Gender differences in under-reporting hiring discrimination in Korea: A machine learning approach

Abstract

Objectives: To examine the gender difference in under-reporting hiring discrimination by building prediction models for the workers who responded 'Not applicable (NA)' to a question about hiring discrimination, although they were eligible to answer. Methods: Using the data from 3,576 waged workers in the 7th wave (2004) of the Korea Labor and Income Panel Study, we trained and tested nine machine learning algorithms using 'Yes' or 'No' responses regarding the lifetime experience of hiring discrimination. We then applied the best-performing model to estimate prevalence of experiencing hiring discrimination among those who answered NA. Under-reporting hiring discrimination was calculated by comparing the prevalence of hiring discrimination between 'Yes' or 'No' group and 'NA' group. Results: Based on the prediction from the random forest model, we found that 58.8% of the 'NA' group were predicted to experience hiring discrimination, while 19.7% of the 'Yes' or 'No' group reported hiring discrimination. Furthermore, we found that female workers were more likely to under-report hiring discrimination than male workers. Conclusion: This study introduces a methodological strategy for epidemiologic studies to address the under-reporting of discrimination by applying machine learning algorithms.