Worker Well‐Being and Quitting: What Can We Learn From Social Media Data?

ABSTRACT We examine the potential to measure worker well‐being using social media data. We utilize more than 345 million tweets from more than 4.1 million users from 2019 to 2022 in 323 cities across 46 US states. We construct a state‐month panel and demonstrate that our work‐related sentiment data predict quits. Our dynamic panel estimates show that a 1 standard deviation decrease in worker well‐being is associated with a 0.41 percentage point increase in the subsequent month's quit rate. We show that social media data can be utilized to provide near‐real‐time information on worker well‐being in a way not possible with traditional survey‐based approaches.