The Psychology of Job Loss: Using Social Media Data to Characterize and Predict Unemployment

WebSci |

Published by ACM

Using data from social media, we study the relationship between the macroeconomic shock of employment instability and psychological well-being. We analyze more than 1.2B Twitter posts from over 230,000 U.S. users who either lost or gained a new job over a period spanning five years, from 2010 to 2015. First we quantify the magnitude and length of effects of  job loss/gain on psychological variables such as anxiety, sadness, and anger. We then define a behavioral macroeconomic model that leverages these changes in psychological state to predict levels of unemployment in the U.S. Our results show that our psychological well-being measures are leading indicators, predicting economic indices weeks in advance with higher accuracy than baseline models. Taken together, these findings suggest that by capturing the human experience of a shock like job loss, social media data can augment current economic models to generate a better understanding of the overall causes and consequences of macroeconomic performance.