Urban Migration of Adolescent Girls: Quantitative Results from Developing Countries

Abstract

This chapter assembles a quantitative portrait of the adolescent girls who migrate to the cities and towns of poor countries, drawing mainly on a large collection of data from demographic surveys and census micro-samples. For adolescent girls and young women, migration puts important urban resources within reach, in the form of access to higher levels of schooling, more varied labor markets and employment opportunities, and multiple levels of health-care institutions. But while the move is underway and until she locates a safe home in her new location, an adolescent girl can confront a range of social and sexual risks that can threaten her well-being and thwart hopes for advancement. Much of the literature on adolescent migration is focused on these risks but neglects the potential benefits. We find that in many countries, significant percentages of urban adolescent girls are recent in-migrants. In characterizing their life-circumstances, we give special attention to indicators of social isolation, the conditions of housing and neighborhood, and school enrollment. We show that adolescent girl migrants are a highly diverse group, advantaged in some respects and disadvantaged in others. Field studies in urban India shed light on the difficulties with which girls must cope as they strive to adapt to urban life.

Contact email: mmontgomery@popcouncil.org. Portions of this chapter draw on Temin et al. (2013), a recent report on migrant adolescent girls.

Appendix

Broadly similar approaches are taken in censuses and DHS surveys to collect data on migration, but the approaches differ in several respects and neither data-gathering mechanism is fully satisfactory. To appreciate the differences in the two approaches, it may be helpful to set them against a third alternative that would collect minimally adequate migration information. Such a mechanism would summarize moves in terms of their origin and destination , with the urban or rural nature of both locations recorded and situated geographically either by point coordinates or (more realistically) in terms of small administrative units. In this way, the data would be aligned with migration theory, which emphasizes the role of spatial differences in current living standards and longer-term life prospects across a range of potential destinations (Lucas 1997). Unfortunately, neither the IPUMS censuses nor the DHS surveys meet these minimal criteria, and they fall short in different ways.

One relatively minor difference between the census and survey approaches has do with the description of a household ’s place of residence at the time of the interview . The vast majority of censuses—although surprisingly, not all—indicate whether the current residence is urban or rural according to the country’s official definition.¹² The official definition is also applied by the DHS to classify the sampling clusters of its surveys. Where the census and DHS programs differ is in supplying geographic context on the administrative units in which the interviewed households reside. This is less a matter of what data are collected than of restrictions on their release into the public domain. To protect respondent confidentiality, the census files made available through IPUMS generally identify locations only by broad administrative region , such as the province of residence or a similar first-level administrative unit. The equivalent of first-level administrative area is also available in most DHS surveys, but quite a number of these surveys additionally supply finer geographic detail in supplemental country-specific variables. In recent years, an increasing percentage of DHS surveys have gone even further in the direction of spatial specificity by collecting longitude–latitude coordinates for their sampling clusters, making these available in the public-domain datasets.¹³

If the DHS program offers greater specificity about current residence , its surveys are generally less revealing than censuses about migration. Most censuses collect information on place of residence 5 years before the date of the census, although a few focus instead on 1 or (in rare cases) 10 years before the census. Migrants are then defined as those whose current residence differs from residence 5 years previous. In focusing on these two points in time, this (conventional) definition overlooks important movements that take place between them: seasonal migrants would not be identified, nor would most short-term “target migrants” who have returned by the time of the census to where they had previously lived (Bilsborrow 1984; Hertrich and Lesclingand 2012). A number of censuses include a question on the length of current residence (coded in years) as an alternative to the 5-years-previous question; and some censuses gather both. When both measures are available, we use the more conventional 5-years-previous measure; if it is not available, we define migration as taking place when the length of current residence is less than 5 years. If more than one move took place over the 5-year period, neither of these measures will record it: they indicate whether any move took place, but not the number of moves.

In the Demographic and Health Survey s , only the length of current residence is generally available (it is also coded in years), and relatively few surveys have provided more detail than that. For a time, in the late 1990s to early 2000s, the DHS program experimented with using monthly calendars as a device to record demographic behavior over the 6 years leading up to the survey , and about 25 countries included migration in these calendars. An examination of the calendars shows that length of current residence as calculated via the calendar is broadly consistent with the standard question on years of residence. Also, relatively few adolescents or adult women are found to have moved more than once over the 6-year span of the calendar (under 10 % in these surveys), suggesting that not much information on the number of recent moves is sacrificed by using length of residence to indicate whether any move took place.

An important difference between the IPUMS -processed censuses and the surveys—one to which we give considerable attention in this chapter—concerns the distance or boundary-crossing criterion that distinguishes a migratory move from a mere change of residence . For current urban residents, the DHS practice has been to define migration as a move that originated outside the city or town in which the respondent currently lives. Since the boundaries of these urban places are difficult to discern, and since neither interviewers nor respondents can be expected to know them precisely, it seems that the DHS interviewers must in some way bring judgement to bear in separating out migration from all accounts of moves given by respondents. It is not at all obvious what criteria are applied in these surveys to define rural migration—is migration entailed in a change of village ?—and possibly in rural areas locational boundaries would be even less evident than in urban areas. Census data-collection efforts typically define migration with greater consistency and transparency, making specific reference to the boundaries of official administrative regions.¹⁴ Some censuses define migration to be a move that crosses a major administrative unit boundary; others allow crossings of minor unit boundaries to count. As Standing (1984, p. 32) wrote nearly thirty years ago in a passage that is still on point today,

Somewhat remarkably, most demographers and other social scientists have let statisticians and survey administrators determine the areas between which moves are classified as “migration”.. . . It has been said that the areas between which moves count as migration are first defined by bureaucrats and later rationalised by social science researchers.

