URBOPP

When working with longitudinal databases aggregated at a fine spatial administrative level, such as census tracts, researchers face a recurring methodological challenge: the boundaries of these units frequently change over time due to demographic growth, administrative reorganization, mergers, or splits. These shifting geographies complicate the analysis of temporal dynamics, particularly when models include spatial fixed effects to control for time-invariant local characteristics. If spatial units are not stable, fixed effects no longer correspond to consistent geographic entities.

To address this issue, we develop a procedure for constructing pseudo–census tracts that remain stable over time. These units are defined as the smallest spatial entities that can be consistently tracked throughout the period of interest. By harmonizing changing boundaries into a set of temporally stable units, researchers can estimate models with spatial fixed effects that credibly capture unobserved, time-invariant local factors, while preserving the finest possible spatial resolution.

The Python tool presented here was initially developed for studying the evolution of urban segregation in France within the ANR project Urban Neighborhoods and Economic Opportunities (ANR-23-CE26-0001, PI: Gabrielle Fack & Miren Lafourcade) and was also used in Garrouste & Lafourcade (2025) "Place-Based Policies: A Path to Opportunity or a Mark of Stigma for Targeted Neighborhoods?". The code is designed as an open-source resource and can be used by other scholars studying the dynamics of spatially aggregated outcomes over time. The input data can also be reused for other research projects in the French context.

In the French statistical system, an IRIS (Îlots Regroupés pour l’Information Statistique) is the smallest administrative geographic unit used by the National Institute of Statistics and Economic Studies (Insee) to disseminate census data in municipalities with more than 10,000 inhabitants. An IRIS typically contains around 2,000 residents and serves as a fundamental building block for analyzing demographic, social, and economic patterns at a fine spatial scale in France.

The program provided below constructs, from these IRIS units, temporally stable pseudo–census tracts (hereafter referred to as IRISX). By default, the tool generates IRISX units for the period 2009–2019, although the time span can be adjusted to any period of interest, subject to certain methodological constraints.

The outputs include:

• IRIS_historiques_IRISX.xlsx: a historical correspondence table mapping each IRIS code (column IRIS) to its associated IRISX identifier (column irisx_id), along with the year of correspondence (annee_corresp_).

• IRISX20092019.shp: a shapefile aggregating IRIS boundaries for a reference year (2019 by default) at the IRISX level. The column irisx_id provides the unique IRISX identifier, CODE_IRIS_ indicates the IRIS code in the reference year (here 2019), and geometry contains the polygon coordinates defining the IRISX boundaries.

References

• Adélaïde, L., Stempfelet, M., Babut, C., Bulté, A., Ehrhart, B., & Jaccy, N. (2023). HistorIRIS: table de passage des IRIS de 1999 à 2022. Technical report, Commissariat Général au Développement durable (CGDD).

• Behrens, K., & Martin, J. Concording large datasets over time: The C3 method. Technical document.

• Garrouste, M., & Lafourcade, M. (2026). Place-based policies: A path to opportunity or a mark of stigma for targeted neighborhoods? Conditionnaly accepted at the Journal of the European Economic Association.