Migration and Fertility in Ghana

1
Migration and Fertility in Ghana

• Stephen Obeng Gyimah, PhD
• Department of Sociology
• Queens University
• Kingston, Ontario
• gyimahs_at_post.queensu.ca
• A paper prepared for presentation at the annual
  meetings of the Canadian Population Society,
  Winnipeg, June 2-5 2004.

2
Research context

• Previous research on migration in sub-Saharan
  Africa has primarily focused on motives,
  determinants and consequences.
• This contrast with the considerable research on
  fertility-child mortality nexus.
• Although studying migration per se brings
  attention to the spatio-temporal aspects of
  population redistribution, a better understanding
  of population dynamics in general may be gained
  if the links between migration and the other
  components of population change are examined in
  unison.

3

• To our knowledge, only a handful of studies have
  empirically examined the relationship between
  migration and fertility in sub-Saharan using
  national level data and these were mostly based
  on data that may not capture recent trends.
• As Lindstrom (2002) argues, fertility and
  migration are generally thought to be affected by
  similar factors and as such, understanding their
  inter-connectedness may provide a setting for
  analyzing fertility response to social and
  economic change.
• Also, the bulk of previous migration-fertility
  research in the developing world has focused
  exclusively on rural-urban migrants.

4

• While such studies, particularly in Asia and
  Latin America, seem justified given the
  overarching volume of the rural-urban stream, the
  same cannot be said of sub-Saharan Africa where
  other migration streams (rural-rural,
  urban-rural, urban-urban) are equally important
  (Oucho Gould, 1993).
• In the context of sub-Saharan Africa thus, the
  multi-dimensionality of the migration-fertility
  relationship may not be adequately captured if
  the other migrant streams are ignored. Perhaps
  the contradictory findings in studies from
  sub-Saharan Africa may be due to this failure.

5

• With the availability of data for much of
  sub-Saharan African through the United States
  Agency for International Development funded
  Demographic and Health Survey (DHS) program, this
  study contributes to the discussion by exploring
  the impact of migration on individual womens
  fertility in Ghana.

6
Theoretical framework

• Exploring the fertility behavior of migrants
  requires an understanding of the underlying
  theoretical mechanisms.
• This study is guided by competing but often
  complimentary theses on migrant fertility
  focusing on the processes of adaptation,
  selectivity, and disruption .
• These hypotheses have received varied empirical
  support in the developing world (see, e.g.,
  Bacal, 1988 Brockerhoff Yang, 1994 Campbell,
  1989 Farber Lee, 1984 Goldstein Goldstein,
  1983 Goldstein, White Goldstein, 1997
  Hervitz, 1985 Lee, 1992 Lee Farber, 1985 Lee
  Pol, 1993 Lindstrom, 2003 Stephen Bean,

7

• In our view, these theoretical perspectives may
  suggest high or low fertility depending on a
  particular migratory pattern.
• In general, whether the fertility of migrants is
  similar or dissimilar to those at destination may
  depend on the extent to which they are subjected
  to the processes of socialization, adaptation,
  selectivity and disruption.
• In this study, we explore the empirical evidence
  for migrant fertility differential and attempt an
  explanation for the observed pattern through
  these paradigms.
• Although the cross-sectional nature of our data
  may limit a rigorous test of these hypotheses,
  they nonetheless provide an organizing framework
  for our ideas.

8
Methodology

• While longitudinal data are ideal for studying
  migrant fertility, such data are not readily
  available in much of sub-Saharan Africa including
  Ghana.
• Following previous research, we used the
  cross-sectional retrospective data from cycles 2
  and 3 (1993 1998) of the Ghana Demographic and
  Health Survey (GDHS) individual womens file.
• The GDHS is a nationally representative,
  stratified, self-weighting probability sample
  survey of women aged15 to 49 years.

9

• Because of the use of the similar sampling
  techniques and survey instruments at both cycles,
  we decided to merge the data to ensure adequate
  sample size for detailed analysis by migration
  stream.
• The analysis in this paper was limited to
  ever-married women who were non visitors.
• The merged file yielded a sample size of 7281
  ever married women composed of 3597 from the 1993
  survey and 3684 from the 1998 sample.

10
Measures

• Previous research has used varied measures such
  as birth timing and children ever born as
  measures of fertility (e.g., Brockerhoff Yang,
  1994 Lee, 1992 Lee Pol, 1993 Lindstrom,
  2003 Trovato, 1987 White et al., 1995).
  Consistent with this, we used children ever-born
  as the measure of fertility.
• The main independent variable, migration status,
  was derived from the questions on migration
  history. Six categories of migrants were
  identified as rural-urban urban-rural
  urban-urban, rural-rural, rural non migrant and
  urban non migrant.

11
Measures

• Given the cross-sectional nature of the data,
  there is the lack of information on the precise
  timing of migration and this is a limitation that
  needs to be recognized
• To refine the analysis, however, information on
  duration of stay in current place was used to
  further disaggregate each of the two main migrant
  streams (rural-urban urban-rural) into recent
  and long term migrants.
• Recent migrants are defined as those who have
  been living in the current area for a period of
  less than 5 years while long term migrants are
  those who have been resident for more than 2
  years
• This was done for only rural-urban and
  urban-rural migrants primarily because such
  movements involve economically and culturally
  distinct settings unlike rural-rural or
  urban-urban which invariably involve movements
  into similar settings.

