those individuals in a perpetual state of unemployment. These results are reported in appendix table 4. In model A1, I exclude all individuals who never move out of the unemployment category. Those who remain in the sample have held a job at one time and arguably could do so again. In model A2, I utilize an even stricter definition of chronic unemployment and exclude individuals who have ever been unemployed. The estimates for the occupational transitions coefficients in model A2 are statistically insignificant. Nonetheless, in both models, the direction 14

• f the sign on the estimates show the core conclusion continues to hold and downward and neutral
• ccupational mobility is associated with higher likelihood of a subsequent migration event. This

suggests the ambiguity of the unemployment category does not unduly drive the results. Another shortcoming of the data is the lack of a true occupational history. As occupational statuses are only measured in the month prior to and month after a migration event, these data cannot explicitly account for any potential changes in occupational status within the same loca-

• tion. However, considering the dearth of surveys in sub-Saharan Africa that measure concurrent

employment and residential trajectories, these data afford us at least a greater degree of insight into how occupational mobility and migration are related over time. The analyses presented here high- light the utility in employing a nuanced view of migrant populations and distinguishing between frequent movers and those that only experience one migration event. This is further illustrated in the tabulations of occupational transitions disaggregated by frequent and infrequent movers (see appendix tables 5 and 6). For first observed occupation, a much larger proportion of infrequent movers, relative to frequent movers, are employed in the top three occupations that require at least semi-skilled labor. Similarly, for the last observed occupation, a smaller proportion of infrequent movers have respondents that end up in the lowest two occupational statuses.

5 Conclusion

Labor migration remains an important means for social mobility across many developing country

• contexts. While migration studies have been focused on estimating the socioeconomic returns to

migration, few studies have explicitly accounted for migration as a complex sequence of multiple moves often involving multiple destinations (Skeldon 1990). Occupational attainment receives great attention as an indicator of social mobility of migrants, yet evaluations of how migrants may dynamically adjust their migration behavior in response to their occupational attainment are scarce. In this paper, I examine one aspect of social mobility (occupational transitions) among black South Africa adults, and model its relationship to migration events. Occupational transitions are a strong predictor of migration events and the results are consistent with the simple job search model I propose. Those that experience upward occupational mobility are much less likely to engage in 15

subsequent migration events. This suggests such individuals successfully translate skills, knowledge and other forms of capital into a job and thus have lower incentive to change locations. Conversely, downward occupational transitions or neutral occupational transitions are associated with a higher

• dds of engaging in a further migration. This is consistent with DaVanzo’s (1983) idea of making

a “corrective” move, where individuals who fail to achieve their desired occupational outcome at their first migration event are more likely to move again. The analysis presented here does contain several data limitations. First, the data do not contain a true employment history. While the occupational status immediately before and immediately following a residential move represents key points along an individual’s migration trajectory, this study was unable to account for the occupational changes that can occur in one location. Another limitation is lack of information about settlement assistance and help finding a job. Social networks and migrant’s social capital are important resources for migrants to draw upon for employment, housing, and other resettlement assistance (Korinek et al. 2005). Finally, it is also likely there are unobservable factors such as the individual’s ambition or determination that affect the success of the migrants (Zuberi 2004). My current model assumes that subsequent migration decisions are not about unobserved characteristics of the individual but rather a response to the occupational transitions experienced. Future studies could incorporate migration intentions to better disentangle whether certain individuals exhibit greater foresight or risk-aversion that affect their migration behavior. Despite these limitations, the present study highlights the utility in distinguishing between the varied employment outcomes migrants may experience. The findings suggest downward and lateral movers may adjust their migration strategy depending on employment experiences in different

• locales. It is thus theoretically important to treat occupational attainment as both an outcome

and a determinant. Especially when migration is considered a social process that can consist of multiple sequences of moves, the occupational attainment may be an outcome in one period but a determinant in the subsequent period. Studies that jointly consider geographic and occupational mobility histories can shed light on whether the most transient individuals are also the most socially and economically vulnerable. 16