Standing and others have noted that because these administrative units vary a good deal in their geographic size, both within and across countries, it is difficult to work out an acceptable method for standardizing estimates so that they are not size-dependent.

Censuses and the DHS program have taken fundamentally different approaches to characterizing the area from which a move took place. Ideally, as we’ve mentioned, a migrant’s origin area would be described not only in geographic but also in urban –rural terms. In reality, neither censuses nor the DHS provides such minimally complete information, at least in general. Censuses do not commonly record whether the community from which the migrant came was urban or rural. The DHS surveys, by contrast, typically do describe the rural–urban status of the origin community, but offer no clues as to its geographic location. Moreover, the basis on which the urban–rural status of the origin is decided is not obvious. It would again appear that the classification is left to the DHS interviewer to decide.

The geographic distance covered by the migrant is not available in either DHS surveys or censuses, and in neither case is the origin described in sufficient detail for distance to be computed very precisely after the fact. Given data on the boundaries of the administrative units recorded in the census (stored in a shapefile or the equivalent), the minimum and maximum possible distances travelled in a move could be calculated, and if additional data were available on the spatial distribution of population within these administrative units, the distance traversed by a migrant could be estimated in a statistical model. This would be a substantial although feasible empirical exercise, but it lies outside the scope of this chapter.

The respondents canvassed by censuses and DHS surveys also differ in ways that could significantly affect migration estimates. Census interviewers collect information from each household member, or at least from those old enough to be eligible for consideration. (Age five is the usual cut-off below which migration questions do not apply.) This information is conveyed to the census-taker by one household member who speaks for the household as a whole. In the DHS survey program, by contrast, migration-related data are collected only from the subset of adults who are selected (at random) for in-depth individual interviews, rather than from all migration-eligible household members, and the interviewees speak for themselves. In most DHS surveys, the respondents are women aged 15–49, although it is becoming more common for men to be interviewed as well, allowing a more representative picture of migration to emerge. An important consideration is that in a number of Asian and North African countries, DHS individual interviews are restricted to ever-married women, a design decision that introduces the potential for selection bias in migration estimates. (We will provide examples below.) The by-proxy census reports of migration may well contain more measurement error overall than if individual members gave their own accounts to the census-taker, but the census data should not be afflicted by marriage-related selectivity bias.

Marriage-Related Selectivity Bias

A number of DHS surveys interview only women who have been married, and because it is through the individual interviews that migration status is ascertained, this practice raises the possibility of selection bias that could distort estimates of migration. Migration questions in censuses are framed without reference to marital status, and unlike DHS surveys, these questions cover all household members who are old enough to be asked. Figure 26.18, based on census samples for Egypt and Vietnam , illustrates how marriage selection effect s can introduce bias. These calculations compare estimates of urban in-migration for all women who were canvassed in the censuses with estimates from the census records of ever-married women. At older ages, by which nearly all women in these countries have married, the migration percentages coincide. At younger ages, however, they differ substantially—but the direction of bias is upward in the case of Egypt and downward for Vietnam. Although census data do not establish the time-sequence of events, it would appear that in Egypt, women tend to migrate just before, upon, or shortly after marriage, so that the migration percentages for ever-married women are well above those for all women. In Vietnam as in much of Southeast Asia , migration is typically undertaken by young unmarried women who move for a variety of reasons—among them, to enjoy a period of relative autonomy away from parents, and to earn incomes that help support younger siblings—and thus an artificial restriction of the sample to ever-married women would depress urban migration percentages.

Fig. 26.18

Urban in-migration percentages for all women and for ever-married women, Egypt (2006) and Vietnam (2009). All-women percentages are depicted in blue lines (beginning age 10) and percentages for ever-married women in green lines (beginning age 15). Migration defined as a cross-governorate move for Egypt and a cross-province move for Vietnam (Source: IPUMS ). (a) Egypt, 2006 census. (b) Vietnam, 2009 census

These census-based examples suggest that migration estimates from surveys restricted to ever-married women will tend not to give an accurate representation of migration overall, especially in the age ranges in which substantial percentages of women are yet to marry. Since the direction of bias as well as its magnitude is situation-dependent, we have opted to exclude from our analyses all DHS surveys limited to ever-married women. This is an unfortunate—Egypt , India , and a number of other large countries have DHS surveys restricted to ever-married women, and some of these countries have been surveyed multiple times—but we see no way to correct statistically for the selection bias.

Moving for . . . What?

Censuses and many surveys (although not those in the DHS program) often ask migrants to describe why they moved. The usual practice is to permit only one “most important reason” to be recorded, which is unduly restrictive given that migration is often motivated by many considerations. If they are limited to one response only, girls and young women may supply the reason that others would be likely to find most socially acceptable. A girl who migrates to join her spouse, but who also holds ambitions to pursue university schooling and gain professional employment, may simply describe her move as being “for marriage” so that her high ambitions remain appropriately cloaked.

As guides to motivation, questions such as these also suffer from a fundamental and irremediable logical flaw: They are asked only of movers. If the desire to be with a spouse is an important consideration in a girl’s choice of location, then a girl who stays home to be with her spouse is never given the opportunity to say that she “stayed for marriage.” It is obvious—and yet the literature seldom remarks upon this obvious point—that questions put only to movers cannot detect which motivations truly guide decisions about location.

If these questions have little value for understanding the considerations that lead some girls to move and others to stay, they have other uses. If a girl says that she moved to the city “for employment”, but has no job at the time of interview , this might be read as a mismatch between her pre-migration expectations and the realities she has faced after the move took place. There is value and the potential for securing insight in this kind of comparison.