An approach to investigate European migration to the UK using the Facebook advertising platform Francesco Rampazzo F.Rampazzo@soton.ac.uk @chiccorampazzo University of Southampton
Migration 2
Migration Data in the UK International Passenger Survey Annual Population Survey Running since 1961, All major airports, sea routes, and train stations, 700,000-800,000 interviews every year of which are 4,000 long-term migrant. Running since 2004, Household Survey, Sample size 320,000, Uses data from Labour Force Survey (LFS). Link: Link: https://www.ons.gov.uk/surveys/informationforhouseholdsandindividuals/ https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/empl householdandindividualsurveys/internationalpassengersurveyips oymentandemployeetypes/methodologies/annualpopulationsurveyapsqmi 3
Migration Data Sources 1. National population censuses 2. Sample surveys 3. Administrative sources Limitations Definitions (e.g. short term long term migrant, unregistered, irregular ) Quality and Comparability Timely information Costs 4
Can we estimate migration with new data sources? 5
IOM, (2018). Data Bulletin: Big Data and Migration.Link: https://publications.iom.int/system/files/pdf/issue_5_big_data_and_migration.pdf Blank, G., Graham, M. and Calvino, C. (2018) "Local Geographies of Digital Inequality", SOCIAL SCIENCE COMPUTER REVIEW. 36 (1) 82-102. Link to the figure: http://geography.oii.ox.ac.uk/?page=internet-use-inbritain 6
The Internet and Demography 1. For improving existing statistics 2. For adding new dimensions 3. For information on hard to reach population Emilio Zagheni s talk at the Plenary Session 4 on Data innovation and big data for migration at the International Forum on Migration Statistics (2018): https://oecdtv.webtvsolution.com/4495/or/plenary_session_4_data_innovation_and_big_data_for_migration_.html 7
Can new sources of data improve migration statistics? Blumenstock, 2012. Inferring patterns of internal migration from mobile phone call records: evidence from Rwanda. Hawelka et al., 2013. Geo-located Twitter as proxy for global mobility patterns. Swier et al., 2015. Using geolocated Twitter traces to infer residence and mobility. Fiorio et al., 2017. Short-term Mobility and Long-term Migration. Zagheni et al., 2017. Leveraging Facebook s Advertising Platform to monitor stocks of Migrants. 8
Hard to Reach Population State et al., 2014. Migration of Professionals to the U.S. Evidence from LinkedIn Data. Pötzschke and Braun, 2017. Migrant Sampling Using Facebook Advertisments: A Case Study of Polish Migrants in Four European Countries. 9
My Contribution Use aggregated and anonymized Facebook s Advertising Data for estimating European Migration to the UK and create a profile of these migrants. Challenge to understand bias and representativeness of this source of data. 10
Facebook s Advertising Platform 11
12
13
Limitations Source: https://www.theguardian.com/technology/2017/sep/07/facebook-claims-it-can-reach-more-people-than-actually-existin-uk-us-and-other-countries 14
Dataset Countries: Great Britain, England, Wales, Northern Ireland, and Scotland Sex Country of Origin: All Facebook profiles in the UK, All British, All Expats, Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Switzerland, Estonia, Poland, Hungary, Latvia, Romania. Age groups: 15+, 15-19, 20-24, 25-29, 30-24, 35-39, 40-49,-50-59,60-54, 65+. Education: All levels of education, Graduated, In college, In High School, In Grad School, Unspecified, No degree, High School. 15
The dataset is made up by 38,000,000 Facebook anonymized, aggregated group of individual in the UK, of which 31,000,000 are British, and 7,000,000 are migrants. 16
European migrants: ONS and Facebook Estimates ONS: Annual Population Survey data for 2017; Facebook: European Migrants estimates for October 2017 age 15+. 17
Facebook distribution of Users within the UK 18
Facebook Age Profile in the UK 19
Facebook Education Levels of the Users in the UK 20
Conclusions This data shows high potential for looking at the characteristics of migrants. It is necessary to understand the bias and representativeness of this data. Next steps: out of sample validation, combination of traditional and new data sets. 21
Thanks! And thanks to my supervisors Agnese Vitali, Jakub Bijak, Ingmar Weber, and Emilio Zagheni! @chiccorampazzo 22