6 Appendix

Table 4: Predicting Repeat Migration (Excluding Chronically Unemployed) Model A1 Model A2 Upward transition (Ref) Downward transition 1.73∗ [0.96; 3.12] 1.39 [0.07; 10.00] Neutral transition 2.03∗∗∗ [1.44; 2.89] 1.85 [0.79; 5.12] Professional-1 (Ref) Professional-2 1.22 [0.69; 2.18] 0.88 [0.39; 1.99] Semi-skilled 1.30 [0.76; 2.28] 0.87 [0.40; 1.95] Unskilled 0.96 [0.56; 1.68] 0.53 [0.25; 1.19] Unemployed 2.71∗∗ [1.48; 5.05] AIC 2387.58 1030.88 BIC 2578.10 1190.26 Log Likelihood −1166.79 −490.44 N 8572 4338

Includes the same set of control variables as model 2 in table 3 Exponentiated coefficients; 95% confidence intervals in brackets Data Source: South Africa Health and Migration Study

∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

Table 5: Tabulation of Occupational Transitions for Infrequent Movers Last Observed Occupation Prof-1 Prof-2 Semi-skilled Unskilled Unemployed Total First Occupation Prof-1 5.6% 0.0% 0.0% 0.1% 0.4% 6.2% Prof-2 0.0% 4.4% 0.0% 0.0% 0.0% 4.4% Semi-skilled 0.1% 0.0% 5.6% 0.0% 0.4% 6.2% Unskilled 0.3% 0.7% 4.5% 11.8% 0.8% 18.1% Unemployed 2.1% 3.6% 14.5% 5.8% 39.2% 65.2% Total 8.1% 8.6% 24.7% 17.7% 40.9% 100.0% 17

Table 6: Tabulation of Occupational Transitions for Frequent Movers Last Observed Occupation Prof-1 Prof-2 Semi-skilled Unskilled Unemployed Total First Occupation Prof-1 1.0% 0.2% 0.0% 0.0% 0.6% 1.7% Prof-2 0.0% 2.1% 0.0% 0.0% 0.8% 2.9% Semi-skilled 0.2% 0.4% 2.5% 1.0% 0.6% 4.6% Unskilled 1.0% 0.4% 1.5% 12.7% 6.5% 22.1% Unemployed 4.4% 5.2% 7.9% 10.4% 40.8% 68.7% Total 6.5% 8.3% 11.9% 24.0% 49.2% 100%

References

[1] Banerjee, Biswajit. 1983. “The Role of the Informal Sector in the Migration Process: A Test of Probabilistic Migration Models and Labour Market Segmentation for India.” Oxford Economic Papers 35(3):399-422. [2] Beauchemin, Cris and Philippe Bocquier. 2004. “Migration and Urbanisation in Francophone West Africa: An Overview of the Recent Empirical Evidence.” Urban Studies 41(11):2245-72. [3] Bell, Martin et al. 2015. “Internal Migration and Development: Comparing Migration Intensities Around the World.” Population and Development Review 41(1):33-58. [4] Berg, Elliot. 1961. “Backward-Sloping Labor Supply Functions in Dual Economies–The Africa Case.” The Quarterly Journal of Economics 75(3):468-92. [5] Byerlee, Derek. 1974. “Rural-Urban Migration in Africa: Theory, Policy and Research Implica- tions.” The International Migration Review 8(4):543-66. [6] Castagnone, Eleonora, Tiziana Nazio, Laura Bartolini, and Bruno Schoumaker. 2015. “Under- standing Transnational Labour Market Trajectories of African-European Migrants: Evidence from the MAFE Survey.” International Migration Review 49(1):200-231. [7] Chattopadhyay, Arpita. 1998. “Gender, Migration, and Career Trajectories in Malaysia.” De- mography 35(3):335-44. 18

