Demographic Segregation in Online Information Ecosystems
Burak Ozturan, David Lazer
Working Paper
|
September 16, 2026
Concern about online echo chambers rests on the premise that the internet sorts people by politics. Research on online ideological segregation has accordingly measured partisanship almost to the exclusion of other social divides, and has measured it mostly within news, which is a small share of online activity (Allen et al., 2020; Muise et al., 2022). Whether partisan segregation is distinctive has therefore rarely been tested against demographic segregation on equal terms. Using a 1.6-million-user Twitter/X panel linked to voter file records (July 2018 to May 2023), we measure age, gender, partisanship and geography with a Theil index, and race more briefly alongside them. First, partisanship is not the dominant division of online attention. When measured across Democrats, Republicans and Independents, partisan segregation over general sharing behaviour is roughly a third of demographic segregation, and age separates users more sharply than party even within news, the content category where partisan segregation is strongest and where the existing literature has looked almost exclusively. Second, segregation is built at the article level rather than the outlet level. People who differ on any social divide we measure converge on the same outlets and diverge on which stories they select from them, which places segregation downstream of the outlet choices that most audience-level research observes. Geography is the exception and segregates at the outlet level, as local news would predict. Third, the five-year decline in demographic segregation is produced by different mechanisms for different divides. Gender’s decline reflects almost entirely who left the platform, partisanship’s reflects how people changed, and age’s is an even mixture, so parallel aggregate trends can rest on unrelated causes. Online attention is segregated along social lines that are largely not political, at a level of granularity that outlet-based measures cannot observe. Existing estimates of partisan echo chambers describe a narrower and less distinctive phenomenon than the polarisation literature has assumed.
citation
Ozturan, B., & Lazer, D. (2026). Demographic Segregation in Online Information Ecosystems [Preprint]. SocArXiv. https://doi.org/10.31235/osf.io/zydup_v1
topics
Online Behavior
Artificial Intelligence
Artificial Intelligence
Artificial Intelligence
related work

Measuring Agency-Amplifying AI Literacy: The HuskyAI Coaching Platform for Workforce Readiness
July 2026
HuskyAI is a web platform on which students develop AI knowledge by working through authentic
Artificial Intelligence
Artificial Intelligence
Working Paper


Measuring Agency-Amplifying AI Literacy: The HuskyAI Coaching Platform for Workforce Readiness
July 2026
HuskyAI is a web platform on which students develop AI knowledge by working through authentic
Artificial Intelligence
Artificial Intelligence
Working Paper


Measuring Agency-Amplifying AI Literacy: The HuskyAI Coaching Platform for Workforce Readiness
July 2026
HuskyAI is a web platform on which students develop AI knowledge by working through authentic
Artificial Intelligence
Artificial Intelligence
Working Paper

.png)
.jpg)