by Leslie D’Monte
The so-called Internet Protocol or IP address typically finds itself in the news when authorities use it to track down cybercafes from which terrorists could have sent potentially harmful emails. IP addresses can be spoofed and camouflaged with anonymiser tools yet every device which sends an e-mail can be located at least at the country level by this internationally standardised code.
Two researchers–Emilio Zagheni from the Max Planck Institute for Demographic Research in Germany and Ingmar Weber from Yahoo! Research in Barcelona–made good use of this IP feature.
Analysing the countries derived from IP addresses for a set of messages sent by 43 million anonymous Yahoo! account holders over three years–between September 2009 and June 2011–the two researches have concluded that the Internet can provide migration data that is much more accurate than the official one provided by governments, which “is outdated and inconsistent” since emigrants tend not to register after they move to a new country or do so very late, according to a press statement.
Where estimates of demographic flows exist, they are often outdated and largely inconsistent, reasoned Zagheni, adding: “Global internet data does not have these drawbacks.”
In addition to the date and geographical origin of each message, they compiled the self-reported birthday and gender of the sender. When a person started sending e-mail from a new location permanently, it was assumed that he or she had changed residence. This way they were able to calculate rates of migration from and to almost every country in the world.
Only anonymized data was used, so identifying individuals was impossible and no information about the recipients, the subject, or content of a message was accessed. The findings have now been published in the ACM Web Science Conference Proceedings.
The results not only are a proof of concept. They also reveal international migration characteristics never seen before. For the US, Zagheni and Weber were able to produce the first curve of emigration by age and sex ever. “In the U.S. many statistics are collected about people who move into the country, but there is no system that keeps track of people who move out,” said Zagheni.
The potential of the e-mail statistics goes far beyond calculating gross country profiles. For instance, the researchers also looked into Mexico-US cross-border mobility. The data reveals how strongly both countries are demographically integrated: most people who moved from Mexico to the United States either spent time in the USA before emigrating north, or went back to visit Mexico soon after moving to the United States. Those in their 30s have the highest rate of mobility across the Mexico-US border, while the least mobile are those 50 and older.
Some groups, like the elderly, use e-mail less or not at all and are thus underrepresented. But the researchers managed to calculate adjustment factors for such groups by gauging their e-mail data against migration numbers from European countries, where official data is fairly reliable.
“What we addressed so far is only the tip of the iceberg,” said Zagheni. With further fine-tuning of the adjustment factors and mining more digital data like Twitter messages, more difficult questions could be tackled. For instance one could keep track of the short and long-term mobility patterns before and after a crisis like that of the Japanese Fukushima reactors.
Unquestionably, digital records give demographers the chance to gain a more accurate picture of population dynamics in regions they can so far only guess about, said Zagheni. “This research has the most potential in developing countries, where the Internet spreads much faster than registration programs develop.”
“Our preliminary results indicate that the profile of out-migration from India is qualitatively similar to the one of other major countries, with a peak of out-migration for individuals who are around 25 years old. The results are still preliminary and we cannot share the numeric estimates yet,” Zagheni said in an email response.
