Using Twitter to Evaluate the Perception of Service Delivery in Data-Poor Environments

Abstract

Evaluating service delivery needs in data-poor environments presents a particularly difficult problem for policymakers. The places where the need for social services are most acute are often the very same places where assessing policy interventions is the most challenging. This paper uses Twitter data to gain insights into service delivery needs in a data-poor environment. Specifically, it examines the development priorities of citizens in the north- western region of Pakistan between 2007 and 2020, using natural language processing techniques (NLP) and sentiment analysis of 9.5 million tweets generated by 20,000 unique Twitter users. The analysis reveals that service delivery priorities in this context are centered on access to education, healthcare, food, and clean water. The findings provide baseline data for future on-the-ground research and development initiatives. In addition, the methodology used in this paper demonstrates both current resources and areas in need of future work in the use of NLP techniques in analyzing social media data in other contexts.

Keywords: Social media, Twitter, natural language processing, sentiment analysis, service delivery

JEL Classification: H70, C81, O12