TL;DR
This paper uses Google Trends data to show that the announcement of the first COVID-19 case in a US state significantly boosts public information-seeking behavior, primarily for basic virus information and symptoms. However, this heightened attention is short-lived and doesn't immediately translate into searches for community-level policies or personal health strategies.
What Problem Does This Paper Address?
As the COVID-19 pandemic unfolded, governments faced the challenge of effectively communicating with the public to encourage necessary behavioral changes and policy adherence. A key question was how the public would react to initial news of local outbreaks. Specifically, would the announcement of the first local case trigger widespread information-seeking, and if so, what kind of information would people seek? Understanding this initial public response is crucial for designing effective communication strategies, especially in the early stages of an epidemic when traditional polling methods are too slow.
Methods
The researchers employed an "event study" framework, a common econometric technique used to measure the impact of a specific event on a time series. In this case, the "event" was the announcement of the first COVID-19 case in a US state. They analyzed daily Google search data for various COVID-19 related terms across all 50 US states and Washington D.C. The core of their analysis involved estimating the change in search volume immediately before and after these announcement events, controlling for state-specific trends and national daily trends. They used Poisson models to analyze the log-expected counts of search queries.
Datasets
The primary dataset used was Google Trends data for various COVID-19 related search terms (e.g., "coronavirus," "coronavirus symptoms," "hand sanitizer," "quarantine," "testing near me," "coronavirus hoax"). This data was collected for all 50 US states and Washington D.C. from January 1, 2020, to March 18, 2020, using a restricted-access Google Health Trends API. The timing of the first COVID-19 case announcements in each state was gathered from media reports. The authors also made their state news data and analysis code publicly available on GitHub.
Key Insights and Ideas
- Immediate Information Scramble: The announcement of a first local COVID-19 case acts as a powerful trigger for public information-seeking.
- Focus on the Basics: People's immediate information needs revolve around understanding the virus itself and its symptoms, rather than immediate policy responses or personal protective measures.
- Short-Lived Attention Span: The surge in information-seeking is a temporary phenomenon, fading within a week or two, even as more significant policy measures are implemented.
- Policy vs. Personal: Public interest in broader community-level interventions (like quarantine or school closures) or specific personal health strategies (like masks or grocery delivery) is not immediately sparked by the first case announcement.
- No Impact on Conspiracy Searches: Official news of a local case did not appear to significantly reduce searches related to "coronavirus hoaxes" or "conspiracies."
Key Results
- Searches for "coronavirus" increased by approximately 36% (95% CI: 27% to 44%) on the day immediately following the first case announcement in a state.
- Searches for "coronavirus symptoms" increased by 52%.
- Searches for "hand sanitizer" increased by 35% and remained elevated for the observable period.
- However, searches for "coronavirus treatment," "testing near me," "quarantine," "isolation," "coronavirus conspiracy," and "coronavirus hoax" did not show a significant immediate increase following the first case announcement.
- The elevated attention to "coronavirus" searches returned to baseline levels in less than one to two weeks.
Limitations and Future Work
The authors acknowledge several limitations:
- Google Trends Representation: Google Trends data only captures the search behavior of a specific segment of the population, not the entire public.
- Short Time Window: While the short study window mitigates issues like long-term data drift, it means the study primarily captures initial reactions and not longer-term evolving patterns.
- Confounding Factors: Unobservable local events or anticipatory behavior could potentially influence search trends, though the researchers attempted to mitigate this through sensitivity analyses.
Future work should focus on tracking evolving patterns of public information-seeking in response to subsequent news and policy announcements. The authors suggest that policymakers could leverage the brief window of heightened public attention following focal events (like first cases or deaths) to deliver clear and actionable advice.
Summary updated