Self-assessed vs. statistical evidence of racial discrimination: the case of indigenous Australians

ABSTRACT

This paper provides new insights on the labour market discrimination faced by indigenous Australians one of the most disadvantaged indigenous populations in developed countries. Combining two large, nationally representative datasets, we decompose the employment gap between indigenous and non-indigenous populations as of 2014–2015, and show that differences in characteristics between the two groups account for only 43% of the employment gap for females, and 23% of the gap for males. We then demonstrate that statistical measures are positively related to discrimination reports of females and negatively related to discrimination reports of males. Our findings underscore the importance of improving transparency in employment processes for addressing the issue of disadvantage of racial minorities.

KEYWORDS:

• Racial discrimination
• employment
• Australia

JEL CLASSIFICATION:

• J15
• J21
• J71

Acknowledgments

This project received support from the United Nations University World Institute for Development Economics Research (UNU-WIDER). We use unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Australian Government Department of Social Services (DSS) and is managed by the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute). The findings and views reported in the paper, however, are those of the authors and should not be attributed to DSS or the Melbourne Institute. We are grateful to Mike Dockery, Clas Weber, participants of the Australian Conference of Economists 2017, and two anonymous referees for valuable comments. All errors are those of the authors.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Kalb et al. (2014) attribute their ability to explain a larger proportion of the gap in labour market outcomes between indigenous and non-indigenous populations to controlling for previously unobserved differences in characteristics of the two groups, such as health and household composition. However, it is not implausible that some differences in these characteristics could be either a direct or an indirect outcome of discriminatory practices in the labour market. Such feedback effects may mean that labour market discrimination is under-estimated in their study (Grimshaw and Rubery 2002).

² A somewhat related literature looks at the relationship between reported beliefs and objective measures applied to contexts such as corruption (Olken 2009) and inequality (Gimpelson and Treisman 2018).

³ It should be explicitly highlighted that the statistical evidence of discrimination considered here is very different to the concept of statistical discrimination whereby employers may consciously discriminate because they lack information about the productivity of individuals and ‘statistically discriminate’ against them on the basis of easily observable characteristics (Phelps 1972; Aigner and Cain 1977).

⁴ This measure focuses on the difference in the returns ${\beta }^A-{\beta }^B$ to individuals’ observed characteristics $X_i$ ignoring the returns to individuals’ unobserved characteristics.

⁵ The DataLab operates as a restricted remote access system where authorized users can view and analyse unit record information, while the files remain in the secure environment. All analytical outputs are checked by the Australian Bureau of Statistics before being provided to researchers.

⁶ In the baseline sample of indigenous Australians, those who are not in the labour force comprise almost 81% of individuals who are not employed.

⁷ While these household variables are not included in the baseline models due to potential endogeneity concerns, they feature in extended robustness analysis.

⁸ These include: NSW major cities (omitted), NSW inner regional, NSW outer regional, VIC major cities, VIC inner/outer regional, QLD major cities, QLD inner regional, QLD outer regional, SA non-remote, WA major cities, WA inner/outer regional, TAS non-remote.

⁹ It is possible that the perceptions of discrimination among individuals who are employed but who have not been looking for a job capture discrimination experienced at work (e.g. when applying for a promotion or a pay rise) more so than discrimination that affects the probability of holding a job. Since these discrimination experiences can be different in nature to those that statistical measure of discrimination generated based on an employment equation might capture, in robustness checks, we analyse the link between perceived and statistical measures of discrimination when this group is omitted from the sample.

¹⁰ Using Schwarz’s Bayesian Information Criterion (Schwarz 1978), Kalb et al. (2014) test the performance of ordered logit against less restrictive multinomial logit, showing that the use of the former in this case is justified.

Additional information

Funding

This work was supported by the United Nations University World Institute for Development Economics Research (UNU-WIDER) (contract ID: 605UU-0000000000000000000001352).