Our Programmes



Sign up to our newsletter.

Follow ODI

Why we need better data on mobile populations in cities – and how to get it

Written by Gemma Hennessey

With nearly all migrants heading for cities, and a (growing) majority of the world’s population living in urban areas, migration to cities can’t be ignored.

Migration is thought to contribute to around half of urban population growth in Asia and Africa – and up to 80% in China and Thailand. These populations are fluid and constantly on the move.

Frequently missing from administrative records and living in informal settlements, migrants are often hidden or invisible in official statistics. In many cases, this means city populations are badly underestimated. In Karachi, for instance, the most recent census put the population at 14.9 million, a figure contested as severely undercounting the true size of a megacity thought to be the seventh-largest in the world.

This is more than a statistical inconvenience – it has critical ramifications for how governments understand – and respond to – growth, change and poverty in their cities.

Why do we need better data on migrants in cities?

Urban migrants are often particularly vulnerable: more likely to be doing precarious work, living in informal settlements in overcrowded or sub-standard housing and lacking access to basic services. But lack of data means we don’t know enough about their specific needs and vulnerabilities.

It also means that migrants are likely to be excluded or underserved by utility providers or governments. For example, WASH service coverage is typically based on official data, and so automatically excludes migrants living in informal settings.

Better data on a city’s population is essential for better urban planning, the wise use of scarce resources and sustainable urban development.

Why are migrants missing from data?

Existing data collection methods struggle to generate representative data on mobile populations in cities. Censuses are tied to a residence and cover people formally registered as permanent residents, excluding temporary or undocumented migrants.

People living in informal settlements – which most urban migrants do – are also usually excluded. Censuses are conducted infrequently, and rapid urban growth means that, by the time the data’s published, it’s already out of date.

The absence or inaccuracy of census (and other administrative) data presents a huge challenge to other surveys as well. Household surveys mostly use censuses as sampling frames to draw random samples and to generate representative data. But, they won’t reflect the actual population since groups like migrants are left out.

One survey in Beijing using an alternative sampling technique found that 45% of the sample – mostly internal migrants – would have been excluded in standard data collection efforts.

Mobile populations tend to be difficult to count, with the potential for high non-response rates. Migrants may be hard for enumerators to access if they’re living in precarious settings that are not mapped, or that enumerators don’t feel safe or comfortable in.

Undocumented migrants may try to conceal their identity for fear of detection by the authorities, or they may be unreachable during the day as they tend to work long hours and are not at home when enumerators call.

How can we collect better data on mobile populations in cities?

Many surveys have given up altogether on trying to collect representative data on mobile populations. Instead, they use other, non-random, sampling techniques, such as snowballing. However, the data generated isn’t generalisable and can’t tell us about the whole city population.

More recently, researchers are starting to use GPS technology to generate representative data on city populations, including mobile and other hidden populations.

ODI is a partner in a new project – Quantifying cities for sustainable development which will use GPS-based methods to create a new sampling approach. We will overcome the challenges involved in generating cheap, timely and accurate representative data by:

  • Using the urban landscape – as it actually is – as a sampling frame to generate representative data of the whole city.
  • Interviewing those present, instead of only formally registered residents, and using a combination of household and street-intercept surveys to overcome challenges in accessing mobile populations.
  • Developing a new, replicable, low-cost methodology for collecting data on urban populations, using statistical reconstruction techniques to keep survey lengths to a minimum.

Our new sampling approach will be tested and refined in Larkana, Pakistan, in 2020 and we will be sharing lessons here as we go along.