The website, Google Trends, enables us to see how the interest in a keyword evolves over time. The authors extracted this time series for 67 keywords semantically related with migration in three languages, i.e. Spanish, English and French. Internet users who are considering leaving their own country will often search for keywords such as the name of the desired country of destination or combine it with terms such as immigration, passport, or salary. Such internet searches are thus driven by migration intentions and typically lead to increased migration flows.

In November 2016, Google Trends recorded a sudden peak in search intensity for the keyword “USA”. This increase in interest coincided with the date of election of the current president of the United States, Donald Trump. It can be assumed that this partly reflected a certain concern among Mexicans about immigration opportunities in anticipation of Trump’s clampdown on immigration. In the face of the fear that entering the United States would become more complicated in the following months, Trump’s election victory may have led many Mexican citizens to seek information on migration to the US. Official data from 2017 from the OECD show that there was also a peak in immigration from Mexico to the USA during that year.

The main origins of migratory movements to Spain between 2004 and 2015 among countries that officially or administratively use one of the three languages studied – Spanish, English or French – were Argentina, Colombia, Ecuador, Morocco, the United Kingdom and Venezuela. 

Predictions based on the model including internet searches follow the actual migration flows quite closely. In all cases, we can observe how the dotted line coincides with marked decreases in immigration to Spain. The key driver behind the decrease in immigration was the economic crisis, which began in 2008 and lasted for several years. The predictions including internet searches also reflect well the slight increase in arrivals towards the end of the study period which corresponds to the rebound of the Spanish economy.

Migration has become one of the most important and most disputed issues in developed and developing countries alike. Traditional prediction models tend to be static and often perform poorly when it comes to detecting turning points in migration flows. Adding an index that reflects Google searches improves the accuracy of these methods.

Prediction based on keywords uses a free source of data that can help to facilitate migration research, especially in developing countries. Furthermore, the approach holds the promise of allowing short-term predictions. Information on migration flows tends to arrive with considerable delays that can be several years long. This new approach is especially promising with regard to delivering better results when studying countries or regions that have a relatively high internet penetration.

This graph shows total immigration flows into OECD countries from Algeria, Australia, Chile, Pakistan, Spain and Togo, as well as their predictions based on the benchmark and augmented model. The graph shows that the Google Trends approach (dotted line) not only works for predictions from one country to another, but also from one country to many. The solid line corresponds to official data from the OECD.

Georeferenced internet search data offer new and practically infinite opportunities to measure and predict human behaviour. Innovative applications beyond the context of migration are feasible, provided that demand for specific online information is indicative of subsequent human behaviour. To help identify turning points in migration, predictions of migration patterns can be further fine-tuned to specific corridors or regions where sudden changes are to be expected.

The approach based on Google Trends could be used to predict migration flows close to real time, ahead of official data release. In this way, it can be useful also for giving fast political responses to emergency situations.

BÖHME, M. H., A. GRÖGER and T. STÖHR (2020): "Searching for a better life: Predicting international migration with online search keywords", Journal of Development Economics, 142, 102347, https://doi.org/10.1016/j.jdeveco.2019.04.002.