Covid-19: A DataScience Approach on Brazil’s Geoeconomic Regions Using Google Mobility Report
Analyzing Brazil’s data for Covid-19 given its three geoeconomics regions and their economic poles, which are “Amazonia, Nordeste, and Centrosul” and “Manaus, Recife and São Paulo”, respectively.
Created in collaboration with Joao Vitor Dias Xavier
Introduction
When we look at the “retail” column in Google Mobility Report’s dataset, we can see the mobility trends of people in places like restaurants, cafes, shopping centers, theme parks, museums, libraries, and movie theaters. This data, in other words, shows the economic recovery of countries, states, and even some cities.
Knowing this, we can divide Brazil into three “Geoeconomic Regions” — which is an unofficial division that groups states with similar economic activities — in order to compare and see what is happening in these three regions and in their respective economic centers. This way, the main objective of this analysis is to assess the economic recovery between the regions of Brazil.
Method
First of all, data was extracted from Google Mobility Report, a free and open base of data that can be accessed in:
The study was made using Python, one of the most used languages for data science for its simpleness and versatility. Besides this, the programming language ranking site PYPL has ranked Python as the number one programming language with a considerable popularity gain in 2020:
Python has a range of libraries that can be used for data analysis, and one of the most famous is pandas. Using pandas, we can, with a simple line of command, do the initial lecture of data, saving the results in a structured table that can be easily manipulated.
How can be seen, the Google Mobility Report contains data of every country, but we will use just Brazil ones. So, data was cleaned to remove unnecessary or dirty lines and the unnused countries, this stage is essential to a more accurate analysis. Then, data was aggregated in new subdivisions, each representing one of the unofficials geoeconomics regions of Brazil: Amazonia, Nordeste and Centrosul. Not only that, three cities were collected, representing the biggest economical centers of each geo economic area, respectively: Manaus, Recife and São Paulo.
As the regions represent more than one city, the values of each region were marked as the average of the cities that compose it. The first result is the following table
The next step was to apply the moving average of the last seven days to all columns in the dataframe — this was done to neutralize possible outliers. That done, the data is now ready. However, the table view is not the best way to show the results, so it was used matplotlib library for plotting graphs with the finded values. We will priorize code developed can be better seen in
The first graph produced is an overview of the “retail” values over time for each place we work. Even if this format offers a better understanding of the data, it is also not clear enough.
So, a second way was developed: for each regions one graph, placed in two lines and three columns, allowing data to be compared both vertically and horizontally — that is, we can compare between regions, the region and its pole city and only cities. A deeper analysis of the information acquired through the chart was placed next session.
Results
As shown in the six generated graphs, we can see some interesting things to talk about. For example, the arrows and percentage give an overview on how much the retail values increased over time. This way, if we observe, for instance, the graphics for “Amazonia” and its biggest economical center, “Manaus”, we can clearly see that the city has been recovering its economy much better than the region in general, which enables us to infer that Manaus has recovered much faster than other cities in the geoeconomic Amazonia region. As shown in the news, specifically this one from “Folha de São Paulo” journal: Recuperação na Zona Franca de Manaus indica que economia já reage, dizem especialistas. Other thing we can see in the graphic is that Manaus’s smallest value wasn’t in early April, like almost every other state.
If we now compare the three regions, we can see that Amazonia is the one with the highest retail recovery value, followed by Nordeste. But this state isn’t the same all the time. If we look closely by middle-end May, we can see that the Centrosul region had a higher retail value than its neighbors. Looking deeper into journals and news, we can see that the Centrosul region really started to return its economical activities, like in this report: Em meio a aberturas, Sul vê número de casos de coronavírus subir 86%, which is a pattern that was happening all over the region. That is probably why its retail value was so high back then.
Looking deeper into the Nordeste region, we can see that its retail value stayed below the retail in the other regions almost all the time. This value started to grow in early June and it is gradually growing, showing a percentage value that became higher than Centrosul’s. From the data collected by Inloco: Segurança e performance para apps, privacidade para pessoas regarding social distancing in Brazil, the Nordeste region had the highest rate back in April, which confirms the data shown in our graphics.
Conclusion
The data available on the internet can tell us many things about our reality, if we can use it properly. The objective of this article was to analyze data given by Google Mobility Report to confirm suspicions about three macro regions in Brazil, the Geoeconomic Regions. As we saw in results, this subdivision of Brazil is strongly related to the retail value in Google’s dataset, showing that our initial idea makes sense and actually happens.
The use of Google Collab and Python language were of extreme help to achieve our goals in this work, making the work easier to do and even easier to show. Also, the material given by professor Ivanovitch Silva as basis, helped us a lot.
Contributions
Conceptualization, Carmem and João; Development, Carmem and João; Introduction, Carmem and João; Methodology, Carmem; Results, João; Conclusion, João.
