# Keep moving, there’s nothing to see here!

Expressis-Verbis

Editor’s note:On 05 March 2021, we forwarded this article to the Luxembourg media, assuming that this information would be of interest to them. We would like to thank Mr François Aulner of RTL for taking up our offer, for his interest and for his time.

On the official Covid19 page of Santé, you can admire a number of statistics on the graphics page [1]. It is noticeable that the start date is different in each case. Some start at the end of February, when the first tests started, others a little later, such as the positive rate of PCR tests, which only starts on 27 March:

### Why is that, actually?

Let’s try to reconstruct the graphic before this point in time. The data is provided by the part of the source code of the page that is responsible for the graphic “Tests COVID-19”. “Graceful degradation”, a well-known technique in web developer circles, ensures that web pages with extended functionality are still rendered in a meaningful way even in less powerful browsers. The data required for rendering the interactive graphics is available in the source code as an (HTML) table.

From this, we can calculate the positive rate from the values for the total number and the number of positive tests (light blue). We have smoothed this data with the moving average over 7 days (dark blue). The area corresponding to the initial area of the previous graph is shaded grey (from 27 March 2020).

In the first wave in March 2020, only symptomatic persons were tested. The time difference between infection and entry into the statistics is given as 8 days [2]. Accordingly, the effect of a non-pharmaceutical intervention (NPI) can become noticeable at the earliest after 8 days in the reported cases.

If we now enter the data for the lockdown (16 March) and also consider the time 8 days later (24 March), we find that the maximum positive rate of the PCR tests already occurred before (approx. 21 March), and that **there can therefore be no causal connection between lockdown and containment of the first wave.**

This is confirmed by a graph shown on Research Luxembourg’s developer platform “GitHub”, which illustrates the calculation of the reproduction number Rt taking this time correction into account [2]:

Research Luxembourg: Real time R

Note: during COVID-19 pandemics, there is a delay from infection to detection during to latency, sampling times and so on. Hence, it is necessary to refer all data to their true infection time. Here, we assume constant shift. By comparison with more advanced nowcasting procedures (e.g. RKI’s), we estimated such shift to be equal 8±1 days. Hence, be aware that we are looking at the past. This is considered in plots below._{t}estimation

The red curve labelled “Smoothed Infected” shows the absolute number of new infections in contrast to the blue curve of the “Smoothed Detected”, which are only registered in the statistics 8 days later. Accordingly, the time-corrected reproduction number (lower graph) had already been falling since at least 9 March and reached the value 1 on 19 March.

The fact that the maximum of the blue curve occurs somewhat later (approx. 26 March) than the maximum of the positive rate (21 March) can be explained by the fact that the number of tests rose sharply during this period and thus more positives were found. While the positive rate already dropped again, the absolute number of positive tests continued to rise for another 5 days. The following graph shows the evolution of these two variables: Positive rate (blue dashed line) and moving average over 7 days (blue continuous), left scale, and absolute number of positive tests (red dashed line) and moving average over 7 days (red continuous), right scale.

The curve of the reproduction number, however, is shown differently on the official Covid19 website [1]. The graph, starting on 18 March, shows the reproduction number which only crosses the value 1 on 30 March!

Why was this artifice necessary to collect the missing data before 27 March in a cumbersome way in the HTML source code of the website, when there are official data sources? **Yes, these sources exist!** However, for some inexplicable reason, the data up to and including 16 March was deleted. The screenshot shows the downloadable file on data.public.lu [3], which is offered in both Excel and CSV format.

For Germany, for example, it could also be shown [4] that the number of reproductive cases had already fallen below one **before the lockdown**. Comparable to Luxembourg, the maximum number of positive cases was reached on 29 March, while the maximum number of corresponding infections is dated as early as 12 March, as it is assumed that there is an average delay of 17 days before the virus is included in the statistics.

In the meantime, there is a whole series of studies that question the benefit of lockdowns. A representative example is a study in which the renowned Professor John P. A. Ioannidis was involved [5]. It compares 11 countries with partly very different NPIs and comes to the conclusion:

While small benefits cannot be excluded, we do not find significant benefits on case growth of more restrictive NPIs. Similar reductions in case growth may be achievable with less-restrictive interventions.

### Conclusion:

In view of these facts, it can therefore be considered probable that:

- the maximum positive rate of the PCR tests had already been reached on 21 March. The development of the infections has thus slowed down on its own and the lockdown was not the cause.
- by publishing the graph of Research Luxembourg’s time-corrected reproduction figure on the GitHub platform, at least the Task Force, and thus in all likelihood the government, is fully aware of the facts described in point 1.
- the representation of the reproduction figure in the official publications was distorted because it was time-delayed.
- the data from before the lockdown was
**subsequently**deleted.

It is pointless to conclude that a wave of infection** that subsides by itself without any significant measures** then naturally deprives all subsequent “nonpharmaceutical interventions” of their raison d’être:

- Mandatory masks,
- Large Scale Testing,
- Curfews etc.

Your Expressis-Verbis Team

### Sources:

[1] Official Covid19-Website – Graphs https://covid19.public.lu/fr/graph.html

[2] Research Luxembourg: Real time R_{t} estimation
https://github.com/ResearchLuxembourg/covid-19_reproductionNumber/blob/master/src/estimation_R_eff.ipynb

[3] data.public.lu: Données COVID19 https://data.public.lu/fr/datasets/donnees-covid19/

[4] Stefan Homburg: Evidenz zur Coronainfektion und der Wirkung des Lockdown Discussion Paper No. 670, ISSN 0949-9962 http://diskussionspapiere.wiwi.uni-hannover.de/pdf_bib/dp-670.pdf

[5] Assessing mandatory stay-at-home and business closure effects on the spread of COVID-19 https://onlinelibrary.wiley.com/doi/epdf/10.1111/eci.13484

This article was written in German and translated into French and English.