Experimental statistics Satellite-based early estimate of short-term economic development



28 November 2022 - Both the availability of, and demand for up-to-date data has strongly increased in the last few years. The need for fact-based economic data for users from politics, science, society, businesses and the media is obvious especially in times of current crises such as the Covid pandemic or the war in Ukraine. The Federal Statistical Office has released the gross domestic product (GDP) as early as 30 days after the end of a quarter (flash estimate) since July 2020 to show short-term economic developments at an early point in time. Estimates are required to achieve an earlier availability of results. Further new digital data will be developed to improve such estimates and provide GDP flash estimates even earlier (from flash estimate to nowcast). The results of the first project phase were published in issue 6/2019 of the WISTA scientific journal under the title "Moving from GDG flash estimates to GDP nowcasts: first results of a feasibility study to further accelerate early GDP estimation". Such data include satellite data. The idea behind using satellite data for short-term economic statistics is that economic activities leave optical traces that can be captured by satellites and quantified on this basis. Today, satellites can provide images of large areas at short intervals. Information from satellite data are available early and at a small-area level. In addition, analyses can be made across administrative boundaries and without breaks in methodology.

It was examined in the project on Smart Business Cycle Statistics (SBCS) to what extent information from satellite images can be used to represent business cycles. The results are described in the EXSTAT article on “Smart business cycle statistics based on satellite data”. The total number of cars parked in front of retail shops in a test area was examined as an example. A limiting factor was the lacking availability of satellite images with the required spatial and time resolution for automated recognition of small objects. An overview of various quality levels and resolutions of satellite images is contained in the EXSTAT article on the SBCS project. In the meantime, both the availability of satellite data and the opportunities of analysis have improved. Consequently, the feasibility study on “satellite-based early estimate of short-term economic development” examines the concrete question of whether information from satellite data can be used to support early estimates of short-term economic development of the Federal Statistical Office.


The project is based on an examination of satellite-based economic indicators produced externally and acquired by the Federal Statistical Office. One of the data sources is the European Space Agency (ESA). During the pandemic, the ESA published the The Rapid Action on Coronavirus and Earth Observation dashboard (RACE). The dashboard shows how remote sensing data, ground-based observations and numerical models can be used to illustrate socio-economic and ecological changes. Indicators on agriculture, health, air quality, water and the economy are shown in the dashboard for various European countries. For Germany, a total of 10 different indicators are published on the dashboard, including an indicator showing the production of vehicles at 15 locations in Germany, in the following referred to as ESA indicator of vehicle manufacturing.

Radar data of the Sentinel-1 satellite of the European Copernicus earth observation programme are used for this indicator. Each of the two satellites Sentinel-1A and Sentinel-1B has a Synthetic Aperture Radar (SAR) that allows information on the Earth’s surface to be provided irrespective of light and cloud cover. The sensors of the satellites send radar beams to the Earth’s surface. The radar beams are backscattered with different intensity, depending on the absorption and reflection properties of the objects hit by them, and received again by the satellite sensors. The idea behind using SAR data is that the backscatter intensity depends on the objects’ material. Consequently, the variation of backscatter intensity can indicate what there is at the location observed.

For the ESA indicator of vehicle manufacturing, parking areas of car factories are monitored. Cars have a high backscatter intensity, which is due to their metallic properties. The more cars are parked on the factory parking areas, the higher backscatter intensities are recorded by the satellites. Consequently, the variation in backscatter intensity can serve as an indicator of how many objects there are on car factory parking areas at a specific time.

Figure 1

To calculate the indicator, the ESA manually selects Areas of Interest (AOI) and extracts them from the Sentinel-1 data for the marked AOI time series of the radar backscatter values. The Sentinel-1 satellites have an overflight rate of 6 - 12 days. This means that the satellites fly over the AOIs every 6 - 12 days and collect data at these intervals. A spatial average per observation date is calculated for every AOI, taking account of potential radiometric artefacts. Data on 15 locations are currently available for Germany for the period 2017 - August 2022. The locations are spread across Germany (distribution among the Länder: 2 x Baden-Württemberg, 3 x Bayern, 1 x Bremen, 1 x Hessen, 2 x Niedersachsen, 1 x Nordrhein-Westfalen, 1 x Saarland, 3 x Sachsen, 1 x Thüringen).

