2019 Workshop CEPR-AMSE on social mobility [segregation]


We bring together mobile phone and geocoded tax data on the three biggest French cities to shed a new light on segregation and spatial frictions. Urban segregation is generally measured through the glance of residential segregation. However, mobility can bring together people from different areas and shape the spatial distribution of income. The infra-day approach we propose takes into account the effect of individual mobility on within-day segregation dynamics. Population flows decomposed by income groups are also used to estimate the heterogeneity in spatial frictions for people on the extremes of income distribution.

That paper proposes an innovative methodology to study segregation dynamics at fine spatial and temporal granularity for both low- and high-income groups. We build infra-day segregation indexes using individual geocoded position records from anonymized mobile phone data. We adopt a Monte-Carlo procedure to estimate phone users’ likelihood of belonging to low- or high-income groups and construct segregation indexes by taking into account co-presence at 500x500 meters cells. We propose robustness checks and compare results with residential segregation indexes derived from tax data. To estimate the effect that distance has on the interactions between neighborhoods, we use a gravity model from a large scale origin-destination matrix. We account for the biases arising from flows selection by building a zero-inflated count model.

As a consequence of mobility, income groups are less concentrated during daytime than during nighttime. Residential segregation represents the acme of segregation: dissimilarity index drops down by 50 percent between its pinnacle (midnight to 4pm) and daytime stable level. Diffusion of low and high-income people out of their residential area during the day makes segregation drop at city level. However, distance plays a key role to limit population flows: the further two neighborhood are, the less likely population inflows and outflows between them will happen. We find out that Marseille is characterized by inequalities in spatial frictions between income groups while differences between income groups are more marginal in Lyon and Paris. Low-income people that live in city center in Marseille face, for instance, high distance cost making them more isolated than other populations.


Communication at the workshop on social mobility organized by CEPR-AMSE-Banque de France

Estimated evolution of low-income population in Paris city:

You can find the presentation here:

This browser does not support PDFs embedding. Please download the PDF to view it: Download PDF.

Lino Galiana
Data Scientist

I am a statistician in the Department of Economic Studies at the French national statistical institute, Insee. I study Big Data and computational methods related to microeconometric and data science fields.