Determinants of International Migration: A Methodological ...

26
Determinants of International Migration: A Methodological Study Based on Egypt Richard E. Bilsborrow, Samir Farid and Qi Zhang Presented at the International Population Conference of the International Union for the Scientific Study of Population, December 5-10, 2021 1. Introduction There is rapidly increasing interest in international migration in many parts of the world, with the population constituted by international migrants (defined as living outside their country of birth) reaching 281 million in 2020, 50% more than in 2000 (and 3.5% of the world total: UN Population Division, 2020). Migration flows, especially from developing to developed countries, have been growing since the 1990s, due to increasing global inequalities and vast improvements in communication systems and reductions in transport costs. This has led to governments in receiving countries to restrict immigration (Esipova et al 2011; Tjaden & Laczko 2019), although the international community recognizes that international migration has positive socio-economic effects in both countries of origin and destination (UN 2006, 2009, Skeldon 2013), culminating in migration targets of the Sustainable Development Goals (2030). Nevertheless, migration continues to be the “poor stepchild of demography” (Kirk, 1960, p. 307): since 1960, the field of fertility has advanced greatly, linked to linked to global programs of large investments in data collection, analysis and publications, in contrast to migration. But internal and international migration flows have come to be the major determinants of population change and growth within and across countries, as fertility, mortality and natural population growth decline (except for Sub-Saharan Africa). Meanwhile, knowledge of the causes of migration and therefore the ability to formulate better policies is compromised by the lack of adequate research, in turn linked to weak data (CGD 2009; UN High Level Dialogues; UN 2018, Global Compact for Safe, Orderly and Regular Migration, Objective 1: “Collect and utilize accurate and disaggregated data as a basic for evidence-based policies.”). Thus, data shortcomings limit analyses, in turn impairing advances in theory and policy (see Massey et al 1993 comments on this linkage between migration theory and policy). The MED-HIMS (Mediterranean Household International Migration Survey) program is one of the main recent regional efforts still ongoing (Bilsborrow 2018b) to respond to the need for better data to improve knowledge about migration, but does not include in-depth analysis. Egypt was the first country to complete a MED-HIMS 1

Transcript of Determinants of International Migration: A Methodological ...

Page 1: Determinants of International Migration: A Methodological ...

Determinants of International Migration: A Methodological Study Based on Egypt

Richard E. Bilsborrow, Samir Farid and Qi Zhang

Presented at the International Population Conference

of the International Union for the Scientific Study of Population, December 5-10, 2021

1. Introduction

There is rapidly increasing interest in international migration in many parts of the world, with the population

constituted by international migrants (defined as living outside their country of birth) reaching 281 million in

2020, 50% more than in 2000 (and 3.5% of the world total: UN Population Division, 2020). Migration flows,

especially from developing to developed countries, have been growing since the 1990s, due to increasing global

inequalities and vast improvements in communication systems and reductions in transport costs. This has led to

governments in receiving countries to restrict immigration (Esipova et al 2011; Tjaden & Laczko 2019), although

the international community recognizes that international migration has positive socio-economic effects in both

countries of origin and destination (UN 2006, 2009, Skeldon 2013), culminating in migration targets of the

Sustainable Development Goals (2030).

Nevertheless, migration continues to be the “poor stepchild of demography” (Kirk, 1960, p. 307): since 1960, the

field of fertility has advanced greatly, linked to linked to global programs of large investments in data collection,

analysis and publications, in contrast to migration. But internal and international migration flows have come to

be the major determinants of population change and growth within and across countries, as fertility, mortality

and natural population growth decline (except for Sub-Saharan Africa). Meanwhile, knowledge of the causes of

migration and therefore the ability to formulate better policies is compromised by the lack of adequate research,

in turn linked to weak data (CGD 2009; UN High Level Dialogues; UN 2018, Global Compact for Safe, Orderly and

Regular Migration, Objective 1: “Collect and utilize accurate and disaggregated data as a basic for

evidence-based policies.”). Thus, data shortcomings limit analyses, in turn impairing advances in theory and

policy (see Massey et al 1993 comments on this linkage between migration theory and policy).

The MED-HIMS (Mediterranean Household International Migration Survey) program is one of the main recent

regional efforts still ongoing (Bilsborrow 2018b) to respond to the need for better data to improve knowledge

about migration, but does not include in-depth analysis. Egypt was the first country to complete a MED-HIMS

1

Page 2: Determinants of International Migration: A Methodological ...

survey, in 2013, has the largest sample as well as by far the largest population of any country in the Middle

East-North Africa (MENA) region, and a long history of emigration, mostly labor migration to the Gulf

Cooperation Council (GCC) and other Arab states (notably Jordan and Libya). The MENA region continues to

attract particular interest following the Arab Spring, civil war in Syria, and large outflows of migrants from

countries in the region and from migrants in transit. In Egypt, most out-migrants are male and “temporary”,

migrating for work and returning, but there has been little study of this type of labor migration, much less its

determinants, including roles of labor recruiting agents, visa policies, migration networks, and return migration.

While all countries have different and often unique circumstances and policy concerns, there has been little

in-depth quantitative research on the factors responsible for international migration from developing countries

due to the lack of adequate data sets at the household level in developing countries (Bilsborrow et al., 1997;

Groenewold & Bilsborrow, 2008; Bilsborrow, 2018, for IOM GMDAC). Many major UN and other global meetings

on international migration have occurred in recent years, documenting the data shortcomings and lack of

research findings on the determinants and consequences of this migration (viz., Global Compact for Safe, Orderly

and Regular Migration, 2018: Objective 1, “Collect and utilize accurate and disaggregated data as a basis for

evidence-based policies.” This paper hopes to help address this important need.

2. Conceptual approach

Theories of migration are commonly traced back to Ravenstein (1989) and have been advanced by social

scientists of all disciplines. At the relevant micro or household level, this includes human capital theory, justifying

individual attributes, such as age, gender, education, work experience, and occupation, as key determinants

(Sjaastad 1962); the sociological push-pull theory of Lee (1966), both evolving to view migration as a household

decision (Mincer 1978). This led to the New Economics of Labor Migration (NELM) theory, which views

households as allocating household labor across time and space based on household needs, human capital, and

risk reduction (Stark 1984; Stark & Bloom 1985; Taylor & Lopez-Feldman 2010). Much sociological work has

found marital status, family relationships, and migration networks fundamental determinants (Massey 1990;

Curran & Rivero-Fuentes 2003; White 2016). Recent disciplinary reviews are found in White (2016) on geography,

including on the importance of distance and “place utility” (Wolpert 1965), economics (Greenwood, on access to

information & labor market opportunities and conditions, Todaro, 1969), and sociology (White & Johnson, on

family ties & life cycle). In addition to all these potential factors, International migration is also affected by

government policies on migration including promoting or preventing, visa availability and costs/requirements,

and border controls, as well as language/culture, historical ties from colonialism and trade, transport costs, etc.

(Massey et al. 1993; Bilsborrow et al 1997).

2

Page 3: Determinants of International Migration: A Methodological ...

Regarding return migration, there has been neither little formal theorizing nor large empirical studies (DeHaas,

et al 2015). While affected by the same individual and household factors as emigration, how might it be

different? One attempt to address this theoretically is Cassarino (2004), who cites as important ties with the

origin household and society and the type of returnee, in turn drawing on Cerase (1974). Cerase studied Italian

migrants returning from the US and classified then in various types: returns of failure, those who had trouble

socially integrating or obtaining work; returns of conservatism, persons achieving target savings; and returns of

innovation, actors aiming to invest in the origin. Cassarino also notes key roles of returnee’s preparedness and

willingness to return and resources available (to cover travel costs) and new skills/education acquired abroad.