[8] Choe, Chung and E.LaBrent Chrite. 2014. “Internal Migration of Blacks in South Africa: An Application of the Roy Model.” South African Journal of Economics 82(1):81-98. [9] Collinson, Mark, Stephen Tollman, Kathleen Kahn, Samuel Clark and Michel Garenne. 2003. “Highly Prevalent Circular Migration: Households, Mobility and Economic Status in Rural South Africa.’” Presentation at conference for African Migration in Comparative Perspective. Johan- nesburg, South Africa. June. [10] Collinson, Mark. 2010. “Striving Against Adversity: The Dynamics of Migration, Health and Poverty in Rural South Africa.” Global Health Action. (3): 5080. [11] Conway, Dennis. 1980. “Step-Wise Migration: Toward a Clarification of the Mechanism.” The International Migration Review 14(1):3-14. [12] Cushing, Brian and Jacques Poot. 2004. “Crossing Boundaries and Borders: Regional Science Advances in Migration Modeling.” Papers in Regional Science 83:317-38. [13] DaVanzo, Julie. 1983. “Repeat Migration in the United States: Who Moves Back and Who Moves On.” The Review of Economics and Statistics. 65(4): 552-559. [14] DaVanzo, Julie and Peter Morrison. 1981. “Return and other Sequences of Migration in the United States.” Demography 18(1): 85-101. [15] Cornwell, Katy and Brett Inder. 2004. “Migration and Unemployment in South Africa: When Motivation Surpasses the Theory.” Moansh University Working Paper. [16] De Jong, Gordon and Marilou C. Legazpi Blair. 1994. “Occupational Status of Rural Out- migrants and Return Migrants.” Rural Sociology (59)4: 693-707. [17] De Jong, Gordon F., Aphichat Chamratrithirong, and Quynh-Giang Tran. 2002. “For Better, for Worse: Life Satisfaction Consequences of Migration.” The International Migration Review 36(3):838-63. [18] De Jong, Gordon and Marilou Legazpi. 1994. “Occupational Status of Rural Out Migrants and Return Migrants.” Rural Sociology 59(4):4693-4707. 19

[19] Flippen, Chenoa. 2014. “U.S. Internal Migration and Occupational Attainment: Assessing Absolute and Relative Outcomes by Region and Race.” Population Research and Policy Review 33(1):31-61. [20] Goldstein, Sidney. 1964. “The Extent of Repeated Migration: An Analysis Based on the Danish Population Register.” Journal of the American Statistical Association 59(308):1121-32. [21] Kok, Pieter, Derik Gelderblom, John Oucho, and Johan Van Zyl, eds. 2006. Migration in South and Southern Africa: Dynamics and Determinants. Cape Town, South Africa: HSRC Press [22] Korinek, Kim, Barbara Entwisle, and Aree Jampaklay. 2005. “Through Thick and Thin: Layers

• f Social Ties and Urban Settlement among Thai Migrants.” American Sociological Review

70(5):779-800. [23] Lindstrom, David. 2013. “The Occupational Mobility of Return Migrants: Lessons from North America.” Pp 175-205 in The Demography of Europe. Ed. By Neyer Gerda, et al. Springer. [24] Lucas, Robert. 2015. “Internal Migration.” Retrieved March 12, 2017 (http://www.knomad.org). [25] Nogle, June Marie. 1994. “Internal Migration for Recent Immigrants to Canada.” International Migration Review 28(1):31. [26] Ogena, Nimfa and Gordon F. De Jong. 1999. “Internal Migration and Occupational Mobility in Thailand.” Asian and Pacific Migration Journal. 8(4):419-446. [27] Reed, Holly. 2013. “Moving Across Boundaries: Migration in South Africa, 1950-2000.” De-

• mography. 50(1): 71-95.

[28] Reed, Holly E., Catherine Andrzejewski, and Michael J. White. 2010. “Men’s and Women’s Migration in Coastal Ghana: An Event History Analysis.” Demographic Research 22(25):771-812. [29] Tienda, Marta and Franklin D. Wilson. 1992. “Migration and the Earnings of Hispanic Men.” American Sociological Review 57(5):661-78. [30] Toussaint-Comeau, Maude. 2006. “The Occupational Assimilation of Hispanic Immigrants in the U.S.: Evidence from Panel Data.” International Migration Review 40(3):508-36. 20

[31] Wilson, Franklin D. 1985. “Migration and Occupational Mobility: A Research Note.” The International Migration Review 19(2):278-92. [32] Zourkaleini, Younoussi and Victor Pich. 2007. “Economic Integration in an Urban Labor Mar- ket: Does Migration Matter? The Case of Ouagadougou, Burkina Faso.” Demographic Research 17:497-540. [33] Zuberi, Tukufu and Amson Sibanda. 2004. “How Do Migrants Fare in a Post-Apartheid South African Labor Market?” International Migration Review 38(4):1462-91. 21