Using Household Surveys to Measure International Migration and Remittances in Developing Countries: Examples and Methodological Issues Richard E. Bilsborrow University of North Carolina Presented at Expert Group Meeting on Measuring Migration, Remittances, and Development Using Household Surveys: An Assessment of Efforts in Developing Countries, US Census Bureau, January 14-15, 2008.

Importance of international migration in the world, demographically 191 million persons living in a country other than that of their birth in 2005, 3% of world population Compares with 75 million, 2.9% in 1960 But much higher, 9.5%, in more developed countries (MDCs) vs. 1.4% in less developed (LDCs) Above is lifetime data: flow data show how rare it is UN data indicate annual in-migration rate during of 0.22% in MDCs and -0.05% in LDCs

Recent migration has increased but is still rare compared to the sending or receiving country The only countries in the world with over 1 million population in 2005 which had a net annual international migration rate over 1% in were: Eritrea, Afghanistan, Kuwait, and United Arab Emirates—positive Liberia, Tajikistan, Georgia, and Oman— negative All values 1-1.9% except UAE 5.0% Only country with over 6.5 million pop is Afghanistan

International migration as a rare phenomenon (continued) Developed countries with highest inflows were Spain and Ireland at 0.97 and 0.98% Others of interest include Canada 0.7, Australia 0.5, USA 0.4, Germany 0.3, UK 0.2, Netherlands 0.2, France 0.1 Albania -0.6, Ukraine -0.3, Armenia -0.7 Iran -0.4; Morocco, Sudan -0.3; Philippines, Pakistan - 0.2; Egypt -0.1 Guatemala -0.5; Mexico, Ecuador, Nicaragua -0.4, Dom. Rep. -0.3; Peru -0.2 Surprisingly few others in world (over 5 million population) over -0.1 (China, India -0.03)

Defining and Measuring Migrants Based on place (country) of birth = foreign born Based on country of previous residence being different from current, and time of arrival (yields fixed-period migration, almost flow statistic) Based on citizenship Duration of residence (or intended residence) in current country and previous country (6, 12 mo. criteria) Sources: Census, current population register, border/admission statistics, registers of foreigners/foreign workers, naturalization stats.

Purpose of data collection: is it to Measure/count international migrants— immigrants, emigrants, return migrants Over some fixed recent time period (cut-off of 1, 2, 5, 10….years) Characterize migrants: age, sex, education, work experience, assets…. Collect data on remittances sent/received Study determinants of migration and/or consequences of migration

Two survey approaches Adapt existing survey—which has major purpose (s) established, sample size and design, questionnaire content, stakeholders, periodicity Advantages: main cost already covered, inexpensive to add a few questions Disadvantages: cannot add many questions, sample of migrants may be inadequate OR Design and implement new specialized survey

Focus is on developing countries of emigration Data on individuals who have left (emigrated) from households can be obtained from household members remaining behind (proxy respondents) Limitations in data that can be obtained Data on whole households that emigrated usually only obtainable in country of destination, indicating major limitation of survey carried out only in origin country

Types of existing surveys to think of adapting Labour force surveys Demographic and Health Surveys (DHS) Living Standard Measurement Surveys (LSMS) Other (e.g., Multi-purpose) surveys Inherent limitations of border statistics, passenger statistics, admission statistics, so not discussed further

Adapting existing surveys to collect more/better data on International Migrants Requirements of surveys: large sample size and/or high prevalence of migrants in country Issue of “rare elements”, especially if focus is on recent migrants vs. lifetime migrants Example, survey of 10,000 households in country with 10% foreign born, 2% recent, yields only 500 adult migrants Most useful are surveys which already have some relevant information, e.g., previous place of residence or birth, employment status, etc. The longer the questionnaire already in use, the more limited new questions can be in general

Labour force surveys—the best prospects Most universal across countries, regular Administered by National Statistics Office Relatively large sample sizes National coverage and representation Already collect some key data on international migrants—age, sex, education, employment,etc. Many ask place of birth, thus identifying foreign born, some also ask previous place of residence x years ago or when came to country