Based on the migration theories above, we formulate a broad conceptual model of the determinants of

emigration (Fig. 1). This model takes the macro-contexts of Egypt and of countries of destination as implicit,

including government policies, economy, visa requirements, border controls, etc. (indicated by dotted lines).

Each other construct includes specific multiple factors identified by theory and previous research to be

potentially important in influencing migration decisions of individuals, nested in households, with both individual

and household characteristics affecting migration (straight-line arrows). Each construct comprises variables to be

tested in the statistical model.

3. Brief background on international migration

from Egypt

We undertook a computer review of the recent literature, finding it concentrated on studies of aggregate

data/flows of migrants, small data sets when households were studied, and immigration/adaptation of migrants

rather than determinants, and a geographic focus on North America and Europe (c.f. also, White, Introduction

2016). We cite here only those most relevant to the proposed models for Egypt, e.g., several finding education

not positively linked to emigration, as found in Egypt and contrary to the usual theoretical expectations (Mayda

2010; Rendall & Parker 2014), and some based on small household surveys: Massey & Espinosa (1997),

Chengrong & Nangia (1998) & Mora & Taylor (2006) on Mexico, Bohra & Massey (2009) on Nepal, Liang &

Chungu (2013) on China, plus others from the Latin America Migration Project and the Migration in Africa Project

3

Page 4: Determinants of International Migration: A Methodological ...

(MAFE). A particularly interesting study is Gonzalez-Ferrer et al (2014) based on the MAFE data, unique in its

combined study of the determinants of both emigration and return migration, outward from three African

countries (Senegal, Ghana, Dem. Rep. of the Congo) to six European countries and back. As the sample sizes of

those actually returning were literally only 12, 15 and 16 adults in the three countries, investigators actually

studied only intentions to return, with 85 of 662 emigrants intending to return to Senegal, 50 of 441 to Congo

and 83 of 472 to Ghana—still fairly small numbers.

This illustrates the severe limitations of sample size in prior studies (also cited in Bilsborrow 2018a, b). On the

determinants of emigration, the usual age, education and current household income effects were found in a logit

model, along with strong effects of (country of) residence of spouse/children. Estimates of factors affecting

return migration were weaker, but included time abroad, work status in Europe, and having a family in origin

(positive on return), and whether the migrant had been sending remittances or made visits while living in Europe

(actually negative on return, perhaps serving as a substitute for returning rather than paving the way for return

(as found in Constant & Massey 2002). In another MAFE study on actual return migration to Senegal, Flahaux

(2017) found those returning to be more likely a repeat migrant or unemployed in the destination, and unlikely

to return if their family was in the destination or if they had visited the origin in person since departing.

In the MENA region, there have been few studies on the determinants of international migration (Fargues &

Venturini (2015). For example, David & Jarreau (2016) used three waves of the Egyptian Labor Market Panel

Survey in 1998-2012 to find emigrants positively selected on household wealth, with unemployment or working

in the informal sector also being incentives to emigrate. And Gang and Bauer (1999) analyzed return migration to

Egypt of 474 males from 6 villages, finding access to information networks with Egypt reducing time abroad

(associated with return).

4. The Data for Egypt

The data used in the present research for Egypt come from the Egypt Household International Migration Survey

(Egypt-HIMS), which was part of the Mediterranean Household International Migration Survey (MEDHIMS)

project. MEDHIMS was initiated via international workshops promoted by the MEDSTAT projects of the European

Commission and Eurostat, starting in 2010. The project involved top technical officials of the statistical offices of

eight countries in the MENA region (Middle East-North Africa) of the Mediterranean basin participating in

MEDHIMS, as well as funding/stakeholder agencies, including the European Commission, Eurostat, the World

Bank, and UNHCR initially, later joined by the International Labour Office, UN Population Fund, International

Organization for Migration, and the League of Arab States. These meetings resulted in the development of

4

Page 5: Determinants of International Migration: A Methodological ...

detailed, user-friendly prototype questionnaires, agreed upon by representatives of the government statistical

offices in the region, plus accompanying manuals on sampling, training interviewers and supervisors, survey

fieldwork organization, and analysis/tabulation plans. There was also an important component of international

technical assistance from consultants, led by Farid and coordinated by Giambattista Cantisani. Participating

countries agreed to use same the basic questionnaires (but could add a few country-specific questions) to ensure

the collection of comparable data in all countries, to facilitate comparisons and generalizations across countries.

The documents and international aid led to large sample sizes and high-quality data in the first two countries to

implement the full survey questionnaires-- Egypt in 2013 and Jordan in 2014. Subsequently, Morocco in 2018

and Tunisia in 2020 have carried out nationally representative surveys on international migration based on a

shorter “light” version of the MEDHIMS questionnaires, also developed and agreed upon by all the participating

MEDHIMS countries and stakeholders.

The Egypt questionnaire comprised six modules: a household screening questionnaire, focusing on household

membership and composition, including key questions to carefully identify international migrants of interest as

well as non-migrants; four detailed and comparable individual-level questionnaires--for current out-migrants

(emigrants), return migrants (from abroad), forced migrants , and non-migrants; and finally, a module on1

housing/household living conditions. The paper here is based on data from the five modules, excluding forced

migrants. Specialized sampling procedures were used. First, the sample frame itself was one of the four

independent large national samples prepared by CAPMAS based on data from the most recent census available

at the time (in 2006), which was updated in 2011. Numbers of enumeration areas (EA) in each province of Egypt

were selected in proportion to the population, then in each sample EA, averaging about 90 households in urban

areas and 80 in rural ones, households were screened in the field to identify all those with one or more

emigrants (called current migrant, in emigrant households), who were automatically selected into the sample,

plus three non-migrants randomly from non-migrant households in the EA cluster. The result was oversampling

households with international migrants (see Bilsborrow et al., 1997; also Bilsborrow et al., 1984). This was done

to ensure that the field work would be economically efficient in collecting data on significant numbers of migrant

households, in contrast to the results from the usual random sampling procedures of household surveys, which

results in small numbers of migrants compared to non-migrants (details on the sampling frames and sampling

procedures for Egypt as well as Jordan and Morocco are found in Bilsborrow (2013).

1 In the case of Egypt, the sample of forced migrants was based on a totally separate, non-probability sample of householdsdrawn from lists of refugees and asylum seekers in the Cairo metropolitan area, provided by UNHCR, and is therefore notincluded in the research reported here.

5

Page 6: Determinants of International Migration: A Methodological ...

The sample size of the Egyptian survey is perhaps the largest ever for a specialized household survey on

international migration in a developing country, with over 90,000 households initially screened using module 1,

providing data for 83,358 households. Of these, 5,259 reported having one or more emigrants living abroad

(defined as a former household member who was aged 15-59 at the time of emigration since 2000, to exclude

dependents as they are not migration decision-makers) and who had not returned by the time of the survey in

2013 (referred to as current migrants or emigrants). Meanwhile, 4,695 households (from among the emigrant

households plus the non-migrant households) were found to have one or more return migrants since 2000, the

reference year). Households with non-migrants were also sampled (at a much lower rate) and interviewed to

collect data to compare with that of migrant households, numbering 3,135. All emigrants and return migrants

were administered a detailed individual questionnaire as was one non-migrant adult in every household, the

total number of households being 11,703, with total individuals interviewed being 22,635 (emigrants, return

migrants, and non-migrants). Because of the oversampling of households with emigrants compared to migrants

in the last-stage sampling units (EAs), it was crucial for supervisors to keep track of the households listed,

sampled, and successfully interviewed for migrant households and non-migrant households, to be able to

properly weight all data for all analyses.

Fieldwork was carried out in 2013 by the government statistics office, CAPMAS (Central Agency for Public

Mobilization and Statistics). Samir Farid, Chief Technical Advisor, helped coordinate various aspects of the project

in Egypt, from questionnaire design (viz., additional questions on health suggested by and funded by the World

Health Organization were adapted and added to the module 1 screening questionnaire for all household

members), to manuals for training interviewers and supervisors, the conduct of fieldwork, data cleaning and data

analysis. He also led the production of a wealth of tabulations, leading to the comprehensive government

descriptive publication (Farid et al. 2016) with over a thousand tabulations, along with additional tabulations on

the CAPMAS website.

The data were subsequently made available to Dr. Farid for us to undertake multivariate statistical analyses

starting in late 2017, which resulted in findings being publicly reported here for the first time. Given the careful

attention to the screening and listing of households in sample enumeration areas and funds available, very large

samples were available, providing an extraordinarily rich dataset for analysis, which promise to yield a

much-improved understanding of the factors behind emigration and return migration in Egypt and the other

participating MEDHMS countries in the (MENA) region, the analysis of which should provide useful information

for policy as well as methodological lessons.

5. Some descriptive results on emigrants vs. non-migrants

6

Page 7: Determinants of International Migration: A Methodological ...

Before we examine the multivariate results, we examine some of the descriptive data and findings, which is

always important to do first. Table 1 shows the characteristics of individuals in all survey households—of

non-migrants, emigrants (out-migrants), and return migrants for whom data were collected for 2000-2013. This

allows us to begin to understand how migrants differ from non-migrants, and return migrants from out-migrants,

providing initial insights also about the factors that may be responsible for emigration and subsequently return

migration.

Table 1. Characteristics of emigrants and non-migrants, Egypt-HIMS 2013

CharacteristicEmigrant(age 15+)

Return migrant(age 15+)

Non-migrant(age 15-59)

1. Survey households• Urban/Rural residenceUrban 19.7 25.8 47.5Rural 80.3 74.2 52.5Total 100.0 100.0 100.0• Household headshipMale 51.0 92.4 85.2Female 49.0 7.6 14.8Total 100.0 100.0 100.02. Characteristics of migrants and non-migrants• Age composition (At first migration) (Current age) (Current age)

15-19 10.7 1.2 22.120-24 33.3 4.8 15.925-29 28.7 11.9 13.330-34 13.2 18.2 10.635-39 7.4 18.9 10.440-44 4.0 14.4 8.345-49 1.9 12.1 8.050-59 0.7 13.2 11.460+ 0.1 5.3 --Total 100.0 100.0 100.0

• SexMale 97.9 89.1 45.7Female 2.1 10.9 54.3Total 100.0 100.0 100.0

• Current educational levelBelow primary complete 21.9 28.7 22.9Primary/Preparatory 14.6 13.8 27.5Secondary 48.3 42.4 37.1Higher 15.2 15.1 12.5Total 100.0 100.0 100.0

• Current occupationManagers 2.1 9.1 4.7Professionals 7.2 10.2 17.5Technicians & associated professionals 3.7 4.9 7.0Clerical support workers 1.2 1.4 5.3Service and sales workers 11.6 5.8 11.8

7

Page 8: Determinants of International Migration: A Methodological ...

Skilled agricultural, forestry, and fishery workers 9.7 26.6 16.1Craft and related trades workers 49.5 26.9 22.6Plant and machine operators and assemblers 8.1 10.8 10.4Elementary occupations 6.9 4.3 4.6Total 100.0 100.0 100.0

Number 5847 5085 11703Note all data in this table refer to all sample households and individuals from 2000 to 2013, and not only the cohort

2010-2013 used in the multivariate analyses below.

Survey household characteristics

Most households with migrants reside in rural areas-- 80 % of emigrant households and 74% of the return

migrant households, in contrast to a bit over half of households in Egypt, showing the selectivity of migrants

from rural areas. Emigrants also tend to come from larger households than non-migrants, in both urban and rural

areas. Among non-migrant households, the traditional pattern of male-headship exists in both urban and rural

areas, with the overall percentage of male-headed households being 85 percent, rising to 92% for households

with a return migrant. A very different pattern is observed among households with emigrants, where 49% are

female-headed. This shows that many of the migrants are actually male household heads, whose absence

creates de facto female headship. Data on return migrant households shows this is usually only a temporary

situation, until the emigrant returns.

In terms of household possessions & assets, most households in Egypt own most modern household appliances

with little variation by migration status or urban rural residence, although rates of ownership are generally

higher among the return migrant households than either emigrant or non-migrant households.

Characteristics of individual migrants and non-migrants

Age-sex distribution

The distribution of emigrants by age at migration is distorted demographically. It shows an inverted U-shaped

pattern with respect to age at first migration, with a low percent among persons aged 15-19 (11) and most first

migrants young at age 20-29 (62%), resulting in a median age at first migration, among those migrants who

moved abroad since the beginning of the year 2000, of 25.1 years. The age distribution of return migrants also

has an inverted U-shape in terms of current age at time of interview in 2013. Only a tiny percent return at age

15-19 years (1.2), with others returning with a broad peak extending over the age range 25-49 years (over 75%).

In contrast, more than half of non-migrants (51%) are in the age range of 15-29 years, with another 22% aged

15-19, and a similar percent in the broad age range of 45 to 59 years.

8

Page 9: Determinants of International Migration: A Methodological ...

The results also clearly show that migration of individuals from Egypt is predominantly male― only 2% of

emigrants aged 15+ being female, along with 11% of return migrants and in contrast to 54% of the non-migrant

population.

Education

A majority of migrants and non-migrants are relatively well educated, but migrants tend to be better educated,

as the percentage who completed secondary or higher education being 64 among emigrants and 58 among

return migrants in contrast to 50 among non-migrants. Among emigrants who first moved abroad during the

years 2010-2013, 51% had completed secondary education while another 19% had university degrees.

Motives for migration

Around 87 percent of the individual emigrants studied moved abroad for economic reasons, with only 10 percent

for social reasons and 3 percent for other reasons. The three most important economic motives for first

migration from Egypt were ‘to improve standard of living’ (34%), followed by ‘income in Egypt was insufficient’

(25%), and ‘lack of employment opportunities’ (11%). Among female return migrants in survey households

(comprised of those few who moved as individuals and returned and overwhelming those who left with husband

and returned with him), social reasons were the main motive for the first migration with 60% moving to reunite

with family and 25% to get married and join the husband abroad.

Who migrates where?

The vast majority of emigrants from Egypt (95%) go to Arab countries, mainly in the Gulf and neighboring Jordan

and Libya, with only 5% going to other destinations outside of Arab region, predominantly to Europe (3%) and

North America and Australia (1%).

Pre-migration contact with recruiters

Around 31% of emigrants had prior contact with a private recruiter to work abroad and facilitate the migration.

