Predicting Volunteers' Decisions to Stay in or Quit an NGO Using Neural Networks

Abstract

[EN] This paper uses non-traditional approaches to predict why volunteers remain in or quit a non-governmental organisation position. A questionnaire featuring 55 predictors was conducted via an online survey mechanism from March to May 2021. A total of 250 responses were received. The subsequent data analysis compared logistic regression and artificial neural network results, using machine-learning interpreters to explain the features which determined decisions. The results indicate greater accuracy for neural networks. According to the logistic regression results, intrinsic motivation, volunteering through an NGO and the age of volunteers influenced the intention to remain. Moreover, NGOs that offered online volunteering opportunities during the COVID-19 pandemic had higher rates of intention to remain. However, the neural network analysis, performed using the Local Interpretable Model-Agnostic Explanations (LIME) method, indicated the need to consider different predictors to those identified by the logistic regression. The LIME method also enables the individualisation of the explanations of predictions, indicating the importance of considering the role of volunteers' feelings in both quit and remain decisions, which is something that is not provided by traditional methods such as logistic regression. Furthermore, the LIME approach demonstrates that NGOs must address both volunteer management and experience to retain volunteers. Nonetheless, volunteer management is more critical to stop volunteers quitting, suggesting that volunteer integration is crucial.