The Pennsylvania State University
The Graduate School
The Department of Human Development and Family Studies
YOUTH MIGRATION AND YOUTH TRANSITIONS:
THE HAITIAN EXPERIENCE
A Dissertation in
Human Development and Family Studies
and Demography
by
Jessica Heckert
 2013 Jessica Heckert
Submitted in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
December 2013
The dissertation of Jessica Heckert was reviewed and approved* by the following:
Rukmalie Jayakody
Associate Professor of Human Development, Family Studies, and Demography
Dissertation Advisor
Chair of Committee
Susan McHale
Professor of Human Development and Family Studies,
Director, Social Science Research Institute
D. Wayne Osgood
Professor of Crime, Law, and Justice and Sociology
Affiliate Professor, Human Development and Family Studies
Kevin J.A. Thomas
Associate Professor of Sociology, Demography, and African Studies
Steve Zarit
Distinguished Professor of Human Development and Family Studies
Head of the Department of Human Development and Family Studies
*Signatures are on file in the Graduate School
iii
ABSTRACT
The transition to adulthood in developing countries is undergoing dramatic
transformations, and many rural youth view migration as a promising strategy for a successful
future. This dissertation in composed of three studies on youth migration in Haiti. The first study
examines two characteristics of youth migration: youth’s motives and parents’ continued
provision of financial support to migrants. Findings reveal that education motivates nearly a
quarter of youth migration episodes. Labor motivated migration becomes increasingly common
between the ages of 10 and 24, and family-tied migration becomes less common. Findings also
reveal that nearly two-thirds of youth migrants receive financial support from their families.
Though female youth are more likely to be migrants, they are less likely to receive financial
support from their families. Youth migration is discussed in the context of the changing labor
market, and as a whole, education migration should be considered part of a continued parental
investment strategy.
The second study examines education migration following the completion of primary
school from a within family perspective. Multilevel logistic regression models compare outcomes
between siblings, and findings suggest that parental perceptions of how smart their children are
and how well their children do in school are strongly associated with education migration
following primary school. There was no evidence to support the hypotheses that having older,
migrant siblings is positively associated with education migration or that having younger siblings,
who can take on the responsibilities of household chores, eases the departure of older siblings.
Findings also suggest that families have reduced the use of education migration as a strategy for
investing in their children’s education as a response to the 2010 Haitian earthquake.
The third study explores the context and timing of early sexual experiences among ruralto-urban migrant adolescents in Haiti by examining their early sexual behaviors, knowledge, and
iv
attitudes. I first examine the timing of sexual initiation as it relates to migration using three
competing hypotheses—adaptation, disruption, and selection. Findings reveal that a disruption
hypothesis is that most plausible explanation among female migrant youth, who are less-likely to
initiate sexual initiation near migration. Findings also reveal weak associations between migration
and sexual initiation among boys. Furthermore, migrant adolescents accumulate less protective
knowledge and endorse premarital sex similarly to non-migrants. This study finds evidence that is
contrary to the overwhelming assumption that migrant youth risk deleterious sexual and
reproductive health outcomes. The high aspirations of migrant youth provide a likely explanation
for these findings.
v
TABLE OF CONTENTS
List of Figures .......................................................................................................................... vii
List of Tables ........................................................................................................................... viii
Acknowledgements .................................................................................................................. ix
Chapter 1 Educationally Motivated Youth Migration in Haiti: Foraging One’s Future
with Continued Family Investments ................................................................................ 1
Introduction ...................................................................................................................... 1
Youth Migration ....................................................................................................... 2
A Family-Based Strategy ......................................................................................... 4
Economic Change, Labor Demands, and Migration ................................................ 5
Methods............................................................................................................................ 7
Data-Haiti Youth Survey 2009 ................................................................................. 7
Variables .................................................................................................................. 8
Descriptive analysis of migration motives ............................................................... 10
Event history analysis of youth education and labor migration ............................... 11
Provision of support ................................................................................................. 12
Results .............................................................................................................................. 14
Migration Motives .................................................................................................... 14
Provision of Financial Support ................................................................................. 18
Discussion ........................................................................................................................ 21
Provision of support ................................................................................................. 22
Limitations ............................................................................................................... 23
Future Research ........................................................................................................ 24
Conclusion................................................................................................................ 25
Chapter 2 The Role of Child Endowments and Sibling Experiences in Determining
Parents’ Investments in Education Migration .................................................................. 34
Introduction ...................................................................................................................... 34
Education opportunities in rural Haiti ...................................................................... 35
Investing in education migration .............................................................................. 36
Sibship characteristics and education migration ...................................................... 38
Education migration responses to event shocks ....................................................... 39
Methods............................................................................................................................ 40
Data .......................................................................................................................... 40
Assessment of generalizability ................................................................................. 42
Variables .................................................................................................................. 43
Analysis .................................................................................................................... 44
Results .............................................................................................................................. 47
Descriptive results .................................................................................................... 47
Parents’ perceptions of children’s endowment: Multivariate models ...................... 48
Sibship characteristics: Multivariate results ............................................................. 49
Earthquake consequences ......................................................................................... 49
Discussion ........................................................................................................................ 50
vi
Limitations ............................................................................................................... 51
Conclusion................................................................................................................ 52
Chapter 3 Gendered perspectives on the intersection of migration and sexual initiation
among Haitian adolescents............................................................................................... 58
Introduction ...................................................................................................................... 58
Theoretical Framework ............................................................................................ 59
Method ............................................................................................................................. 63
Haiti DHS-Data ........................................................................................................ 64
Haiti DHS- Analysis................................................................................................. 66
HYTS- Data ............................................................................................................. 67
HYTS Variables ....................................................................................................... 69
HYTS-Analysis ........................................................................................................ 70
Results .............................................................................................................................. 71
Descriptive Results: Timing of Sexual Initiation ..................................................... 71
Event-History Results: Timing of Sexual Initiation ................................................. 72
Descriptive Results: Knowledge and Attitudinal Change ........................................ 74
Growth Curve Results: Knowledge and Attitudinal Change: .................................. 75
Discussion ........................................................................................................................ 76
Limitations and Future Research.............................................................................. 79
Conclusion................................................................................................................ 80
References ........................................................................................................................ 88
vii
LIST OF FIGURES
Figure 1-1. Migration Motives by Age and Gender................................................................. 26
Figure 1-2. Proportion of youth who experience education of labor migration by gender. ..... 27
Figure 3-1. Sexual and Reproductive Health Knowledge Patterns During Migration............. 81
Figure 3-2. Change in Endorsement of Premarital Sex During Migration .............................. 82
viii
LIST OF TABLES
Table 1-1. Descriptive Statistics: Characteristics of Education and Labor Migrants
Compared to the Complete Sample in the 2009 HYTS ................................................... 28
Table 1-2. Discrete-Time Event-History Analysis Predicting Education Migration ............... 29
Table 1-3. Discrete-Time Event-History Analysis Predicting Labor Migration ...................... 30
Table 1-4. Descriptive Differences Among Non-Migrants, Departed Youth, and Youth
Who Live Separately from Their Parents......................................................................... 31
Table 1-5. Heckman Probit Models Predicting Provision of Support for Youth Who
Departed the Household within the Past Three Years...................................................... 32
Table 1-6. Heckman Probit Models Predicting Provision of Support for Youth Who Live
Separately from Still-Living Parents................................................................................ 33
Table 2-1. Comparisons between HYTS respondents and the 2005-06 Haiti DHS ................ 53
Table 2-2. Characteristics of HYTS Respondents ................................................................... 54
Table 2-3. Tests of Parents’ Perceptions of their Children’s Endowments in Multilevel
Logistic Regression Models Predicting Education Migration. ........................................ 55
Table 2-4. Tests of Sibship Structure Characteristics Endowments in Multilevel Logistic
Regression Models Predicting Education Migration. ...................................................... 56
Table 3-1. Description of DHS Analytic Sample .................................................................... 83
Table 3-2. Migration Variable Predicting First Sex Among Female Adolescents................... 84
Table 3-3. Migration Variable Predicting First Sex Among Male Adolescents ...................... 85
Table 3-4. Descriptive Statistics HYTS Respondents ............................................................. 86
Table 3-5. Results of Growth Curve Analysis HYTS .............................................................. 87
ix
ACKNOWLEDGEMENTS
I am grateful to the many sources that helped fund data collection for this dissertation.
The National Science Foundation Graduate Research Fellowship Program (Award No. DGE0750756) and a predoctoral Traineeship from the Eunice Kennedy Shriver National Institute of
Child Health and Human Development (T-32HD 007514) funded my time and early research
trips. Data collection was funded by the Africana Research Center, an Early Career Award from
the Center for Global Studies, the College of Health and Human Development, and the Ruth
Ayers-Givens award, all from the Pennsylvania State University.
Without my friends and colleagues in Haiti, the fieldwork that formed several key
components of this dissertation would likely have not been possible, and surely would not have
been as enjoyable. I am particularly thankful to the Mayoral Council of Anse-a-Pitres for their
permission to conduct this research. Francia Valescot, Alpacite Brene, St. Jeanne Noel, and Jean
Jacques Thiebaut crossed mountains paths and streams on foot, motorcycle, and donkey in search
of every last respondent. My data entry assistants, Mario and Yasariz, along with Francia and
Brene, carefully turned audio recordings and the pages of completed surveys into analyzable data.
During my time in Haiti, I was thankfully never without a home. Micheline, Jeanne,
Josette, Marie Carmel, Carline, and Nancy ensured that I had a bed and a full stomach at the end
of every day. Above all, I am deeply appreciative of Mateo Ramon Ramon, the director of
ProSaluif, who consistently solves seemingly insurmountable challenges with his calm,
optimistic, and determined demeanor.
Finally, I am also grateful to my advisor Ruk Jayakody who helped refine my vision for
this project and to my committee members Susan McHale, Wayne Osgood, and Kevin Thomas
who provided thoughtful and constructive insight.
1
Chapter 1
Educationally Motivated Youth Migration in Haiti: Foraging One’s Future
with Continued Family Investments
Introduction
The transition to adulthood in developing countries is undergoing dramatic
transformations, and many rural youth1 view migration as a promising strategy for a successful
future. Burgeoning globalization exposes youth to new opportunities and ideas through the
increased availability of transportation and communication infrastructures that facilitate
geographic mobility and exposure to novel ideas (Behrman & Sengupta, 2005). As a result, many
youth view non-agricultural labor opportunities and educational attainment as part of the formula
for successful adulthood (Mensch, Grant, & Blanc, 2006). For rural youth who lack these
opportunities locally, migration often facilitates a path toward achieving these goals.
Despite the high prevalence of youth migration and the age-graded life course
experiences that make youth’s migration experiences distinct from adults’, there has been little
work conceptualizing youth migration and identifying what differentiates youth migration from
migration during other life course phases (Tienda, Taylor, & Moghan, 2007). Youth marks a time
of transition from household dependent to an interdependent contributor to the household’s wellbeing; however, youth often still depend on their parents’ financial support and social networks
during this time.
In this paper I examine two of the primary features that differentiate youth migration:
first, youth’s migration motives, and second, parents’ continued provision of financial support.
1
I define youth as those aged 12 to 24.
2
Haiti provides an ideal context to study this experience: the escalating demand for education,
cultural values that encourage migration, and land pressures that limit farming opportunities act
jointly to encourage geographic mobility among young people (Bredl, 2010; Mintz, 2010;
Schwartz, 2009). Historically, male labor migration played a facilitative role in the transition to
adulthood; male youth demonstrated their independence and earned the financial capital to begin
farming and construct a homestead—key markers of marriageability (Schwartz, 2009). But social
and economic changes have transformed the motivations and experiences of youth migrants. Key
among these is that youth increasingly migrate for educational opportunities—a pattern that has
become prevalent throughout developing countries (McKenzie, 2008).
In the following sections, I provide background information on youth migration based on
the current literature. I describe why youth might be motivated to migrate and how youth
migration functions as part of families’ strategies to adapt to social and economic change. I then
describe Haitian migration in the context of the changing Caribbean economic system, which has
altered the skills required in the labor market.
Youth Migration
Migration is one of many strategies that families use to adapt to the demands of changing
global contexts. Youth participate in these strategies, not only as those who accompany their
families, but as individuals who pursue migration with their own aspirations (Punch, 2007;
Yaqub, 2009a). Multiple factors encourage youth migration, including direct motives, such as
employment and education, as well as cultural and psychosocial factors in which direct motives
are embedded. The primary migration pattern among developing country youth is from rural to
urban areas (Yaqub, 2009b), and the unequal distribution of resources and opportunities between
these areas underpins the flow of migrants. The flow of rural-to-urban labor migrants in particular
3
is maintained by the potential for more lucrative employment prospects in urban areas (Todaro,
1969). In Haiti, where population growth, limited agricultural land, and land’s diminished
production potential reduce the possibility of earning a living through farming, youth perceive
higher returns to the urban-sector jobs available in Port-au-Prince and large towns (Mintz, 2010).
The assumption that most migration is motivated by labor opportunities permeates the
academic literature. However, youth may also migrate to seek education opportunities (Brauw &
Giles, 2008; Crivello, 2009). Whereas most rural youth in developing countries live within
walking distance of primary schools, secondary schools are much more distant, and even when
schools are available, many parents perceive that urban schools provide superior education
(Lloyd, 2004). Youth and their families have high expectations for the potential returns to
education, and when rural areas lack these opportunities, social mobility and geographic mobility
are inextricably linked (Bjarnason & Thorlindsson, 2006).
Labor and education opportunities directly influence rural-to-urban migration, but
migration may be part of a coming-of-age experience that is embedded in cultural and
psychosocial motivations (Punch, 2007). Circular patterns characterize intra-Caribbean migration,
and many laborers are motivated to earn money so that they can return home with a higher status;
in doing so, they demonstrate economic viability and set themselves aside as adults in the eyes of
Haitian society (McElroy & de Albuquerque, 1988; Schlesinger, 1968). Return migrants may be
perceived as more mature, experienced, and ready for adult roles, and the relative status of
returned migrants further incentivizes migration among youth. Returned and visiting youth
transmit migration-related values through the possession of material goods and worldly
knowledge to their peers who may be especially influenced by older and higher-status peers
(Castellanos, 2007; Collins & Steinberg, 2006). In sum, the presence of returned and circular
migrants contributes to the supply of eager young migrants who seek to legitimize their adulthood
and gain status through migration.
4
A Family-Based Strategy
Children and youth’s roles within family-based migration strategies can be
conceptualized in three ways: those who migrate with their families, those who are left behind
when parents migrate, and those who migrate independently from their parents (Tienda et al.,
2007). The likelihood of a young person migrating separately from his or her parents begins an
upward trajectory at age 12 and increases most steeply between ages 15 to 17 (Yaqub, 2009b).
Despite the increasing prevalence of migration during this age and the saliency of youth’s own
migration aspirations, current research has given limited attention to independent youth migration
due to the lack of appropriate data and theoretical frameworks that pertain specifically to youth
(Yaqub, 2009a).
Oftentimes, youth migrate as part of their connection to a larger family system (though
the phrase independent youth migrant suggests otherwise). They may migrate with friends or kin,
reside with the same friends or kin or others at their destinations, and continue to receive financial
support from their families (Thorsen, 2010). Parents may also continue to exert authority over
migrant youth via expectations and monitoring by other family members (Castellanos, 2007). The
combination of these practices reinforces existing social ties among extended family members at
the origin and destination (Hareven, 1982).
Among youth migrants, parental financial investments may play a critical role in
maintaining their well-being. Households in Port-au-Prince, Haiti’s capital city, are often a
conglomeration of extended kin and friends, and one-third of Port-au-Prince households receive
regular remittances from the countryside—though relationships between senders and recipients
was not identified (Manigat, 1997). The need for financial investments may especially be the case
among education migrants who require sustained economic investments for school fees (which
must be paid even for public schools), school materials, and living costs. Thus, especially for
5
education migrants, resource transfers occur primarily from rural parents to their urban children.
This pattern differs from the commonly accepted expectation in the literature that resources will
flow from the migrant to their rural family after an initial brief investment period (Massey et al.,
1993).
Gender also plays a critical role in how families manage youth migration—both whether
youth migrate and their experiences at the destination. Families may resist investing in girls’
opportunities if the labor market is less hospitable to women (Buchmann, 2000), and girls’ more
intensive household contributions may tie them to the home (Hsin, 2007). However, the factors
that may discourage parents from allowing their daughters to migrate may encourage urban
families to receive girls, especially those that will provide domestic service (Moya, 2007). From
girls’ own perspectives, they may view migration as an opportunity to break from genderedrestrictions at home while maintaining other family obligations through migration; they may
accomplish this by sending home remittance and reinforcing relationships with kin at the
destination (Castellanos, 2007; Thorsen, 2010).
Economic Change, Labor Demands, and Migration
The opportunities available to potential youth migrants depend on the social and
economic characteristics of the larger region. Within the traditional Haitian agrarian society,
circular labor migration played a role in the transition to adulthood by facilitating economic
viability and marking adulthood status (Schwartz, 2009). Haitian migrants have long fulfilled a
need for laborers in physically intensive agricultural production (e.g. sugar cane harvesting) in the
Caribbean region and its periphery (Gammage, 2004). Intra-Caribbean migratory patterns follow
wage labor opportunities, and Haiti is the region’s largest producer of migrant laborers (Ferguson,
6
2003). This phenomenon is due both to Haiti’s relatively large population, its political instability,
and relative poverty.
During recent decades, as sugar prices dropped, and Caribbean countries lost their
position as Europe’s preferential choice for warm climate agricultural products; the Caribbean
economy transformed from agricultural production with reliance on sugar, bananas, and coffee,
into a service-based economy with tourism and assembly as its primary pillars (Palmer, 2009).
These economic transformations have largely excluded Haiti, due to political instability and
economic embargos (Metz, 2001). However, the economic changes in nearby countries directly
influenced Haiti via changing skill demands for Haitian migrants. Whereas decades ago the
typical young migrant was male and spent physically laborious days cutting sugar cane, those
seeking jobs in tourism are often expected to speak multiple languages, operate computers, or
produce art to sell. These varieties of service-sector jobs also provide increased opportunities to
women. Additionally, low-skill labor demands have diversified. Inexpensive Haitian labor
supports the construction industry, and as women in more industrialized Caribbean countries
enter the formal labor market, they require domestic workers, a role oftentimes filled by Haitian
women and adolescent girls (Palmer, 2009).
