Trails of Data: Three Cases for Collecting Web Information for Social Science Research

Abstract

As the availability of online data grows rapidly, researchers are confronted with a pressing question: How should social scientists collect Internet data for research? This study focuses on one of the most commonly used data collection techniques: web scraping. Going beyond canned approaches by leveraging a general framework of data communication, this study illustrates how online information can be systematically queried and fetched for reproducible research. To generalize our approaches, we additionally explore the variations in site security and architecture that analysts may encounter during the scraping process before they are given access to the desired data. The approaches we introduce do not rely on any proprietary software and can be easily implemented on any computing platform with programming languages such as Python or R. The methodological discussion in this study is meant to be applicable to current web-based research efforts. We include three examples with complete Python implementation. We also present an integrated workflow that enables researchers to produce analytical data sets that are traceable and thus verifiable for analysis or replication. Lastly, options related to the validity and efficiency of data are discussed, and we highlight the ongoing debate surrounding the ethics of online data collection, ultimately advocating for the fair use of online data.