Also regarding whether the migrant had a visa before emigrating, it should be pointed out that Egyptian citizens

do not need a visa or work permit to enter several Arab countries, including Iraq, Jordan and Libya, which were

the first destination for nearly two-fifths of the emigrants. Thus virtually all of these emigrants who first moved

to these three countries did not have a pre-migration work permit or contact with a labor recruiter and started

looking for a job upon arrival, often contacting labor recruiters and migrant networks in the destination country.

This contrasts with the common procedure of working through labor recruiters operating in Egypt for emigrants

going to the Gulf region countries apart from Iraq.

9

Page 10: Determinants of International Migration: A Methodological ...

The role of networks in migration

Around two-thirds (65%) of emigrants had networks in their country of destination before departure from Egypt.

These migration networks were mostly extended family members and close friends, mostly prior male migrants.

More than four-fifths of these migrants who had a network at destination received assistance whether before

the move and /or upon arrival in the destination country.

Employment situation of migrants before leaving and after

Around three-fourths (74%) of the emigrants were employed in the 3-month period preceding the migration,

with the remaining 26% who were not working evenly divided between those seeking work and those not

seeking work. One in two emigrants (54%) had a job waiting for them upon arrival in their destination, with 98

percent of male migrants currently working abroad compared with 30% of female migrants. The economic sector

of work was often different in the country of destination than in the country of origin, and often did not match

their skills in Egypt. Most migrants in the Arab region ended up in the construction sector (47%), followed by the

wholesale and retail trade (12%), agriculture (11%) and manufacturing (7%).

For emigrants from rural areas of Egypt, the main occupational change was from agriculture to trading, with

others in unstable or casual employment. Migrants from urban areas in Egypt, by contrast, had more diversity in

occupations abroad, with 30% in upper-level occupations in managerial, professional, and technical positions,

but a majority in the lower echelons of the occupational structure. Most of the highly skilled migrants of older

ages were involved in occupations similar to those they had before migration, while most of the younger

migrants got involved in craft and related trades and in services occupations, far from their specialization,

resulting in skill wastage. In terms of benefits provided to migrants by current employer abroad, most Egyptian

migrants are not provided with any form of benefits, with only 29% receiving housing benefits, 24% payments for

overtime, 21% health insurance, and 18% paid annual leave.

Remittances sent by current migrants

Around 70% of emigrants sent money to their origin households in the 12-month period preceding the survey,

on average six times, with 72% submitted via the formal financial system in Egypt.

Migration Intentions

About 11% of the non-migrants interviewed intend to move to another country, while most (70%) intend to

remain in Egypt with the remaining 19% undecided. Substantial differences are reported by gender, with 17% of

men but only 6% of women saying they intend to migrate. Among prospective migrants, 85% intend to migrate

10

Page 11: Determinants of International Migration: A Methodological ...

for economic reasons, 8% for social reasons (mainly to obtain more education), and 7% for other reasons. The

economic motives include income-related reasons (54%) and work-related reasons (31%).

6. Methods of statistical analysis

We used a logistic model with the dependent variable being whether the household member 15-59 emigrated

since 2000 and continued to live abroad, to explore factors associated with the person’s most recent emigration.

A wealth of potentially relevant variables at individual and household levels is available, including situation

before emigration (work status; education, whether married or with children, and countries of residence);

contact with labor recruiters; migration networks; etc. Data were pooled from migrants and non-migrants to

create the population at risk of emigration for each year for each individual and sample household.

Non-migrants include those who had never migrated for more than three continuous months to another country,

plus those who had but had returned (return migrants from abroad living in the sample household at the time of

the survey in 2013). Together, these three groups constitute the population at risk, or the appropriate

comparison group (Bilsborrow et al. 1997). The dependent variable is then whether the person emigrated or not,

assigned a value of 1 if the person emigrated in the most recent years of 2010-13, in order to focus on the most

recent time period, for two reasons, first, to minimize recall error (see Som, 1973), and second, to minimize

endogeneity risk from certain potentially important household variables (e.g., household incomes and assets)

having been collected only at the time of the survey rather than at the time of the emigration (decision). The

result of this was to limit our estimation to the most recent time period, but with such a large sample available,

this still left the analysis to be based on a large sample of 10,133 individuals (about half the total)—perhaps still

the largest ever for a multivariate statistical investigation of the determinants of emigration from a developing

country, from a specialized migration survey based on a nationally representative sample.

The dependent variable being dichotomous—whether the person emigrated or not--the appropriate statistical

approach is a logistic model. The results are then interpreted as follows: odds ratios above 1.0 indicate the

explanatory variable had a positive impact on emigration, while values below 1.0 indicate a negative effect.

Larger odds ratios indicate larger effects. Statistical significance is indicated by P values (models estimated using

Stata). While the usual criterion for judging statistical significance is the 0.05 level, with this large sample the

0.01 level or better is more appropriate.

7. Multivariate results

Table 2 below shows the results on the determinants of emigration of individuals from Egypt in 2010-13, with

hypothesized signs (see Appendix for actual regression results). These results are very strong statistically (most at

11

Page 12: Determinants of International Migration: A Methodological ...

the 0.001 level), and mostly consistent with theoretical expectations. This includes results for important variables

rarely investigated in prior empirical research due to lack of data, including variables capturing the characteristics

of the individual (migrant) at (or just before) the time of migration and household variables also measured at the

time of the migration decision. Among the significant individual variables are the migrant’s age, marital status,

urban vs. rural residence in childhood, work situation before emigration, and knowledge of a foreign language.

Significant household variables measured at the time of migration included urban/rural residence, household

size, and whether the household had another emigrant still living abroad or a return migrant from abroad living

in the household. While the discussion of the results below focuses on those for all age groups together (15-59 at

time of migration), major differences across larger 15year age groups will also be briefly described (15-29, 30-44,

45-59).

a. Results for individual-level variables

Overall, individual-level characteristics were very strong, and much stronger than household factors, which adds

to the evidence accumulating from other recent research finding individual characteristics generally stronger

than household factors (ADD REFS), suggesting that migration is mostly an individual decision although often

affected to a lesser degree by household factors. Thus the effects of the following individual characteristics were

all highly significant at the 0.001 level: age, sex, whether lived in an urban area when growing up, education,

knowledge of another language besides Arabic, marital status, and whether had own children under age 15 living

in the household at time of migration decision. The economic activity of the person prior to emigration is also a

powerful factor, captured in three independent variables reflected different dimensions of economic activity:

first, whether the person ever worked before the move, or work experience (years of work could have been even

more useful, but was not available); second, whether was actively looking for work a month before; and third,

type of occupation, based on the first digit of the detailed occupation code of the International Labour Office.

These results are described in more detail below, beginning with age.