Though the opportunities available to labor migrants have changed, contemporary youth
still conclude that migration is a pathway to a productive future. Furthermore, because of the
skills required in some of the sectors where migrants aspire to work, educational attainment
beyond what is commonly available in rural areas is also necessary, and many youth migrate to
first seek educational opportunities. This shift means longer investment periods for families,
especially those of educational migrants.
7
Methods
Building on this literature, I examine youth migration with attention to youth’s motives as
well as families’ provision of support for youth migrants. The analyses draw on data from the
nationally representative 2009 Haiti Youth Survey (HYS). I first examine migration motives and
how these motives differ between the ages of 10 and 24. Two primary motives are education and
labor opportunities, and I then examine how time spent as an education or labor migrant is
associated with youth’s characteristics using a discrete-time event history analysis approach.
Specifically, I consider gender, region of birth, parent vital status, school enrollment timing,
current place of residence, wealth of their current household, their current activities (school and
work), and their educational attainment. I then explore how families provision support for youth
migrants and examine the characteristics of youth and households that are associated with
parents’ provision of support for migrant youth using HYS data and Heckman probit models.
Data-Haiti Youth Survey 2009
The 2009 HYS is a nationally representative household survey that sampled youth aged
10 to 24 years (Lunde, 2009). Selected households were identified through a multi-stage,
stratified random design. Each household survey included (i) background information on the
entire household (ii) a household roster with detailed information on residents aged 10 to 24 years
(iii) a detailed individual interview, including a migration history module, with one randomly
selected youth from the household, and (iv) information on all youth who left the household
within the past three years including their current age, age at migration, and current activity at the
destination. In some contexts, where boarding schools are common for secondary school students
and labor migrants reside in factory dormitories, a household survey would not be adequate to
8
study youth migrants. However, neither boarding schools nor factory dormitories are common in
Haiti, and a household survey provides an adequate sample.
Using these data, I identify migrant youth through three approaches: (i) the migration
histories of youth selected for an in-depth interview (one per household) and (ii) youth who
departed from the household during the past three years according to the primary household
respondent, and (iii) youth who reside without either of their living parents (separated youth)
according to the household roster. Among separated youth, a portion may be migrants. However,
some may also be left behind by migrant parents or have established their own households, and I
consider this possibility when using data on separated youth. Additionally, because marriage is
most often associated with the formation of a new household, married youth are not included in
the analysis of the provision of support.
I use the migration histories from 1,318 youth to compare migration motives between the
ages of 10 and 24, calculate the proportion of youth who have ever migrated for education or
labor reasons across this same age span, and model the likelihood of being an education or labor
migrant in any particular year. In all cases, the migration event had to exceed three months in
order to be included. I then examine the provision of support for the 269 youth who have left the
household during the past three years and the 725 youth who reside separately from both living
parents. In both cases, I use a Heckman probit model, which simultaneously models (i) the
provision of support among migrant or separated youth and (ii) the probability of being a migrant
or separated youth compared to the 1,971 youth living in a household with at least one parent .
Variables
In migration histories, youth reported their primary reason for migrating, their primary
activity at the destination, and the dates for each migration event. I classify these responses into
9
four categories. Education migrants were those who reported that they migrated in order to attend
school, attended school at the destination, and did not migrate in order to reunite with family.
Labor migrants worked or looked for work at the destination. Family migrants reported that their
primary reason for migrating was to follow or rejoin their family. A number of migrants did not
work, look for work, or attend school, nor did they migrate to join family; I classify these
episodes as idle/other. I calculate their age in years for each event. Only events that occurred after
the age of 10 are analyzed.
Background information on household members was reported by the primary household
respondent. Information included gender, current age, region of birth, parental vital status, age at
primary school enrollment, per capita income, current activities (i.e., work, school), and
educational attainment.
For this analysis, the country is divided into five regions. Port-au-Prince includes the
capital city and the surrounding metro area. The Southeast includes the West2 and Southeast
departments. The North includes the North and Northeast departments. The South includes the
Grande-Anse, Nippes, and South departments. The Central Region includes the Artibonite,
Northwest, and Central Plateau departments. Timing of first school enrollment is divided into ontime (age 5 or 6), moderately delayed (age 7 to 9), and severely delayed or never (age 10 or
older). The earnings of all household members were reported; per-capita income was reported and
divided into quintiles (see Lunde, 2009). The educational attainment of each household member
was reported, and I include a variable for whether any household adult (over age 25) has
completed primary school. I divide youth’s own education attainment into three categories: has
not completed primary school, completed primary school, and completed lower secondary school.
2
The West Department is in the central part of the country, surrounding Port-au-Prince, not in the western
part of the country. The name likely carries over from when Haiti comprised a larger portion of the island.
10
With regard to the provision of support, it is a binary variable that describes whether the
family provides money (to live or to travel) or gifts to the departed or separated youth. In the case
of both departed and separated youth, this information was reported by the primary household
respondent. The analysis of financial support also accounts for gender, urbanicity of current
household, youth age, whether the youth is working or in school, whether an adult in the
household has completed primary education, and the per-capita income quintile. Additionally,
with regard to separated youth, the household respondent reports whether the youth was born in
the current household; to an extent this helps control for youth who are left behind by migrant
parents.
With regard to departed youth, the household respondent reported whether the youth
returns to visit, whether the youth went to an urban or rural destination, and the time in months
since he/she departed. For youth who visit their families, resource exchanges might instead occur
during visits, rather than being sent. The few international departures were coded along with
urban departures. Additionally, youth who have been gone for a shorter amount of time have had
less time for parents to send money.
Descriptive analysis of migration motives
First, I calculate the proportion of migration episodes motivated by education, labor,
family migration, and idle/other by age group (10-12, 13-15, 16-18, 19-23). Age refers to the age
when each unique episode began, and these calculations include all migration events reported by
youth.
Then, because of the specific importance of labor and education migration as youth
transition into their adult roles, I further examine when youth experience their first education and
labor migration events. I calculate the Kaplan-Meier survivor estimates for experiencing a first
11
event separately for education and labor migration. I present these calculations as the proportion
who have ever experienced an education or labor migration episode (Pr(migration) = 1 –
Pr(survival)). I use the log-rank test for equality of survivor functions to determine if the survivor
function differs by gender.
Event history analysis of youth education and labor migration
Migration is often cyclic: migrants may return home and may experience multiple
migration episodes. Therefore it is important to consider what factors are associated with
becoming and remaining a migrant. To model this approach I use a discrete-time event-history
framework to predict the log-odds of being a migrant in any given year. Respondents enter the
risk set at age 10, when there is a notable increase in education and labor migration events. Even
after an initial migration episode, they remain at risk for migrating; thus they are not censored
until the age of survey administration.
Logistic regression is used to model an approximation of the hazard of being a migrant in
any given year using the basic form of the equation:
(
)
I conduct these analyses separately for time spent as en education migrant and time spent
as a labor migrant. For each outcome, Model 1 includes characteristics of the individual that can
likely be attributed to occurring before age 10 when respondents begin to experience the risk of
migration. These include gender, time-varying categorical age variables, region of birth, parent
vital status, and school enrollment timing. Model 2 includes additional characteristics that
represent current their current circumstances: type of place of residence (rural or urban), current
region of residence, income quintile of current household, whether an adult in the household has
12
completed primary education, currently in school, currently working, and current educational
attainment.
Provision of support
After modeling the characteristics that are associated with education and labor migration
among youth, I then examine how youth’s families provide them with financial support. The
available data afford two different approaches for analyzing parents’ provision of financial
support: reports of youth who departed the household during the past three years and youth who
reside separately from both of their still-living parents. Data from departed youth provide
information on the sending household and include all youth, regardless of their destination.
However, families can only provide limited information about these youth at their destination.
Data from youth residing separately provide more detail on their current circumstances,
but this approach cannot include information about the sending household, and by nature of the
survey design, excludes any youth who migrate internationally. Additionally, not all youth
residing separately from their parents are migrants; separated youth may also be left behind by
migrant parents. To analyze the provision of support, I separately model data from both departed
and separated youth to provide a more comprehensive picture.
I model the provision of support using a two-part Heckman probit model, which is
appropriate for when a dependent variable (provision of support) is only observed among a nonrandom group of respondents (departed or separated youth) (Van de Ven & Van Praag, 1981).
The first stage models selection using a probit model describing the underlying latent variable
(propensity to be departed or separated):
13
The dependent variables is only observed among those for whom the propensity to
migrate is greater than zero. The second stage models the binary dependent variable and describes
the underlying relationships between the explanatory variables and the propensity to provide
financial support.
This two-equation approach also allows for different independent variables to predict the
probability of providing support and the probability of departing or living separately. Using data
on departed youth, I predict selection into departure as a function of youth gender, youth age,
youth’s current school enrollment and employment status, whether the sending household is a
rural household, whether an adult from the sending household has completed primary education,
and the income quintile of the current household. Then, among departed youth, I model the
receipt of financial support as a function of youth gender, youth age, youth’s current school
enrollment and employment status, whether an adult from the sending household has completed
primary education, the income quintile of the current household, whether the youth visits the
household, the number of months since leaving, and whether the youth went to an urban
destination.
Using available data on separated youth, I model the selection equation using youth
gender, youth age, whether the current location is rural, youth’s current school enrollment and
employment status, and whether the youth was born in the household where he or she currently
reside. Then, among separated youth, I model the receipt of financial support as a function of
these same variables, plus whether an adult in the receiving household has completed primary
education and the wealth quintile of the household.
14
Results
Migration Motives
I first examine the primary migration motives (education, labor, family dependent, or
idle/other) of the 769 migration events reported between the ages of 10 and 24 (Figure 1-1).
Across all four age groups and both genders, educational migration accounts for nearly a quarter
of all migration episodes. An exception is among males aged 19 to 23, for whom the percent of
educationally motivated migration episodes is somewhat smaller (11.7%). Labor migration
becomes a more common migration motive across this same age span. Among 10 to 12 year olds,
labor migration is almost twice as common among boys (17.4% of episodes) compared to girls
(9.3%). It becomes increasingly more common with age, and is the most common reason to
migrate among male and female youth aged 16 and 23.
Whereas labor migration is increasingly common across the age groups, family migration
becomes less common. Among the youngest (aged 10 to 12), both boys and girls are most often
tied migrants whose primary motivation is to accompany or join family members (57.4% of boys
and 67.4% of girls). The percent of episodes motivated by accompanying or rejoining family
members declines from the youngest to the oldest age groups. And among the oldest age group
(aged 19 to 23), only 24.6% of moves among male youth and 11.8% among female youth are to
accompany or rejoin family members.
A number of youth also described themselves as idle. This accounts for a miniscule
number of those aged 10 to 12. However, among the 19 to 23 year olds, 17.4% of male youth and
25.8% of female youth migrate for reasons that are not adequately captured in this survey. One
explanation for these migrants is that they may be returning migrants or those who migrate
primarily to forge their independence with no more specific intentions.
15
To further examine how widely education and labor migration are experienced, Figure 2
shows the weighted proportion of youth who have ever experienced an education or labor
migration episode, divided by gender, using Kaplan-Meier survivor estimates. The proportion of
youth who have ever migrated to attend school increases steadily between ages 10 and 20 and
then tapers off. Education migration is slightly higher among female youth, but the difference is
not statistically significant. The proportion of youth who have ever experienced labor migration
increases more slowly than education migration. The proportion experiencing labor migration
surpasses educational migration around age 20 for males and around age 23 for females. The
differences by gender are trivial.
Next I describe the background characteristics of the sample that I use to examine the
characteristics associated with spending time as an education or labor migrant during between the
ages of 10 and 24. I first compare characteristics of youth who have ever experienced education
or labor migration, accounting for weighting and complex survey design (Table 1-1). Both
education migrants (58.2%) and labor migrants (53.1%) are somewhat more likely to be female,
compared to the complete sample (51.0%), but the difference was not statistically significant.
Both were older (education = 19.3; labor = 20.8) compared to the complete sample. Education
migrants are less likely to have been born in the greater Port-au-Prince area (10.9%) and GrandAnse, Nippes, and South (13.2%), compared to the complete sample (19.6% and 25.2%
respectively); they are more likely to have been born in the North and Northeast (26.6% vs.
16.0%) and the Artibonite, Central Plateau, and Northwest (31.0% vs. 23.1%). Both education
and labor migrants have more commonly experienced the death of a parent. Education migrants
are most likely to have started school on time, and labor migrants less commonly began school on
time and more commonly experienced a moderate (between ages 7 and 9) or severely delayed
school start (after age 10) or never started.
16
Next, I describe the current characteristics of youth who have ever experienced education
or labor migration. Education migrants are more likely than others to live in the greater Port-auPrince area (44.6% vs. 30.1%) and less likely to live in the Central Region (7.4% vs. 24.4%). This
is likely the result of the unequal distribution of secondary schools, which are heavily
concentrated in Port-au-Prince. Those ever experiencing labor migration are geographically
distributed similarly to the complete sample. Education migrants are more likely to be living in
the wealthiest households (24.4% vs. 13.3%) and more commonly live in a household with an
adult who has completed primary education (50.1% vs. 33.3%). Labor migrants, on the other
hand, do not live in households whose wealth differs from the complete sample, and they less
commonly live in a household with an adult who completed primary education (22.8%).
Education migration is associated with being currently enrolled in school (80.9% vs. 69.1%), but
ever experiencing labor migration is negatively associated with current school enrollment
(18.2%). Those ever experiencing labor migration, however, are much more likely to be currently
working (50.3%) compared to the complete sample (12.6%). Educational migrants also have
overall higher educational attainment than the complete sample; 66.2% (vs. 40.5%) have
completed lower-secondary school (9th grade).
I then move to the multivariate analysis and report the results of discrete-time event
history models that predict being an education (Table 1-2) or labor migrant (Table1-3) in any
given year. For each outcome Model 1 includes the characteristics that could reasonably be
attributed to the time before migration, and Model 2 includes current characteristics. For both
education and labor migrants, being a migrant is more common among older age groups. Earlier
analyses described when youth were likely to leave home for the first time; these findings
demonstrate that they are more likely to remain away from their natal homes at later ages.
Youth born outside the greater Port-au-Prince area are more likely to become education
migrants than those born in Port-au-Prince, but place of birth is not associated with labor
17
migration, except that those born in the Central Region are less likely to be labor migrants in
Model 2.
After controlling for other factors, parental death is not associated with education
migration. However, paternal orphans have somewhat higher odds of labor migration (O.R.= 1.48
p ≤ 0.10), and maternal orphans experience significantly higher (O.R.= 1.82; p ≤ 0.05).
Turning to migrant youth’s current living situations, education migrants are much less
likely to live in rural areas (O.R.= 0.34; p ≤ 0.01). After accounting for urbanicity, their
distribution does not differ significantly from one region to another, and their concentration in
urban areas is likely due to the absence of secondary schools in rural areas. Those who are
currently or have ever been labor migrants are more likely to live in the Southeast than in the
Port-au-Prince metro area. After controlling for other factors, there are no differences in the
current household wealth of education migrants, but education migrants are somewhat more
likely to live in a household with an adult who has attained primary education (O.R.= 1.48; p ≤
0.1). Labor migrants more commonly reside in the fourth quintile of household income (wealthy).
This may be a result of income or resources that they have brought into the household, or because
they reside with their employers.
Education migration is strongly associated with current educational attainment. Education
migrants are twice as likely to currently attend school (O.R.= 2.02; p ≤ 0.01) and more likely to
have completed primary school (O.R.= 3.55; p ≤ 0.01) and lower secondary school (O.R.= 3.37; p
≤ 0.01). In contrast, labor migration among youth is associated with lower educational attainment.
They are less likely to be in school (O.R.= 0.29; p ≤ 0.001), more likely to be working (O.R.=
3.78; p ≤ 0.001), and less likely to have completed lower secondary school (O.R.= 0.36; p ≤
0.05).
18
Provision of Financial Support
To examine the provision of financial support from families to migrant youth, I compare
the characteristics of the non-migrant reference group (n = 1,971), youth who departed the
household within the past three years (n= 276), and youth living separately from their still-living
parents (n = 726) ) (Table 1-4). Among those who departed the household, 68.0% receive
financial support, as do 64.0% of separated youth. The difference between these two groups is
small and in-line with reporting bias that would encourage households to over-report sending
money and under-report receiving money. Female youth are significantly more likely to have
either departed the household with the past three years (58.7% vs. 45.6%) or be separated from
their parents (53.8%). This finding occurs even though married youth are excluded from the
analyses.
The non- migrant group is predominantly rural (66.9%), and 70.4% of departed youth
come from rural households, whereas significantly fewer separated youth live in rural areas
(48.0%). Both departed youth and separated youth are more likely to be from the older age groups
(i.e., 19-21, 22-24), and less likely to be from the younger age groups (i.e., 10-12, 13-15).
Regarding school enrollment and employment status, departed youth are much less likely
to be in school (32.4% vs. 80.5%) and much more likely to be working (28.5% vs. 7.8%). Similar
findings apply to separated youth, but the differences, though significant, are less pronounced;
70.6% are in school, and 12.4% work.
Among departed youth, the household characteristics refer to the household from which
they departed, and among separated youth, these characteristics refer to their current residence.
Departed youth more often come from the wealthier income quintile and less often from the
poorest income quintile, and those living without their parents are more often living in the
wealthiest households and less often in the poorest households. Separated youth were also much
19
less likely to have been born in their current household of residence (23.9% vs. 62.2%),
suggesting that many of them may in fact by youth migrants, not just those who have been left
behind by migrant parents.
Among departed youth, 45.1% visit the household regularly, 74.7% left for an urban area,
and they have been gone an average of 16.5 months.
Examining the multivariate results of whether families provide financial support to
migrant youth. The Heckmann probit model produces two equations: a selection equation that
estimates the probability of departing the home or living separate from one’s parents and an
equation the estimates the probability of this individuals receiving support. The value of Rho is
the correlation between the error terms in the two equations of the model.
