“Brexodus?”: A Longitudinal Approach to Investigate European Migration to the UK Using the Facebook Advertising Platform

Agnese Vitali, University of Southampton
Emilio Zagheni, University of Washington
Ingmar Weber, Qatar Computing Research Institute
Jakub Bijak, University of Southampton
Francesco Rampazzo, University of Southampton

International migration to the UK has become a hot topic both in research and in the media. The British Office of National Statistics (ONS), in their August quarterly report, have highlighted a decrease in net migration of EU citizens. Detailed data on migrants’ characteristics are much needed for producing accurate statistics and informing policy. We want to use online advertising data to provide a clearer picture of these migrants. This paper has two aims. First, it aims to create the first longitudinal geo-located dataset from Facebook’s Advertising Platform through a weekly collection from mid-September 2017 onwards for two years. Secondly, it aims to give a weekly picture of immigration from the EU following the EU referendum. The dataset will be stratified by characteristics such as age, gender, country of origin, education level, and employment field. The analysis will complement traditional data sources provided by the ONS and IPS, with which we will compare with our estimates. Then, a time series analysis approach will be used to analyse the data. It will be important to observe the differences by country regarding the trends, cycles, and seasonal changes in the data. With this approach, it will be possible to make projections of future trends in migration using Facebook data. The preliminary results from Facebook show proportions close to the ONS estimates for 2016. This is the starting point for this research.



Measuringinternational migration across countries has traditionally proved challengingfor demographers. In many countries migration registers do not exist, makingthe estimation of international stocks and flows considerably more challengingthan with other demographic events.

In order to informmigration policies of the calibre of the Brexit debate, it is crucial to haveaccess to valid sources of data on international migration. In this paper, we investigate whether the digital traces that individuals leave onFacebook can be used to estimate migration flows of people migrating to the UK,using the Facebook Advertising Platform. Importantly, the Facebook AdvertisingPlatform allows us to distinguish between three types of Facebook users: thosewho reside in the UK, those who have recently been in the UK and those who arevisitors. In addition, our estimates will count for each region of the UK(England, Wales and Scotland and Northern Ireland), the number of internationalmigrants by sex, age group, educational level attained, educational enrolment,field of employment, and country of origin. We will limit our attention to migrants/movementsfrom the European sending countries. Such estimates will complement existing,but inaccurate, official estimates of migration flows.

Consideringthe UK as a case study for the special novelty of its time, this research hastwo main aims. Firstly, we want to collect data from Facebook’s AdvertisingPlatform at weekly intervals, creating an aggregated time series of the migrantpopulation stocks in the UK. Secondly, we aim to provide a picture of theEuropean immigrants in the UK one year after the EU Referendum took place,breaking down the stocks by age, gender, country of origin, education level,and employment field. To sum up, this paper will build a socio-demographicprofile of European migrants in the UK through their Facebook profiles

Facebook’sAdvertising Platform

Facebook’s Advertising Platform[1]permits advertisers to create specific advertisements to be selectively shownon certain Facebook profiles, matching criteria specified by the advertiser.Facebook provides an aggregated estimate of the immigrant population within acountry through the category “Expat(*)”. We will usethis category for estimating the number of European immigrants in the UK and wewill then stratify for a selection of variables (age, gender, education, andemployment field). It is important tohighlight that Facebook gathers these estimates based on information from sitesother than just facebook.com, such as sites that use Facebook Like or sharefunctionality, which means a connection to facebook.com.

Variablesfor Analysis

The variables forthe analysis are divided into:

1.    Territories: UK, England, Scotland, Wales,Northern Ireland.

2.    Expat: All people in the UK, All withoutExpats, All Expats, Austria, Belgium, Finland, France, Germany, Greece,Ireland, Italy, the Netherlands, Portugal, Spain, Switzerland, Estonia, Poland,Hungary, Latvia, Romania, Non-European, and Asia.

3.    Gender: All, Male, Female.

4.    Age: 15+, 15-19, 20-24, 25-29, 30-34,35-39, 40-49, 50-59, 60-64, 65+.

5.    Education Level: All, Graduated, InCollege, In High School, In Grad School, Unspecified, No Degree, High School.

6.    Employment Field:

- All;

-       Group 1:Administration, Business and Financial, Government Employees, Legal,Management, and Retail and Sales;

-       Group 2:Architecture and Engineers, Computers and Mathematicians, IT and technical;

-       Group 3:Health Care and Medical, Nurses, Professional Care;

-       Group 4:Arts, and entertainment, Education and library, Life, Physical, and SocialSciences;

-       Group 5:Cleaning and Maintenance, Construction and Extraction, Food Preparation andServices, Installation and Repair, Production, Transportation and Moving,Farming and Fishing.


With thedownloaded data, we are creating a weekly time series of migrant stocks to theUK divided by age, education level, and employment field. Our objective is tounderstand whether this data is close to reality, to infer selection bias bynationalities, and at the same time summarize the data into consistentinformation about migration trends to the UK.

In our analysis,the first step will be to compare our data with the migration estimates fromthe ONS (Census data) and the IPS. After having validated the data, we willthen use a time series analysis approach to analyse the data. It will beimportant to observe the differences by country regarding the trends, cycles,and seasonal changes in the data. With this approach, it will be possible tomake projections of future trends in migration using Facebook data.


The two graphs below show the proportions of European immigrants livingin the UK. The graph on the left portrays the estimate made by the ONS for2016. The graph on the right represents the proportion of European Facebookusers living in the UK at September 25th, 2017. We can see that theproportion of Romanians is similar, though Facebook reports a higher proportionof Romanians and Italians compared to the ONS estimates. This is the startingpoint for our research.


Presented in Session 1070: Data and Methods