For age, older adults are less likely to emigrate than younger adults (using 15 as a common lower age for adults

in low-income countries), as is generally expected for migrants, given there is always some element of risk going

abroad, usually far away, to work. The data for the three age groups further confirm this, with a positive

coefficient for age for the youngest age group (so those in later 20s are more likely to migrate than those in

lower 20s or teens), while the odds-ratios for age are negative for 30-44 and especially for 45-59. Altogether,

Table 2. Results from Model of the Determinants of Emigration from Egypt, 2010-2013, from Egypt

MED-HIMS Household Survey, 2013 #

Independent Variable Hypothesized Effect Exp. Result

12

Page 13: Determinants of International Migration: A Methodological ...

sign sign

Gender (Males=reference) Females migrate less - -**

Age (15+) Older less likely to emigrate - -**

Urban-Rural residence

(reference=U)

Rural households in Egypt tend to have more

emigrants+ +**

Residence in childhood (ref=U)Urban residence in childhood leads to less emigration

later- -**

Education (years completed) More education usually leads to more emigration + -**

Knew another language before

emigrationLanguage knowledge contributes to out-migration + +**

Married at time of migration Married usually leads to less emigration - +**

Had own children < age 15 in

dwelling before migrationLess likely to migrate if living with own children - -**

Had ever worked before (last)

emigration

More likely to migrate since has work experience and

interest+ +**

Was looking for work just

before emigrationMore likely to migrate since most migration is for work + +**

Occupation not managerial or

professional

More likely to migrate since weaker occupational

ties/lower incomes in Egypt+ +**

Household size (number)

before person emigrated

Larger size means greater need to generate income to

support and more possible emigrants+ -

Persons per room More crowded house encourages someone to leave + +

Household head’s education

More educated head associated with higher

consumption aspirations and/or more access to

information that may encourage emigration

+ +**

Household owns houseHaving house indicates less need to send out member

as emigrant, or may provide security for emigration+/- +*

Household owns farmlandHaving farmland provides source of income and

demand for labor, reducing emigration- -**

Household has bank accountHaving bank account indicates having savings to

finance move or may result from emigration+ +

13

Page 14: Determinants of International Migration: A Methodological ...

Household has previous

migrant living abroad

Previous emigrant provides networks and possible

assistance+ +**

Household has return migrant

from abroad

Previous migrant returned to household provides

network & example and/or encouragement+ +**

*significant at .01 level; ** significant at .001 level N = 10,133

these results show a curvilinear relationship, rising with age to around 30, then falling. Some previous empirical

studies have captured this with more than one age variable, such as age squared.

For gender, relating to cultural values and traditions, females are far less likely to migrate as individuals than

males, in fact more than seven times less likely (odds ratio of .013), and not much different across the three age

groups.

Urban residence in childhood of the emigrant, as reported by the household respondent, has a strong effect on

reducing emigration but only at the .01 level, so not as strong as most of the other individual characteristics

reported here. A likely interpretation is that this gives the person tastes for and/or satisfaction from urban living

reducing interest in going abroad to work, after controlling for all the other individual traits including economic

activity. The results were not much different across the three broad age groups.

The results for education were similar, with more education having a smaller but statistically more significant

(.001) effect reducing emigration, and effect which, however, becomes quite weak for older persons 45-59 (odds

ratio less below 1.0 than for the other age groups, and only marginally significant at the .08 level).

Knowledge of another language besides Arabic (foreign) at the time of emigration, usually English, was

associated with about a 40% increased likelihood of migration at a significance level of nearly .001, with the

effect non-existent for younger adults but increasingly powerful among older adults. Perhaps the lack of a

difference among the young may be because most of them know English due to big increases in enrollment and

graduation levels over time, as has occurred in most developing countries in recent decades, leaving language

knowledge not a significant differentiating factor compared to the situation among older migrants. Curiously,

language skills were most powerful in the older age group precisely in the age group for which education was

much less important.

The results for marital status were even more fascinating, overall being married strongly contributing to

emigration overall, being really strong for younger persons (eight times as likely to emigrate if married): this

could often be even before having children, the newly married man goes abroad to earn money to save and

bring back as a “nest egg”. There is then no effect at all for middle-aged males and even a strong negative effect

14

Page 15: Determinants of International Migration: A Methodological ...

among older males (3 times as likely to not emigrate as single males at 45-59 at a modest significance level

better than .05.

It is also important to control for having dependent children at home, which strongly deters emigration overall,

super strongly for young males: 50 times less likely to emigrate if have dependents vs. 3 times less likely if aged

30-44 and no effect at all for those 45+.

Finally, for economic activity, those with work experience were nearly 5 times as likely to emigrate overall, ceteris

paribus, but this effect declines with age becoming non-existent for those 45+. A much stronger and universal

effect is found for whether looking for work at the time of emigration (almost 9 times as likely to emigrate),

which is to be expected from the descriptive data above on motives for migration, with 87% identifying

economic motives as primary, and the simple fact that 98% were males leaving, almost all for work.

For occupational code, the results are much more complex, likely due to smaller numbers especially when

dividing by both occupation and age group. Note the reference (omitted) occupational code was “managers and

professionals while ILO codes 3 and 4 are also high level technicians/skilled workers, with most emigrating

workers in the lower-status, lower numbered occupational codes 5+ (with lower incomes) who are more likely to

migrate (odds ratios above 1.0). This does contrast to that many other countries, but clearly reflects the

tendency of individual migrants to be relatively young males from rural areas and secondary education seeking

unskilled/semi-skilled work in other Arab countries.

b. Results for household-level factors

Among the significant household variables is region of residence before the move (in the results shown in the

Appendix, in five categories, capturing dimensions of both urban-rural and region--upper vs. lower Egypt, the

latter with the large urban agglomerations of Cairo, Alexandria, etc., constituting the reference region). The

results show effects especially of people migration more than those in the two big metro-agglomerations, from

rural area as well as smaller urban areas, and a bit more from northern (Upper) than southern Egypt, likely due

to more access to information and transportation access to nearby Arab countries. It is curious that the effects of

these geographic differences in origin disappear for older adults.

The second variable is household size just before emigration (defined correctly to include the emigrant, which

has not always been done in previous empirical work). The hypothesis is that a larger household makes it easier

to release a household member from household needs for helping with childcare, farm work, etc., to migrate

away, supported by its alleviating household consumption needs. This has been observed in some previous

research (CITES?). However, the effect in Egypt is negative, though very small and not statistically significant

except for the older age group. A negative effect might suggest that the household has such a low income that it

15

Page 16: Determinants of International Migration: A Methodological ...

cannot afford to lose an able-bodied male from remaining to help around the house or farm and/or the cost of

financing the emigration.

Related to household size is whether there is any crowdedness on living conditions, which can be reflected in

persons per room. In fact, there is such a positive effect on emigration, albeit small (25%) and of marginal

significance overall (.05). Looking at the relationship for age groups shows again that the effect is only for

younger emigrants (76% effect, at .001 level!). This could reflect younger generations being more concerned

about having privacy and their “private space” than older persons.

A third household variable and one found positive and significant in much research is household head’s

education. Each year above the mean results in about 40% more likelihood of emigration, other things equal,

though once again the effect dissipates among those 45+. This overall effect seems logical, as a more educated

head with more access to information and perhaps higher consumption aspirations for the household would be

more likely to encourage younger than older adults in the household to take advantage of opportunities abroad.

This effect is highly significant at the .001 level.

Ownership of house is positively linked to the household having an emigrant (at the .01 level), but only for a

young emigrant (.001 level), there being no significant effect at all for the older ages above 30.

In contrast, owing farmland reduces emigration, as anticipated from theory and previous research: land provides

opportunities to produce food to support the household. Note this effect exists and is strongly significant as a

deterrent for both the older two age groups but is the opposite for young adults, a wealth effect facilitating

emigration (almost 50% more likely and at a significance level of .011). It should be noted that the variable does

not measure amount of farmland but just whether it owns any.