For both models, female youth have a higher probability of leaving the household
(b=0.36; p≤0.001) or living separately from their parents (b=0.18, p≤0.05). However, they have a
lower probability of receiving support than their male counterparts (departed: b=-0.27, p≤0.01;
separated: b=-0.16, p≤0.05). Youth experience a higher probability of departing rural households
(b=0.33; p≤0.001), but whether they go to an urban or rural area does not change the probability
of whether they receive support. Among separated youth, fewer live in rural areas (b=-0.18,
p≤0.05), but those who live in rural areas have a higher probability of receiving support (b=0.28,
p≤0.01). It is possible that youth living without their parents in rural areas have more commonly
been left behind by labor migrant parents who send remittances.
These models reiterate findings presented earlier in this paper that older youth are more
likely to migrate; however, whether parents provide support does not differ significantly across
the age groups. Both models suggest that departed youth and separated youth have a lower
probability of being enrolled in school (departed: b=-1.25, p≤0.001; separated: b=-0.29, p≤0.001),
and being enrolled in school is associated with a higher probability of receiving support
(departed: b=1.39, p≤0.001; separated b=0.41, p≤0.001). The effect of currently working was not
20
statistically significant, but both coefficients were in the opposite direction from those for school
enrollment. Separated youth are less likely to have been born in the household (b=-0.89,
p≤0.001), but those who were born in the household were more likely to receive support (b=0.81,
p≤0.001). Again, this is suggestive of a group of youth whose migrant parents send remittances.
The effects for whether an adult in the household completed primary education were not
significant.
Youth experienced a higher probability of departing from the households in the upper
three income quintiles, but their families less often send support after accounting for other factors.
Among youth who lived separately from their parents, the wealth of the current household where
the youth resided was not associated with receiving financial support. The lack of sending support
among wealthier families presents a paradox. It is possible that youth from wealthier families
need less support if they are afforded work opportunities by connections or educational
attainment. It is also likely that these families send their children to other, equally well-off
families with whom they have established family alliances. Hosting these youth and providing for
their daily needs may be part of an extended cycle of exchanges. Furthermore, accepting financial
support from the youth’s sending family would, in fact, mean that the receiving family would be
less able to request a favor of the child’s parents in the future. Thus, by not accepting support,
they leave open the doors for further exchange of favors and non-financial support.
Among departed youth, whether the youth visits the household is associated with a lower
probability of receiving support (b=-0.25, p≤0.05). This finding may occur because migrant youth
who return home may obtain support during their visits home rather than as a result of being sent
money or goods. Additionally, more time (in months) since departure is associated with a higher
probability of receiving support (b=0.01, p≤0.05).
21
Discussion
For many youth and their families, migration is a strategy for gaining access to
educational training and labor force opportunities. Previous research focuses on labor migration
opportunities (Tienda et al., 2007). However, this paper presents evidence that education
migration is also a prevalent strategy for many youth and their families: approximately onequarter of migration events among youth are with the specific intention of attending school, and
nearly 15% of youth migrate at least once to attend school.
Education migration should be conceptualized as an extension of families’ strategies to
further their children’s education. Those who enrolled in primary school later than their peers are
less likely to migrate for school, and education migration is strongly associated with continued
school enrollment and more advanced educational attainment. In contrast, youth labor migrants
more often began primary school late and are less likely to have completed lower secondary
school. Education migration may be one way that families compensate for a weak national
education system whereby rural youth often have very few opportunities at the completion of
primary school.
Moreover, another way that youth migration is unique from migration among adults is
that parents often continue to provide financial support for migrant youth. Continued financial
support and current school enrollment are strongly and positively correlated. And, again, this
suggests that education migration is part of families’ continued investments in education.
A notable findings from this paper’s analysis of financial support for migrant youth is
that although girls are more likely have left the home or be separated from their parents, they also
have a lower probability of receiving financial support from their parents. This finding prompts
the question of why this occurs. Parents may intentionally choose to invest less in their daughters
if they believe there will be fewer payoffs in the labor market (Buchmann, 2000); however the
22
increased availability of service-sector jobs suggests that opportunities are not unfriendly to
women. An alternative explanation may be that receiving households are more willing to take in
girls, because of girls’ household labor contributions, and parents may not be expected to send
additional financial support in this case. However, this may mean girls are burdened with
excessive household responsibilities in these homes, which may prevent school enrollment or
conflict with school attendance.
This study also encourages one to consider the role of youth migration with regard to
long-term migration strategies during the life course. Short-term migration intentions may lead to
a more permanent status, and hopes of long-term migration may be cut short by numerous
obstacles (Agadjanian, Nedoluzhko, & Kumskov, 2008). Or, shorter, internal migration events
may lead to further distances migrated, or develop into repeated seasonal migration. Moreover, in
the case of education migration, which facilitates skill development, it may lead to improved job
opportunities at more distant locations.
Provision of support
This study provides evidence that families continue to support migrant youth with money
and other goods. To shed some light on how this process occurs, I draw on ethnographic evidence
and open-ended survey responses from youth who migrated and complted a survey after
migration. These data were collected as part of the Haitian Youth Transitions Study in 2011 and
2012, which recruited a cohort of sixth graders from a designated geographic area in rural
southeast Haiti; 224 youth participated and 55 migrated.
Of the 55 migrant youth, 89% receive either goods and/or money from their families, and
80% of youth received something at least once a month. These remittances contribute
substantially to youth’s well-being, and 82% receive groceries and 73% receive money. In
23
contrast, only 20% of youth send anything home to their families, and most of these remittances
were token gifts. Only two reported sending money and four reported sending home food
products (oil and rice).
Parents distribute these goods to their migrants through their social networks and when
they travel to the location where their child is living. A common practice is for parents to recruit
the help of truck drivers on public transportation routes. A large sack of groceries may be placed
in the care of the driver, and when the vehicle arrives at its destination, the recipient or a member
of his/her household retrieves the sack. Additionally, parents may need to venture to the city
where their child is enrolled in school. Frequent, short trips to urban destinations are common
among women who sell in the markets and need to replenish their wares. During these trips,
parents also take time to distribute money or provisions to their migrant children.
Limitations
Despite the unique contributions of this study, it also suffers from limitations. One of the
primary concerns is that there is no way of examining the distance of these migration events.
Closer moves might be more common in the case of marriage or seeking improved or more
affordable housing. However, in the case of migrating specifically to attend school or seek work,
the effort and costs of migration would need to be outweighed by substantially better access to
these opportunities.
Another limitation is the inability to connect data from both sending and receiving
households. The lack of these data limit what conclusions can be drawn with regard to how
migration may reinforce or alter youth’s life course trajectories and about the relations between
sending and receiving households.
24
Future Research
One of the primary conclusions of this paper is that youth migration should be explored
as a unique phenomenon that differs substantially from migration among adults. Thus, it opens
many doors for future research on youth migration. This research should examine the processes
that lead youth to migrate and should also consider the consequences of youth migration on
individual life course trajectories.
The experiences of youth migrants are unique, and many advocacy and rights groups are
concerned for migrant youth’s well-being (Lane, 2008; Tienda et al., 2007). Migration
experiences may differ for youth because they are physically weaker than adults, have not
reached adult levels of psychological development, have accumulated less knowledge and fewer
life experiences, and participate in different age-graded socially-constructed roles (Yaqub,
2009a). Youth migrants also encounter new environments, ideas, and peers when exposure to
novel experiences may be especially influential (Collins & Steinberg, 2006). Furthermore youth
still depend on adults to fulfill physical resources, expand their social networks, and provide
social and emotional support. These characteristics may leave them more vulnerable to negative
experiences during migration. However, empirical evidence that justifies concerns about youth
migrants and suggests possible ways to ameliorate migrant youth’s well-being is limited.
One of the primary limitations face in studying youth migrants is the lack of appropriate
data. Youth migrants are difficult to identify in population-level data. Often they are not counted
in household rosters, particularly if they are domestic workers, live in the street, or misidentify
their residence as their natal home, and many surveys are purposefully timed to miss circular
migrants (Yaqub, 2009b). Moreover, collecting detailed data on migrants that disaggregates
individuals by age and companionship, in is beyond the scope of most developing countries’
current population monitoring and data collection approaches.
25
Conclusion
In sum, this paper highlights two unique features of youth migration: youth’s motives and
families’ provision of financial support. Youth migration has received limited empirical focus,
and education migration should be conceptualized as one way that families continue to invest in
their children’s education. It is important that future research continue to focus on both the causes
and consequences of youth migration and that data to answer key research questions in these
domains be generated.
26
Figure 1-1. Migration Motives by Age and Gender.
Migration Motives by Age and Gender
(n=769 migration events)
35.7
57.4
1.5
17.4
67.4
59.8
4.2
11.5
19.4
5.5
13.9
37.3
31.7
20.7
24.6
24.3
21.5
22.9
male
female
male
female
male
female
aged 13 to 15
Education
Labor
aged 16 to 18
Idle
11.8
25.8
23.8
aged 10 to 12
24.6
51.3
5.0
2.6
9.3
31.6
Family
17.4
34.8
46.4
27.7
11.7
male
female
aged 19 to 23
27
Proportion who ever migrated
Figure 1-2. Proportion of youth who experience education of labor migration by gender.
Proportion of youth who experience education or
labor migration by gender
0.25
0.20
0.15
0.10
0.05
0.00
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Age
Male Ed
Female Ed
Male Labor
Female Labor
28
Table 1-1. Descriptive Statistics: Characteristics of Education and Labor Migrants Compared to
the Complete Sample in the 2009 HYTS
N (unweighted sample size)
Female
Age: mean (SE)
Region of birth
Greater Port-au-Prince
Southeast
North
South
Central Region
Father is dead
Mother is dead
Enrollment timing
On time (age 5 or 6)
Moderately delayed (7-9)
Severely delayed (10+)
Rural
Region of residence
Greater Port-au-Prince
Southeast
North
South
Central Region
Income quintile
Poorest
Poor
Middle
Wealthy
Wealthiest
Currently in school
Currently working
Adult in HH with primary ed.
Current ed. attainment
Did not complete primary
Completed primary
Completed lower sec.
Complete
sample
1318
51.0
16.7 (.18)
Ever
Experienced
Education
Migration
149
58.2
19.3 (.28)
Ever
Experienced
Labor
Migration
130
53.1
20.8 (.30)
19.6
16.1
16.0
25.2
23.1
16.8
10.5
10.9
18.3
26.6
13.2
31.0
24.1
19.8
16.0
18.4
26.0
21.5
18.1
34.6
20.4
35.5
37.0
27.5
57.6
43.7
33.9
22.5
32.1
18.3
26.4
55.4
62.1
30.1
19.9
13.5
12.0
24.4
44.6
22.4
12.6
13.0
7.4
31.1
19.6
14.4
14.7
20.2
21.6
21.7
22.4
21.0
13.3
69.1
12.6
33.3
14.4
13.0
26.4
21.9
24.4
80.9
6.9
50.1
17.2
25.6
19.6
27.8
9.7
18.2
50.3
22.8
30.7
28.7
40.5
4.7
29.1
66.2
37.7
21.4
41.0
29
Table 1-2. Discrete-Time Event-History Analysis Predicting Education Migration
Female
Age Groups (time varying): Age 10-12
Age 13-15
Age 16-18
Age 18-21
Age 22-24
Region of birth
Port-au-Prince metro
Southeast
North
South
Central
Father is dead
Mother is dead
Enrollment timing: on time
Moderately delayed
Severely delayed or never
Rural
Region of residence
Port-au-Prince metro
Southeast
North
South
Central
Income quintile: Poorest (ref)
Poor
Middle
Wealthy
Wealthiest
Currently in school
Currently working
Adult in the HH with primary education
Current educational attainment
Has not completed primary school
Completed primary
Completed lower secondary
Intercept
Outcome: Education Migration
Model 1
Model 2
OR
SE
p
OR
SE
1.23
(.24)
1.24
(.25)
p
2.14
3.77
4.69
4.51
(.15)
(.45)
(.73)
(1.06)
***
***
***
***
2.10
3.91
5.50
5.40
(.18)
(.54)
(.99)
(1.36)
***
***
***
***
2.42
3.48
1.38
3.55
1.14
1.54
(1.07)
(1.19)
(.60)
(1.33)
(.32)
(.48)
*
***
5.05
5.05
3.35
4.85
1.13
1.84
(2.97)
(2.11)
(1.68)
(2.12)
(.32)
(.59)
**
***
*
***
.52
.34
(.13)
(.09)
**
***
.70
.72
.34
(.18)
(.20)
(.12)
1.40
.74
1.30
.64
(.60)
(.44)
(.55)
(.39)
.66
1.18
.69
1.10
2.02
.70
1.48
(.25)
(.43)
(.24)
(.35)
(.54)
(.26)
(.31)
3.55
3.37
.00
(1.51)
(1.35)
(.00)
.02
(.01)
Person-years
10,279
F-test
12.30
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
***
***
***
10,279
10.18
†
**
**
†
**
**
***
***
30
Table 1-3. Discrete-Time Event-History Analysis Predicting Labor Migration
Female
Age Groups (time varying): Age 10-12
Age 13-15
Age 16-18
Age 18-21
Age 22-24
Region of birth
Port-au-Prince metro
Southeast
North
South
Central
Father is dead
Mother is dead
Enrollment timing: on time
Moderately delayed
Severely delayed or never
Rural
Region of residence
Port-au-Prince metro
Southeast
North
South
Central
Income quintile: Poorest (ref)
Poor
Middle
Wealthy
Wealthiest
Currently in school
Currently working
Adult in the HH with primary education
Current educational attainment
Has not completed primary school
Completed primary
Completed lower secondary
Intercept
Outcome: Labor Migration
Model 1
Model 2
OR
SE
p
OR
SE
1.10
(.25)
1.28
(.34)
2.25
4.10
7.02
9.48
(.26)
(.60)
(1.14)
(1.89)
1.41
1.16
.77
.52
1.48
1.82
(.49)
(.47)
(.30)
(.22)
(.33)
(.51)
2.03
3.96
(.68)
(1.37)
.01
(.00)
Person-years
10,234
F-test
17.36
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
***
***
***
***
2.23
3.89
5.98
8.26
(.29)
(.64)
(1.14)
(1.95)
†
*
1.09
1.20
.54
.21
1.19
1.64
(.75)
(.52)
(.35)
(.11)
(.29)
(.54)
1.50
2.20
1.20
(.49)
(.80)
(.39)
3.14
1.10
.67
.92
(1.63)
(.98)
(.35)
(.64)
1.70
1.76
2.22
1.38
.29
3.78
1.53
(.63)
(.69)
(.88)
(.55)
(.09)
(.97)
(.42)
.72
.36
.02
(.22)
(.12)
(.01)
*
***
***
***
10,234
8.48
p
***
***
***
***
**
*
*
*
***
***
**
***
***
31
Table 1-4. Descriptive Differences Among Non-Migrants, Departed Youth, and Youth Who Live
Separately from Their Parents
Unweighted sample size
Sends support
Nonmigrants
living with
parent(s)
1971
Departed
household
during past
three years p1
269
N/A
68.0 %
Female
Rural
45.6 %
66.9 %
58.7 %
70.4 %
***
Age 10-12
Age 13-15
Age 16-18
Age 19-21
Age 22-24
26.1 %
25.6 %
23.8 %
14.3 %
10.3 %
7.9 %
14.7 %
21.4 %
28.7 %
27.4 %
***
***
In school
Works
80.5 %
7.8 %
Adult with primary edu. in HH
Poorest wealth quintile
Poor income quintile
Middle income quintile
Rich income quintile
Richest income quintile
Born in current household
Living
without
parents
725
p1
64.0 %
53.8 %
48.0 %
**
3
***
***
†
***
***
16.4 %
21.9 %
25.2 %
20.4 %
16.1 %
32.4 %
28.5 %
***
***
70.6 %
12.4 %
***
**
31.9 %
31.0 %
2
39.4 %
3
*
24.9 %
21.6 %
23.7 %
19.4 %
10.5 %
62.2 %
15.4 %
16.3 %
28.2 %
22.3 %
17.8 %
2
15.4 %
19.7 %
20.8 %
22.8 %
21.3 %
23.9 %
3
**
2
**
2
2
2
2
**
Visits household
45.1 %
Urban destination
74.7 %
Time since departed (months)
16.5
1
Refers to t-test comparing sub-sample to non-migrants living with parents
2
Refers to household departed from
3
Refers to current household of residence
***
***
3
3
3
3
***
***
32
Table 1-5. Heckman Probit Models Predicting Provision of Support for Youth Who Departed the
Household within the Past Three Years
Female
Rural
Age 10-12 (reference group)
Age 13-15
Age 16-18
Age 19-21
Age 22-24
Currently in school
Currently working
Adult in HH has complete primary ed
Income quintile: Poorest (ref)
Poor
Middle
Wealthy
Wealthiest
Visits
Months since departure
Urban destination
Intercept
Whether household
sends financial
support to youth
B
SE
p
-.27
(.12) *
.12
.10
-.26
-.23
(.27)
(.25)
(.23)
(.23)
1.39
-.17
-.04
(.19)
(.16)
(.14)
-.11
-.45
-.44
-.52
-.25
.01
.01
1.57
(.22)
(.18)
(.24)
(.23)
(.12)
(.01)
(.14)
(.30)
Censored observations
1971
Uncensored observations
269
F test (15, 99)
6.94
rho
-1.00
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
***
*
†
*
*
*
***
***
(.00)
Selection equation
B
SE
p
.36
(.08) ***
.33
(.09) ***
.19
.31
.56
.63
-1.25
.12
.04
(.14)
(.14)
(.16)
(.17)
(.12)
(.12)
(.10)
.20
.41
.45
.85
(.14)
(.13)
(.15)
(.18)
**
**
***
-1.57
(.19)
***
*
***
***
***
33
Table 1-6. Heckman Probit Models Predicting Provision of Support for Youth Who Live
Separately from Still-Living Parents
Female
Rural
Age 10-12
Age 13-15
Age 16-18
Age 19-21
Age 22-24
Currently in school
Currently working
Born in household
Adult in HH has compiled primary ed
Income quintile: Poorest (ref)
Poor
Middle
Wealthy
Wealthiest
Intercept
Whether youth
receives financial
support from parents
B
SE
p
-.16
(.07) *
.28
(.11) **
-.05
.14
.14
.12
.41
-.12
.81
-.08
(.09)
(.10)
(.10)
(.13)
(.09)
(.12)
(.09)
(.08)
-.02
-.02
.07
.02
.63
(.12)
(.12)
(.13)
(.12)
(.16)
Censored observations
1971
Uncensored observations
725
F test (14, 100)
13.84
rho
-1.00
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
(.08)
(.08)
(.09)
(.12)
(.09)
(.11)
(.08)
†
**
***
*
***
***
.14
.22
.28
.29
-.29
.17
-.89
***
-.24
(.13)
†
***
***
(.01)
Selection equation
B
SE
p
.18
(.06) **
-.17
(.08) *
***
34
Chapter 2
The Role of Child Endowments and Sibling Experiences in Determining
Parents’ Investments in Education Migration
Introduction
In many low- and middle-income countries, families bare a significant portion of the
direct (e.g., fees) and indirect costs (e.g., uniforms, books) of their children’s education
(Buchmann & Hannum, 2001). At the same time, families face limited access to educational
opportunities after primary school, especially in rural areas (Lloyd, 2004). When investing in
their children’s education, rural families must weigh the additional costs of children migrating to
urban areas to attend school with the potential payoff of higher educational attainment; added
costs include higher fees, transportation, and the higher cost of living, in addition to the loss of
children’s household contributions. For most rural families, the resources necessary to support
education migration are scarce. In this paper I compare education migration outcomes after
primary school among siblings to examine how families allocate limited resources among their
children.
