Using machine learning to translate applicant work history into predictors of performance and turnover

Area of Science:

• Industrial-Organizational Psychology
• Machine Learning Applications
• Human Resources Management

Background:

• Traditional work history assessment lacks systematic methods for predicting future job success.
• Resumes and applications are common screening tools, but their predictive validity is debated.
• Developing interpretable measures from work history data is crucial for effective applicant screening.

Purpose of the Study:

• To apply machine learning to job application data for creating interpretable measures of work experience.
• To predict future work outcomes using novel metrics derived from past employment.
• To evaluate the model's effectiveness in improving selection processes and reducing adverse impact.

Main Methods:

• Utilized machine learning techniques on job application data, including job descriptions and reasons for job changes.
• Developed interpretable measures: work experience relevance, tenure history, and turnover history (avoiding bad jobs, approaching better jobs).
• Empirically tested the model on a large longitudinal sample of 16,071 public school teaching applicants.

Main Results:

• Work experience relevance and a history of seeking better jobs positively predicted work outcomes.
• A history of avoiding bad jobs was associated with negative work outcomes.
• The model demonstrated potential to enhance selection quality and reduce adverse impact compared to conventional methods.

Conclusions:

• Machine learning offers a systematic approach to translating work history into predictive measures of job performance.
• Specific patterns in work history, such as relevance and proactive job seeking, are key indicators of future success.
• This approach can improve hiring decisions in fields like education, leading to better performance and reduced turnover.