Predicting Job-Hopping Motive of Candidates Using Answers to Open-ended Interview Questions

TL;DR

This study demonstrates that language analysis of interview responses can predict job-hopping motives, with Glove embeddings showing the strongest correlation, and also links language use to personality traits.

Contribution

It introduces a novel approach using text representation methods to predict job-hopping motives from interview responses, outperforming traditional methods.

Findings

Glove embeddings achieved the highest correlation (r=0.35) with job-hopping motives.

Language use correlates with personality traits, notably Openness (r=0.25).

Open-vocabulary text representations outperform closed-vocabulary approaches.

Abstract

A significant proportion of voluntary employee turnover includes people who frequently move from job to job, known as job-hopping. Our work shows that language used in responding to interview questions on past behaviour and situational judgement is predictive of job-hopping motive as measured by the Job-Hopping Motives (JHM) Scale. The study is based on responses from over 45,000 job applicants who completed an online chat interview and self-rated themselves on JHM Scale. Five different methods of text representation were evaluated, namely four open-vocabulary approaches (TF-IDF, LDA, Glove word embeddings and Doc2Vec document embeddings) and one closed-vocabulary approach (LIWC). The Glove embeddings provided the best results with a correlation of r = 0.35 between sequences of words used and the JHM Scale. Further analysis also showed a correlation of r = 0.25 between language-based job-hopping motive and the personality trait Openness to experience and a correlation of r = -0.09 with the trait Agreeableness.