How Google Trends can help us plan and respond better to future pandemics

Investigating how Americans responded to the socioeconomic uncertainties of COVID-19

Surgo Ventures
8 min readJun 1, 2021
A person conducts a Google search on a laptop
Photo by Benjamin Dada on Unsplash

Since the beginning of 2020, there’s been extensive reporting on how the COVID-19 pandemic has upended life in the US. Understanding how Americans have responded is important to shape the public health and policy initiatives that are essential to turning the tide of the pandemic.

What do we know about the questions Americans asked during the first wave of the pandemic — about the virus, about behavioral changes like social distancing and wearing face masks, and about the huge economic shocks wrought by the COVID-19 outbreak? How did these questions change over time, and what differences were there among the states?

In a new paper published in JMIR, Surgo Ventures analyzed data from Google Trends, which shows the relative popularity of search terms used on Google. We looked at 38 search terms connected to ‘coronavirus’ (the most common term used to search for COVID-19 in Q1 2020) and tracked their relative popularity week by week, from January 1 through April 15, 2020 — looking at trends and patterns at both national and state level.

Our findings have implications that could help the government respond more quickly and effectively to people’s needs and worries — whether for this pandemic, or a future one.

What we found:

  1. High demand for information was driven by increasing searches for COVID-19 from multiple news outlets, regardless of the ideological leaning of the news source
  2. Changes in information seeking around major concerns like masks and social distancing often happened well in advance of action by the federal government
  3. Actual unemployment claims were highly correlated with the previous week’s Google searches for unemployment
  4. Alongside the increase in searches for information on COVID-19 care, there was a decrease in searches for other health behaviors (urgent care, doctor’s appointments, health insurance/Medicare/Medicaid)
  5. Search terms were more popular in some regions than in others, suggesting that concerns vary across the country.
  6. States most vulnerable economically or in terms of the social safety net (such as Mississippi, Louisiana, Alabama, Arkansas, and Kentucky) had fewer searches for information on both COVID-19 news and and interventions like social distancing

Google searches capture the health, economic, and social impact of, and response to, COVID-19

Researchers have used Google Trends as proxies for changes in human behavior during major events. These searches reflect what people are in need of, thinking about, or planning for: in March 2020, for example, COVID-related searches were five times more prevalent than generic searches for “weather.”

We used Google Trends to investigate how Americans responded to the uncertainties posed by the COVID-19 pandemic — the evolving information about the virus, changes in how we behave, and the economic shocks. We asked four key questions in the context of Google Trends and COVID-19:

  • How has information seeking changed over time?
  • How does information seeking vary between regions and states?
  • Do states have particular and distinct patterns in information seeking?
  • Does search data correlate with — and even precede — real-life events?

Google provides the relative search values (RSVs) for each query on a scale from 0 to 100, representing a normalized value. We generated data sets at the national level (covering Jan 1, 2016 — April 15, 2020) to compare trends over time, and at the state level (covering Jan 1, 2020 — April 15, 2020) to focus on the first wave of the pandemic. We grouped 38 search terms related to COVID-19 into six themes: Social & Travel; Care Seeking; Government Programs; Health Programs; News & Influence; and Outlook & Concerns.

This allowed analysis of changes and patterns in information-seeking over time and in different parts of the country.

Finding 1: High demand for information as searches for COVID-19 news spiked, regardless of the ideological leaning of the news source.

High demand for information corresponded with increasing searches for “coronavirus” linked to news sources. This was true regardless of the ideological leaning of the news source (as gauged by a Pew Center survey). This highlights the importance of public health authorities working with the media to ensure that information is correct.

Comparison of Google Trends nonpharmaceutical intervention-related queries and actual government and public health nonpharmaceutical interventions between March 1 and April 15, 2020
Fig 1: National trends for Google Trends nonpharmaceutical intervention–related queries compared to actual government and public health nonpharmaceutical interventions at the US federal level between March 1 and April 15, 2020. CDC: US Centers for Disease Control and Prevention; EU: European Union; SD: social distancing; WHO: World Health Organization.

Finding 2: Changes in information seeking on major concerns often happened well in advance of action by the federal government.

[Fig 1] shows how changes in information seeking often preceded action by the federal government and international health organizations such as the World Health Organization (WHO). For example, the popularity of “social distancing” searches increased rapidly from March 8 — that’s five days before the government’s national emergency declaration, and eight days before social distancing guidelines were introduced. Likewise, “How to make a coronavirus mask” was popular by March 23–12 days before CDC advisory promoting masks for the general public.

Data like this can show public health authorities what interventions people are aware of and might accept — and when the best time is to present them to the public and/or take policy actions. This allows them to catch the optimal window for action in a fast-evolving situation that requires a finger on the pulse of public sentiment.

Finding 3: Actual unemployment claims were highly correlated with the previous week’s Google searches for unemployment.

