Probably like everyone else in this year, I was thinking this would go away by summer, or leave us alone at least until September, when lockdown started in April.
I’m a Japanese man living in the eastern part of Europe. I’m not an expat, I’m living here with 3 cats. By looking at the ambulances outside the window, I thought things were going to be ok, but it didn’t as we all know. We suffered lockdown, but some countries managed to tame the virus while some countries are having the second wave. Sweden didn’t implement lockdown but their situation doesn’t seem so different from the UK. Why ? What did make the difference ? It wasn’t that I didn’t have much to do during the lockdown, but I didn’t want to go out. In mid May, I started learning coding, mainly Python and R so I can see the world better without going out so much. Then I started analyzing the Covid19 related data. I did data analysis in my first job, but it was well over 10 years ago. That time everybody was using Excel. It took me some time to get used to coding and try to find the answer to the why.
There are many types of data, such as new cases, total deaths, etc.. Among all, Google mobility reports tell us something about lockdown. Google is tracking our GPS data unless we turn it off from Google map. For the purpose of public health, Google is releasing this massive data in public for free. so, I thought it might be insightful if I compare it with the new cases and new deaths data going around the internet everyday. In Google mobility reports, Google sums up the mobility (how many people visited) in 6 categories of places in each country, state and county. These are retail and recreation places, grocery and pharmacy stores, parks, transit stations, workplaces and residential areas. Mobility in residential areas are the time spent in those areas as Google explains. This is a massive set of data because it’s updated everyday. I looked into the countries (65 nations), where Covid19 data is available, and are not in the extreme circumstances, in monthly basis.
There are loads of discussions going around about Covid19. Especially when it’s about how it’s transmitted to other people, we still have a lot of questions. Weather, culture, population density, public health investment, numbers of factors must be involved in it, however, I found at least two significant connections related to lockdown. Most countries entered lockdown or de facto lockdown around early April. What we did before lockdown actually affected the virus spread after lockdown.
One of them is how many people visited parks. The maximum mobility in parks in March 2020 is in the negative correlation with how many new cases increased from April to May. The correlation coefficient is -51%. This means it has a relatively strong negative correlation. In short, the more people visited parks in March (compared to the local baseline), the less people were infected from April to May. This sounds against our intuition. I also believed staying at home must be the best choice and I don’t know how it helps us. Maybe the vitamin D thing or maybe not. Either way, this is what the data says. Actually plots are in a curve so the correlation can be stronger than -51%. In fact, the most successful countries such as Denmark, Norway and Finland have been letting people visit parks while other countries barricade parks. Greece is on the right end on the plot. They are doing very well too, but they had a spike of park mobility in March and it was low for the rest of the month. What was compared to was how much the new cases increased from April to May. Every country has different population for sure. I therefore divided the new cases confirmed in May by the ones in April (%) so we know how fast the infection spread toward the end of May.
In short, the more people visited parks in March (compared to the local baseline), the less people were infected from April to May.
The other one is how many people worked at office. This tells us lockdown / remote work was crucial as expected. The scatter plot above tells us the mobility in workplaces in March has a certain connection to the new cases increase from April to May. There is a relatively strong correlation there and the correlation coefficient is 49%. In short, the more people worked at office in March (compared to the local baseline), the more people were infected from April to May. Unlike the one about parks, this is median value in mobility. I think this is because people go to the office on the daily basis. I do not know why this takes effects after such an interval (almost 2 months) but maybe an infected family member infects his/her family member plus incubation period etc.. Some Asian countries frantically tried to contain the virus before April in addition to the recommendation of remote work. These factors are making the correlation less visible, but we can still see office can be a major transmission center. Y axis is the same as above.
In short, the more people worked at office in March (compared to the local baseline), the more people were infected from April to May.
After April, the correlations diversify depending on the methods. Probably because too many extra factors start being involved after lockdown and it became chaotic. I also found a positive correlation between time spent in resident area and the new cases, but the situations are diverse within one country and I need to improve the model.
As a conclusion, going out to a park seems helpful to fight Covid19. However, working from home should be still encouraged if possible. In case of a second lockdown, I hope the authorities will allow people to go to the parks with proper protective measures at least during the weekend.
The data about the new cases is from Our World in Data supported by the University of Oxford.
