Analyzing the Factors Shaping Teleworking Decisions

Teleworking has become the new standard because of its sustained widespread adoption in recent times in compared to the period before covid pandemic. Thus, identifying the factors correlated with workers’ preferences for teleworking is important for transport policymakers. While previous studies mainly focused on socio-demographic factors, the study broadens the analysis by incorporating weather conditions and contextual variables. Specifically, it investigates the impact of 18 different factors on the propensity to telecommute in Quebec City, Canada. Due to the severe class imbalance, with only 7% of respondents reporting telecommuting in the pre-COVID Origin-Destination survey, the dataset was counterbalanced using under-sampling. Ten robust ensemble and machine learning algorithms were employed to predict whether the workers in Quebec City would choose teleworking or in-person work. Findings indicate that extreme gradient boosting outperforms all models, achieving approximately 94.23% accuracy on a test dataset and an F1 score close to 94.21%. The Shapley additive explanation (SHAP) is employed to capture the determinants of teleworking, and the results suggest that distance to work, age, temperature, household income, walkability of home location, and car per adult in the household have the most substantial relative influence on teleworking decisions. Further, Partial Dependency Plots are utilized to illustrate the direction influence of variables on teleworking decisions, and their outcomes show that the higher distances, age of 40–49, negative temperatures, higher household incomes, walk scores of over 90, and no access to cars are associated with a higher probability of teleworking preferences.