Developing the ESA indicator might enable conclusions to be drawn about output in vehicle manufacturing. Vehicle manufacturing accounts for a large part of the industry. Turnover from the production of vehicles, trailers and semi-trailers accounted for 22.8% of the total turnover of all industrial establishments in Germany in 2020, according to a press release of 22 February 2021 of the Federal Statistical Office. The ESA indicator of vehicle manufacturing, which shows the number of cars on car factory parking areas, could thus support the early estimation of short-term economic development as an early economic indicator. It is therefore examined below whether the ESA indicator can represent the official production index in vehicle manufacturing.


To be able to compare the official index released on a monthly basis with the satellite-based indicator, the latter has to be aggregated to the same time interval first of all. For the purpose, a monthly average is calculated across all 15 locations for which the ESA releases the indicator. The development of the monthly aggregated time series is shown in Figure 2a (ESA index) and Figure 2b (official production index). A general downward trend in the period 2017 - August 2022 can be seen for both time series.

Figure 2a


Figure 2b


Figure 3 shows that the development of change rates is similar for the satellite-based indicator and the official production index of vehicle manufacturing. A possible connection is visible. A strong decrease in spring 2020, which was caused by the consequences of the Covid-19 pandemic, and the resulting sharp increase in the subsequent year are also obvious.

Figure 3


Figure 4 illustrates that the changes in the number of vehicles on car factory parking areas, based on satellite data, generally develop in line with the events occurring in production, such as short-time work or production stops.

Figure 4


A correlation between the satellite-based indicator and the official index is suggested not only by a graphical comparison but also by the results of a bivariate correlation analysis. Both the absolute values and the change rates are examined in the comparison. Figure 5a and 5b provide a graphical representation of the statistical correlations.

As is shown by Figure 5a and 5b and the correlation coefficients in Table 1, a statistically significant correlation can be identified between the satellite-based indicator of the number of cars and the official production index of vehicle manufacturing. This applies to the comparison of absolute values and that of change rates. It also turns out that the statistical correlation is maintained if only the period before the massive restrictions caused by Covid regulations is examined. This suggests that the correlation is not just produced by outliers in the pandemic period. The correlation analysis confirms the first impression from the graphical comparison that the satellite-based indicator might be suited to represent the official production index of vehicle manufacturing.

Figure 5a


Figure 5b


Table 1: Correlations between satellite-based ESA indicator and official production index of vehicle manufacturing
PeriodAbsolute values Rate of change on the same month a year erlier1
1: Rate of change starting in January 2018. The value for January 2018 describes the change compared with January 2017.
As all change rates refer to the previous year, there are no values for 2017.
January 2017 to August 20220.740.59
January 2017 to February 20200.480.54


Initial results suggest that there is a correlation between the examined ESA indicator of vehicles on car factory parking areas based on satellite data and the official index of production. As, however, some constraints have to be taken into account, the satellite-based indicator examined here cannot be used as a supplementary experimental indicator of short-term economic development yet. So far, only the original (not seasonally adjusted) values have been compared with each other. In many cases, however, short-term indicators are considerably affected by seasonal effects which may obscure the real short-term movements. Overall it did turn out that the ESA indicator might have the potential to support an early estimation of short-term economic development on an experimental basis if the current constraints can be eliminated and the results obtained so far are confirmed in further studies.

The approach of using satellite data for the production of early indicators of the short-term economic development will be pursued as the first results are positive. Also, this will serve to reduce the above constraints. For this purpose, the analyses will be extended and improved in the next steps. This will involve, first, seasonal adjustment of the satellite-based data, thus improving comparisons between these data and the adjusted series of the production index. Second, the ESA indicator will be compared with other indicators such as on storage and output.

Based on the positive results obtained so far, it will also be examined whether the satellite-based indicator – in addition to representing individual official indices – is suited to concretely support the estimation of economic indicators, for instance, in the GDP nowcasting project.

Basierend auf den bisherigen, positiven Ergebnisse soll zudem untersucht werden, inwiefern der satelliten­basierte Indikator nicht nur einzelne amtliche Indizes abbilden kann, sondern auch konkret zur Unterstützung der Schätzung von gesamt­wirtschaftlichen Indikatoren geeignet ist, beispiels­weise im Rahmen des BIP-Nowcast Projekts.