1. Using Internet Advertising Data for Studying International Migration
Ingmar Weber
November 30, 2017
European Commission – Joint Research Centre, Ispra

2. Facebook’s Audience Estimates

3. Audience Estimate = Digital User Census
Facebook, Twitter and LinkedIn all provide “how many users match criteria X” data
Typically used for ad campaign planning
FB supports ad campaign targeting for
Age and gender (self-reported)
Home country (inferred from “home city”)
Education level (inferred from “education history”)
Interests (inferred from “Likes”)
Mostly US: income, ethnic affinity, political leaning

4. Emilio Zagheni, Ingmar Weber, Krishna Gummadi
Leveraging Facebook’s Advertising Platform to Monitor Stocks of Migrants
Published in Population and Development Review

5. Expats Across US States
2014
FB underestimates
FB overestimates
2017

6. Expats Across Countries
2015
FB underestimates
FB overestimates
2017

7. Age-Specific Selection Biases

8. Bias Reduction via Model-Fitting
z = age-gender group
i = country of birth
j = US state of residence
Mean out-of-sample absolute percentage error 37%, down from 56% without origin-age bias correction Adj. R^2 = .70 Does not use GDP, language, internet penetration, …

9. Measuring Assimilation of Interests in Syrian Refugees in Germany
Antoine Dubois, Kiran Garimella, Ingmar Weber, Emilio Zagheni
Measuring Assimilation of Interests in Syrian Refugees in Germany
Ongoing work

10. Do Refugees Share German Interests?
Germans in DEU
FB Interests:
Football (90%)
Max Planck (70%)
Sauerkraut (40%) …
Arabs in DEU
FB Interests: ?
Arabs in MENA
FB Interests:
Quran (80%)
Ibn Al-Haytham (60%)
Falafel (60%) …
What interests to consider? Everybody likes “Music” and “Technology”.
How to interpret the score? High/low compared to European migrants?

11. Pros and Cons of FB Ad Audience Data
+ International reach (~214M of ~244M expats) + Supports disaggregation (age, gender, …) + Very timely data (~ weekly updates) + Supports measuring “soft” attributes - Selection bias (can be corrected, if understood) - Built on black box (but possible to validate) - Relies on FB’s existence (but idea more general) - Fake accounts (unclear if this affects models)

12. Related Ongoing Projects
Studying European Migrants to the UK
Francesco Rampazzo, Jakub Bijak, Agnese Vitali, Ingmar Weber, Emilio Zagheni
Using FB Advertising Data to Track the Global Digital Gender Gap
Masoomali Fatehkia, Ridhi Kashyap, Ingmar Weber
Male Fertility: What Can FB Advertising Data Tell Us?
Francesco Rampazzo, Francesco Billari, Maria Rita Testa, Ingmar Weber, Emilio Zagheni
Gender Skill Gaps on LinkedIn Across the UK
Karri Harranko, Kiran Garimella, Ingmar Weber, Emilio Zagheni

13. Key Collaborator Emilio Zagheni Associate Prof. at Univ. of Washington
Collaborating since 2011
9 joint peer-reviewed publications so far
Always happy to explore collaborations!

14. Thanks. Ask me about data viz for Doha during break http://fb-doha