Migrants in Vienna tend to live in neighborhoods with others of the same nationality, and even within just a few blocks, the demographic makeup can change dramatically. This is the key finding of a pilot study conducted by researchers at the Austrian Academy of Sciences (ÖAW), which utilized a synthetic population register created with artificial intelligence (AI).
According to urban researcher Robert Musil of the ÖAW, the synthetic register—developed by Vienna’s municipal statistics office (MA 23) in collaboration with the company Mostly AI—represents “an extraordinarily innovative approach to working with microdata,” he told APA. Normally, access to such data is costly and heavily restricted due to privacy concerns. The study therefore not only explored patterns of segregation in Vienna’s population but also tested the usefulness of AI-generated data for research.
The synthetic register is, in essence, an AI-generated duplicate of Vienna’s real population register—though it contains no information traceable to actual individuals. The use of such datasets is a novel development in Austria, Musil noted. In countries like the United States, the UK, and Scandinavia, similar synthetic data are already being used in fields such as medical research and finance.
High and Granular Segregation
Researchers used the synthetic dataset to examine, among other things, people’s country of birth and nationality. The results show that people with German backgrounds are more likely to live in inner districts like the 1st, 5th, 6th, or 8th, while individuals with Turkish or former Yugoslav backgrounds are more commonly found in outer districts along the Gürtel, such as Ottakring, Rudolfsheim-Fünfhaus, or Favoriten.
What surprised the researchers was how localized the segregation was. “It’s most pronounced at the smallest scale we looked at—down to individual city blocks,” said Musil. “Usually, segregation is analyzed at broader levels like census tracts, but that misses much of the social variation and dynamics.”
These highly localized patterns can be explained in part by rental price differences within districts. For instance, rents near the Gürtel in the 7th District are significantly lower than just a few blocks away. Another reason could be so-called voluntary segregation, where people prefer to move into areas where their own community is already established.
Significant Differences Between Groups
The study also found wide variation in how segregated different nationalities are. For example, Ukrainians often live in organized accommodations, resulting in higher levels of segregation. People with Polish backgrounds, on the other hand, are more dispersed throughout the city—possibly due to greater socioeconomic diversity, allowing access to a wider range of housing options.
There are also clear differences across segments of the housing market. People of Turkish descent, who have lived in Vienna longer, often move into municipal housing. Recent arrivals such as Germans are more likely to reside in high-priced areas. The overall pattern of segregation is shaped significantly by the distribution of housing types in the city—rental apartments in historic buildings, private ownership, public housing, and non-profit housing all come with different barriers to access and are unevenly distributed geographically.
Synthetic Register Promising, But Needs Fine-Tuning
As for whether the synthetic register delivers results comparable to real data, the research team has mixed findings. While the AI-generated data aligned well with the real register across Vienna as a whole—based on a benchmark by the city’s statistics office—there were notable discrepancies at the local level. In particular, the AI data consistently underestimated levels of segregation.
“We recommend that future iterations of the synthetic register factor in elements that highlight spatial variation, like micro-level housing prices or major transport routes such as the Gürtel,” said Musil.
For city planning and urban research, adapting the synthetic register in this way could be “a huge advantage.” Better data at a finer geographic scale would help identify segregation hotspots early and guide targeted improvement efforts. “Also, just because someone yells the loudest doesn’t mean they’re facing the biggest problems,” Musil added. “This tool could help bring a more objective tone to public debate.”
