The cultural similarity between countries and international migration patterns can be measured quite reliably using Facebook data, a new study reports.

“Cultural hotspot” isn’t the first thing that pops into mind when thinking about social media for most of us. However, new research from the Max Planck Institute for Demographic Research in Rostock, Germany shows that data from Facebook can be used to gauge cultural closeness between countries, and overall migration trends.

And the way to do it is to track ads for food and drink on the platform.

We are what we eat

“[A] few years ago, after reading a work of a colleague using data from the Facebook Advertising Platform, I was surprised to find how much information we share online and how much these social media platforms know about us,” said Carolina Coimbra Vieira, a Ph.D. student in the Laboratory of Digital and Computational Demography at the Max Planck institute and lead author of the research, in an email for ZME Science.

“After that, I decided to work with this social media data to propose new ways of answering old questions related to society. In this specific case, I wanted to propose a measure of cultural similarity between countries using data regarding Facebook users’ food and drink preferences.”

For the study, the team developed a new approach that uses Facebook data to gauge cultural similarity between countries, by making associations between immigration patterns and the overall preference for food and drink across various locations.

They employed this approach as migrants have a very important role to play in shaping cultural similarities between countries. However, they explain, it’s hard to study their influence directly, in part because it is hard to ‘measure’ culture reliably. The traditional way of gauging culture comes in the form of surveys, but these have several drawbacks such as cost, the chances of bias in question construction, and difficulties in applying them to a large sample of countries.

The team chose to draw on previous findings that show food and drink preferences may be a proxy for cultural similarities between countries, and build a new analytical method based on this knowledge. They drew on Facebook’s top 50 food and drink preferences in various countries — as captured by the Facebook Advertising Platform — in order to see what people in different areas liked to dine on.

“This platform allows marketers and researchers to obtain an estimate of the number of Facebook monthly active users for a proposed advertisement that matches the given input criteria based on a list of demographic attributes, such as age, gender, home location, and interests, that can be customized by the advertiser,” Vieira explained for ZME Science. “Because we focus on food and drink as cultural markers, we selected the interests classified by Facebook as related to food and drink. We selected the top 50 most popular foods and drinks in each one of the sixteen countries we analyzed to construct a vector indicator of each country in terms of these foods and drinks to finally measure the cultural similarity between them.”

In order to validate their findings, the team applied the method to 16 countries. They report that food and drink interests, as reflected by Facebook ads, generally align with documented immigration patterns. Preferences for foreign food and drink align with domestic preferences in the countries from which most immigrants came. On the other hand, countries that tend to have few immigrants also showed lower preferences for foreign foods and drinks, and were interested in a narrower range of such products more consistently.

The team cites the example of the asymmetry between Mexico and the U.S. as an example of the validity of their model. The top 50 foods and drinks from Mexico are more popular in the U.S. than the top 50 U.S. foods and drinks are in Mexico, they explain, aligning well with the greater degree of immigration coming from Mexico into the U.S. than the other way around.

All in all, the findings strongly suggest that immigrants help shape the culture of various countries. In the future, the team hopes to expand their methodology to include other areas of preference beyond food and drink, and see whether these align with known immigration patterns.

“The food and drink preferences shared by Facebook users from two different countries might indicate a high immigrant population from one country living in the other. In our results we observed that immigration is associated with higher cultural similarity between countries. For example, there are a lot of immigrants from Argentina living in Spain and our measure showed that one of the most similar countries to Spain is Argentina. This means that foods and drinks popular between Facebook users in Argentina are also really popular in Spain,” she adds.

“The most surprising aspect of this study is the methodology and more precisely, the data we used to study culture. Differently from surveys, our methodology is timely, [cost-effective], and easily scalable because it uses passively-collected information internationally available on Facebook.”

Overall, the researchers say, this study suggests that immigrants indeed help shape the culture of their destination country. Future research could refine the new method outlined in this study or repurpose it to examine and compare other interests beyond food and drink.

“I would like to see our proposed measure of cultural similarity being used in different contexts, such as to predict migration. For instance, it would be interesting to use our measure of cultural similarity to answer the question: Do the migrants prefer to migrate to a country culturally similar to their origin country?” Vieira concludes in her email.”More generally, I hope our work contributes to increasing the development of research using social media data as an alternative to complement more traditional data sources to study society.”

The paper “The interplay of migration and cultural similarity between countries: Evidence from Facebook data on food and drink interests” has been published in the journal PLoS ONE.