I draw on evidence from rural Haiti. I first describe the educational context of rural Haiti
and potential explanations for how parents invest in their children’s education and education
migration in particular. I then use data from the Haiti Youth Transitions Study (HYTS) to predict
whether rural youth who had recently completed their sixth grade exams undergo education
migration. To do so I use multilevel logistic regression models that compare siblings who
experienced the same life course transition (primary school completion). In these models, familylevel effects are fixed between siblings; this strategy accounts for family characteristics that
35
siblings have in common and highlights the sibling differences associated with education
migration. Using this approach, I examine how child specific characteristics—specifically
parents’ perceptions of children’s endowments and sibling experiences relative to their own —are
associated with education migration.
Education opportunities in rural Haiti
Formal schooling in Haiti begins with entry into first grade at age six. A child who makes
consistent progress would be on track to complete the six years of primary education by age 12.
However, most children are older than the prescribed age for grade, due to late school entry,
grade repetition, and extended absences (Demombynes, Holland, & Leon, 2010). Whereas most
rural youth in developing countries live within walking distance of primary schools, secondary
schools are much more distant, and many parents perceive that urban schools provide superior
education (Lloyd, 2004). Similarly, after primary school there are few formal schooling
opportunities in rural Haiti, and vocational training opportunities for non-agricultural
employment are also scarce.
A few local schools attempt to offer lower secondary school (grades 7 to 9) to students.
However, often their status is unofficial; teachers frequently go unpaid; and actual enrollment and
attendance are extremely low. For rural families who live closer to small towns, it is possible to
make a daily commute to school. However, transportation is expensive and difficult to find, and
most students making these daily journeys commute several hours foot. This leaves them little
time to study, and because secondary school is generally offered in the afternoon session, students
may return home well after dark, prompting safety concerns.
As a result of the limited opportunities available in rural areas, youth often migrate to
urban areas to attend school (Brauw & Giles, 2008; Crivello, 2009). Expectations and the promise
36
of potential returns are high among youth, and in areas with few opportunities, social mobility
and geographic mobility are inextricably linked (Bjarnason & Thorlindsson, 2006). For Haitian
youth, their primary destinations are regional towns and cities and Port-au-Prince, the national
capital. Whereas in some countries boarding schools for rural students are common, this is not the
case in Haiti. Migrant youth most commonly reside with extended family or friends, or they may
rent a room in someone’s home. Those who attend school nearer to their home villages can often
manage a weekend trip home. But those who access the best opportunities, available only in Portau-Prince or other large cities, can often only make the trip home for Christmas and summer
vacation.
The cost of education in Haiti is high, and even more so when it includes that additional
costs of living in urban areas. The Haitian state provides some support for public education, but
even public schools charge hefty fees, and books, uniforms, and other materials add additional
costs. Furthermore, there is limited space available in public schools, and the majority of students
attend private schools (Salmi, 2000). The cost of education and education migrants are relatively
large compared to a rural household’s income, but evidence suggests that families prioritize
educational costs within the funds they have available (Bredl, 2010).
Investing in education migration
Previous research provides consistent evidence that children are more likely to enroll in
and attend school if their parents have higher educational attainment and their families are
financially more well-off, because of both the costs of education and the indirect pathways
through which family resources operate (Bjorklund & Salvanes, 2010; Buchmann & Hannum,
2001). For education migrants, the associated costs and reliance on parents’ networks are even
37
higher, and youth still rely primarily on the social and financial capital that they access via their
parents.
Previous research on parents’ investments in their children’s education highlights the
differences among families. Implicit in these studies is that families invest similarly in all their
children. However, when resources are scarce, families may distribute their investments
unequally among their children, such to maximize their investments. For families with multiple
children and limited resources, they are faced with whether to spread their investments thinly
across multiple children, or to invest more heavily in children with particular characteristics who
may provide better returns for their investment. One example of differential investments among
children is that parents may be more likely to invest in boys’ education, especially if parents
perceive that labor market conditions are inhospitable to women (Buchmann, 2000).
Likewise with school investments, the migration of an individual is often part of a
household or family strategy to promote the families’ long-term financial stability (Stark &
Bloom, 1985). Families hope their children will succeed and support them in the future, but the
returns to education migration are uncertain and can be a risky investment. Thus a family’s
willingness to support education migration may depend on the perceived likelihood that the child
will garner returns from education that will benefit the family. I therefore hypothesize that
differences in siblings’ education migration outcomes will be associated with parent’s perceptions
of how well the child does school, intelligence, potential for finding good employment, and the
likelihood that he/she will provide parental old-age support, because parents will view them as
better investments.
38
Sibship characteristics and education migration
Children within the same household may experience different education migration
outcomes because of individual attributes or endowments. Additionally, education migration may
depend on family characteristics that are unique to each child, such as previous experiences of
older siblings and relative position in the sibship structure. To generate potential hypotheses in
this domain, I draw on ethnographic fieldwork that was conducted in preparation for and in
conjunction with the survey data analyzed in this paper.
First, younger siblings may have an easier migration trajectory if their older siblings have
already migrated. This may occur for two potential reasons. Parents may have already negotiated
the logistics of school enrollment and housing for older siblings, and younger siblings may join
households and schools alongside their older siblings. Furthermore, if they are in the same
location, parents often trust that younger siblings will be cared for by older ones. This hypothesis
is supported by discussions in which parents explained that they would send younger siblings to
live with older siblings, that siblings are safer together, and that older siblings might do the
groundwork in identifying a potential school.
An alternative explanation for why having an older sibling that migrated might matter is
that older siblings have forged migration as an acceptable path, and younger siblings might view
older siblings as an ideal of what is possible. Youth often described their future migration plans in
the framework of what their older siblings had accomplished. Thus, having an older sibling who
has ever migrated may indirectly facilitate education migration.
An additional hypothesis is that parents may resist the departure of younger children.
School-aged children play an important role in maintaining the household through chores, such as
cooking, cleaning, and fetching water, and families may suffer the loss of household labor when
youth leave the household. Having younger siblings in the household, particularly younger
39
sisters, who can fulfill the household labor needs, has the potential to ease the departure of older
siblings.
Based on these qualitative observations, I propose three hypotheses. First, having an
older sibling in a Haitian city will be positively associated with education migration, because the
older sibling directly facilitates the migration process. Second, the presence of an older sibling in
any potential migration destination will be positively associated with education migration,
because the older sibling has already forged the migration path. Finally, the presence of a younger
sister in the sending household will be positively associated with education migration, because
she can absorb the potential migrant’s work load
Education migration responses to event shocks
Families may also reshape their migration strategies and migration intentions for their
children in response to events that change the risks and rewards of migration. In January 2010
Haiti experienced a 7.0 magnitude earthquake, a large-scale natural disaster that devastated many
aspects of the country’s physical, social, and economic infrastructure. Jacmel and Port-au-Prince
were the two areas most devastated by the earthquake, and they are also two of the principal
destinations for youth in the region where this study was conducted.
Environmental stressors and natural disaster that damage infrastructure and lead to the
loss of livelihoods also reshape migration decisions (Renaud, Dun, Warner, & Bogardi, 2011). As
a result of this devastating natural disaster, Port-au-Prince and Jacmel lost some of their appeal as
migration destinations. Additionally, throughout the country, costs of living rose substantially,
and families were less able to afford schooling costs and the added costs of education migration.
Therefore I ask: does a major national event reshape families’ migration decisions? If so, do
40
families universally reduce education migration, or is there evidence that decisions related to
child-specific characteristics (i.e., child endowments , relative sibship structure) also change?
Methods
To examine the key hypotheses that I proposed in the previous section, I use data from
the HYTS. I compare respondents from a cohort of students who recently completed sixth grade
to an older sibling who completed sixth grade prior to the 2010 earthquake using multilevel
logistic regression in which family-level characteristics are fixed across sibling pairs.
Data
The HYTS recruited a cohort of youth who had just completed their primary school
exams (6th grade) from within a designated geographical area in rural southeast Haiti during the
summer of 2011. Target youth were first interviews between August and September 2011, and a
parent or guardian was also interviewed. Parent responses provided background information
about the target youth and household, and a portion of the interview elicited detailed responses
about an older sibling who experienced the same life course transition (completion of sixth grade)
prior to the 2010 earthquake. Detailed contact information was recorded, and youth respondents
were contacted and interviewed again between February and March 2012.
Local officials identified all area primary schools, and school directors provided a list of
students who completed their sixth grade exams. During the first survey wave, 199 youth were
interviewed along with a parent or guardian. Additionally, 16 parents whose children were not
available, but met the recruitment criteria were also interviewed. During the second survey wave,
the 10 youth whose parent was interviewed during the first wave but who were not able to
41
participate themselves completed the second survey wave. Only respondents who were initially
recruited during the first survey wave were eligible to be added during the second wave, and this
approach helped reduce non-response by youth who were away from their primary households
during the first wave. To decrease non-response bias, interviewers made extensive efforts to
recruit all potential participants regardless of their ease of access and accommodate the schedules
of youth who were working (primarily seasonal agriculture) or visiting relatives. All who were
contacted agreed to be interviewed. These efforts successfully located 92 percent of
families/youth identified by school directors. The complete sample includes 215 family units, and
of these families, 140 provided information on an older comparison sibling who completed
primary school prior to the 2010 earthquake.
The end of primary school is a point at which limited opportunities lead many youth to
migrate, and all respondents, including migrant youth (23.8% of target youth), were reinterviewed at the destination. This design affords comparisons between migrants and nonmigrants from the same point of origin and offers advantages over designs that simply compare
rural-to-urban migrants to rural non-migrants.
Each youth interview lasted approximately one hour, and parent interviews lasted
approximately 30 minutes. Participants were interviewed in Haitian Creole, their native language,
by a trained interviewer. In the case of respondents who could not be interviewed face-to-face
during the second wave, eight interviews were conducted by telephone during the second wave.
Data on the next oldest sibling comes from the parent or guardian’s report. Parents were
first asked to identify the target child’s next oldest sibling who finished primary school prior to
the 2010 earthquake—June 2009 or earlier. Full siblings, along with half, step, or adopted
siblings who lived in the same household, could qualify as the comparison child. The parent was
asked about the older sibling’s educational history, specifically where he or she lived and what he
or she did in the year after completing sixth grade. Parents were also asked to compare the target
42
youth directly to the older sibling, reporting who did better in school, who was more intelligent,
who would likely get a better job, and who was more likely to support the parent(s) in old age.
All research procedures followed the ethical treatment of human subjects were approved
by the Pennsylvania State University Institutional Review Board. Respondents aged 18 and older
provided verbal consent; verbal assent and parents’ verbal consent was sought for younger
respondents. After each youth interview, the respondent received a small incentive that was
valued at less than two U.S. dollars.
Assessment of generalizability
To assess the generalizability of the HYTS, I compare key respondent characteristics to
rural respondents from the 2005-06 Haiti Demographic and Health Survey (DHS) (Table 2-1).
The DHS is a nationally representative survey, and comparing the two surveys reveals the extent
to which HYTS households are similar to other households in Haiti.
The mean age of HYTS respondents is 15.9 years, and in the DHS, rural sixth graders are
on average 16.5 years old. In both surveys rural sixth graders are well-over their age for grade.
HYTS respondents are someone younger, and this may be a consequence of improved school
enrollment during the five years between the two surveys.
HYTS households were somewhat larger than households in the DHS (6.4 vs. 4.7
members). However, the DHS sampling frame is all households, and the HYTS only interviewed
households where a sixth grader was present. This finding reflects the fact that smaller
households were less likely to be part of the HYTS by virtue of the fact that they had a lower
probability of having a sixth grader.
Additionally, household wealth was compared between the two surveys. The DHS
produces a wealth index based on a principal components analysis of key household assets that is
43
divided into quintiles (Rutstein & Johnson, 2004). HYTS households were matched with DHS
households according to asset ownership and key indicators of housing and sanitation quality.
Rural DHS households were distributed among the five quintiles from wealthiest to poorest in the
following proportions: 5%, 10%, 34%, 30%, 31%. HYTS households were distributed as 6%,
10%, 6% 71%, 8%. In both cases, the rural household surveyed were notably poorer than the
national population.
Variables
The key variables in this study come from the two waves of youth interviews and the
parent interviews. Youth’s primary place of residence and whether he/she was attending school at
the time of second survey wave was used to determine whether the target youth was an education
migrant. Parent responses provide a comparable set of responses for the older sibling; they
reported the child’s primary activity (e.g., attended school) and place of residence (e.g., same
village, name of city migrated to) to determine if he/she was an education migrant at the end of
sixth grade. They also reported the comparison sibling’s gender.
Parents were also asked to compare the target child to his/her older sibling after he/she
had just completed the sixth grade. Specifically, they were asked (i) Between <name of target
child> and <name of comparison sibling>, who does better in school? (ii) Between <name of
target child> and <name of comparison sibling>, who is smarter? (iii) Between <name of target
child> and <name of comparison sibling>, who is more likely to find good work? and (iv)
Between <name of target child> and <name of comparison sibling>, who is more likely to help
you in your old age? Parents were unable or unwilling to provide responses for only 8% of these
items. For each child the responses were coded 0/1 to indicate that the parent perceived the child
to be more endowed in that domain.
44
Youth respondents provided a comprehensive household roster that included the age,
gender, and degree of relatedness to all household members. They also reported information on
their siblings who lived outside the home, including age, gender, and current place of residence.
From this information, I was able to identify if the target youth and the comparison sibling each
had older siblings living in a Haitian city, older siblings living at any potential migration
destination (Haitian city or outside of Haiti), and a younger sister living in the household.
Parents provided information on ownership of assets, which included housing
characteristics, ownership of durable goods, land ownership, and livestock ownership. They also
indicated whether a household member had a bank account or received remittances. These data
were combined into an asset index factor score using principal components analysis. This asset
index differs from the one that compares households to DHS households in that in includes
livestock ownership, bank accounts, and remittance receipt. It is also standardized to better
distinguish among the rural households in the HYTS.
Parent respondents also reported on family size, which is the number of still living
children. They also reported on their own migration histories, which are used to identify whether
parents had ever lived in a Haitian city.
Analysis
To examine how child-specific characteristics are associated with education migration, I
draw on multilevel logistic regression model to compare siblings within families (Raudenbush &
Bryk, 2002). In these models, family-level effects are fixed. Thus, they account for unobserved
family characteristics (e.g., parents’ attitudes toward education) that are constant between sibling
pairs. In doing so, it is possible to differentiate between the characteristics of siblings in the same
family. To explain the benefits of this modeling approach, I use the example of youth with an
45
older migrant sibling. In descriptive data, youth with older migrant siblings may be more likely to
migrate, but this may attributable, at least in part, to their origin in the same family with more
resources or more positive attitudes toward migration. However, once fixing constant the familylevel effects that siblings share, it is possible to examine whether having an older migrant sibling
is associated with education, not just whether siblings commonly share the same migration
destinies.
These models take the basic form,
(
)
In this equation, the logit of undergoing education migration for sibling i in family j is a
function of both individual attributes and characteristics, as well as family-level characteristics
that siblings in family j hold in common; µj is the family-specific level-two error term.
The first set of models test if parents’ perception of children’s attributes (i.e., better in
school, smarter, more likely to find good work, and more likely to help them in old age) are
associated with education migration following primary school. Model 1 begins with only the
control variables: gender at the individual level, family assets, family size, and parent migration
history. Model 2 adds a binary variable at the individual level that indicates if the child is the one
that the parent perceives does better in school. Target children whose parents did not identify a
comparison sibling, did not have a score on this variable; therefore, I also includes a binary
variable to indicate if the target child did not have a comparison sibling.3 Models 3 to 5 are
structured in the same way as Model 2, and add the parent’s perception of who is the smarter
child, who is more likely to find good work, and who is more likely to help him/her in old age.
Significant effects for any of these characteristics would suggest that parents consider these
attributes when considering potential migration strategies.
3
I also ran the models excluding children who did not have a comparison sibling and only including those
with a paired sibling, and there were no differences in the substantive findings.=
46
The second set of models tests if older and younger siblings are associated with
educational migration. Model 1 includes an individual indicator of whether the child had an older
sibling living in a Haitian city. A significant positive effect would suggest that there is a specific
household where this child could potentially live and attend secondary school. Model 2 includes
an individual indicator of whether the child had a migrant older sibling (one living in a Haitian
city or outside Haiti). A significant positive effect of this variable would indicate that the
experience of the older sibling had somehow paved the way even if his/her presence did not
directly facilitate a potential destination for the migrant youth. Model 3 includes an individual
indicator for whether a younger sister was in the home. A significant positive effect would
suggest that a younger sister may take on the household responsibilities of older siblings, thus
freeing him or her to migrate. Again, the models control for gender at the individual level, family
assets, family size, and parent migration history.
