The Evolution of the Pandemic Through the Lens of Google Searches
A Global Dashboard for Monitoring COVID-19
When we fall ill,
many of us turn to Google
to understand our symptoms and treatment options.
Using data from February 1 - Jaunary 31, 2021, this dashboard illustrates how search interest for specific symptoms
strongly matches - and often preceeds - trends in COVID-19 cases.
Trends in search interest in COVID-19 symptoms should not replace
administrative data on cases. The relation between the two is strong
but not perfect.
Search interest can be driven by
news or other events,
and the usefulness of search interest data depends on geographic characteristics such as internet access.
However, Google data can supplement official data.
This is particularly true in circumstances
when testing or data may not be widely available. Moreover, given that
Google trends information is updated in real time, sudden increases in
search interest can warn of potential growth in COVID-19 cases.
Google Trends can also be used to understand impacts of COVID-19
and search interest around prevention measures. Consequently, in addition
to showing search interest in COVID-related symptoms, the dashboard also
shows search interest related to
mental health keywords
(e.g., anxiety and loneliness), other potential consequences (e.g., unemployment and debt),
prevention measures (e.g., face masks) and treatment measures (e.g., teleworking
Determining when the correlation between Google search interest and COVID-19 is strongest
In all figures and analysis, we use the number of new daily COVID-19 cases or deaths
and a 7 day moving average of Google Search Interest. We compute how strongly different
search terms correlate with COVID-19 cases and deaths. In addition, we determine whether
search interest can help predict future cases or deaths
or whether search interest responds or comes after cases/deaths. To determine this, we shift COVID-19 cases/deaths
by up to 21 days from its actual date. We calculate the correlation between
the shifted COVID-19 and the search interest (this approach follows
on COVID-19 and Google Trends). Using all these estimated correlations, we determine the
Lead/Lag Days: The number of days COVID-19 cases/deaths was shifted to obtain the maximum correlation.
Negative values mean that search interest comes before COVID-19, helping to predict cases/deaths while
positive values indicate the search interest reacts to COVID-19 cases/deaths
Correlation between COVID-19 cases (shifted) and Search Term Interest
We access COVID-19 Cases and Deaths from the
data for all countries. In all analyses, we
use a 7 day moving average of Google Search interest. To protect privacy, Google
only releases search interest data when there is a large enough search volume for a
specific search term. We translate search terms from English into
each country's most widely used language using Google Translate.
The below table shows which language is used for each country.
This dashboard builds off of a literature that
uses Google Trends to provide insight into COVID-19.
Articles that were used to inform the dashboard include
The dashboard and analytics were produced by Robert Marty, Manuel Maqueda,
Nausheen Khan, Arndt Reichert and Bibind Vasu of the
Development Impact Evaluation (DIME)
Group at the World Bank. The research team has published a
about the analysis and has also conducted analysis at the subnational level using
Brazil as a case study.
The findings, interpretations, and conclusions expressed in this dashboard are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.