12
Findings
13
Findings (descriptive Table 1)

• Panel A focuses on mean fertility by migrant
  stream while Panel B disaggregates the
  rural-urban and urban-rural streams into recent
  and long term migrants.
• Overall, there is evidence of significant
  differences in fertility. Mean fertility is
  highest among women currently living in rural
  areas regardless of migration stream.
• Such differences are more pronounced among older
  women (aged 35 years and above) where the mean
  fertility is 6.2, 5.7 and 5.6 for rural-rural
  migrants, rural non migrants and urban-rural
  migrants respectively compared with 4.3, 4.4 and
  4.7 for urban non migrants, urban-urban migrants
  and rural-urban migrants respectively.
• It is also clear that fertility of migrants tends
  to be similar to those at the destination area,
  perhaps an indication of the integration
  hypothesis.

14
Descriptive findings
15
Multivariate analysis

• Because children ever born (the outcome variable)
  is a non negative count variable, count data
  models are the natural choice for the regression.
  As Long (1997) points out, using linear
  regression for count outcomes can result in
  inefficient, inconsistent and biased estimates.
  While some argue that the birth function follows
  a Poisson distribution, others suggest a Negative
  Binomial distribution.
• The problem one faces with standard Poisson model
  is that the equi-dispersion assumption which
  underlies its estimation approach is often
  violated.
• To make the choice between Poisson and Negative
  Binomial Regression Models, we tested but found
  no evidence of over dispersion4, suggesting a
  preference for the Poisson model.

16
Multivariate analysis

• In our case, there was the need to go beyond the
  basic Poisson model to account for possible
  unobserved heterogeneity.
• While observed heterogeneity in the form of
  covariates is usually accounted for in most
  studies, the possible unobserved selectivity of
  migrants is often less explored.
• There are potential unobserved factors that could
  influence both migration and fertility processes,
  and not accounting for that may cause the error
  terms to be correlated with the explanatory
  variables leading to inconsistent estimates and
  wrong conclusions.

17

• In the context of the present study, for
  instance, unmeasured upward mobility aspirations
  may not only predispose some women to migrate but
  may also lower their fertility. In other words,
  the decision to migrate may become part of ones
  fertility strategies.
• To account for unobserved heterogeneity, we used
  the Conditional Fixed-Effect Poisson Regression.
• For meaningful interpretation, the coefficients
  have been transformed by exponentiation (exp b)
  and can be interpreted as incidence ratios. For
  a categorical variable, a risk ratio
  significantly greater than one indicates that
  women with this attribute have higher fertility
  than those in the reference category while the
  reverse is true if the ratio is less than one.

18
Conditional Fixed Effect Poisson Results
19
conclusions

• This paper sought to explore migrant fertility
  differentials in Ghana and provide an explanation
  for the observed pattern.
• We found significant differentials in fertility
  with two distinct patterns.
• We noticed a significantly lower fertility for
  rural-urban, urban-urban and urban non migrants,
  and higher fertility for rural-rural, urban-rural
  and rural non migrants. The low fertility of
  rural-urban in particular contrasts the argument
  that city-ward migration will increase fertility
  as a result of the breakdown of traditional
  fertility inhibiting practices .
• There was also evidence that the rural-urban and
  urban-rural streams were not homogenous in terms
  of fertility behavior. In the multivariate
  models, the fertility of recent migrants was
  found to be considerably low fertility regardless
  of stream.

20
conclusions

• For urban-rural migrants in particular, we
  observed that such differences diminish over
  time, leading to a convergence with rural
  fertility. This checkered fertility behavior of
  urban-rural migrants has been explained through
  disruption and integration theses.
• Revisiting the major hypotheses, it is evident
  that migrant fertility differentials in Ghana
  cannot be accounted for solely by any of the
  theses.
• While selectivity explains some of the
  differential fertility of migrants, disruption
  and integration are also important. The lower
  fertility of recent rural-urban migrants may be
  attributed to selectivity and disruption effects
  than integration since such migrants have not
  lived in the urban area long enough to imbibe the
  low fertility norms in such a locality.

21
conclusions

• Overall, there was also evidence of convergence
  in migrant-native fertility. We found that in
  the long run, migrants adapt to fertility levels
  prevalent at destination regardless of being of
  urban or rural origin.
• Migrants moving from urban to rural areas
  exhibited fertility similar to rural residents
  while those from rural to urban areas have
  fertility levels similar to urban natives. While
  the imprecise measure of the timing of migration
  lends some support to each of the theoretical
  perspectives, the overall findings are more
  consistent with the adaptation hypothesis

22

• This study contributes to the literature on
  migrant fertility in a number of ways.
• First, we used a statistical model that allowed
  us to model unobserved heterogeneity and test the
  main theoretical perspectives.
• Again, we provide empirical evidence that the
  major hypotheses may predict different fertility
  outcomes for different streams of migrants.