1 / 27

Surveying Migrant Households: A Comparison of Census-Based, Snowball, and Intercept Point Surveys

Surveying Migrant Households: A Comparison of Census-Based, Snowball, and Intercept Point Surveys. David McKenzie World Bank (with Johan Mistiaen). Motivation. Increasing research and policy attention being given to migration

betsy
Download Presentation

Surveying Migrant Households: A Comparison of Census-Based, Snowball, and Intercept Point Surveys

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Surveying Migrant Households: A Comparison of Census-Based, Snowball, and Intercept Point Surveys David McKenzie World Bank (with Johan Mistiaen)

  2. Motivation • Increasing research and policy attention being given to migration • But few detailed and representative surveys of households of migrants exist • Many specialized surveys of migrants are non-probability samples, making it hard to generalize from them => Paper reports on experiment in Brazil to compare three different survey methods

  3. Context • World Bank survey of Brazilians of Japanese descent (Nikkei) • Approximately 1.2-1.9 million of them among Brazil’s 170 million population • Japanese law allows third-generation and below unrestricted access to labor market • Estimated 265,000 Brazilian migrants in Japan, sending approximately $US2billion in remittances annually. • Survey methods used here equally applicable to attempts to survey migrants in destination countries.

  4. Surveying methods • In most countries, surveying migrants is survey of rare element • Three approaches used in practice: • Stratified sampling using population census • Chain-referral methods such as snowball sampling • Sampling at locations where immigrants tend to cluster. Time and space/intercept sampling. => Question as to how well other methods perform compared to random sample.

  5. Our experiment • Compare three different methods in Brazil • Use the same questions and same survey firm • Wish to compare characteristics of households with and without migrants, estimate proportion of households receiving remittances and with migrants in Japan • Context: population predominately urban; crime a concern => Communications with community organizations, use where possible of Nikkei interviewers.

  6. Method 1: Stratified random sample • Restrict to Sao Paulo and Parana provinces where 80% estimated to live (still 47 million people) • Use 2000 Brazilian Census to classify households as Nikkei or not (approximately) • Census doesn’t ask ethnicity, instead asks: race (yellow), place of birth, whether individual has lived elsewhere in last 10 years • Second complication: only 10% long form of census asks this.

  7. Method 1: Stratified random sample • Steps: • Select 75 census tracts, with stratified random sampling • Carry out door-to-door listing of households in these census tracts • Return and administer questionnaire to households identified as Nikkei • Follow-up round to re-try and interview those who refused on first set of interviews, use shorter interview if refuse longer one.

  8. Listing • Letters sent out to 150 Nikkei associations with bases in the areas chosen, to explain survey • Census tract averages 301 houses • 42 interviewers in Sao Paulo and 24 in Parana used for 50 and 25 tracts. • Three attempts made to interview household; proxy-reporting used if no one at home.

  9. Summary of listing • Listed 22, 539 dwellings • Among these, detected 839 Nikkei households – 528 interviewed in person, 311 by proxy-reporting • Initial phase of interviewing interviewed 247 Nikkei households • Returned and interviewed another 156 (45 long interview, 111 short interview) • Found some of those identified as Nikkei by proxy were not • Estimate in the end: • Parana: 76% interviewed, 12% refuse, 10% absent • Sao Paulo: 45% interviewed, 31% refuse, 17% absent

  10. Comparing households in 2 phases • Main difference is that households that required the additional effort to survey are much more likely to refuse to say how much they receive in remittances. • Look similar in many other respects to those who answered first time.

  11. Snowball survey • Contacted 25 Nikkei organizations and asked each to provide 3 names as seeds. • Targeted sample of 300 households • Received 67 seed names • Two main problems: • Some of seed households refused • Many households refused to provide referrals • In the end have sample of 100 households, 60 seed households, 40 referrals.

  12. Intercept survey • Consulted with local researchers, Nikkei organizations, and Sudameris officers to select broad range of locations which Nikkei community frequents • Chose 9 fixed points, and 6 events • Sports club, metro station in Liberdade neighborhood, two Feiras (Sunday marketplaces), hospital, grocery stores, language school, outside branch of bank. • Japanese film event, large cultural festivals, Japanese food festival, Japanese art exposition, Christmas concert and music festival. • Short questionnaire – 62 questions, 7 minutes.

  13. Intercept survey • At each location, 2 interviewers used, one to count number passing through location, one to interview • Interviewers there for 129 total hours over 2 weeks (~8.5 hours/location). • Each person asked how often in past 2 weeks had frequented any of the other locations – used to reweight answers. • Only 19% had visited only one location – on average individuals had visited 3.18 out of 15 locations during the two week period. • Weight:

  14. Fourth method? • Use phone book to select individuals with minority names? (Osili with Nigerians in Chicago) • 90% have landline phone (most of rest have cellphones) • With intermarriage, 74.8% have recognizably Japanese surname • Only 24% appear in the phone book • Only 20% in phone book with Japanese name!

  15. Comparison of different methods • Hypotheses: H1: intercept and snowball households will be more closely connected to Nikkei community than random sample. H2: weighting intercept survey should bring it closer to random sample H3: snowball and intercept surveys will overestimate proportion of households with migrant experience. H4: refusal rates for questions about remittances will be higher for intercept survey, since they take place in public location.

  16. Results: Connection to Japan

  17. Results: Migration and Remittances

  18. Remittances • Proportion refusing to report how much they receive is much higher (82%) in intercept survey than in household survey (31%). • Fear of crime likely to lead to more reticence in public places.

  19. Do they give different results in regressions • Probit regressions to consider determinants of being a return migrant • Stratified and intercept survey give quite similar results • Snowball gives quite different picture – gives different answers to education and gender selectivity of migrants.

  20. Comparison of costs • Total cost per household interviewed: • Stratified survey: US$212 ($2 per dwelling listed; $80 per interview) • Snowball survey: US$100 • Intercept survey: US$30 • Recall survey lengths differ: • Stratified and Snowball: 36-page questionnaire, 1 hour to complete • Intercept: 3-page questionnaire, 7 minutes to complete

  21. Conclusions • Snowball and Intercept methods oversample individuals more closely connected to community • Intercept method does seem to give results closer to random sample when reweight data – suggests multiple location sampling important • Snowball, often seen as cheapest and easiest, may be just as difficult to carry out in some applications.

  22. Conclusions • Choice of survey method does make large and significant differences to estimates of means of interest, and to regression analysis. • Cost considerations may suggest: • Intercept survey for exploratory analysis, or where target population attends community locations • Stratified survey for representative analysis.

  23. Discussion: when are these methods applicable • Methods here easily applicable in migrant destinations (e.g. Salvadoreans in D.C.) • Are methods applicable in sending countries when group is not ethnically different? • Listing based methods clearly are • Snowball used by sociologists • Intercept – festivals, transportation hubs, money-transmitting branches, churches, social support networks, …

More Related