The last indicator of household socio-economic status is whether anyone in household had a bank account,

which is especially likely to be endogenous since it is often the result of someone migrating (rather than only a

measure of socio-economic status that could cause or facilitate emigration via financing it, etc.). Thus if the

household did not have such an account before, having an emigrant could very well stimulate it to get one to

facilitate receiving cash transfers from abroad. This is a good example of the problem of endogeneity that could

also be present with the variables on ownership of a house or farmland, as having an emigrant could lead to

remittances making it possible for the household to acquire a house or farm. To minimize but not totally

eliminate this, our whole empirical model is estimated here only for the very recent emigrants, some migrating

only days or months before the survey up to 3.5 years before. Most emigrants take some months to settle and

earn enough to save, before they can send remittances, and many take a few years, so our approach reduces but

16

Page 17: Determinants of International Migration: A Methodological ...

does not eliminate this risk, which pertains to any variables relating to household wealth. The solution would

have been to have collected data on possession of major asset in the household prior to the emigration decision

being studied (see also discussion in lessons learned, in the conclusions to the paper below).

The last two household variables included in the model and discussed here are included to capture the migration

experience of the household (Bernard and Perales 2021), in lieu of asking about migration networks or

relatives/friends known by household members living abroad. Thus it is well-known that migrant networks

established from previous household members (or other relatives or friends) emigrating tend to are crucial

factors in subsequent migration. We capture this by determining whether at the time just prior to the migration

decision (to migrate or not, referring to the person who did and the randomly selected non-migrant person who

did not, together comprising the population at risk of migration), the household had a previous household

member living abroad (current emigrant) or a current household member who had lived abroad but returned to

the household, viz., a return migrant. Creating these variables for every household required, first, reconstituting

the household composition (having already done this for household size for each year since 2000) to identify for

each household each year since 2000 when it had a prior emigrant still living abroad and when a return migrant

in the household, and how many. We are not aware of any previous study of international migration based on

household-level data that has controlled for both of these potentially important influences on new migration.

And indeed, both were found to be independently significant and the most powerful household variables,

rivaling in explanatory power the strong effects of the individual variables discussed above. Thus a household

with a current migrant living abroad is over 9 times as likely to have a new migrant in the reference period, even

stronger for those 15-44 (both overall and for the two younger age groups at better than the .001 level), but

having no effect on older persons emigrating. Similarly, if the household has a return migrant, that multiplies the

likelihood of some member migrating by 2.5 overall as well as among those 15-29 and by far more for those

30-44, but again not at all for those 45-59 at the age of emigration. The stronger effect of emigrants than return

migrants seems expected, as they sometimes will be in the same country of destination chosen by the recent

emigrant and could provide substantial assistance in housing and finding work, and even help finance the move.

We also collected data on several other variables rarely available in empirical studies of international migration

except in cases of specialized migration surveys to explore their possible effects, including contacts with labor

recruiters, whether they had a work contract before leaving, possession of a visa, etc., none of which was

statistically significant once all the other variables above were controlled for. An explanation for this for one of

these variables is that Egyptians do not need visas to migrate to the two neighboring Arab countries, Jordan and

Libya (Farid et al., 2016). Another may be that work contracts are arranged not by labor recruiters per se but

often by government or private employment agencies. It is most interesting that some knowledge of another

17

Page 18: Determinants of International Migration: A Methodological ...

language besides Arabic, usually English, was linked with emigration to destinations in general and not just to

non-Arab countries, as we anticipated, which may show the benefits of access to information and/or a link to

higher consumption aspirations.

8. Conclusions

There are a number of important methodological conclusions. To begin with, the rich empirical results described

above demonstrate the benefits of collecting and analyzing data on international migration from a specialized

survey on international migration in contrast to what can be learned from the many existing studies based on

census data or data from labor force or other household surveys developed for another primary purpose. Linked

directly to this, the results also demonstrate the value of a sample design specifically oriented to focus data

collection on migrants, via a. stratification of geographic or administrative areas in the country based on (if

available from any source) the proportion of migrants of interest in the population, and then oversampling areas

with higher proportions of migrants in the population, as well as at the last sampling stage of clusters or

enumeration areas oversampling households with migrants, as they otherwise tend to be relatively rare

elements, especially for recent migrant. This ensures that field data collection will yield a sufficient number of

migrants as well as non-migrants.

Second, the careful development of questionnaires comprising household modules including an initial screen

model to identify households with migrants of interest, and individual modules pertaining to emigrants (for

whom data will usually have to be collected from the most knowledgeable household member as a proxy

respondent), and one or more non-migrants in every household sampled. To study the determinants of

out-migration/emigration, data on migrants should refer to their situation at or just prior to the time of

migration. Ideally, such data should also be collected on the situation of the migrant’s household at that time, as

well as for non-migrant individuals and their households (see Bilsborrow et al., 1997, and also limitations

discussed below).

Third, technical assistance in the design of the sample, questionnaire modules, data processing/entry and

cleaning and data analysis will often be crucial in many countries to produce quality results. Fourth, the data

resulting should be made available for detailed analysis both in the country and globally. International funding

support for the various phases of the project as noted above can facilitate both the data collection, a large

sample size, and detailed analyses that take advantage of the data to improve understanding of migration

processes in the country, some with policy implications.

18

Page 19: Determinants of International Migration: A Methodological ...

Building on the analysis here of the determinants of emigration, with such a large and nationally representative

sample, it is possible to disaggregate the data to run the regression model to estimate the determinants of

emigration for different sub-populations, such as for 5-year age groups (viz., 15-19, 20-24,…,55-59); by gender,

for men vs. women (in the case of Egypt with such a small percentage of women, the absolute number is still

sufficient to model); type of region of origin (urban vs. rural, northern vs. southern Egypt, and combinations

based on both dimensions); type of destination country (Arab vs. other, etc.).

The descriptive data in section 5 above also provide a taste of directions for extending the analysis, to estimate

the determinants of return migration and of migration intentions from the data in this survey in Egypt. For

example, for return migration, the population at risk is the stock of emigrants currently living abroad plus those

who have returned to the household (who can be interviewed directly, as done in this Egypt case). Since several

questions were also asked in the individual questionnaire of the selected non-migrant in every sampled

household on whether they intend to migrate or not, it is possible to estimate the determinants of intentions to

emigrate for a large sample here, with many covariates and possibilities for disaggregating as above by gender,

age, education, etc. Finally, it would then be possible to compare the results of the determinants of return

migration and the determinants of migration intentions with those here for the determinants of emigration, to

compare and contrast the determinants, to elucidate better than has been done before what the differences

area and what this implies for net past and possible future population loss, brain drain, and migration policy.

Some limitations and suggestions for the future

In the case of the Egyptian survey, there are some limitations, beginning with the big one inherent in any such

survey based only in a country of origin (as is true of any population census). Whole households that have left

leave no one behind to report on their composition or emigration and are hence omitted. It is suspected that this

may account for about a third of the emigration from Egypt. To collect data on them would require coordination

with and permission from the main countries of destination of Egyptians abroad (such as Jordan, Libya, Iraq,

Saudi Arabia, Kuwait, the United Kingdom, etc., which is complicated.

More specific limitations arise from the questionnaire design, in which data were collected only for the individual

attributes of the emigrant at the time of departure (of the migration decision), and not on the situation of the

migrant’s household, nor from the non-migrant individual questionnaire on the situation of the non-migrant at

the time of emigration of the person from the household, and least of all on the non-migrant randomly selected

for the individual interview in the non-migrant household. In the absence of a reference year for retrospective

data on the non-migrant household, the mean year of emigration in the study population (or a randomly

19

Page 20: Determinants of International Migration: A Methodological ...

selected year with the approximate expected distribution of yeas for the emigrants should be used (which is

feasible to do, though never tried to date). Finally, there was insufficient data collected on potentially important

economic variables pertaining to the time of migration, including individual and household incomes and

measures of wealth (including even the limited ones collected in this case—ownership of house or a farm or a

bank account, but further data on these such as estimated value and the time of their acquisition as well as of

other significant assets could be asked.

