29 November 2022 - Note: The indicators will no longer be updated from 1 January 2023.
Stand: 29.11.2022 - Measures to contain the Covid-19 pandemic have brought public life largely to a standstill in most countries of the world. In Germany, as in many other countries, the closure of shops, schools and universities, workplaces, cultural institutions and much more aim at reducing social contacts.
Based on mobile network data available in the short term, the Federal Statistical Office compiles regional indicators that show the development of individual mobility at administrative district and municipality levels.
This article presents first results from an ongoing project of an experimental nature. The results and conclusions that can be drawn from the experimental indicators should therefore be interpreted with caution. As with the other experimental statistics on our website, the indicators differ in maturity and quality from official statistics.
Mobility changes
In order to depict the change in general mobility behaviour, all entries to, and mobility within the administrative districts are being evaluated for each district. Following this, the daily mobility determined in this way is compared with an average value for the corresponding weekday from the same month of 2019. A value of -20, for example, shows that mobility is 20% lower than in the respective month of 2019.
Furthermore, an adjustment is made for public holidays, as the usually lower mobility on public holidays distorts the reference values. For non-holidays, public holidays are therefore excluded from the calculation of the reference value. Rates of change for public holidays are in turn calculated in comparison with the corresponding public holiday in 2019, i.e. regardless of the day of the week. Further details on public holiday adjustment are provided under the section "Methodological challenges".
Figure 1
The cartographic representation in Figure 1 shows the daily change in mobility for each district over the past 31 days to the current end.
Figure 2
In addition to the cartographic representation, Figure 2 compares the change in mobility in an aggregated form for the individual Länder. In contrast to Figure 1, the calculated results in this figure go back to January 2019. This allows the changes in daily mobility to be shown right from the beginning of the COVID-19 pandemic.
Small-area mobility
The regional and nationwide restriction measures taken because of the COVID-19 pandemic aim at reducing personal contacts and thus the risk of infection. Measures such as the ban on large events and the restriction of catering for out-of-home consumption, as well as the temporary regional curfews, have a particular impact on mobility in the evening, at night time hours and on weekends. To analyse the change in mobility over the course of the day, the Federal Statistical Office makes use of hourly mobility values at the small-area level.
The small-scale grid cell data are visualised cartographically in the form of interactive grid maps. The calculated mobility rate is shown for each grid cell, provided that the cells are filled for the current point in time and for the reference period of the previous year and that they are not subject to anonymisation. The grid width, i.e. the edge length of a cell, can vary. In the map application, the grid width is 10 kilometres for the whole of Germany. For regions with higher movement figures, grid cells with edge lengths of 5 kilometres, 1 kilometre or 500 metres are also displayed.
©Geoportal Berlin, Esri, HERE, Garmin, FAO, METI/NASA, USGS - Datenquelle: © Statistisches Bundesamt (Destatis) |©Teralytics
Figure 3
Means of long-distance passenger transport
The mobile network data allow conclusions to be drawn about the use of different means of transport on a daily basis, which means that changes in long-distance travel can be mapped in a timely manner. To identify the means of transport, information about the course of road and rail routes as well as patterns of synchronous connections of several mobile devices is used. Only trips with distances over 30 kilometres are analysed. For local traffic, it is not possible to distinguish between the different means of transport, whereas for long-distance journeys, more than 80% of the movements could be assigned to specific means of transport.
The results should be interpreted with caution for two reasons: first, the available mobile network data measure all movements and thus do not allow to distinguish between long-distance passenger and freight transport. Second, double counting cannot be ruled out, as not all SIM cards installed in motor vehicles can be identified and eliminated from the calculation.
Figure 4
Note: Due to an adjustment of the data basis on 30 June 2021, only values up to this date can be provided in the figure.
Methodological challenges
The above-mentioned results are based on first experimental analyses. Calendar adjustment of the data is still pending, which makes comparability between two annual values difficult and can lead to biases in some cases. Currently, the comparative values for 2019 are summed up for each weekday over all days of the respective month and divided by the number of weekdays. Public holidays are not considered here. Furthermore, reasons for outliers, especially in results on the detailed regional level, need to be investigated. Currently, the calculated change rates of commuter mobility are being manipulated by excluding the upper and lower percentile at the municipality level. Using this method, a large part of the implausible change rates can be eliminated. Nevertheless, plausibility checks are still pending.
Another methodological challenge is the granularity of regions, as the regional breakdown of the underlying basic data, provided by Teralytics, is not homogeneous. For example, on the one hand, basic data, especially for urban regions, are available even at the town district level, whereas, on the other hand, basic data for rural regions are only available aggregated over several municipalities or districts. This may lead to biases, especially when calculating commuter mobility. These and other challenges have to be analysed and solved in the further course of the project.
Holiday adjustment
By implementing public holiday adjustment, the influence of public holidays is compensated for in the calculation of reference values. This is done by calculating weekday averages of the reference months excluding public holidays. For example, in May 2019, 1 May (Labour Day) is not included in the calculation of the average Wednesday in May. In contrast, rates of change for public holidays are calculated in comparison to the respective public holiday of the reference year, so for example, mobility on 1 May 2020 is compared to the mobility on 1 May 2019. Days treated as public holidays in this context are the statutory holidays as well as 24 and 31 December, which play a special role.
Figure 5 shows the effects of public holiday adjustment by comparing the calculated rates of change on 2019 based on adjusted and unadjusted reference values. In the unadjusted data, clear outliers of public holidays (such as New Year's Day, Easter, Labour Day, Day of German Unity on 3 October and Christmas).
Figure 5
Data basis
To map mobility based on mobile network data, the Federal Statistical Office uses data from mobile phones accessing the network of the Telefónica telecommunications company. The movement data are being processed and provided by the private service provider Teralytics AG. Three mobile communications companies, including Telefónica, share the German mobile phone market almost equally. However, the market shares apparently vary regionally and partly quite notably. In order to make statements about the entire German population, the data provider uses an extrapolation algorithm. The extrapolation is based primarily on geographically differentiated local market shares. The extent to which the extrapolation allows statements to be made about the total population or only about mobile network users will have to be investigated in further methodological analyses.
Note: As of 15 July 2021, all data presented in Figures 1 and 2 are based on a changed data basis, due to methodological inconsistencies found in a subset of the data. Change rates in comparison to 2019 could have been distorted during the summer months in regions dominated by tourism. All affected figures were corrected, the updated values may therefore differ from previously downloaded data.
Teralytics works exclusively with anonymised data, which are aggregated in order to obtain information about the mobility of the population. The data allow an overview of the number of mobile devices that carry out a certain movement. Teralytics removes all information from the data set that contains less than five movements for data protection reasons. Furthermore, the data do not contain any personal information and it is not possible to draw conclusions about individual persons. For this purpose, Telefónica has developed a Data Anonymisation Platform (DAP) in consultation with the Federal Commissioner for Data Protection. A three-stage anonymisation procedure is being used. You can find further data security information from the company here .
The data provided contain the number of movements within a certain period of time (day or month), which are identical with regard to the regions of origin and destination (district or municipality). A movement appears, when a mobile device switches from one radio cell into another. When the mobile phone remains in a cell for at least 30 minutes, the target region of a movement is reached. Movements can therefore also be detected within a region, provided that the mobile phone changes from one radio cell to another.