A large meta-analysis of rental-market field experiments has found persistent discrimination across several protected or socially salient characteristics, while also showing why headline averages can mislead. The study, released as IZA Discussion Paper 18788 in July 2026, synthesises 475 experimental estimates from 31 countries published between 2002 and 2024.
The researchers compare discrimination linked to ethno-racial origin, gender, sexual orientation, disability and social origin. Vrije Universiteit Brussel describes the broader evidence base as spanning 114 studies and more than 500,000 rental applications or contacts.
Field experiments are particularly useful for discrimination research because they can hold most candidate characteristics constant while systematically changing one signal. Two otherwise comparable prospective tenants might differ in a name associated with a particular ethnic background, for example, allowing researchers to compare reply rates. This design does not capture every stage of a rental decision, but it directly measures unequal treatment at an early gatekeeping point.
In the random-effects pooled estimates, the paper reports substantial discrimination across all five grounds, with the largest penalties for ethno-racial origin, social origin and disability. Men and same-sex couples show smaller average penalties. Within ethno-racial comparisons, applicants perceived as Arab, North African or Turkish face the strongest disadvantages in the pooled evidence.
Geography also matters. Contrary to an assumption that North American rental markets would necessarily show larger gaps, the authors find higher discrimination in Europe than North America for ethno-racial origin and sexual orientation. They do not find a consistent decline over time for most grounds, with social origin as the main exception.
The most scientifically important caveat comes from publication-bias analysis. For several discrimination grounds, correcting for small-study publication bias attenuates the pooled estimate toward no discrimination. Smaller studies with large or statistically striking effects can be more likely to appear in the literature, which can inflate a naive average. The correction does not invalidate the evidence of discriminatory treatment; it changes how confidently researchers can interpret a single pooled magnitude.
Heterogeneity is similarly central. Effects vary by target group, country, market structure and research design. Real-estate agents discriminate less than private landlords for ethno-racial origin in the pooled data, but that difference does not generalise cleanly across all other grounds. Mode-of-contact effects are mixed rather than uniform.
The paper's publication status also deserves precision. The IZA Discussion Paper Series is designed to circulate new research before it appears in refereed journals. That makes the study accessible and citable, but it is not the same as saying the paper has already completed journal peer review.
For future research, the meta-analysis points toward a more demanding question than whether discrimination exists. Researchers need to identify which mechanisms create the observed gaps and which interventions reduce them. Standardised application systems, enforcement, audit testing and professionalisation of rental management can all be studied experimentally.
The paper therefore offers two findings at once. Discriminatory treatment remains measurable across a large international evidence base, especially for some groups. But the size of that treatment cannot responsibly be compressed into one universal percentage without accounting for publication bias and the substantial variation across places and designs.