Such things complicate data collection and tend to be difficult to sell to government statistical offices. Also useful

would have been to collect further information on migrant networks, on the residence abroad of close relatives

(e.g., only parents, spouses, children, siblings). A case could be and has often been made for the collection of

data at a level above the household, on the community or town, to gauge the effects of contextual factors on

migration (c.f., e.g., Bilsborrow et al.1984; Massey 1990).

Egypt/CAPMAS plans to conduct another large survey on international migration in 2022. Hopefully some of

these extensions will be considered for inclusion, to lead to another leap in the richness of the data and in the

resulting analyses and findings, both substantive and methodological.

References cited

Bernard, A., and F. Perales. 2021. Is migration a learned behavior? Understanding the impact of pastmigration on future migration. Population and Development Review 47(2): 449–474.

Bilsborrow, R. 2013. Sampling Plans for MED-HIMS Surveys (Programme of Mediterranean HouseholdInternational Migration Surveys). Paris: ADETEF MEDSTAT III, pp. 66.https://circabc.europa.eu/faces/jsp/extension/wai/navigation/container.jsp

Bilsborrow, R., 2016. Concepts, Definitions and Data Collection Approaches. In M. White, ed., Handbook ofMigration and Population Distribution. Springer, Ch. 7, pp. 109-156.

20

Page 21: Determinants of International Migration: A Methodological ...

Bilsborrow, R., 2018a. Household surveys: A key potential resource for data on migration. Blog on IOMGlobal Migration Data Analysis Centre portal, May 31.

____, 2018b. Specialized migration surveys: Six regional examples improving research. Blog on IOM GlobalMigration Data Analysis Centre portal, August 31.

Bilsborrow, R., A. Oberai, & G. Standing. 1984. Migration Surveys in Low-income Countries: Guidelines forSurvey and Questionnaire Design, London: Croom-Helm, for the International Labour Office.

Bilsborrow, R., G. Hugo, A. Oberai, & H. Zlotnik. 1997. International Migration Statistics: Guidelines forImproving Data Collection Systems. Geneva: International Labour Office.

Bohra, P., & D. Massey. 2009. Processes of Internal and International Migration from Chitwan, Nepal.International Migration Review, 43(3):621-651.

Cassarino, J.P., 2004. Theorising return migration: The conceptual approach to return migrants revisited.International Journal on Multicultural Societies, 6(2): 253-279.

Center for Global Development, 2009. Migrants Count: Five Steps Toward Better Migration Data. Report ofthe Commission on International Migration Data for Development Research and Policy, WashingtonDC: CDG, Michael Clemens, Project Director, pp. 28.

Cerase, F. 1974. Expectations and Reality: A Case Study of Return Migration from the United States toSouthern Italy. International Migration Review, Special Issue: Policy and Research on Migration:Canadian and World Perspectives (Summer, 1974), 8(2): 245-262.

Chengrong, D., & P. Nangia, 1998. Emigration from Kerala: Characteristics of Migrants and theirContribution to Household Economy, IIASSI Quarterly (Indian Association of Social ScienceInstitutions) 16(3 & 4): 226-257.

Constant, A. and Massey, D.S., 2002. Return migration by German guestworkers: Neoclassical versus neweconomic theory. International Migration Review, 40(4), pp.5-38.

Curran, S., & E. Rivero-Fuentes. 2003. Engendering migrant networks: The case of Mexican migration.Demography, 40(2), 289-307.

David, A., & J. Jarreau. 2016. Determinants of Emigration: Evidence from Egypt. Cairo: Economic ResearchForum, Working paper No. 987.

De Haas, H., T. Fokkema, & M. Fassi Fihri. 2015. Return Migration as Failure or Success? Journal ofInternational Migration and Integration, 16(2): 415-429.

21

Page 22: Determinants of International Migration: A Methodological ...

Esipova, N., J. Ray & R. Srinivasan. 2011. The World’s Potential Migrants—Who they are, where they wantto go and why it matters. Washington C: Gallup, Inc.

Farid, S., Nour El-Deen, A., & El-Batrawy, R., eds. 2016. Egypt Household International Migration Survey2013: Volume I: Determinants and Consequences of International Migration. Cairo, Egypt: CentralAgency for Public Mobilization and Statistics, pp. 384.

Farid, Samir, Tarek Abou Chabake, Richard E. Bilsborrow, Giambattista Cantisani, and Ingrid Ivins. 2015.Mediterranean Household International Migration Survey (MED-HIMS). Manual 1: ModelQuestionnaires. Manual 2: Survey Design and Organization. Manual 3. Supervisor´s Manual.Manual 7. Tabulation Plan. Paris: ADETEF – MEDSTAT III, available at EU CIRCA Portalhttps://circabc.europa.eu/faces/jsp/extension/wai/navigation/container.jsp

Flahaus, M.L. 2017. The role of migration policy changes in Europe for return migration to Senegal.International Migration Review, 51(4):868-892.

Gang, N., Bauer, T. 1999. Temporary Migrants from Egypt: How Long Do They Stay Abroad? IZA DiscussionPaper No. 3. Available at SSRN: https://ssrn.com/abstract=164553, 26 pp.

González-Ferrer, A., Baizán, P., Beauchemin, C., Kraus, E., Schoumaker, B. and Black, R., 2014. Distance,transnational arrangements, and return decisions of Senegalese, Ghanaian, and Congolesemigrants. International Migration Review, 48(4), pp.939-971.

Groenewold, G., & R. Bilsborrow. 2008. Design of Samples for International Migration Surveys:Methodological Considerations and Lessons Learned from a Multi-country Study in Africa andEurope. In C. Bonifazi, M. Okolski, J. Schoorl, & P. Simon, eds. International Migration in Europe:New Trends and New Methods of Analysis, Amsterdam University Press, pp. 293-312.

Kirk, D. 1960. Some Reflections on American Demography in the nineteen sixties. Population Index, 27:305-310.

Kish, Leslie. 1965. Survey Sampling. New York: John Wiley & Sons.

Lee, E. S. 1966. A theory of migration. Demography, 3(1), 47-57.

Massey, D. 1990. Social structure, household strategies, and the cumulative causation of migration.Population Index, 56(1), 3–26.

Massey, D., J. Arango, G. Hugo, A. Kouaouci, A. Pelligrino, & J.E. Taylor. 1993. Theories of Internationalmigration: A Review and Appraisal. Population and Development Review, 19(8): 431-466.

22

Page 23: Determinants of International Migration: A Methodological ...

Massey, D.S. and Espinosa, K.E., 1997. What's driving Mexico-US migration? A theoretical, empirical, andpolicy analysis. American journal of sociology, 102(4), pp.939-999.

Mincer, J. 1978. Family Migration Decisions. Journal of Political Economy, 86(5): 749-773

Mora, J., & J. E. Taylor, 2006. Determinants of Migration, Destination, and Sector Choice: DisentanglingIndividual, Household, and Community Effects. In International Migration, Remittances, and the BrainDrain, eds. Ç. Özden & M. Schiff. New York: Palgrave Macmillan, pp. 21-52.

