Can we identify the highest priority industry & county combinations for the Department of Labor to proactively and most efficiently investigate potential wage violators with its limited resources?
Good Stuff Happens in 1:1 Meetings: Why you need them and how to do them well
Predicting Wage Violations - Bayes Hackathon 2014
1. Predicting Wage Violations
for the Department of Labor
Xtreme Aquatic Rhinos
Aaron Keys*
Ankit Jain
Nick Handel
Samantha Fernandez
* Our weird team name was his idea
2. Billion $ Problem of Wage Theft
Sources: Working “Off the Clock” – How Employers Steal Wages
* Based off of 2009 study of an average of 4,000 low wage workers in New York City, Los Angeles and Chicago
15% of earned
annual wages lost*
ADVERSELY AFFECTS THE POOR
IMPACTS THEIR SPENDING ON VITAL SERVICES
DIFFICULT TO IDENTIFY AND ENFORCE
Low Investigation Likelihood
1 in 1,250 chance
for a garment factory
$2,634 lost of a $17,616
annual salary
~1,000 Dept. of Labor
investigators
3. Real Life Examples
Noe
Line Cook
Hilda
Garment Worker
Felipe
Car Wash Worker
Source: Working “Off the Clock” – How Employers Steal Wages
4. Can we
identify the highest priority
industry & county combos
for the DoL to proactively
and most efficiently
investigate potential wage
violators with
its limited resources?
5. Prediction Methodology
• Data Set:
– Wage & Hour Compliance Action Data from 1989 – 2013 featuring
~188,000 complaint investigations
• Methodology:
– Calculate Net Monetary Impact (Total Backwages + Total Civil
Monetary Penalties) for each wage investigation
– For each particular investigation evenly distribute the NMI across the
years the NMI was incurred
– Split data by State + Industry
– NMI split by Year for each Company
– Accumulate by State + Industry
– For a given State + Industry, use Time Series Analysis to Forecast NMI for
next year (2015)
– Distribute the predicted NMI of a State to Counties by weighting % of
working population
– Map Forecasted NMI
7. Key Predictions:
Top Areas for Investigation
• Construction in CA and FL
• Waste Management in NY, CA, MD and FL
• Transportation and Warehousing in CA
• Manufacturing in CA
• Health Care and Social Assistance in FL
• Oil & Gas Extraction in TX
8. Potential Future Analyses
• Future analysis on the employer level
• Explore additional data to better
understand the nature of the risks
– For example is the risk related to underpayment,
medical screening violations, discrimination or
other?
• When violations occur
– For example, do wage violations increase during
recessions or after natural disasters?