The third set of models examines how the 2010 earthquake is associated with education
migration. The first model includes a dummy variable for whether the child is the younger sibling
(after the earthquake) or the older sibling (before the earthquake). Then the subsequent models
test variables describing the multiplicative interactions between earthquake timing and childspecific characteristics (i.e., child endowments, relative sibship characteristics).
In all cases data were analyzed using xtmelogit in STATA 12.0. This package handles
unbalanced data sets (families that did not have an older comparison sibling), but cases with
incomplete responses would be dropped from the analysis. Therefore, multiple imputation
procedures were used to account for incomplete responses, and standard error adjustments reflect
the use of multiply imputed data.
47
Results
Descriptive results
I first describe the characteristics of the target youth and their older siblings (Table 2-2).
Overall, 29% of youth migrated to an urban area to attend school the year after they completed
the sixth grade, and this experience was more common among older siblings than among the
current cohort (36% vs. 24%). Of these youth, 53% were male, and there were no gender
differences by sibling cohort or migration experience.
After identifying the comparison child, the parent was asked which of the two did better
in school, was smarter, was more likely to find a good job, and was more likely to provide oldage support. (For 35% of youth, the parent did not identify a comparison child.) In all four cases,
parents were more likely to attribute these characteristics to the older child. Also, those identified
as better in school and smarter more often migrate to attend school following primary school.
Regarding migrant siblings and siblings in the home, 42% of youth had a sibling in a
Haitian city, 61% had an older sibling in a migration destination (Haitian city or outside Haiti),
and 56% had a younger sister in the household. The target youth were more likely than their
comparison siblings to have an older sibling in a Haitian city or at a migration destination and
less likely to have a younger sister in the household. This is a function of their birth order. Youth
with older siblings in Haitian cities were more likely to undergo educational migration (51% vs.
38%). There was a non-significant association (p < 0.10) between migration and having a sibling
at a potential migration destination. There was no significant association between migration and
having a younger sister in the household4.
4
I explored alternative definitions of this variable using siblings of both genders, restricting the definition
to children older than age six (i.e., capable of helping with household chores), and broadening the
definition to anyone in the household regardless of relatedness. No alternative definitions were significant.
48
In regard to the family characteristics, youth who migrated to attend school immediately
after sixth grade were more likely to be from wealthier families. The average family size of 6.6
children did not differ by sibling cohort or migration status. Approximately 20% of the parents or
guardians interviewed had previously lived in a Haitian city, and youth who migrated were more
likely to have a parent who had lived in a Haitian city than youth who did not migrate (32.3% vs.
15.8%).
Parents’ perceptions of children’s endowment: Multivariate models
To further explore how parents’ perceptions of children’s characteristics are associated
with educational migration at the end of primary school, I turn to the multilevel logistic
regression models (Table 2-3). The first model includes only the control variables. Children from
families where the parent previously lived in a Haitian city experience 2.7 times higher odds of
migrating at the end of primary school. This finding is consistent across all the models in Tables
2-3 and 2-4. Children from families with more assets also experience 1.3 times higher odds of
migrating, but this finding is only marginally significant and is not significant in all subsequent
models. Child gender and family size are not significant.
The second model, which includes the parent’s determination of the child who does better
in school, suggests that the child that the parent perceives to be smarter experiences 2.4 times
higher odds of migrating. The control variable for children who did not have a comparison sibling
is not significant. The third model, which includes parents’ perceptions of who is smarter, finds
that the smarter child has twice the odds of educational migration after primary school. In the
fourth model, parents’ assessment of who is more likely to find good work, the potential to find
good work is associated with 1.7 times higher odds of migrating, but this effect is only marginally
49
significant. The fifth model suggests that the perception that the child will help them in old age is
not significantly associated with educational migration.
Sibship characteristics: Multivariate results
I, then, examine how having siblings who migrated and younger siblings within the
household are associated with education migration following primary school (Table 2-4). Neither
having a sibling in a Haitian city, having a sibling at any migration destination, nor having a
younger sister in the household is significantly associated with migration once they are included
in the multivariate models. The effect sizes for having a sibling in a Haitian city or at a migration
destination are in the expected direction (positive), but they are not significant. The coefficient for
having a younger sister in the home is not in the expected direction (it is negative).
The results for the family-level characteristics and child gender are similar to those
presented in the models in Table 2-3. Thus, those findings are not repeated.
Earthquake consequences
Next I examine how the 2010 earthquake may have changed these family migration
strategies (Table 2-5). In the first model, a dummy variable indicating earthquake timing (child
finished primary school before the earthquake) was tested. Findings reveal that youth who
completed primary school before the 2010 earthquake experienced 2.1 times higher odds of
education migration than their own siblings who completed primary school after the earthquake.
In the subsequent models interaction terms were tested. None of the proposed interaction
terms were significant. Therefore these models are not shown or discussed.
50
Discussion
This paper explores how families allocate resources among their children by examining
the differences in siblings that are associated with education migration following primary school.
Investing in migration is one strategy that families use to potentially maximize their long-term
benefits (Stark & Bloom, 1985). But as one father described education migration, “Investing in
your kid is like playing the lottery.” In examining how parents attempt to minimize this risk, I
first explore how parental perceptions of children’s endowments are associated with education
migration.
Findings demonstrate strong positive effects for associations between education
migration and parents’ perception that the child is smarter and does better in school. A weak
positive trend was also revealed between education migration and the perception that the child is
more likely to find good work. The association between education migration and the perception
that the child will help the parent in old age, however, was not significant and in the negative
direction. The positive effects support the idea that education migration is part of a continued
investment period.
To an extent these findings are consistent with the perspective that migration is part of a
family-based strategy to maximize the long-term financial stability (Stark & Bloom, 1985).
However, as a whole, these findings suggest that investments may be based more on short-term
outcomes (intelligence and school performance) than long-term ones (employment and old-age
support). This may occur because parents may also take pride in and derive utility from the
children’s educational accomplishments, and these immediate benefits may be more important in
maintaining their social position than any potential long-term benefits. An additional explanation
is that parents may be basing their investments on who would benefit more from that specific type
of investment, rather than showing preferences for a particular child. To this end, smarter children
51
who do well in school are rewarded with additional schooling opportunities, whereas those who
show merit in other domains may be rewarded in other appropriate ways.
The second part of the study examines the role of the sibship structure. There is no
evidence to suggest that after controlling for similar characteristics that the presence of older
sibling at a potential migration destination helps facilitate the migration process. There is also no
evidence to suggest that the presence of a younger sister in the household helps ease the migration
of older siblings. These findings are in contrast to the qualitative findings that suggested these
hypotheses. An explanation for these findings may be that families are able to draw on kinship
networks that extend beyond older siblings, or that parents might find that the favors called upon
for older siblings have already been exhausted.
Finally, the third part of the study examines if differences in these processes are
associated with a large-scale natural disaster that devastated key youth migration destinations and
the country’s economic infrastructure. Evidence suggests that fewer youth underwent education
migration in the younger cohort compared to their older siblings. Education programs should
expand their support of post-primary school education in rural areas.
Limitations
Despite this paper’s contributions, it also suffers from limitations. One limitation is that
parents are interviewed retrospectively about the older sibling and compare their memory of the
older sibling to the current characteristics of the younger sibling. Bias is inherent in this type of
comparison: a parent might rather praise the talents of the child who is already migrated than
claim they suffer buyer’s remorse. However this strategy likely suffers less error than rating each
child’s characteristics during interviews that occur several years apart.
52
The study is also limited by its ability to completely determine the sibling characteristics
of the older comparison sibling. Information on the oldest child’s siblings is derived from the
household and sibling rosters produced by the younger sibling during the first interview. This
strategy first assumes that the sibling pairs share all other siblings in common. When in fact,
Haitian family structures are relatively unstable, and most individuals have half siblings.
However, the age proximity of the two siblings improves the likelihood that the two are in fact
full siblings with all other siblings in common. The strategy for classifying siblings also assumes
that older siblings are relatively stable in their residential locations and that if they are older than
the older sibling in the sibling pair that they were in their current location prior to when the older
of the two siblings being compared would have migrated. The difficulty in accurately measuring
the sibship characteristics of the oldest child may be one reason that the effects of these variables
are not significant.
Conclusion
In conclusion, this paper examines the factors that lead to different education migration
outcomes within sibling pairs by comparing differences within siblings using multilevel logistic
regression models. Findings suggest that parental perceptions of how smart their children are and
how well their children do in school are strongly associated with education migration following
primary school. There is also strong evidence that families have reduced the use of education
migration as a strategy for investing in their children’s education as a response to the 2010
Haitian earthquake.
53
Table 2-1. Comparisons between HYTS respondents and the 2005-06 Haiti DHS
Characteristics of
HYTS
Respondents
Age of 6th graders (mean)
Household size (mean)
15.9
6.4
Distribution across DHS wealth quintiles
Wealthiest
Wealthy
Middle
Poor
Poorest
6%
10%
6%
71%
8%
Characteristics of
Rural Households
in 2005-06 Haiti
DHS
16.5
4.7
5%
10%
34%
30%
31%
54
Table 2-2. Characteristics of HYTS Respondents
All
n=
Child Characteristics
Migrated for school
Male child
Does better in school3
Is smarter3
More likely to find good work3
More likely to help parent in old age3
Sibling in a Haitian city
Sibling who migrated
Younger sister in household
Does not have a comparison sibling
%
355
28.8
52.9
50.0
49.6
51.4
53.8
41.7
61.4
55.5
Target
Child
%
215
Older
Sibling
%
140
23.8
55.3
37.7
40.7
39.8
48.0
52.5
71.2
51.2
34.9
36.4
49.3
68.9
63.4
69.1
62.7
25.1
46.4
62.1
Migrated
p1
%
102
Did Not
Migrate
%
253
47.4
61.6
59.5
58.1
58.8
50.9
68.6
52.9
55.2
45.3
45.6
48.6
51.8
38.0
58.5
56.5
p2
*
***
***
***
*
***
***
*
Family Characteristics
Family assets (mean factor score)
.05
(.05)
.00
.14
.28
Family size ( mean # living children)
6.55
(.12)
6.45
6.71
6.51
Parent migration history
20.56
19.07
22.86
32.32
1
P-value of Student's t-test comparing the characteristics of target children and older siblings
2
P-value of Student's t-test comparing education migrants with others
3
Values do not all equal 50% because not all target children had a comparison sibling.
-.04
6.57
15.82
**
*
*
†
**
***
55
Table 2-3. Tests of Parents’ Perceptions of their Children’s Endowments in Multilevel Logistic Regression Models Predicting Education
Migration.
Model 1:
OR
SE
Model 2:
p
OR
SE
Model 3:
p
OR
SE
Model 4:
p
OR
SE
Model 5:
p
OR
SE
(.44)
(.33)
1.55
.64
.69
(.33)
(.43)
(.33)
p
Individual Characteristics
Does better in school
Is smarter
More likely to find a good job
More likely to help in old age
No comparison sibling
Male child
.72
Family Characteristics
Family assets (factor score)
Family size (living children)
Parent migration history
1.34
1.00
2.70
(.17) †
(.07)
(.42) *
1.32
1.00
2.77
(.19)
(.08)
(.46) *
1.31
.99
2.67
(.18)
(.08)
(.44) *
1.31
.99
2.64
(.18)
(.08)
(.43) †
1.32
.99
2.63
(.18)
(.08)
(.43) *
.28
1.25
2.84
(.58) *
(.37)
*
.18
1.41
2.71
(.71) *
(.40)
*
.21
1.34
2.51
(.68) *
(.39)
*
.26
1.29
2.26
(.65) †
(.38)
*
.27
1.41
2.09
(.66) *
(.38)
†
Intercept
Level 2 Random effects
F-test
n = 355; N = 215
2.36
(.32) **
1.99
(.30) *
1.66
(.32)
.68
.76
(.46)
(.34)
.69
.73
(.44)
(.33)
.67
.70
(.30) †
56
Table 2-4. Tests of Sibship Structure Characteristics Endowments in Multilevel Logistic Regression Models Predicting Education Migration.
Model 1
Level 1: Individual Characteristics
Sibling in a Haitian city
Sibling who migrated
Younger sister in household
Male Child
Level 2: Family Characteristics
Family assets (factor score)
Family size (living children)
Parent migration history
Intercept
Level 2 Random Effects Parameters
F-test
n = 355; N = 215
OR
SE
1.43
(.32)
Model 2
p
OR
SE
1.61
(.32)
Model 3
p
0.73
(.30)
0.73
(.31)
1.30
1.00
2.54
(.17)
(.07)
(.40)
1.32
0.98
2.70
(.17)
(.07)
(.41)
†
0.26
1.07
2.70
(.55)
(.40)
0.24
1.15
2.70
(.58)
(.37)
*
*
*
*
*
*
OR
SE
p
0.75
0.70
(.32)
(.32)
1.34
1.00
2.76
(.17)
(.07)
(.42)
†
0.33
1.22
2.43
(.59)
(.37)
†
*
*
57
Table 2-5. Test of Earthquake Timing in Multilevel Logistic Regression Models Predicting Education Migration
OR
Model 1
SE
Individual Characteristics
Before earthquake (2010)
Male child
2.12
.76
(.30)
(.33)
Family Characteristics
Family assets (factor score)
Family size (living children)
Parent migration history
1.33
.99
2.79
(.18)
(.08)
(.45)
.21
1.37
3.07
(.63)
(.39)
Intercept
Level 2 Random effects
F-test
n = 355; N = 215
p
*
*
*
**
58
Chapter 3
Gendered perspectives on the intersection of migration and sexual initiation
among Haitian adolescents
Introduction
There is growing interest in how migration during adolescence may influence sexual and
reproductive health (Luke, Xu, Mberu, & Goldberg, 2012; Mberu & White, 2011). As young
people mature—both physically and socially—and begin leaving their natal communities, many
are still unmarried and initiating their sexual trajectories. Consequentially, there are concerns that
youth who migrate to more-urban areas (from villages to towns or towns to cities) are at risk of
deleterious outcomes as a result of separation from their family, unfamiliarity with their new
environment, the desire to improve their social standing, and less knowledge of protective
strategies (Dhapola, Sharan, & Shah, 2007; Zuma et al., 2005).
Scholars and public health professionals have long scrutinized the sexual experiences of
migrants; their geographic mobility and larger number of sexual partners may increase the
prevalence and geographic breadth of HIV and other STIs (Brockerhoff & Biddlecom, 1999;
Dhapola et al., 2007; Wolffers, Fernandez, Verghis, & Vink, 2002; Zuma et al., 2005). However,
most previous research of the sexual behaviors of migrants emphasizes the experiences of
married adults for whom return to a spouse following an extra-marital partner—thus increasing
the likelihood of HIV and STI transmission—is the primary concern. Adolescents’ experiences
differ from adults’, and many are still initiating their sexual careers. As part of a life course
approach to sexual development, it is important to examine the associations between migration
59
and sexual behaviors, especially during adolescence— a unique life course phase (Carpenter,
2010).
In this paper I draw on evidence from Haiti to unravel the intersection of sexual initiation
and migration and further develop the nascent body of literature on this topic. As a country where
internal migration among male and female youth is frequent (Lunde, 2009), age at first marriage
is relatively late (26.3 for males, 20.4 for females), and young people commonly initiate sex prior
to marriage (Cayemittes et al., 2007), Haiti provides an ideal case to examine this experience. I
begin with a review of the relevant literature, which integrates perspectives on migration, gender,
and adolescent development. I then use data from the Demographic and Health Surveys to
examine competing hypotheses for how migration may influence the timing of first sex. Next, I
draw on longitudinal data from the Haiti Youth Transitions Survey (HYTS) to examine how
sexual and reproductive health knowledge and attitudes develop differently among migrants and
non-migrants, spanning the time from shortly before migration until six months later. This study
extends previous work on migration and sexual initiation by considering the gendered
components of sexual and reproductive health among migrants and accounting not just for a
single event—first sex—but also how the accompanying knowledge and attitudes change.
Theoretical Framework
Using a gendered lens approach to examine the intersection of migration and sexual
initiation, gender may first contribute to if and how adolescents migrate. Families may resist
investing in girls’ opportunities if the labor market is less hospitable to women (Buchmann,
2000), and girls’ more intensive household contributions may tie them to the home (Hsin, 2007).
However, the factors that may make parents unwilling to let their daughters leave home may
encourage urban families to receive girls, especially those that will provide domestic service
60
(Moya, 2007). In turn, girls may view migration as an opportunity to break from genderedrestrictions while maintaining family obligations; they are able to reinforce social ties by sending
remittance home and residing with kin at the destination (Castellanos, 2007; Thorsen, 2010).
Gender also paints the backdrop on which early sexual experiences occur. Families’
concerns about girls’ physical health and safety heighten following puberty (Sommer, 2010), and
families often believe that girls who live away from their parents and who are new to urban areas
are at risk for early and unprotected sex (Bruce & Hallman, 2008). Romantic and sexual
relationships may penalize female adolescent’ reputations and educational opportunities because
of pregnancy, negative stigma, and by boyfriends who restrict their autonomy. In contrast, these
relationships are far less damaging to male adolescents; male sexual prowess is often rewarded,
and few expectations are made of young fathers (Barrow, 1999). Moreover, culturally specific
sexual scripts teach girls to market their sexuality, which is often rewarded through gifts and
favors (Schwartz, 2009).
Previous literature that examines the relationship between migration and the timing of
sexual initiation invokes three primary explanations: (i) adaptation—that migrants’ behaviors
change as they adapt to circumstances in the new location, (ii) disruption—that the migration
experience itself may transform individuals, and (iii) selection—that migrants are a select
population that tend toward different behaviors (Brockerhoff & Biddlecom, 1999; Goldstein &
Goldstein, 1981). I elaborate these three hypotheses and explain how they may operate differently
by gender.
Central to the adaptation hypothesis are the stark differences in the social contexts and
opportunity structures for young people in rural, town, and urban areas. Rural Haitian life is
organized around the lakou, a cluster of several small houses of extended kin (Smith, 2001).
