Movement In Georgia During Covid-19

6 min read

Facebook recently released its data about people’s movement worldwide. The dataset which was previously only available for the research or other types of organisations, is now publicly accessible. Its aim is to aid governments fine tune their policies since the data shows how people react to them. It can answer such questions as: Do people comply with the rules of a lockdown? Do they move around equally every day or is there variation during the week? Do the particularities of different regions cause different patterns in the movement of their residents? and so on.

Methodology
(skip this part if you like riding a bike but don’t want to know how it works)
As stated on the methodology page of Facebook, the data is taken from the people who have have option in for tracking location history on their phones. However, Facebook adds some noise to the data so the individuals cannot be traced by any third party. The whole world in divided into 600 by 600 metre tiles and they count the number of users visiting each tile every day.

We have two metrics — a relative movement and Stay Put index. The former compares the movement of the people compared to a baseline which is a whole month of February 2020. This means that each weekday is compared to the similar day in February 2020. That is why it is a relative measure — it does not tell us an absolute number of the visited tiles. While Stay Put index directly shows what proportion of people (users who use location services) stayed at the same tile all day long. For this reason the latter might be a more telling of people’s whereabouts rather than the relative movement index. All the charts below are based on these two types of metrics.

Covid-19 and Movement in Georgia

The government of Georgia had taken four major steps in response to the pandemic (see full report). From early March the government entered the second phase of its plan, which was slowing down the spread of the virus. This was the time when the restrictions with different degrees were introduced, such as closing down the schools and borders for the neighbouring countries. But before examining such details, let’s take a look to the overall trend on the figure 1.

Figure 1. Mean movement and mean Stay Put across the country

Movement and Stay Put are almost mirror images of each other as staying home is the opposite of moving around. And at the same time this symmetry validates each graph. The general trend is very obvious — since early March movement begins to decline rapidly, reaches its lowest during late April and begins to increase in May, making a U shaped concave curve. During the summer it more or less levels and begins to vary near the end of August and in September. This increase during summer is self-explanatory. First, the restrictions were removed (not all of them) and the pandemic was well managed and second — this is a relative movement compared to the baseline of the February data. Which means people usually move more during summer than in winter. We can examine the details on figure 2.

Figure 2. Movement data is plotted against the number of Covid-19 daily cases and the major government actions are highlighted by the vertical lines

As is shown, Covid-19 daily cases were below 50 almost well before September. But in April cases were rising and by the governments predictions without restrictive measures it could have gone out of hand. So on 31 March they announced curfew, banned gatherings of more than three people and shut down public transport, among many other restrictions.

There is an interesting spike at both sides of the lowest datapoint in April. On April 15 several big municipalities, Tbilisi and Rustavi among them, came under lockdown and two days later movement by vehicles was banned. Right before the municipalities’ lockdown and transport prohibition, people suddenly began to move much more. Then it dropped as expected and there was another big spike as the cars were allowed to roam the streets again, after which people reduced their level of mobility. Our guess is that when the government announced its plans to ban cars from the streets and completely lock municipal borders, people rushed to do everything that required relocation by a vehicle. Then their needs amassed as under the lockdown big section of economy was stalled, contact was extremely limited and when pedestrians with big bushes on their heads spilled on the streets to make Tbilisi look like a perfect venue for Woodstock festival. Those needs were to be met in a short amount of time — in a day or two — hence the second spark as after 27 April cars rolled out on the roads.

Smaller fluctuations denote the difference between working days and weekends. For working days movement was reduced to -0.0752 on average compared to -0.0556 on weekends.

Regional Differences

Municipalities differ from each other in many ways. For example in some municipalities workplace might be concentrated within its borders, while in others people travel over longer distances. The video below shows the changes in movement in different regions on a timeline. Some municipalities lack enough data for some months to reach the threshold (300 hundred eligible users) and is omitted from the data. The redder the area, the more reduction in movement is implied. Overall, a pattern is clearly visible.

But line charts for each municipalities show the data in more detail. Figure 3 shows the both metrics for each municipality. Most obvious conclusion is that each municipality is almost a perfect copy of the country level pattern in movement: movement declines in march, that U shaped curve is replicated almost everywhere, than people move more than in February and in September it begins to decline yet again.

Figure 3. The movement and Stay Put indices collapsed into municipalities

But closer examination reveals variation between them. For example, during summer months Batumi hits the highest for obvious reasons. Tbilisi and Mtskheta are ducking below the baseline or keep quite close to zero. While places like Tsalka and Martvili fluctuates a lot well above zero. Marneuli, Tsalka and Ninotsminda, for example, does not show that distinctive U curve that is probably car-related.

Figure 4. This box-plot represents the variation between municipalities as for how much people changed their behaviour during this time period

Figure 4 above retells the same patterns shown in the line charts. Where on the line chart line oscillates a lot and goes very low and up very high, here is represented by a long boxes and whiskers. The box in the middle shows 50% of the data and the other 50% is represented by whiskers. Since boxplots of Batumi, Tbilisi and Mtskheta are quite long, it means people behaved differently during different time, while in Dedoflistskharo, Chkhorotsku and Tsalenjikha locals behaviour did not change that much.

This data shows how Georgians reacted to the actions taken by the government of Georgia during Covid-19. Personal or social experience and media can be relied on to represent the reality but data on this granular level shows every nook and cranny of people’s actions. For example, the fact that people rushed before and after the restrictions on vehicles were imposed and that this was more so in Tbilisi and Mtskheta than in Chkhorotsku and Akhalkalaki is revealed in this kind of analysis.

As Countries continue to struggle to deal with the problems caused by the pandemic, detailed big data issued by Facebook and Google might help governments to better predict their citizens’ behaviour before they take any measure.

Authors