Measurement Error in Issue Salience: The Potential of Google Trends Data

The proliferation of online data sourced from either unknown populations or non-probability-based sampling frames provides both opportunities and pitfalls. Recent studies have tentatively endorsed the capacity of one such source – Google Trends data – to measure issue salience in the United Kingdom, Spain and the United States. Google Trends data report the prevalence of search terms entered to Google on a weekly basis, and are freely available via Google. Using data from a time series of Australian opinion surveys – the ANUpoll series of public opinion – this Paper has two aims. First, a deductive examination of the content validity of Google Trends data in the Australian case against probability-based survey data, following existing studies explores the potential for replacing conventional measures of issue salience with incidental search engine data. Second, the study uses search term data to identify measurement error inherent in conventional measures of political issue salience. Taking a grounded approach provides insight into the measurement validity and potential information loss when verbatim responses to the question of ‘most important issue’ are coded into pre-existing categories by interviewers. It is hypothesised that the deductive and inductive analytical techniques will indicate that Google search data cannot meaningfully predict survey measures of issue salience, but that this failure can be at least partially attributed to measurement error in the coding of issue salience responses. This study has substantial implications for the external validity of attitudinal survey research, refinements to measurement and understanding of issue salience, and the use of incidental and other non-probability-based social data.