New research paper: How to Best Predict the Daily Number of New Infections of COVID-19

Our team of CA-SG researchers (Lukas Jürgensmeier M.Sc., Kevin Stowe Ph.D., Prof. Dr. Bernd Skiera, and Prof. Dr. Iryna Gurevych) has just submitted our first research paper related to the Corona crisis. We compare the ability of several data sources, in particular Johns Hopkins University (JHU), Google search data and Twitter data, to predict the official number of new infections of Covid-19 and examine the need to complement the official numbers with additional predictions.

Using Germany as an illustration, our most important findings are:

  1. The widely popular predictions from Johns Hopkins University (JHU) deviate for Germany on average by 79% from the official numbers.

  2. Using a simple regression to adjust the prediction from Johns Hopkins University (JHU) reduces the prediction error to 35%.

  3. Google search and Twitter data predict the official much better than the unadjusted predictions of John Hopkins University. The adjusted prediction of John Hopkins University does better Google search data and Twitter can each predict three days ahead of time.

  4. The official numbers in Germany are reported “correctly” with several days of delay, which means that the number of new infections on a particular day, say March 26, 2020, will be reported on the following day but updated again for several additional days, e.g., until March 30, 2020, or even later.

  5. Even worse, the official numbers of Germany suffer from an underreporting on weekends in the area of more than 40%.

Our main conclusion is that there is a strong need for complementing the official numbers in Germany (and, probably, in other countries as well) with other predictions, such as those that build upon Google search and Twitter data. Such alternative data sources could also help to better understand the spread of COVID-19 in those countries where other information sources might not exist or concerns about the validity of the official data are often expressed.

Feel free to download the paper https://bit.ly/34ir6t5 and provide us with your comments and suggestions.

Photo by Markus Spiske on Unsplash