Using Google Search engine data to predict current migration

2021 December 15

Predicting the current and future migration trends has become of greater importance to policymakers and the research community for planning public funds, labour supply and demand, infrastructure and well-being, as well as evaluating policies. However, traditional migration data sources are not always accurate and up to date. The latest FUME article explores a relatively new approach in migration studies – “now-casting” of the current migration trends from Romania to the United Kingdom with new forms of data.

Romanians in the UK: two stories

According to the UK’s data from censuses and the Labour Force Survey, the number of Romanians living in the UK has increased almost 57 times between 2001-2018. Due to large emigration, Romania is now considered one of the fastest shrinking countries in the whole world.

However, such significant change in the country has not been visible in official records. One of the reasons is that at least prior to Brexit, Romanian citizens were not legally required to register their departure to any formal authority.

Given the challenges emerging from the traditional data sources to study the migration patterns, in the past years, demographers are focusing on alternative methods, for example, by using publicly available data from social media, or search engines.

Google Trends analysis in migration studies

The use of timely reported new forms of data methods enables researchers to detect sudden changes in migration trends. This could be especially useful for short-term forecasting (or now-casting) of migration.

According to Andreea Avramescu and Arkadiusz Wiśniowski from the University of Manchester, using Google Trends data to study migration patterns is not as common as in the sectors of finance or tourism. They argue that by combining search keywords related to employment and education, as well as using the lexical approach, one can improve the very short-term migration forecasts.

Find out more about the study here: