The language of discrimination: assessing attention discrimination by Hungarian local governments

Abstract In our study we assess the responsiveness of Hungarian local governments to requests for information by Roma and non-Roma clients, relying on a nationwide correspondence study. Our paper has both methodological substantive relevance. The novelty is that treat discrimination as classification problem what extent emails written clients can be distinguished, which in turn serves metric general. We show it possible detect textual data an automated way without human coding, machine learning (ML) may features coders not recognize. To best knowledge, first attempt using ML techniques. From point view, focuses linguistic algorithm detects behind discrimination. models worked significantly better compared random (the accuracy was 61%), confirming differential treatment clients. most important predictors showed answers sent ostensibly are only shorter, but their tone less polite more reserved, supporting idea attention discrimination, line with results Bartos et al. (2016). A higher level detectable against male senders, smaller settlements. Also, interpreted digital sense Edelman Luca (2014) use this term.