Desktop version

Home arrow Economics arrow Children and Forced Migration: Durable Solutions During Transient Years

Methodology

Data

This study relies on data collected by Samuel Hall Consulting in conjunction with the Maastricht Graduate School of Governance (MGSoG) for the independent evaluation of the UN High Commissioner for Refugees (UNHCR) shelter assistance program from 2009 to 2011. The household survey took place in late 2012, across 15 provinces of Afghanistan. As the original purpose of the survey was to evaluate shelter assistance programs for return migrants and internally displaced persons (IDPs), the sampling reflects the general distribution of shelter assistance by international organizations while also taking into account local security restrictions. The sample therefore cannot be considered strictly representative across the country. Still, measures were taken to increase representativeness, including the selection of at least one province within each of the country’s eight regions. Within provinces, one or more districts were selected for cluster sampling, with villages then randomly selected in light of a general record of shelter assistance beneficiaries’ locations. Within these villages, both beneficiary and non-beneficiary households were surveyed at random (MGSoG and Samuel Hall Consulting 2013).

Because the objective of the survey was to capture information for return migrants and IDPs, the questionnaire allows for direct identification of internally displaced households in comparison to those who have never moved.[1] Moreover, because respondents were asked two distinct questions about why they decided to move from their community of origin and why they decided to move to their current place of residence, it was possible to identify households that were involuntarily displaced because of conflict, insecurity, persecution, or natural disaster—arguably exogenous shocks resulting in minimal choice in the decision to move.[2] In addition, identification of those who chose their destination location without consideration for the outcomes in question (e.g., services including education, food, and health) was possible. This allows for a greater distinction between involuntary and voluntary movement as well as exogenous location choice, helping to minimize any potential bias in the estimates because of endogeneity. Ultimately, the sample is made up of 1020 household-level observations of which 40 % are considered involuntarily and exogenously internally displaced.

Even though any bias because of selection is arguably minimized by only taking into account involuntary migration, it is still possible that some households may be systematically more exposed to violence, and therefore displacement, given inherent characteristics (e.g., wealth levels prior to displacement), leading to inconsistent estimates (Kondylis 2010). Even though it is not possible to fully account for such selection bias using advanced econometric techniques because of the limitations of the dataset, the estimates should be considered lower bounds for two reasons. First, one could readily suppose local violence in general is likely to be targeted towards those wealthier members of the community as they arguably have more local authority. In Afghanistan, however, there is little evidence to support such a conjecture because much of the violence has been noticeably indiscriminate against the civilian population (see Human Rights Watch 2007).

The data here corroborates such a lack of targeted violence, as only 16 % of those households considered internally displaced are so because of personal, family, or ethnic persecution. The majority of internally displaced households, 83 %, are so because of general insecurity and conflict. Nonetheless, although it is not possible to categorically discount targeting, supposing true would mean displaced households should have been systematically better off before migration relative to nondisplaced households in terms of general well-being. Second, the migration journey itself, even internally, is not without cost. Similarly, if those who migrated were able to assume such costs while those who stayed behind were not, one can assume that displaced households were again better off before migration took place. As such, the estimates presented here, if in fact imprecise, should in principle be underestimating any true effect that shows a negative difference between displaced and nondisplaced households.

As the goal here is to measure the extent to which internal displacement has an effect on livelihood prospects later on in life, the focus is on child-related outcomes connected to human capital formation— namely, education and nutrition. Both outcomes are widely recognized as fundamental determinants of future well-being and have been the focus of myriad of studies in the development literature (see Alderman et al. 2006; Rosenzweig 2010). Regarding education, the outcome of interest is the categorical variable of school attendance of school-age children differentiated by whether no children attend school, all children attend school, or only boys attend school.[3] As for nutrition, the outcomes of interest include the categorical variable of food insecurity differentiated by whether the household never has problems satisfying food needs, rarely has problems (one or two times per month), or often has problems (greater than three times per month); this is in addition to the continuous variable of dietary diversity proxied by the number of times the household has eaten meat during the last week.

Table 1 presents summary statistics that report the sample mean and standard deviation, not only for the entire sample but also differentiated by whether the household is considered internally displaced or not, including a simple means difference test in the far right column. In terms of the outcomes of interest, of 1020 respondents who knew whether their school-age children were in school or not, 35 % reported no attendance whatsoever. On the other end of the spectrum, 41 % said that all children in the household attended school, and 20 % reported only boys were in school. There is a high statistically significant difference in responses when comparing nondisplaced and displaced households, apart from when the response was only boys attended school.

As for food insecurity, 20 % of households responded as having no problems satisfying food needs, while some 44 % rarely had problems (one to two times per month) and another 37 % often had problems (greater than three times per month). Nondisplaced households were much more likely to never have problems and displaced households more likely to often have problems. Finally, the average times households had eaten meat during the prior week, an indication of dietary diversity, is around once. Though, this masks the fact that 46 % of the sample reported having eaten no meat over the last week. Between groups, there is a statistically significant mean difference with nondisplaced households eating meat more often than their displaced counterparts.