Ravenstein, E. G. 1889. The laws of migration. Journal of the Royal Statistical Society, 52, 241–302.

Rendall, M., & S. Parker. 2014. Two Decades of Negative Educational Selectivity of Mexican Migrants to theUnited States. Population and Development Review, 40(3): 421-446.

Schwartz, A. 1976. Migration, age and education. Journal of Political Economy, 84(4), 701–720.

Sjaastad, L. 1962. The costs and returns of human migration. Journal of Political Economy, 70(5), 80–93.

Skeldon, R., 2013. Global migration: Demographic aspects and its relevance for development. New York:United Nations.

Som, R.K., 1973. Recall Lapse in Demographic Enquiries. New York: Asia Publishing House.

Stark, O. 1984. Rural-to-urban migration in LDCs: A relative deprivation approach. Economic Developmentand Cultural Change, 32(3), 475-486.

Stark, O., & D. Bloom. 1985. The new economics of labor migration. The American Economic Review, 75(2),173-178.

Taylor, J. E., & A. Lopez-Feldman. 2010. Does migration make rural households more productive? Evidencefrom Mexico. Journal of Development Studies, 46(1), 68-90.

Tjaden, J., & F. Laczko. 2019. Linking migration intentions with flows: evidence and potential use.International Migration, 57(1): 36-57.

Todaro, M.P. 1969. A model of labor migration and urban unemployment in less developed countries. TheAmerican Economic Review, 59(1): 138-148.

United Nations, 2006, 2009, 2013, 2016. High Level Dialogue on Migration and Development. New York: UNGeneral Assembly. Recurring global meeting of government leaders.

United Nations. 2006. International Migration and Development, Report of the Secretary General.(A/60/871). New York: Department of Economic and Social Affairs.

23

Page 24: Determinants of International Migration: A Methodological ...

United Nations Population Division. 2020. International Migration Stock 2020. New York: Department ofEconomic and Social Affairs.

United Nations. 2018. The Global Compact for Safe, Orderly and Regular Migration. Meeting of UN Heads ofState and High Representatives, Morocco, July 11 Final Draft.

White, M., ed., 2016. Handbook of Migration and Population Distribution. Springer.

Wolpert, J. 1965. Behavioral aspects of the decision to migrate. Papers in Regional Science, 15(1),159–169.

24

Page 25: Determinants of International Migration: A Methodological ...

Annex Table 1. Determinants of Emigration, Egypt, 2010-13

All ages Age:15-29 Age:30-44 Age:45-59

VariableOddsRatio

StdErr P>|z|

OddsRatio

StdErr P>|z|

OddsRatio Std Err

P>|z|

OddsRatio Std Err

P>|z|

Age 0.9380.00

6 0.000 1.335 0.034 0.000 0.934 0.0160.00

0 0.840 0.0320.00

0

Female (0/1) 0.0130.00

3 0.000 0.004 0.002 0.000 0.007 0.0040.00

0 0.007 0.0090.00

0

Region (1_Urb-Gov) [ref]

(2_Low-Urb) 2.1100.40

4 0.000 3.332 1.029 0.000 2.165 0.7340.02

3 0.800 0.3590.61

9

(3_Low-Rur) 1.7790.40

6 0.012 3.342 1.217 0.001 2.542 0.9960.01

7 0.391 0.2290.10

8

(4_Upp-Urb) 2.1680.45

0 0.000 3.244 1.017 0.000 2.606 0.9920.01

2 0.878 0.4980.81

9

(5_Upp-Rur) 3.4650.80

2 0.000 7.170 2.774 0.000 3.557 1.3920.00

1 1.565 0.8900.43

1

Lived in urban area as child (0/1) 0.6080.11

4 0.008 0.541 0.170 0.051 0.638 0.1970.14

6 0.626 0.2920.31

5

Education (years) 0.7950.01

5 0.000 0.771 0.022 0.000 0.736 0.0290.00

0 0.863 0.0740.08

4

Knows other language (0/1) 1.4150.16

5 0.003 0.886 0.151 0.477 1.926 0.4250.00

3 4.055 1.7770.00

1

Married or signed contract (0/1) 3.2820.56

0 0.000 7.980 2.800 0.000 1.185 0.3730.58

9 0.321 0.1640.02

6Has children in home under 15(0/1) 0.221

0.028 0.000 0.019 0.006 0.000 0.353 0.082

0.000 1.551 0.495

0.169

Worked before move (0/1) 4.7620.79

9 0.000 3.461 0.980 0.000 2.951 1.4280.02

5 1.394 0.7610.54

2

Looking for work (0/1) 8.6161.60

6 0.000 3.497 0.870 0.000 12.074 4.9570.00

0 13.477 12.6120.00

5

(Occupation-3) [ref. = 1&2] 0.4900.11

4 0.002 0.271 0.134 0.008 0.485 0.1810.05

3 1.017 0.6270.97

8

(Occupation-4) 0.2210.11

5 0.004 0.281 0.351 0.310 0.247 0.2100.10

0 0.414 0.3460.29

1

(Occupation-5) 1.6130.30

3 0.011 0.792 0.273 0.499 1.534 0.4500.14

5 0.911 0.8590.92

1

Page 26: Determinants of International Migration: A Methodological ...

(Occupation-6) 3.0340.49

3 0.000 2.030 0.613 0.019 4.170 1.0190.00

0 3.131 1.7300.03

9

(Occupation-7) 4.9380.79

6 0.000 1.857 0.558 0.039 6.675 1.6000.00

0 8.619 4.4970.00

0

(Occupation-8) 1.9320.36

4 0.000 0.856 0.298 0.655 2.143 0.5970.00

6 2.507 1.5680.14

2

(Occupation-9) 0.9350.26

5 0.812 0.380 0.225 0.101 1.012 0.3950.97

5 1.076 0.8770.92

9

HH size before left 0.9580.03

6 0.253 1.076 0.060 0.188 0.931 0.0520.19

9 0.792 0.0740.01

3

HH has prior emigrant (0/1) 9.4132.47

8 0.000 25.783 9.564 0.000 12.422 7.3920.00

0 1.113 0.9070.89

5

HH has return-migrant (0/1) 2.4820.51

6 0.000 2.905 0.823 0.000 15.577 9.5300.00

0 1.447 1.3780.69

8

HH head's education (years) 1.3940.02

5 0.000 1.490 0.041 0.000 1.384 0.0520.00

0 1.140 0.0910.09

8

HH owns house (0/1) 1.6170.27

5 0.005 2.303 0.627 0.002 1.159 0.3300.60

5 1.486 0.6930.39

6

Persons per room 1.2440.13

9 0.050 1.761 0.303 0.001 1.162 0.2510.48

8 0.474 0.1860.05

7

HH owns farmland 0.7670.07

6 0.008 1.454 0.214 0.011 0.681 0.1180.02

7 0.336 0.1220.00

3

HH has bank account 1.1910.18

1 0.250 0.885 0.230 0.638 3.269 1.0050.00

0 1.575 0.6320.25

8

[_Constant] 0.1850.07

7 0.000 0.000 0.000 0.000 2.285 2.1040.36

9 6651.41 15214.60.00

0

Model evaluation All ages Age:15-29 Age:30-44 Age:45-59

Chi2 1524.25 746.82 489.85 153.4

Prob>Chi2 0 0 0 0

R2 0.604 0.745 0.612 0.441

pseudolikelihood -2444.3 -957.43 -672.3 -258.15

Observations (N) 10133 5872 2955 1306