Labour force surveys (cont.) Some developing countries already include modules on international migration Examples include Pakistan, Costa Rica, Ecuador, Poland, UK, Armenia, Philippines, Egypt, Mexico, on immigrants and/or emigrants Most ask motives for migration, date of departure/arrival, remittances received Only a few ask about household composition and characteristics, migrant’s work, education, etc., before departure (or arrival)

Example of Thailand 2006 LFS Carried out since 1963, quarterly since 2001, 80,000 households/quarter, 26.7K monthly Covers all 76 provinces as PSUs, has 5,796 Secondary Sample Units, 3/5 sample is urban though population is only 30% urban so must weight data Uses normal LF questionnaire, including wages (but no housing or assets data), in 4 th quarter asks 19 questions on immigration, including duration of residence, whether registered, plans to remain or not, if worked in month before coming to T, whether sends $, amount, use.

Experimental module on IM used in 4 th quarter, 2006 (ILO/WB) 22 questions, for every member, referring to previous 12 mos. Ask if x received any $/goods from elsewhere, times, amount, relationship to head, usage, etc. Asks if x is citizen of T, if not, where When 1 st came to T, last time, from where Asks education, occupation, industry, etc., at time of arrival (can compare with current)

2006 Thai module (cont.) Missing items-- Thai language ability on arrival and now, marital status then and now, with whom came Only 0.6% of pop constituted by immigrants, mostly from 3 neighboring countries, estimated less than 500 persons (150 hhs?) born abroad Tabulations in report based on these numbers multiplied by national inflation factor of 823, obscuring real numbers; this is reason for presenting raw numbers in reports No data on emigration, nor remittances received

Ecuador, Survey of Employment, Underemployment and Unemployment 2005 national sample, 19,596 “dwellings” Module with 33 questions per emigrant– when left, why, work status/occupation, marital status, etc. Some on current status—country of residence, age, education, work Remittances previous mo., 12 mos., no. times, use including investing, sector Independently, every person in household 18+ is asked if received remittances ever, when, whether invested Gets current education but not at time of emigration, and marital status at emigration but not current Migration intentions asked only in hhs with emigrants

Survey on Overseas Filipinos Conducted annually by NSO since 1987, every October, sample currently 41,000 households Asks about those who left to work or travel overseas in past 5 years, characteristics now, and remittances received and by what means Results on web and in Yang et al. find 14% rise in number working abroad in 2006 vs. 2005, slightly more women than men, 17% rise in remittances, 95% in cash, 79% sent thru banks

National Survey of Occupation and Labour, 2007 (INEGI) 120,260 “dwellings” representative of states…. Measures immigrants from questions on place of birth, previous residence, why came Measures emigrants by asking if any former member has left to live elsewhere, and why In 4 th quarter of 2002, Module on Migration included with more questions, sample 80,000, asks residence 5 years ago, country/state of current residence, why/when left, means of transport, remittances sent Questions also for return migrants, including if went to USA to work, whether with legal docs

DHS surveys Advantages—moderate sample sizes, K households Already have detailed demog. data including on children: ask residence Also, accustomed to orphanhood questions, so ask residence of siblings, etc Examples of Ecuador, Colombia

Ecuador DHS, 2004 Explosion of emigration since 1995 ENDEMAIN IV covered 28,908 households (CEPAR), with module on emigration Asks if anyone left in previous 5 years, when, characteristics of person when left, work in year before leaving, why left, current residence, remittances sent in past 12 months, frequency, amount, use Shows about 7% of households (over 1500) have an emigrant since 1999, 62% of which received remittances in past 12 months Could have asked about residence of absent children, migration intentions

LSMS surveys Advantage: Rich economic data sets, so can use to better measure economic situation of migrants and non-migrants Flexibility: Countries can add modules of interest Disadvantages: Small sample size, long questionnaires already so hard to add to Examples of Ghana, Peru, Ecuador, Armenia, Albania

