Racial Discrimination Analysis in US Job Market - Statistical Inference
MY ROLES
TIMELINE
2020
TOOLS USED
Statistical Inference, Hypothesis Testing, Python, Pandas, Social Science, Employment Analysis, Discrimination Research, Data Visualization, P-value Interpretation
DESCRIPTION
This project analyzed racial discrimination in the US job market by examining how identical resumes with either Black-sounding or White-sounding names affected callback rates from employers. Using statistical hypothesis testing, the analysis found strong evidence of racial bias in hiring practices, with White-sounding names receiving significantly more callbacks than identical resumes with Black-sounding names.
OUTCOMES
- Demonstrated statistically significant difference in callback rates between identical resumes with different racial identifiers
- Quantified the extent of racial discrimination in hiring with near-zero p-value
- Provided evidence-based recommendations for addressing unconscious bias in hiring practices
FEATURES
Permutation testing, difference of means analysis, statistical significance testing, demographic trend visualization, contextual interpretation of findings
CHALLENGES
Isolating race as a factor when multiple variables could affect hiring decisions. Designing a robust statistical approach to test the hypothesis. Interpreting results in a socially meaningful context while acknowledging limitations of the analysis.
APPROACH
Analyzed a dataset where resumes were randomly assigned Black-sounding or White-sounding names to isolate the impact of perceived race. Conducted permutation testing to determine statistical significance. Found strong evidence (p-value near 0) rejecting the null hypothesis that race has no effect on callback rates. Acknowledged limitations by noting that while race was a significant factor, the analysis couldn't determine if it was the most important factor without examining other variables.