Google Trends were highly positively correlated with unemployment claims. There was a correlation of 0.96 between the previous week’s Google Trends RSV for unemployment applications and the actual weekly initial unemployment claims normalized to 0–100 (seasonally adjusted as recommended by the Bureau of Labor Statistics to eliminate seasonal spikes and enable easy detection of nonseasonal anomalies such as COVID-19–related spikes). Tracking these searches could be used to alert the government about coming surges in claims, as has been pointed out by other researchers, e.g. Hyungyoung Choi and Hal Varian, 2009.

Trend lines of actual unemployment claims compared to Google searches for unemployment applications.
Fig 2. Trend line for Google searches for unemployment applications and actual initial weekly unemployment claims normalized from 0–100 over time (with seasonal adjustment).

Finding 4: While searches for COVID-19 ‘care’ spiked, searches for other health behaviors dropped.

The increase in searches for information on coronavirus care was paralleled by a decrease in searches related to other health behaviors, such as urgent care, doctor’s appointment, health insurance/Medicare/Medicaid.

Bar chart showing the change in monthly average relative searches for health concerns from the beginning of January and the end of March, 2020.
Fig 3: Bar chart showing the change in monthly average relative searches for health concerns from the beginning of January and the end of March, 2020.

This supports evidence that during pandemics, patients with other serious prospective illnesses reduce their health-seeking behaviors in fear of catching the disease or because the hospitals have less capacity to deal with other ailments outside of the pandemic.

The decline in searches could be an early-warning indicator for authorities to target vulnerable populations with serious underlying health risks with messages about the importance of continuing to seek care.

Finding 5: Search terms were more popular in some regions than in others, suggesting that concerns vary across the country.

Google Trends can represent relative search patterns across states, allowing us to form a unique “search persona” for each state that indicates the primary concerns of its populace. For example, in March/April 2020 searches for “Bars and restaurants nearby” were relatively more popular in Florida, Texas, Arizona, and Missouri, while “Stimulus check” were relatively more popular in Arkansas, Kentucky, West Virginia, and Mississippi. This information could help state governments understand the needs and concerns of their populations as they emerge.

Google search comparing one salient query from each search category during March/April, 2020.
Fig4. Google search interest across all states for one salient query in each category, measured by Relative Search Value (RSV) during March/April 2020.

Finding 6: States most vulnerable economically or in terms of the social safety net had fewer searches for information on both COVID-19 news and social distancing.

To illustrate each state’s relative search persona, we clustered search trends to discover natural patterns. As we show in[Fig 5], states furthest to the right (Mississippi, Louisiana, Alabama, Arkansas, and Kentucky) had fewer searches for COVID-19 information from news sources, or for information on social distancing, while also showing more searches indicating economic or social vulnerability (such as the receipt of Medicaid, stimulus checks, or food stamps).

Scatterplot of search terms clustered into regional personas, with arrow directions correlated with the x and y-axis, explaining the variation in data.
Fig 5: Scatterplot of state principal component analysis loadings and scores for the first two components, with the top queries shown as arrow vectors. Each arrow represents the relative weight of each query, and the direction indicates the points to the states that most exhibit this search pattern. The arrow direction measures correlation; arrows in the same direction are highly positively correlated, while divergent arrows in opposite directions are highly negatively correlated. Component 1 (x-axis) and Component 2 (y-axis) explain 25% and 18% of the variation in the data, respectively.

Our analysis suggests that these states needed a multifaceted stimulus to combat the pandemic. First, an economic stimulus would help address concerns about making ends meet. Second, an information stimulus may have helped to combat disinterest and noncompliance.

This information could be used to increase awareness of social distancing and its benefits, while targeting and prioritizing resources to support these states and help people who are otherwise unable to adopt behavior change to protect themselves from the pandemic.

How can public health authorities use these Google Trends insights?

Our research holds important lessons for both federal and state government leaders and public health officials, in a fast-evolving situation that requires a finger on the pulse of public sentiment:

  1. People of all political persuasions were hungry for COVID-19 information early in the pandemic, so it’s incumbent on policymakers and public health officials to align critical public health information across a diverse range of news sources for correct and consistent messaging.
  2. Tracking the topics people are seeking information about can help authorities understand their concerns, and also what interventions they are aware of and might be ready to accept — often earlier than the government might anticipate.
  3. In a future health and even economic crises, spikes in searches for information on unemployment benefits could be used as a sentinel to alert the government about coming surges in claims.
  4. Noticing declines in searches for information about other diseases could warn health authorities to encourage people to continue to seek care, especially in areas with serious underlying health risks.
  5. Regional variations in interest in COVID-19 and social distancing could help prioritize information and resources to states that are most interested in economic and social support. For example, the government can ramp up social distance awareness campaigns and increase the amount of stimulus checks in these states.

This is Part 1 of a three-part series. Read Part 2 and Part 3.

This work was made possible by everyone at Surgo Ventures, including but not limited to (in alphabetical order): Daisy Chung, Aaron Dibner-Dunlap, Vincent Huang, Hannah Kemp, Tich Mangono, Sema K. Sgaier, and Peter Smittenaar. Thank you for additional graphic design support from Katie Armstrong.

--

--

Surgo Ventures

We use all the tools available from behavioral science, data science, and artificial intelligence to unlock solutions that will save and improve people’s lives.