Regarding other covariates, the analysis takes into account an array of household-level factors including the age of the adult respondent (i.e., household head or spouse); whether the person had received any formal education; whether the household head was married; the number of chilTable 1 Summary Statistics

Full sample

Nondisplaced

Internally

displaced

Mean

SD

Mean

SD

Mean

SD

t-test

Outcomes

School attendance

None

0.3539

0.4784

0.2946

0.4562

0.4425

0.4973

***

All school- age children

0.4078

0.4917

0.4550

0.4984

0.3374

0.4734

***

Only

school-age

boys

Food

insecurity

0.2029

0.4024

0.2111

0.4084

0.1907

0.3933

Never

0.1951

0.3965

0.2357

0.4248

0.1345

0.3416

***

Rarely (1-2 times/month)

0.4382

0.4964

0.4239

0.4946

0.4597

0.499

Often (> 3 times/month)

0.3667

0.4821

0.3404

0.4742

0.4059

0.4917

*

Dietary diversity (times eaten meat/week)

0.9725

1.3221

1.1178

1.4573

0.7555

1.0543

***

Household-level covariates

Age of respondent

37.74

14.09

37.64

14.12

37.89

14.06

No formal education

0.8265

0.3789

0.8200

0.3845

0.8362

0.3706

Married

0.8510

0.3563

0.8167

0.3872

0.9022

0.2974

***

Number of children

5.20

3.07

5.25

3.04

5.11

3.11

Disabled

0.2225

0.4162

0.2013

0.4013

0.2543

0.4360

Log of household income

3.5692

0.6106

3.5972

0.6206

3.5275

0.5935

Received

assistance

Ethnicity

0.4500

0.4977

0.3797

0.4857

0.5550

0.4976

***

Pashtun

0.6255

0.4842

0.6072

0.4888

0.6528

0.4767

Tajik

0.0990

0.2988

0.1358

0.3429

0.0440

0.2054

***

Hazara

0.0725

0.2595

0.0835

0.2768

0.0562

0.2307

Other

0.2029

0.4024

0.1735

0.3790

0.2469

0.4318

**

Location-based FE

Rural

location type

0.7784

0.4155

0.7005

0.4584

0.8949

0.3071

***

{continued)

Table 1 (continued)

Full sample

Nondisplaced

Internally

displaced

Mean

SD

Mean

SD

Mean

SD

t-test

Province

Kabul

0.0520

0.2221

0.0524

0.2230

0.0513

0.2210

Parwan

0.0216

0.1453

0.0164

0.1270

0.0293

0.1690

Bamyan

0.0069

0.0826

0.0049

0.0700

0.0098

0.0985

Laghman

0.0598

0.2372

0.0917

0.2888

0.0122

0.1100

***

Nangarhar

0.3147

0.4646

0.4124

0.4927

0.1687

0.3749

***

Balkh

0.0333

0.1796

0.0491

0.2163

0.0098

0.0985

***

Faryab

0.0725

0.2595

0.0360

0.1865

0.1271

0.3335

***

Jawzjan

0.0402

0.1965

0.0524

0.2230

0.0220

0.1469

**

Sari Pul

0.0118

0.1079

0.0180

0.1331

0.0024

0.0494

**

Kunduz

0.0245

0.1547

0.0245

0.1549

0.0244

0.1546

Takhar

0.0559

0.2298

0.0164

0.1270

0.1149

0.3193

***

Helm and

0.0931

0.2908

0.0409

0.1983

0.1711

0.3771

***

Kandahar

0.0706

0.2563

0.0491

0.2163

0.1027

0.3039

**

Paktya

0.0314

0.1744

0.0475

0.2128

0.0073

0.0854

***

Herat

0.1118

0.3152

0.0884

0.2841

0.1467

0.3542

**

Note: *0.10, **0.05, ***0.01

dren in the household; whether the household had a physically or mentally disabled member; the log of monthly income per capita; and whether it had received formal assistance from a nongovernmental organization (NGO), government, or international organization. Furthermore, the model controls for ethnicity of the household and considers location- based fixed effects, including rural location type and province.

Table 1 also illustrates the fact that nondisplaced and internally displaced households are similar in a number of ways. These similarities include the fact that most adult respondents have no formal education, they have around the same number of children, and similar wealth levels based on the log of monthly income. On the other hand, these families differ in that displaced households are slightly more likely to have a married household head, as well as more likely to have received formal assistance. As for ethnicity, the majority of the sample is Pashtun, similar across subgroups. Moreover, the Tajik are less likely to be internally displaced while “other” ethnic groups are more likely to experience displacement. Regarding location, most households are located in rural areas, with this figure noticeably greater for the internally displaced. Last of all, in terms of provincial location, internally displaced households appear to be concentrated, in much greater numbers than nondisplaced households, in Faryab, Takhar, Helmand, Kandahar, and Herat, while the opposite is true in Laghman, Nangarhar, Balkh, Jawzjan, Sari Pul, and Paktya.

  • [1] Return migrants are excluded from the sample in order to minimize any potential selection bias.
  • [2] Any displacement in general can be considered involuntary. However, restricting the sample tothose who moved because of these four reasons helps strengthen the argument that displacement inthis case is influenced by exogenous factors, leaving less potential for selection bias. The other reasons for displacement include: no land or housing, no access to arable or pastoral land, no access tofood and water, no access to health services, no access to education, and no access to employmentopportunities.
  • [3] The category “only girls attend school” is excluded because of the low number of observations.Moreover, households where no school-age children are present are not included in the analysis.
 
Source
< Prev   CONTENTS   Source   Next >

Related topics