Predicting Wage Theft by Locale and Employer #6
Description
Background: Wage theft is estimated to cost US workers $50 billion annually. Wage theft is also widespread - affecting 17% of all low wage workers. The amount recovered by a Department of Labor investigation is often just a fraction of what is stolen from workers (e.g. from 2017-2020 only $3 billion was recovered) https://www.epi.org/publication/wage-theft-2021/
Goal: Use previous reports and convictions of Wage Theft, as well as social media data (reddit, instagram, X, etc where people complain about wage theft) to train a model that can take as an input some social media feed, and produce as an output a likelihood estimation of wage theft violations by a particular employer.
Stretch goal: Use prediction, and actual case settlement amount to test the economic viability of Liz Ford's proposal to create Wage Recovery Funds (WRF) "...a WRF is a pool of funds housed at a government agency or community organization. Employees who are victims of wage theft could approach the WRF; if the WRF accepts the case, it would make the worker whole upfront— before the employer has paid—and then take assignment of the worker’s claim. The WRF would then pursue wages, interest, and penalties through administrative enforcement proceedings. Money recovered from employers would then be returned to the fund to support the next case"
Data:
- All completed investigations and wage amounts since 2005 https://enforcedata.dol.gov/views/data_summary.php
- pushshift.io for reddit and gram data
Previous / Related work:
- Johnson, T., Forest Peterson, M., Myers, R. S. T., & Fischer, M. (2018). Predicting, Analyzing, and Educating on Wage Theft with Machine Learning Tools. Center for Integrated Facility Engineering Technical Report, 229. https://stacks.stanford.edu/file/druid:mx396wr3611/TR229.pdf
Related to #4