Parenting responsibilities and resources are shared, and young people customarily receive
parental support from multiple adults (Edmond, Randolph, & Richard, 2007). In contrast, this
61
system wanes in towns and even more so in cities, which are less hospitable and where violence
is sometimes common (Kolbe & Hutson, 2006). Towns and cities also provides alternatives to
traditional attitudes about sexuality (Carpenter, 2010), and adolescents may encounter these novel
ideas through media sources and urban peer groups, at a life course stage when the ideas are
particularly appealing and their influences are salient (Arnett, 2002; Collins & Steinberg, 2006).
These characteristics paint an urban environment where—for adolescents in particular—sexual
messages and sexual opportunities are more abundant, which may accelerate the timing of sexual
initiation.
Cities and towns also have distinct reward systems where the payoffs of education and
labor are higher, and the opportunity costs are comparatively harsher (Easterlin, 1975). Though
young urbanites may be exposed to a more overt sexual environment, lost educational and
employment opportunities as a consequence of pregnancy, may prompt girls in particular to
postpone sexual activity. Additionally, she may postpone sexual and romantic relationships if she
believes that knowledge of these relationships would inhibit her opportunities. In contrast, the
urban cash economy and the cost of courtship for male youth may mean that urban male youth
have more difficulty securing sexual partners. In fact, the interaction of gender and urban status is
apparent in the timing of first sex. Whereas age at first sex occurs later among urban women
(18.1 vs. 17.8), it occurs earlier among urban men (15.9 vs. 16.6) (Cayemittes et al., 2007).
An alternative explanation, the disruption hypothesis, suggests that the migration
experience itself—not just the new environment—may foster divergent outcomes. One factor
may be youth migrants’ living situations. Youth migrants often reside with extended kin or
siblings only a few years their senior (Yaqub, 2009b). These home environments may lack
supervision and emotional support. Recipient households may also deliberately leverage
migrants’ well-being to bargain for additional resources (Gibbison & Paul, 2006). These
circumstances could potentially encourage early and/or high risk sexual relationships (Bruce &
62
Hallman, 2008). The need for status attainment at the destination may also encourage sexual
encounters, whether through making the encounters known—more likely the case among male
youth—or through the acquisition of gifts, rides, or other favors that could potentially convey
status—more likely the case among female youth. Additionally, migrant youth may become more
marketable as sexual partners during visits home (Dhapola et al., 2007). Moreover, the migration
experience itself may alter whether youth are prepared to protect themselves in their early sexual
encounters; rural youth also receive less sexual and reproductive health education and
demonstrate less knowledge in this domain (Cayemittes et al., 2007). Thus, upon migrating to
urban areas, they may have fewer skills to safely negotiate bring potential sexual and romantic
encounters.
These explanations of the disruption effect imply that migration fosters risky
environments for youth’s sexual and reproductive well-being. However, the disruption effect may
also be protective. The opportunities afforded by migration, and the risk of losing these
opportunities, can also protect adolescents by delaying sexual debut and motivating safer sex
practices. Although education and the urban environment weaken the traditional sexual values
that postpone sex, education and labor may opportunities also provide youth an incentive to delay
sex. School enrollment in and of itself protects against sexual initiation (Lloyd & Mensch, 2008),
and many youth migrate with educational goals. Among migrants with high aspirations, perceived
opportunities for education or employment may serve to delay sexual initiation. This may
especially be the case among girls for whom pregnancy would compromise opportunities.
Finally, selection may explain why migrant and non-migrant adolescents demonstrate
differing sexual initiation patterns. If this is the predominant explanation, migration itself has no
direct effect on sexual and reproductive health outcomes, and instead, external factors drive
change in both domains. An often offered explanation for migrants’ larger number of sexual
partners—particularly among male adults—is that they may be prone to riskier lifestyles
63
(Dhapola et al., 2007). Also potentially owing to selection effects would be youth who initiate sex
before migrating. Though higher rates of premarital and unsafe sex among rural-to-urban youth
migrant factory workers in Nepal were widely thought to be a consequence of their living and
working circumstances, semi-structured interviews established that most migrants, both male and
female, initiated sex and developed their more-permissive sexual attitudes prior to migration (Puri
& Busza, 2004).
In light of the previous literature examining migration, sexual initiation, and gender I
examine four specific questions. Does migration influence the timing of sexual initiation via an
adaptation, disruption, or selection effect? Does this processes differ for male and female youth?
Do sexual and reproductive health knowledge and the endorsement of premarital sex differ in
their levels between migrant and non-migrant youth? Do these factors transform differently for
migrant and non-migrant youth over the migration period?
Method
I address the key research questions with two data sources: the DHS (women 2005-06,
men 2000) and the HYTS. With DHS data I employ discrete-time event history analysis to test
whether an adaptation, disruption, or selection hypothesis best fits the observed pattern of how
the timing of first sexual experience intersects with migration. I compliment the analysis of
nationally representative cross-sectional data with analysis the HYTS—a cohort of rural youth
who were first interviewed after recently completed primary school and interviewed again after a
22% migrated to an urban area. Using these data I conduct growth curve analyses to examine
changes in knowledge and attitudes during this life course transition. Drawing on two
complimentary data sources provides insight into both the overall trends in the population and
how these processes may unfold during the migration period.
64
Haiti DHS-Data
The Haiti DHS is a nationally representative household survey. Target households are
identified via multi-stage stratified random design, and the survey includes detailed interviews
with men and women of reproductive age. I draw on 2005-06 data for female respondents.
However, key migration questions were not asked in the 2005-06 male modules, and I instead
draw on the 2000 survey for male respondents. For this study, I limit the sample to respondents
aged 15 to 20 at the time of the survey who are regular residents of the selected household, and I
limit the analysis to those who initiated sex between the ages of 12 and 20. In some contexts a
household survey that excludes group quarters from the sampling frame would be a poor choice
for studying migrants. However in Haiti, both schools and factories that board youth are
extremely rare.
Migration information is ascertained indirectly by responses to the length of time an
individual has lived in their current location, their previous type of place of residence (i.e., rural,
town, or city), and their current type of place of residence. From this information I determined the
age at migration and the migration pattern (e.g., rural-to-city, town-to-city, rural-to-town). After
creating a person-year file, I coded two sets of time-varying covariates. One set of covariates
classified each person-year into one of five responses: city non-migrant, town non-migrant, rural
non-migrant, city post-migration, or town post-migration. The second set of covariates includes a
binary indicators for whether or not the individual migrated in the particular person-year, one
year post-migration, and two years post-migration.
Migration events, which reoccur and are often circular, are difficult to account for in
cross-sectional data, and these data only describe the most recent migration episode. To limit the
sample to correspond to the study’s focus on migration to more-urban areas and to account for
smaller residential changes and circular migration, I restrict inclusion in the analysis using several
65
criteria. I excluded lateral migrants (e.g., one rural location to another). This eliminated 13.4% of
the potential female sample and 6.7% of the male sample. I also excluded those who migrated to
less dense areas (4.8% of the potential female sample and 2.3% of the male sample). The final
sample includes 1,215 female youth (6,185 person-years) and 829 male youth (3,586 personyears).
Respondents report their age at first sex. Accurately accounting for sensitive information
plagues survey research, and most notably, female adolescent more often underreport their sexual
experiences, whereas male adolescents exaggerate them (Zaba, Pisani, Slaymaker, & Boerma,
2004). However, this same study reports that the largest reporting discrepancies are in the number
of sexual partners, whereas the age at first sex is less-subject to biased reporting. Moreover,
particular survey formats may facilitate more accurate reporting. Survey formats, such as this one,
that for the age at first sex without first asking if the respondent has ever had sex, provide more
plausible responses (Curtis & Sutherland, 2004). Furthermore, it is unlikely that responses bias
differs by migration status. Therefore, these data’s usefulness in addressing this paper’s primary
research questions addressed is not severely limited.
Other characteristics may potentially confound the association between the timing of first
sex and migration. Affiliation with protestant or evangelical religious groups may be associated
with later sexual initiation, and I control for protestant/evangelical vs. other (primarily Catholic).
Educational attainment is associated with the liberalization of sexual values, but it may also
provide an incentive to postpone early sexual activity. Thus, I control for educational attainment.
Many respondents are still enrolled in school, and thus educational attainment is limited to
whether they completed primary school. The economic and educational status of the family may
also be associated with the timing of first sex, and I control for household wealth and parental
education. The household wealth index is calculated using a scale of assets reported by the
66
household head (Rutstein & Johnson, 2004). Female respondents reported their parents’
educational attainment, which I classify by whether they completed primary school.
Haiti DHS- Analysis
I analyze data on age at first sex and migration using discrete-time event history analysis,
which models the probability that an event (first sex) will occur at any given age (Singer &
Willett, 2003). The basic form of the odds of experiencing first sex in a given person-year is:
(
)
is a vector of migration variables, which varies across the three models to reflect the
adaptation, disruption, or selection effect.
12-14, 15-17, 18-20).
is a vector of dummy variables for age groups (ages
is a vector of individual characteristics that may potentially confound
the association between migration and timing of first sex: religion, educational attainment,
household wealth, and parent education (female sample only).
In each of three models, I define
differently to align with one of the three proposed
theoretical explanations of how migration influences the timing of first sex. Model 1includes a
time-varying vector of dummy variables that indicates whether the individual is currently living
in a city, town, or rural area (reference group) as a non-migrant or is living in a city or town as a
migrant. These coefficients describe the likelihood experiencing first sex in any particular year
compared to rural residents. I then compare the coefficient for non-migrant city residence to the
one for city residence post-migration and the coefficients for non-migrant town residence to the
one for town residence post-migration. The difference in these values describes the differences in
the odds of first sex between migrants and non-migrants in cities and towns. If the values for
migrants and non-migrants in each location are similar to one another, the data support the
67
adaptation hypothesis, and if they differ significantly, the observed data support the disruption
hypothesis.
Model 2 includes three time-varying binary indicator variables for the year of migration,
one year post migration, and two years post migration. If these effects are significant, it provides
support for the disruption hypothesis. Model 3 includes binary indicators for each of three
different migration patterns (rural-to-town, rural-to-city, town-to-city) for those who migrated
after age 10; non-migrants are the reference group. Significance among these effects supports the
selection hypothesis
HYTS- Data
After examining how migration is associated with the timing of sexual initiation, I
examined if sexual and reproductive health knowledge and endorsement of premarital sex
transform as young migrants are adapting to their new environments. To do so I analyzed
longitudinal data from the HYTS, which recruited a cohort of rural youth as they completed
primary school (6th grade). The end of primary school is a point at which limited opportunities
lead many youth to migrate, and all respondents, including migrant youth (21.5% of the sample),
were re-interviewed at the destination. This design affords comparisons between migrants and
non-migrants from the same point of origin and offers advantages over designs that simply
compare rural-to-urban migrants to rural non-migrants.
Participants were recruited from school records that documented all students who
completed their sixth grade exams in the summer of 2011, within a specific geographical area in
rural southeast Haiti. Initial interviews occurred between August and September 2011, when a
parent or guardian was also interviewed and background information was collected. The second
interview occurred between February and March 2012. Youth who met selection criteria during
68
the first wave and could not be located were invited to participate in the second wave. A total of
224 youth participated; 213 were interviewed in the first wave and 215 in the second. Youth who
were eligible, but not available during the first wave were contacted again for the second wave;
thus 11 youth were added to the sample. In the case of respondents who could not be interviewed
face-to-face during the second wave, eight interviews were conducted over the telephone during
the second wave. Overall, 91% of respondents were interviewed in both waves, and multiple
imputations are used to account for the remaining respondents with incomplete data on the key
indicators.
Each interview lasted approximately one hour, and participants were interviewed in
Haitian Creole, their native language. Conducting research on sensitive topics with youth creates
the challenges of gaining accurate information such to protect the confidentiality of respondents
and ensure parents of their child’s welfare. In response to such concerns, interviews were
conducted in the view of family members at a distance where responses would not be audible
(e.g., the far end of the yard). To develop rapport with the respondent, the interviewer held the
guide in sight of the respondents so that he/she could follow along with the information that was
being recorded. While piloting the sexual and reproductive health knowledge component, it was
observed that respondents occasionally read from the available responses in the interview guide.
Therefore the original approach was modified, and interviewers recorded open-ended responses
to questions. To further ensure confidentiality and build rapport, cards that respondents could
touch to indicate their responses were used for categorical and ordinal responses. The cards
showed the responses in written format along with images of different quantities of beans
corresponding to the responses. Respondents were encouraged to touch their preferred response,
which the interviewer recorded, though many still responded orally.
69
HYTS Variables
To examine if sexual and reproductive health knowledge and endorsement of premarital
sex transformed during the early migration period, I constructed two scales that are measured
identically in the two waves. Sexual and reproductive health knowledge—which focused on how
to prevent HIV and pregnancy—was assessed by coding responses to open ended questions.
Youth were asked, (i) Please tell me what you know about HIV and AIDS and (ii) What can a
person do to prevent HIV and AIDS? (iii) Sometimes a person may want to have sex without
becoming pregnant. Can you tell me what someone can do to avoid becoming pregnant? (iv)
What do you think you might do in order to have the number of children you want? (v) When you
hear ‘family planning,’ (planin in Haitian Creole) what is it?
For each question, responses were recorded by the interviewer, and after each response,
the interviewer probed the respondent for additional responses to the same question. Responses to
the questions were coded by the author to reflect the knowledge domains that are typically taught
in national sexual health curriculums for youth.
Scores could range from zero to 12. A point was awarded for each of the following
components of HIV knowledge (i) identifying that HIV was an incurable illness (ii) knowing that
HIV could be transmitted by sex (iii) or by blood/needles, (iv) knowing that HIV could be
prevented by using condoms, (v) maintaining a mutually monogamous sexual relationship, (vi) or
by avoiding sexual activity, and (vii) for not providing any incorrect information (e.g, HIV is
transmitted by mosquito bites). A point was also awarded for each of the following components
of pregnancy prevention knowledge (i) identifying condoms as a means of pregnancy prevention,
(ii) recognizing that family planning was a means of preventing pregnancy, (iii) knowing that a
doctor could provide you with means of preventing pregnancy, (iv) identifying any specific
family planning method, and (v) for not providing any incorrect information (e.g., that pregnancy
70
could be prevented by not having during menstruation). Initially, HIV and pregnancy prevention
knowledge were two different scales, but they behaved similarly in the analysis and were
therefore combined.
To assess respondents’ endorsement of premarital sex, respondents were asked the extent
to which they agreed with a series of seven statements on a four-point Likert scale using the touch
cards. The items were generated based on beliefs articulated during preliminary fieldwork.
Respondents were asked if it was acceptable for a female youth to have sex with a friend,
someone she loved, and with a boyfriend, the same questions from the perspective of a male
youth, and if was acceptable for female youth to buy condoms (T1 α = .72; T2 α =.78).
Whether the respondents migrated between the two waves was determined from the lifehistory calendar. Having had a boyfriend or girlfriend and exposure to media sources were
additional factors that could influence the key outcome variables and vary by migration status.
Whether he/she had ever had a significant other at each time point was determined from the lifehistory calendar. Whether the individual had watched television during the previous week was a
response to a yes or no question. In both rural and urban areas, television is often watched in
public events where a television owner uses a generator and charges admission to see a movie or
show. In many urban areas, television service is more widely available, and adolescents may
watch television at home or at a friend’s home.
HYTS-Analysis
To examine whether there were differences in the overall level of knowledge and
attitudes, between migrants and non-migrants, whether they change between the two time points,
and whether change in these domains occurred differently for migrants and non-migrants, I fit a
71
multilevel model (i.e., growth curve) for examining change over time (Raudenbush & Bryk,
2002). The basic form of the model is:
represents the dependent variable—either sexual and reproductive health knowledge
or endorsement of premarital sex—at time i for person j. β1 is a time-varying covariate for survey
wave (0 or 1). The effect of this variable indicates whether there is growth or a decline in the
outcome between the two survey waves.
is a value for whether the individual ever migrated
and is consistent across the two time points. The effect of this variable indicates whether migrants
differ from non-migrants on the overall value level of the dependent variables.
is a migration
by time interaction (both values centered at 0.5 before calculating the interaction), and its
coefficient indicates whether the growth between the two survey waves differs between migrants
and non-migrants.
is the effect of the time consistent variable for gender.
represents the
effects of two time-varying covariates—whether the individual has ever had a boyfriend or
girlfriend and whether the individual watched television during the previous week. These are
included to control for factors that may confound that relationship between migration and the
dependent variables.
Results
Descriptive Results: Timing of Sexual Initiation
I first describe the characteristics of the DHS samples, which I used to analyze the timing
of sexual initiation as it relates to migration (Table 1). The reported values account for weighting
and complex survey design and are displayed separately for the male and females samples.
Starting with the age at sexual initiation, fewer female adolescents than male have initiated sex in
72
the sample (43.6% vs. 57.3%). This is consistent throughout adolescence, and by age 20, 68.8%
of female adolescents and 81.9% of male adolescents have experienced sex.
Turning to migration experiences, 21.6% of female and 9.9% of male adolescents in the
analytic sample migrated to a more-urban area between the ages of 10 and 20, and the mean ages
of these migration events were 15.2 and 15.1 respectively. The majority of non-migrants has
always lived in a rural area (45.8% among female and 61.4% among male respondents). Among
female adolescents, the most common migration pattern is town-to-city (9.8%), and fewer are
rural-to-city (4.1%) and rural-to-town (4.1%) migrants. Among male adolescents, rural-to-city
migration is most common (4.7%), and fewer are rural-to-town (2.7%) and town-to-city (2.6%)
migrants.
The current age of the respondents was similarly divided among the three age groups.
The sample is similarly divided by the current age of respondents. Evangelical affiliation was
reported by 48.8% of female and 39.1% of male respondents, and primary school completion was
reported by 49.4% of female and 40.8% of male respondents. Among female respondents, they
reside more often in households in the higher wealth quintiles, whereas male respondents are
more evenly distributed. The unequal distribution of female respondents may be a result of welloff households being more willing to receive girls who more-often contribute domestic help.
Among female respondents 42.4% of their mothers and 54.5% of their fathers completed primary
school. (Male respondents were not asked about parent education.)
Event-History Results: Timing of Sexual Initiation
Next, I describe the results of the discrete-time event history analysis for female
respondents (Table 2), beginning with the control variables. Across all the models, sexual
initiation occurred more often between the ages of 15 and 20 than between the ages of 12 and 14.
