Convincing Evidence? a Meta-Analysis on Field Experiments on Ethnic Discrimination in the Housing Market

Tuesday, 12 July 2016: 09:00
Location: Hörsaal 26 (Main Building)
Oral Presentation
Andreas SCHNECK, LMU Munich, Germany
Katrin AUSPURG, LMU Munich, Germany
There exists a long research tradition of audit and correspondence studies in the US and European countries stating that ethnic discrimination limits the scope of available housing options. Despite being the gold-standard method to examine discrimination, experiments might nevertheless be biased: While the experimental set-up provides obviously a high internal validity, the generalizability and external validity of results is questionable. Threats to external validity stem for instance from small (convenience) samples that often cover only specific (and very small) geographical areas or suffer from reactive data-collection strategies in audit studies (see the critique by Heckman). Furthermore only specific sets of stimuli are in most studies implemented that may be confounded (e.g. names with social class).

So far, effects of these design features, are, however, mostly speculative. Meta-analyses allow exploring the heterogeneity of different experiments identifying systematic patterns (that might be caused by different methodological settings) using the high statistical power of pooled data analyses. But surprisingly so far not any meta-analysis on field experiments in the housing market exists.

We present a meta-analysis that sums up experimental evidence from nearly 50 publications covering a timeframe from 1973 to 2014 containing more than 60 experiments and 800 estimates of effect sizes of (conditions of) ethnic discrimination in the housing market. Is there a robust effect of discrimination in the housing market controlling for time as well as geographic location? How do findings covary with methodological settings, such as using audits or correspondence tests, small or large sample sizes? In our meta-analyses we will demonstrate several robustness checks, including recent tools to detect publication bias that have be found to overcome limitations of classical meta-regression approaches.