U.S. Occupational Wage Analysis

Glassdoor: Decode U.S. wage inequality using hypothesis testing and regression analysis. Python, Scipy, Statsmodels, Matplotlib.

Glassdoor's economic research team benchmarks the compensation data reported on its platform against public labor market statistics. This project works with the U.S. Bureau of Labor Statistics Occupational Employment and Wage Statistics (OEWS) release as that public reference point.

Assignment Overview

The goal of this project is to perform a multi-faceted analysis of occupational wage data across the United States. The dataset includes information on employment totals, average wages, occupation categories, and geographic distribution by state and region.

Your task is to analyze wage disparities, employment patterns, and statistical relationships across different job sectors and states.

Objectives:

• Investigate wage differences across geographies and job types
• Test hypotheses using inferential statistics (T-tests, Z-tests, F-tests)
• Build regression models to understand employment-wage relationships
• Visualize wage distributions, top/bottom states, and occupation trends

Data Description

Scope: Annual occupational wage and employment data
Geography: 50 states + territories
Fields:

• OCC_TITLE: Occupation name
• TOT_EMP: Total employment
• A_MEAN: Average annual wage
• PRIM_STATE: State abbreviation
• O_GROUP: Occupation category (Major, Total, Detailed)
• Additional wage percentiles and employment estimates

Analytical Techniques

The analysis covered the following components:

Key Questions

1. Which U.S. states have the highest and lowest annual mean wages across all occupations?
2. Which occupation groups provide the best wage-to-employment efficiency?
3. Within California, how does employment vary among major occupation categories?
4. Is there a significant wage difference between management and production occupations? (Two-sample T-test)
5. Does higher employment predict lower wages? (Simple linear regression)
6. Is there a statistical correlation between total employment and mean wage? (Pearson correlation)
7. Do tech occupations (SOC code 15) earn significantly more than all others? (Z-test)
8. Do healthcare and education occupations show significantly different wage variances? (F-test)

Business Implications

• For HR & Policy: Understand where pay is uneven and which roles carry regional importance.
• For Analysts: Identify which categories provide deeper value when segmented.
• For Decision-Makers: Align compensation frameworks with actual labor dynamics and statistical evidence.

Takeaway:
Asking the right questions reveals hidden layers in occupational wage data. Statistical and visual tools enable smarter labor planning and compensation design.

Deliverables

• Python notebook (data cleaning, plots, statistical testing)
• Visualizations for each analytical method
• Markdown summaries of key questions and interpretations

Estimated Completion Time: 6–8 hours

Tips & Considerations

• Use errors="coerce" with pd.to_numeric() to handle wage conversion
• Handle missing values before plotting or aggregating
• Use seaborn or matplotlib for visual clarity
• Group occupations carefully using .groupby() and .agg()
• Anchor your questions in real-world labor policy or HR needs

FAQ

Q: Can I add more statistical tests?
A: Yes — include any analysis that enhances insights.

Q: Is predictive modeling required?
A: No, the goal is exploratory and inferential analysis.

Q: Can occupation groups be reclassified?
A: Yes, but explain any re-mapping logic in the notebook.

Completing this project demonstrates your ability to extract questions from raw data and use statistical methods to inform labor market insights.