73
Evangelical identification lowered the odds of sexual initiation, and primary school completion
was not significant. Those living in households from the fourth wealth quintile were more likely
to initiate sex. The father’s completion of primary school decreased the odds of sexual initiation,
whereas mother’s was not significant.
Turning to the migration variables, results from Model 1, which included time-varying
covariates for residential location, first show that time-varying residence in a city as a nonmigrant is associated with higher odds of sexual initiation (O.R. = 1.58, p<.05). Other values do
not differ significantly from rural residence. The difference between the effects for town
residence as a non-migrant and town residence post-migration were not significant but did
indicate a trend whereby residence in a town as a non-migrant was associated with earlier sexual
initiation. The results describing the differences between city residence as a non-migrant and city
residence post-migration paralleled those described for town residence. In both of these cases, the
differences for migrants and non-migrants were insignificant; therefore, an adaptation explanation
is feasible. However, given that there was an observable trend, it is also possible that a disruption
effect accounts for these patterns.
Results from Model 2 reveal that female adolescents have lower odds of initiating sex in
the year that they migrate (O.R. = 0.46, p<.05); there is no significant effect for one year postmigration, and the effect of two years post-migration demonstrates that they have marginally
significant lower odds (O.R. = 0.44, p<.1) of initiating sex at that time. These findings garner
support for the disruption hypothesis. Counter to the belief that migrant youth are at risk, they
suggest that migrating to a more-urban area may in fact protect youth. Results from Model 3
reveal that the timing of sexual initiation for rural-to-town and rural-to-urban migrant female
adolescents do not differ significantly from non-migrants. However, the effect for town-to-city
migrants is marginally significant and positive (O.R. = 1.35, p<.1), suggesting that this group
may be somewhat more likely to initiate sex than non-migrants (who are primarily life-time rural
74
residents) at any given age. As a whole, Model 3 suggests that it is unlikely that the selection
hypothesis accounts for the pattern of sexual initiation among migrant adolescents.
I now turn to the results from male sample (Table 3), again beginning with the control
variables. Across all the models sexual initiation was most common between the ages of 15 and
17, but still higher between the ages of 18 and 20 than 12 to 14. Evangelical identification
delayed the odds of sexual initiation, whereas having completed primary education hastened it.
The effects of living in households of differing wealth were not significant.
Turning to the migration variables, notably, none of the migration variables are
significant after controlling for potentially confounding factors. Among rural-to- city migrants
(Model 3), rural-to-town migrants had marginally significant lower odds of initiating sex at any
given time compared to non-migrants. As a whole, these findings suggest that the association
between migration and the timing of first sex among male adolescents is trivial.
Descriptive Results: Knowledge and Attitudinal Change
Among respondents in the HYTS (Table 4), 21.5% experienced a migration episode
between the two waves of data collection. The mean sexual and reproductive health knowledge
score among all respondents in both waves of data was 7.64. This value was higher among nonmigrants than migrants and increased between the two waves of data collection. The mean score
of the endorsement of premarital sex was 14.62. These values were somewhat higher among nonmigrants and increased slightly between the two waves of data collection. Across the two waves
of data collection, 32.93% had a significant other; this experience was more common among nonmigrants and increased between the two waves of data collection. Watching television during the
previous week was reported 56.71% of the time, and this value was highest among those who
75
migrated to towns and cities after they migrated. Fifty six percent of the sample was male—
52.1% of migrants and 57.1% of non-migrants.
Growth Curve Results: Knowledge and Attitudinal Change:
I next describe the development of sexual and reproductive health knowledge during this
period (Table 5). These results reveal that knowledge increased between the two waves of data
collection (b = 0.50; p < .05) and that boys express higher levels of knowledge (b = 0.91; p <
.001). These findings are consistent with the expectation that sexual and reproductive health
knowledge will increase over time among adolescents and that boys might be more willing to
articulate their knowledge of sexual and reproductive health information. Moreover, those who
migrated had lower overall levels of sexual and reproductive health knowledge (b = -0.53; p <
.05). Migrants and non-migrants had similar scores at the first survey wave, and the difference in
the overall levels of knowledge is the consequence of diverging patterns by migration status (b =
-0.94; p < .05). Whereas sexual and reproductive health knowledge increased for non-migrants, it
remained stable for migrants (Figure 1). Further analyses were conducted to examine if these
patterns further differed by gender, and interactions with gender were not significant and are
therefore not reported. The time-varying effect of having a significant other was marginally
significant, whereas watching TV during the previous week was not.
Turning to adolescents’ endorsement of premarital sex (Table 5), it is again observed that
the acceptability of premarital sex increases between the two waves of data collection (b = 0.81; p
< 0.05) and that overall levels are higher for boys than girls (b = 3.25; p < .001 ). These findings
are consistent with expectations that adolescents are more willing to endorse these values as they
mature and that boys are more willing to endorse them than girls. Moreover, these attitudes did
not differ between migrants and non-migrants, nor was the interaction between time and
76
migration significant (not shown). Figure 2 shows the predicted values of endorsement of
premarital sex as they vary by time, gender, and migration status.
Discussion
Current perspectives assume that migrant adolescents who are new to urban areas risk
early and unsafe sex, because migration exposes them to more liberal sexual values and
psychosocial vulnerabilities (Bruce & Hallman, 2008; Lane, 2008). The findings from this study
largely present contrary evidence. Through a gendered lens I explored the timing of first sex,
sexual and reproductive health knowledge, and the endorsement of premarital sex among Haitian
youth who migrate to more-urban areas (e.g. rural-to-town, town-to-city).
I first examined how migration was associated with the timing of first sex, and whether
the observed patterns provided evidence for one of three hypotheses—adaptation, disruption, or
selection. The analytic sample included those who never migrated and those who migrated to
more-urban areas; it did not include lateral migrants or those who migrated to more-rural areas.
Results provided some support for the adaptation hypothesis and more convincing support for the
disruption hypothesis among female adolescents. Contrary to the prevailing assumption that
migrant adolescents are at risk, migrant female adolescents may in fact be delaying sexual
initiation during the year that they migrate and the years afterward. The findings among male
adolescents, however, present a different story: migration variables were not significantly
associated with the timing of their sexual debut.
Secondly, findings reveal that sexual and reproductive health knowledge trajectories
diverge for migrant and non-migrant adolescents during the initial period following migration.
The observed patterns indicate that knowledge increases between the survey waves among nonmigrants, but it remains constant among migrants. Though female adolescents had an overall
77
lower level of knowledge, the observed pattern did not differ by gender. Endorsement of
premarital sex also increased between the two waves of the data. Overall levels were higher
among male adolescents, but they did not significantly differ between migrants and non-migrants,
nor did they have differing patterns by migration status.
To adequately interpret these findings, I first carefully consider the population to which
the findings are generalizable. The Haiti DHS sample, which was used to analyze the timing of
sexual debut, compared those who were non-migrants to those who migrated to more-urban areas.
It excluded lateral migrants (e.g., from one rural household to another), many of whom may move
to nearby residences (commonly observed patterns for marriage and relocating to a new home)
and those who migrated to less-urban areas (likely to be return-migrants). In the HYTS, which
was used to examine change in knowledge and attitudes, the respondents were a cohort of rural
youth as they finished the sixth grade. Given the educational attainment necessary to meet the
inclusion criteria, those from the most disadvantaged households were more-likely excluded.
However, the exclusion criteria were not drastic, as among younger cohorts 70% of men and 60%
of women have completed primary school (Cayemittes et al., 2007). Thus, these findings are
generalizable to large portion of rural-to-urban adolescent migrants. However, they do not extend
to international migrants, displaced persons, or repeat migrants.
Given that a disruption effect best describes the observed patterns, what are the
explanatory mechanisms? One explanation for female adolescents’ lower odds of initiating sex
near the timing of migration is that their aspirations and preference for labor opportunities that
involves advanced education may motivate them to avoid romantic and sexual relationships.
Though their aspirations may be unrealistic, they shape how female adolescents plan for their
futures (Frye, 2012), and migration may further shape their aspirations and identity (Wolffers et
al., 2002). Moreover, in a context where pregnancy often quickly follows first sex (Cayemittes et
al., 2007), boyfriends often assume the right of sexual access, and sexual violence is
78
commonplace (Gómez, Speizer, & Beauvais, 2009), adolescent girls may perceive that avoiding
relationships is part of an effective strategy for completing school and achieving a respectable
job. For male adolescents, however, the risks of sex and pregnancy’s ability to jeopardize future
opportunities are largely irrelevant in a cultural context that places few demands on unwed
teenage fathers (Barrow, 1999).
An alternative explanation is that adjusting to the destination environment may hamper
their opportunities for sexual relationships. They may have fewer friends as a new migrant and
encounter fewer opportunities for sexual relationships. However, smaller social networks and
fewer sexual opportunities are likely not a complete explanation, as these explanations would
likely be similar among male and female youth. Instead, this study, along with previous work,
finds that migration factors were not significant among boys (Luke et al., 2012).
These findings—whereby female adolescents less-often initiate sex during the migration
year and migration and first sex are unrelated in male adolescent—suggest that, among the
adolescent migrants studied, they may not be more vulnerable to high-risk sex in the period
following migration. Findings from the second part of this study, however, reveal more subtleties.
The diverging trajectories of sexual and reproductive health knowledge between migrants and
non-migrants is perplexing, especially because urbanites generally demonstrate higher levels of
sexual and reproductive health knowledge (Cayemittes et al., 2007). Meanwhile, rural-to-urban
migrant adolescents endorse premarital sex similarly to rural non-migrants, and the level of
endorsement for these two groups increases similarly between the two survey waves. This occurs
even though migrant youth may experience more intense exposure to sexual topics in their new
urban environments. The stagnant knowledge growth trajectories, along with the growing
acceptance of premarital sex, raise concern about the well-being of those who are sexually active.
Given their lower knowledge adolescent migrants who are sexually active may be more likely to
engage in unsafe sex.
79
The diverging knowledge trajectories prompt the question: why might migrant
adolescents fail to continue acquiring important sexual and reproductive health knowledge? An
explanation may be their relationships with their peers. Though sexual education occurs formally
in schools and through media campaigns, much of this information is shared and discussed in
informal social groups (Bongaarts & Watkins, 1996). As migrant adolescents adapt to their new
environment, they may be absent from peer groups that reinforce this knowledge. Furthermore
migrant adolescents, who may be highly focused on reaching their educational and career
aspirations, may intentionally avoid sexual and reproductive health topics with their peers.
Limitations and Future Research
This study informs ideas about sexual and reproductive health among adolescent
migrants, but it is not without limitations. A primarily limitation is that the data in nationally
representative surveys—such as the DHS data used to examine the timing of sexual initiation—
do not allow for the incorporation of thorough migration histories, and migration information is
ascertained indirectly. The HYTS, which overlaps with the migration period, provides insight into
how knowledge and attitudes change; however, many have not yet begun sexual activity. In order
to further examine the intersection of migration and sexual and reproductive health, it is
important to continue gathering data that include both of these life course experiences.
The study is also limited in its attention to the characteristics of early sexual partnerships,
such as condom use and partner age. Previous findings demonstrate that rural-to-urban migrant
youth were less-likely to use condoms in recent sexual encounters than urban non-migrants
(Sambisa & Stokes, 2006). However, it was not possible to examine condom use in the Haiti
DHS data, because reported condom use at first sex was extremely low. As condom use at first
sex becomes more prevalent among adolescents, this may become a worthwhile research path.
80
Conclusion
A life-course approach and a gendered lens provide a useful framework for examining
the intersection of migration and sexual initiation. I tested competing hypothesis that may account
for the timing of first sex—adaptation, disruption, and selection. Previous work using this
framework to explain the fertility outcomes of adult women finds that adaptation is the most
likely explanation (Kulu, 2005). However, in this study, which examines age at first sex, a
disruption hypothesis is the most plausible explanation for the timing of first sex among female
adolescent migrants, although some support was found for the adaptation hypothesis. The
evidence suggests that female migrants are delaying sexual debut, and their aspirations may be a
key explanatory factor. Furthermore, the observed patterns of sexual initiation and migration
among male adolescents did not support any of the three hypotheses.
Additionally, migrant adolescents endorse premarital sex similarly to non-migrant
adolescents, and at the same time, migrant adolescents fail to continue accumulating protective
knowledge. The combination of these findings suggests that although female migrant youth may
be less likely to initiate sex, the context for doing so may be riskier. Sexual and reproductive
health education should address the unique needs of rural-to-urban migrant youth and take into
consideration their lower proclivity to engage on these topics. Additionally, because migration
from rural areas is common, rural adolescents health curriculums should help develop skill sets
for negotiating the urban environment.
81
Figure 3-1. Sexual and Reproductive Health Knowledge Patterns During Migration
Sexual Health Knowledge
Sexual and Reproductive Health
Knowledge During Migration
9
8
7
6
5
Time 1
Time 2
Female migrant
Female non-migrant
Male migrant
Male non-migrant
82
Figure 3-2. Change in Endorsement of Premarital Sex During Migration
Endorsement of Premarital Sex
Endorsement of Premarital Sex During Migration
17
16
15
14
13
12
11
10
Time 1
Time 2
Female migrant
Female non-migrant
Male migrant
Male non-migrant
83
Table 3-1. Description of DHS Analytic Sample
Female
Adolescents
43.6
26.1
62.3
68.8
Male
Adolescents
57.3
42.7
72.8
81.9
Ever migrated (btw. ages 10 and 20)
Mean age at migration
21.6
15.2
9.9
15.1
Residential pattern
Rural (never migrated)
Town (never migrated)
City (never migrated)
Rural to town migrant
Rural to city migrant
Town to city migrant
45.8
12.3
24.1
4.1
4.1
9.8
61.4
6.1
22.6
2.7
4.7
2.6
Current age
Age 15-16
Age 17-18
Age 19-20
36.1
33.9
30.0
34.7
32.2
33.1
Evangelical
Completed primary school
Household wealth quintile
Poorest
Poor
Middle
Wealthy
Wealthiest
48.8
49.4
39.1
40.8
14.3
17.3
18.5
23.5
26.4
22.7
19.1
16.2
19.5
22.4
Mother completed primary
Father completed primary
42.4
54.5
NA
NA
Sample size
1215
829
Has initiated sex
Sex by age 15
Sex by age 18
Sex by age 20
84
Table 3-2. Migration Variable Predicting First Sex Among Female Adolescents
Model 1:
Time-Varying
O.R.
SE
Time-varying residence
Rural
Town
City
Town post-mig.
City post-mig
Time-varying migration
Migration year
One year post-mig.
Two years post-mig.
Migration type
Non-migrant (ref.)
Rural to town
Rural to city
Town to city
Analysis age: 12-14
15 -17
18-20
Evangelical/protestant
Completed prim.edu.
Household wealth
Poorest
Poor
Middle
Wealthy
Wealthiest
Mother completed prim.
Father completed prim.
Intercept
Test of differences
Town - Town post-mig.
City - City post-mig.
1.00
1.10
1.58
.83
1.17
(.17)
(.30)
(.18)
(.24)
Model 2:
Migration Timing
O.R.
SE
*
.46
1.08
.44
1.00
2.86
2.55
.64
.85
(.31)
(.46)
(.06)
(.10)
1.08
1.04
1.00
1.57
1.12
1.11
.70
.10
(.19)
(.17)
1.32
1.35
(.28)
(.28)
(.26)
(.21)
(.15)
(.08)
(.02)
Model 3:
Selection
O.R.
SE
***
***
***
**
**
**
1.00
2.88
2.63
.65
.84
1.04
1.00
1.00
1.72
1.33
1.10
.73
.06
Person-years
6185
6185
F-test
9.8
*** 10.9
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
(.18)
(.33)
(.21)
(.31)
(.47)
(.06)
(.10)
*
†
***
***
***
(.18)
(.16)
(.27)
(.22)
(.14)
(.09)
(.01)
1.00
.88
.68
1.35
1.00
2.85
2.57
.65
.85
**
***
1.04
1.01
1.00
1.68
1.22
1.10
.72
.06
***
6185
10.9
***
†
(.17)
(.21)
(.24)
(.30)
(.45)
(.06)
(.10)
†
***
***
***
(.18)
(.16)
(.26)
(.21)
(.14)
(.09)
(.01)
***
**
***
***
85
Table 3-3. Migration Variable Predicting First Sex Among Male Adolescents
Model 1:
Time-Varying
O.R.
SE
Time-varying residence
Rural
Town
City
Town post-mig.
City post-mig
Time-varying migration
Migration year
One year post-mig.
Two years post-mig.
Migration type
Non-migrant (ref.)
Rural to town
Rural to city
Town to city
Analysis age: 12-14
15 -17
18-20
Evangelical/protestant
Completed prim.edu.
Household wealth
Poorest
Poor
Middle
Wealthy
Wealthiest
Intercept
Test of differences
Town - Town post-mig.
City - City post-mig.
1.00
1.16
.99
1.43
1.10
Model 2:
Migration Timing
O.R.
SE
(.23)
(.23)
(.33)
(.23)
1.55
.78
.77
1.00
2.52
1.57
.79
1.74
Model 3:
Selection Model
O.R.
SE
***
*
*
***
1.00
2.53
1.60
.79
1.75
(.23)
(.31)
(.02) ***
.78
.78
1.00
1.14
1.26
.07
(.29)
(.36)
(.10)
(.22)
.81
.79
1.00
1.10
1.26
.07
(.16)
(.15)
.81
.90
(.19)
(.18)
Person-years
3586
F test
8.7
***
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
3586
9.6
(.57)
(.42)
(.45)
***
*
†
***
1.00
1.01
.63
.90
1.00
2.55
1.62
.80
1.71
(.22)
(.25)
(.01) ***
.78
.78
1.00
1.19
1.35
.07
(.29)
(.36)
(.10)
(.22)
(.16)
(.15)
***
3586
10.1
(.24)
(.17) †
(.34)
(.29)
(.37)
(.10)
(.22)
***
*
†
***
(.16)
(.15)
(.24)
(.28)
(.01) ***
***
86
Table 3-4. Descriptive Statistics HYTS Respondents
Grand Mean
Mean
Outcomes
Sexual and reproductive health knowledge 7.64
Endorsement of premarital sex
14.62
Time-varying covariates
Had a significant other
Watched TV during previous week
Individual Characteristics
Migrants
Male
SE
(.09)
(.25)
Wave 1 (Pre-migration)
Migrants
Non-migrants
Mean
SE
Mean
SE
7.14
13.87
(.32) 7.27
(.87) 14.27
Percent
32.9
56.7
Percent
22.9
43.2
Percent
27.4
55.9
21.5
56.1
52.1
57.1
(.11)
(.38)
Wave 2 (Post-migration)
Migrants
Non-migrants
Mean
SE
Mean
SE
7.21
13.91
Percent
33.3
80.0
(.29) 8.28
(.81) 15.38
Percent
41.1
54.8
(.15)
(.40)
87
Table 3-5. Results of Growth Curve Analysis HYTS
Sexual and Reproductive
Health Knowledge
B
S.E.