Key issue in surveys of migration: use of appropriate comparison groups To study either the D or C of migration, need data for both migrants and non-migrants (M & N) For IM, to study the D, need data for both the M and the rest of the at-risk pop of N in the country of O, which serves as the appropriate comparison group But for whole hhs migrating, need data from hhs in D that migrated from O plus hhs in O that did not migrate Furthermore, need detailed data on the situation prior to the migration, of the individual and hh context. For individual M, get data from proxy respondent. For indiv N, from person. Time reference should be just before mig for migrants, and at the mean time of migration for N, in neither case, at the time of the survey!

In sum, to study migration’s determinants In country of origin: survey households with and without recent emigrants In country of destination: survey recent immigrants originating in the country of origin Compare: migrants at time of departure from origin with non-migrants at origin at the average time of departure of migrants

Sampling Migrants in specialized surveys of international migration Absolute need for probability sampling Define survey purpose and hence appropriate comparison groups at outset (and countries!) Based on budget, define survey domain and target sample size and distribution Recognize that migrants are rare elements Find sampling frame—for migrants and non- migrants Calculate proportions of population which are migrants of interest in PSUs, form strata

Sampling rare elements: disproportionate sampling from strata, two-phase sampling Example of country of Origin, select migrants and non- migrants Sample Primary Sampling Units such as provinces using disproportionate sampling, oversampling from strata with high proportions of international migrants Continue in subsequent stages, oversampling In final sample Ultimate Area Units (UAUs), conduct screening operation, use 2-phase sampling, first listing occupied households to identify those with migrants Oversample those with migrants, but also select households without; conduct interviews

Examples of specialized surveys of international migration NIDI Argentina IOM Surveys in Guatemala Others—Morocco, Maghreb (2007), Malaysia, Colombians in Ecuador

NiDi 7-country Push-Pulls Project on Causes of Migration to EU, countries of O, Italy and Spain of D Common survey purpose, sampling methodology, questionnaires; completed sample sizes Defined migrant households in O as those with member who left to live abroad without returning in previous 10 years In D defined as having someone come from one of two specific countries of O Rare elements, so used stratification, etc. Used appropriate comparison groups

NIDI-Eurostat Survey Project, Five countries of origin: Turkey, Egypt, Morocco, Senegal and Ghana Two destinations: Italy and Spain Common methodology: samples, questionnaires Sample sizes range from Migrant: person leaving/arriving in past 10 years

NIDI-Eurostat Survey Project, Migrants at destination: Those originating from either of two countries of origin For immigrants from Egypt and Ghana in Italy, appropriate comparison groups are available: non-migrants interviewed in Egypt and Ghana For immigrants from Morocco and Senegal in Spain, appropriate comparison groups are available: non-migrants in Morocco and Senegal

Complementary Survey on International Migrants in Argentina Survey in of migrants from 5 neighbouring countries, in main areas of residence of each (2 to 6 for each, total of 18), not national sample Households interviewed if contained someone born in any of the 5 countries, so not focused on recent immig Used 2001 census to create sampling frame, two-stage sample in Buenos Aires region, one-stage elsewhere Sample sizes of 13,296 and 8,222, respectively Asks antecedents to coming, residence history in Arg., whether sent or received $, residence of close relatives, property in origin country now, migration intentions, etc.

Example of IOM Surveys in Guatemala, 2002-present Annual surveys, national sample of 3,000 households, based on 3-stage sample using 2002 census to create sampling frame PSUs are districts (municipios) selected with PPES, ……, UAUs census sectors Listing operation performed on all 55,000 dwellings in sample UAUs, then hhs with migrants “selected with equal probability” in country to interview Wealth of data obtained, many pamphlets published, with results, including on remittances, use for investment, effects on poverty of receiving households

Examples of questionnaire modules in country of emigration To identify emigrants On work activity of emigrant prior to emigration On reasons for emigration On work activity in country of destination On migration intentions On remittances On return migrants