Level-1: Time-varying Characteristics
Time (wave 1 = 0, wave 2 =1)
Had a significant other (time varying)
Watched TV previous week (time varying)
.50
.33
-.02
(.21)
(.18)
(.18)
*
†
Level-2: Individual Characteristics
Male
Migrated to town/city
.91
-.53
(.17)
(.22)
***
*
Cross-level Interaction
Time x Migration
-.94
(.41)
*
6.90
(.21)
***
Intercept
† p ≤ 0.10, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001
Endorsement of
Premarital Sex
B
S.E.
.81
.49
.04
(.40)
(.54)
(.50)
*
3.25
-.74
(.55)
(.66)
***
12.37
(.55)
***
88
References
Agadjanian, V., Nedoluzhko, L., & Kumskov, G. (2008). Eager to Leave? Intentions to
Migrate Abroad among Young People in Kyrgyzstan1. International Migration
Review, 42(3), 620–651. doi:10.1111/j.1747-7379.2008.00140.x
Arnett, J. J. (2002). The psychology of globalization. American Psychologist, 57(10),
774.
Barrow, C. (1999). Family in the Caribbean: themes and perspectives. Marcus Wiener.
Behrman, J., & Sengupta, P. (2005). Changing contexts in which youth are transitioning
to adulthood in developing countries: converging toward developed economies?
In The Changing Transition to Adulthood in Developing Countries: Selected
Studies (pp. 13–55). Washington, DC: The National Academies Press.
Bjarnason, T., & Thorlindsson, T. (2006). Should I stay or should I go? Migration
expectations among youth in Icelandic fishing and farming communities. Journal
of Rural Studies, 22(3), 290–300. doi:10.1016/j.jrurstud.2005.09.004
Bjorklund, A., & Salvanes, K. G. (2010). Education and Family Background:
Mechanisms and Policies (SSRN Scholarly Paper No. ID 1631137). Rochester,
NY: Social Science Research Network. Retrieved from
http://papers.ssrn.com/abstract=1631137
Bongaarts, J., & Watkins, S. C. (1996). Social interactions and contemporary fertility
transitions. Population and Development Review, 22(4), 639.
doi:10.2307/2137804
89
Brauw, A. D., & Giles, J. (2008). Migrant Opportunity and the Educational Attainment of
Youth in Rural China. SSRN eLibrary. Retrieved from
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1096849
Bredl, S. (2010). Migration, remittances and educational outcomes: The case of Haiti.
International Journal of Educational Development, In Press, Corrected Proof.
doi:10.1016/j.ijedudev.2010.02.003
Brockerhoff, M., & Biddlecom, A. E. (1999). Migration, Sexual Behavior and the Risk of
HIV in Kenya. International Migration Review, 33(4), 833–856.
doi:10.2307/2547354
Bruce, J., & Hallman, K. (2008). Reaching the girls left behind. Gender & Development,
16(2), 227–245. doi:10.1080/13552070802118149
Buchmann, C. (2000). Family Structure, Parental Perceptions, and Child Labor in Kenya:
What Factors Determine Who Is Enrolled in School? Social Forces, 78(4), 1349–
1378. doi:10.1093/sf/78.4.1349
Buchmann, C., & Hannum, E. (2001). Education and Stratification in Developing
Countries: A Review of Theories and Research. Annual Review of Sociology, 27,
77–102. doi:10.2307/2678615
Carpenter, L. M. (2010). Gendered Sexuality Over the Life Course: A Conceptual
Framework. Sociological Perspectives, 53(2), 155–178.
doi:10.1525/sop.2010.53.2.155
Castellanos, M. B. (2007). Adolescent Migration to Cancún: Reconfiguring Maya
Households and Gender Relations in Mexico’s Yucatán Peninsula. Frontiers,
28(3), 1.
90
Cayemittes, M., Placide, M. F., Mariko, S., Barrère, B., Sévère, B., & Alexandre, C.
(2007). Enquête Mortalité, Morbidité et Utilisation des Services, Haïti, 20052006. Calverton, Maryland: Ministère de la Santé Publique et de la Population,
Institut Haïtien de l’Enfance et Macro International Inc.
Collins, W. A., & Steinberg, L. (2006). Adolescent development in interpersonal context.
In Handbook of Child Psychology (6th ed., Vol. 3, Social, emotional, and
personality development). Hoboken, NJ, USA: John Wiley & Sons, Inc.
Crivello, G. (2009). Becoming somebody’: Youth transitions through education and
migration–evidence from Young Lives, Peru. Young Lives Working Paper, 43.
Retrieved from
http://www.dfid.gov.uk/R4D/PDF/Outputs/YoungLives/WP43_Summary.pdf
Curtis, S. L., & Sutherland, E. G. (2004). Measuring sexual behaviour in the era of
HIV/AIDS: the experience of Demographic and Health Surveys and similar
enquiries. Sexually Transmitted Infections, 80(suppl 2), ii22–ii27.
doi:10.1136/sti.2004.011650
Demombynes, G., Holland, P., & Leon, G. (2010). Students and the Market for Schools
in Haiti (SSRN Scholarly Paper No. ID 1726252). Rochester, NY: Social Science
Research Network. Retrieved from http://papers.ssrn.com/abstract=1726252
Dhapola, M., Sharan, M., & Shah, B. (2007). Migration, Youth and HIV Risk: A Study
of Young Men in Rural Jharkhand. Economic and Political Weekly, 42(48), 40–
47. doi:10.2307/40276716
Easterlin, R. A. (1975). An Economic Framework for Fertility Analysis. Studies in
Family Planning, 6(3), 54–63. doi:10.2307/1964934
91
Edmond, Y. M., Randolph, S. M., & Richard, G. L. (2007). The lakou system: A cultural,
ecological analysis of mothering in rural Haiti. Journal of Pan African Studies,
2(1), 19–32.
Ferguson, J. (2003). Migration in the Caribbean: Haiti, the Dominican Republic and
Beyond. London, UK: Minority Rights Group International.
Frye, M. (2012). Bright Futures in Malawi’s New Dawn: Educational Aspirations as
Assertions of Identity. American Journal of Sociology, 117(6), 1565–1624.
doi:10.1086/664542
Gammage, S. (2004). Exercising Exit, Voice and Loyalty: A Gender Perspective on
Transnationalism in Haiti. Development and Change, 35(4), 743–771.
doi:10.1111/j.0012-155X.2004.00378.x
Gibbison, G., & Paul, C. (2006). Economic Incentives for Fostering Jamaican Children.
The Journal of Developing Areas, 39(2), 29–39.
Goldstein, S., & Goldstein, A. (1981). The Impact of Migration on Fertility: an `Own
Children’ Analysis for Thailand. Population Studies, 35(2), 265–284.
Gómez, A. M., Speizer, I. S., & Beauvais, H. (2009). Sexual Violence and Reproductive
Health Among Youth in Port-au-Prince, Haiti. Journal of Adolescent Health,
44(5), 508–510. doi:10.1016/j.jadohealth.2008.09.012
Hareven, T. K. (1982). Family and industrial time: The relationship between the family
and work in a New England industrial community. University Press.
Hsin, A. (2007). Children’s Time Use: Labor Divisions and Schooling in Indonesia.
Journal of Marriage and Family, 69(5), 1297–1306. doi:10.1111/j.17413737.2007.00448.x
92
Kolbe, A. R., & Hutson, R. A. (2006). Human rights abuse and other criminal violations
in Port-au-Prince, Haiti: a random survey of households. The Lancet, 368(9538),
864–873. doi:16/S0140-6736(06)69211-8
Kulu, H. (2005). Migration and Fertility: Competing Hypotheses Re-Examined.
European Journal of Population / Revue Européenne de Démographie, 21(1), 51–
87. doi:10.2307/20164290
Lane, C. (2008). Adolescent refugees and migrants: a reproductive health emergency.
Watertown, MA: Pathfinder International.
Lloyd, C. B. (2004). Growing up global: The changing transitions to adulthood in
developing countries. National Academies Press. Retrieved from
http://books.google.com/books?hl=en&lr=&id=4EGF6vKb2_8C&oi=fnd&pg=P
A1&dq=growing+up+global&ots=xu6rXEq0CR&sig=U005TQg__8rDBOVQVR
sSfZqur8M
Lloyd, C. B., & Mensch, B. S. (2008). Marriage and childbirth as factors in dropping out
from school: An analysis of DHS - data from sub-Saharan Africa. Population
Studies: A Journal of Demography, 62(1), 1. doi:10.1080/00324720701810840
Luke, N., Xu, H., Mberu, B. U., & Goldberg, R. E. (2012). Migration Experience and
Premarital Sexual Initiation in Urban Kenya: An Event History Analysis. Studies
in Family Planning, 43(2), 115–126. doi:10.1111/j.1728-4465.2012.00309.x
Lunde, H. (2009). Haiti Youth Survey 2009. Volume I: Tabulation Report. Oslo: Fafo.
Manigat, S. (1997). Haiti: The Popular Sectors and the Crisis in Port-au-Prince. In The
Urban Caribbean: transition to the new global economy. Baltimore, Maryland:
Johns Hopkins University Press.
93
Massey, D. S., Arango, J., Hugo, G., Kouaouci, A., Pellegrino, A., & Taylor, J. E. (1993).
Theories of International Migration: A Review and Appraisal. Population and
Development Review, 19(3), 431–466. doi:10.2307/2938462
Mberu, B. U., & White, M. J. (2011). Internal migration and health: Premarital sexual
initiation in Nigeria. Social Science & Medicine, 72(8), 1284–1293.
doi:10.1016/j.socscimed.2011.02.019
McElroy, J. ., & de Albuquerque, K. (1988). The impact of migration on mortality and
fertility in St. Kitts-Nevis and the U.S. Virgin Islands. Caribbean Geography: A
Journal of Geography for the Region, 2(3), 177–194.
McKenzie, D. J. (2008). A Profile of the World’s Young Developing Country
International Migrants. Population and Development Review, 34(1), 115–135.
doi:10.1111/j.1728-4457.2008.00208.x
Mensch, B. S., Grant, M. J., & Blanc, A. K. (2006). The Changing Context of Sexual
Initiation in sub‐Saharan Africa. Population and Development Review, 32(4),
699–727. doi:10.1111/j.1728-4457.2006.00147.x
Metz, H. C., & The Library of Congress (US) Federal Research Division. (2001).
Dominican Republic and Haiti: Country Studies (3rd ed.). Washington, DC:
Library of Congress.
Mintz, S. W. (2010). Three Ancient Colonies: Caribbean Themes and Variations.
Cambridge, MA: Harvard University Press.
Moya, J. C. (2007). Domestic Service in a Global Perspective: Gender, Migration, and
Ethnic Niches. Journal of Ethnic and Migration Studies, 33(4), 559–579.
doi:10.1080/13691830701265420
94
Palmer, R. W. (2009). The Caribbean economy in the age of globalization. New York,
NY: Macmillan.
Punch, S. (2007). Migration projects: children on the move for work and education. In
Workshop Independent Child Migrants: Policy Debates and Dilemmas,
Development Research Centre on Migration, Globalisation and Poverty,
University of Sussex and UNICEF Innocenti Research Centre. Westminster,
London. Retrieved from
http://www.childmigration.net/files/Punch_migration_paper.pdf
Puri, M. C., & Busza, J. (2004). In forests and factories: sexual behaviour among young
migrant workers in Nepal. Culture, Health & Sexuality, 6(2), 145–158.
doi:10.1080/13691050310001619653
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and
data analysis methods. Thousand Oaks, CA: Sage Publications.
Renaud, F. G., Dun, O., Warner, K., & Bogardi, J. (2011). A Decision Framework for
Environmentally Induced Migration. International Migration, 49, e5–e29.
doi:10.1111/j.1468-2435.2010.00678.x
Rutstein, S. O., & Johnson, K. (2004). The DHS Wealth Index. Calverton, Maryland:
ORC Macro.
Salmi, J. (2000). Equity and Quality in Private Education: The Haitian paradox.
Compare: A Journal of Comparative and International Education, 30(2), 163–
178. doi:10.1080/03057920050034101
95
Sambisa, W., & Stokes, C. S. (2006). Rural/Urban Residence, Migration, HIV/AIDS, and
Safe Sex Practices among Men in Zimbabwe*. Rural Sociology, 71(2), 183–211.
doi:10.1526/003601106777789684
Schlesinger, B. (1968). Family Patterns in the English-Speaking Caribbean. Journal of
Marriage and Family, 30(1), 149–154.
Schwartz, T. (2009). Fewer Men, More Babies: Sex, Family, and Fertility in Haiti.
Lanham, MD: Lexington Books.
Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis: Modeling
change and event occurrence. Oxford University Press, USA.
Smith, J. M. (2001). When the hands are many: community organization and social
change in rural Haiti. Cornell University Press.
Sommer, M. (2010). Where the education system and women’s bodies collide: The social
and health impact of girls’ experiences of menstruation and schooling in
Tanzania. Journal of Adolescence, 33(4), 521–529.
doi:10.1016/j.adolescence.2009.03.008
Stark, O., & Bloom, D. E. (1985). The new economics of labor migration. The American
Economic Review, 75(2), 173–178.
Thorsen, D. (2010). The Place of Migration in Girls’ Imagination. Journal of
Comparative Family Studies, 41(2), 265–280.
Tienda, M., Taylor, L., & Moghan, J. (2007). New Frontiers, Uncertain Futures: Migrant
Youth and Children of Migrants in a Globalised World. In Background Paper for
Zurich Migration Workshop (Vol. 17). Retrieved from
http://crcw.princeton.edu/migration/files/new_frontiers_uncertain_futures.pdf
96
Van de Ven, W. P. M. M., & Van Praag, B. M. S. (1981). The demand for deductibles in
private health insurance: A probit model with sample selection. Journal of
Econometrics, 17(2), 229–252. doi:10.1016/0304-4076(81)90028-2
Wolffers, I., Fernandez, I., Verghis, S., & Vink, M. (2002). Sexual behaviour and
vulnerability of migrant workers for HIV infection. Culture, Health & Sexuality,
4(4), 459–473. doi:10.1080/13691050110143356
Yaqub, S. (2009a). Independent child migrants in developing countries: unexplored links
in migration and development. UNICEF Innocenti Research Centre. Retrieved
from http://ideas.repec.org/p/ucf/inwopa/inwopa09-62.html
Yaqub, S. (2009b). Child migrants with and without parents: Census-based estimates of
scale and characteristics in Argentina, Chile and South Africa. UNICEF
Innocenti Research Centre. Retrieved from
http://ideas.repec.org/p/ucf/indipa/indipa09-4.html
Zaba, B., Pisani, E., Slaymaker, E., & Boerma, J. T. (2004). Age at first sex:
understanding recent trends in African demographic surveys. Sexually
Transmitted Infections, 80(suppl 2), ii28–ii35. doi:10.1136/sti.2004.012674
Zuma, K., Lurie, M. N., Williams, B. G., Mkaya-Mwamburi, D., Garnett, G. P., & Sturm,
A. W. (2005). Risk Factors of Sexually Transmitted Infections among Migrant
and Non-Migrant Sexual Partnerships from Rural South Africa. Epidemiology and
Infection, 133(3), 421–428. doi:10.2307/3865658
1
Curriculum Vitae
Education
Ph.D.
The Pennsylvania State University, University Park, PA
2013
Human Development & Family Studies and Demography
Dissertation: Youth Migration and Youth Transitions: The Haitian Experience
Specializations: Family Demography, Research Methods
M.S.
The Pennsylvania State University, University Park, PA
Human Development & Family Studies and Demography
Thesis: Ethnic Group Disparities in Academic Skill
Development across Four Economically Developing Countries
2011
B.S.
Greensboro College, Greensboro, NC
Psychology, Summa Cum Laude, Honors Program Degree
2003
Publications
Heckert, J. & Fabic, M.S. (2013). Survey Data Needs Concerning Women’s Empowerment in
Sub-Saharan Africa. Studies in Family Planning, 44(3): 319-344.
Smith-Greenaway, E. & Heckert, J. (2013). Does the Orphan Disadvantage “Spill Over”? An
analysis of whether living in an area with a higher concentration of orphans is associated with
children’s school enrollment in sub-Saharan Africa. Demographic Research, 28(40): 1167-1198.
Nau, C. & Heckert, J. (2012). Integrating perspectives on child health: Where do we go now? In
N. Landale, S. McHale, & A. Booth (Eds.), Families and Child Health (pp. 213-226). New York,
NY: Springer.
Jayakody, R., Heckert, J., and Anh, D.N. (2010). Social change and premarital sexual behavior
and attitudes in Vietnam. In R. Silbereisen & X. Chen (Eds.), Social Change in Human
Development: Concepts and Results (pp. 227-244). London, UK: Sage.
Under Review
Heckert, J., Boatemaa, S., & Altman, C. (Under Review). Migrant Youth’s Emerging Dietary
Patterns in Haiti: The Role of Peer Social Engagement. at Public Health Nutrition
Awards & Fellowships
Graduate Research Fellowship, National Science Foundation
Predoctoral Traineeship in Family Demography, NICHD
Ruth W. Ayers-Givens Award, Penn State
Early Career Development Award, Center for Global Studies, Penn State
Research Grant, Africana Research Center, Penn State
University Graduate Fellowship, Penn State
2010-2013
2008-2010
2012
2011
2011
2007-2008