What Google’s mobility data can tell us about COVID-19 lockdowns in India?

1. Background

In order to help public health experts Google has published aggregated, anonymized mobility data across 6 categories.

• Grocery & pharmacy
• Parks
• Transit stations
• Retail & recreation
• Residential
• Workplaces

The data shows how visitors to (or time spent in) categorized places change compared to the baseline days. A baseline day represents a normal value for that day of the week. The baseline day is the median value from the 5‑week period Jan 3 – Feb 6, 2020.

For each region-category, the baseline isn’t a single value—it’s 7 individual values. The baseline day is the median value from the 5‑week period Jan 3 – Feb 6, 2020.

For each region-category, the baseline isn’t a single value—it’s 7 individual values. The same number of visitors on 2 different days of the week, result in different percentage changes.

India was the 48th country to start a National-level lockdown to control the spread of COVID19, which was extended multiple times. For our analysis we have following phases to look at:

• Pre-lockdown (before 25th March)
• Phase 1 (25 March – 14 April)
• Phase 2 (15 April – 3 May)
• Phase 3 (4 May – 17 May)
• Phase 4 (18 May – 31 May)

As phase-4 is ongoing we do not have the data on it but Google’s raw dataset does contain information from 15th Feb (pre-lockdown) till the end of 3rd lockdown.

2. Methodology

Google published mobility data on most of the countries in the world. But as our focus is India, we filter out the rest of the data. Then we apply a state-level cutoff of 5000+ cases (as of 23rd May) to avoid our graphs getting incomprehensible. It gives us 7 worst affected states in India:

• Maharashtra
• Tamil Nadu
• Gujarat
• Delhi
• Rajasthan
• Madhya Pradesh
• Uttar Pradesh

Then we simply plot the data across each of the 6 place categories for all these 7 states in an interactive plotly graph. (available here if you want to play around)

3. Findings

Retail & recreation Mobility: covers places like restaurants, cafes, shopping centers, theme parks, museums, libraries, and movie theaters.

Phase-1

• As expected we see a significant drop of 68% to 89% in mobility.
• There is a clear distinction between how much the states were able to restrict the mobilty, Delhi being the most restrictive and Uttar Pradesh is the least restrictive.

Phase-2

• This phase seems better planned. All states were more restrictive compared to the phase-1 and also less variation in state trends.
• However relative ranking was maintained as that of phase-1 (Delhi most restrictive and Uttar Pradesh the least). Towards the second half of phase-1 Rajasthan became least restrictive

Phase-3

• It was not as restrictive as phase-2 but less variance compared to phase-1. Basically shows the uniform implementation of guidelines but these were slightly less restrcitive compared to the phase-1.

Grocery & pharmacy Mobility: covers places like grocery markets, food warehouses, farmers markets, specialty food shops, drug stores, and pharmacies.

• A lot more variance compared to Retail & recreation, especially in phase-3
• Uttar Pradesh, Rajasthan and Tamil Nadu are almost back to the baseline (12% to 14% drop) whereas Delhi still was 50%

Parks Mobility: covers places like local parks, national parks, public beaches, marinas, dog parks, plazas, and public gardens.

• Perhaps Tamil Nadu could have done more to restrict the mobility. For example in phase-2 when Delhi's mobility had dropped by 98%, Tamil Nadu noted drop of 40%-45%.
• As per Ministry of Home Affairs (MHA) Order No. 40-312020-DM-l (A) dated 15th April, 2020 Parks were closed. That's the drop we see in lockdown 2.0 for states like Delhi, Rajasthan, Maharshtra. States like Uttar Pradesh and Tamil Nadu could not enforce it effectively as the data shows.
• We also see some weekly patterns here for states. Perhaps people tried to spend a day per week outside their house.

Transit stations Mobility: covers places like public transport hubs such as subway, bus, and train stations.

• All states maintain a clear relative ranking throughout all 3 phases. This is the second most restricted place category after retails and recreations.
• The mobility around transit stations is steadily increasing but even after 3 lockdowns it is more than 50% down. This sudden drop and prolonged restrictions in all states might the root of ongoing migrant worker crisis (they can not earn in the current state but can not go back to their home state).

Workplaces Mobility: covers places of work. We should expect to see significant drop here.

• A very interesting weekly spike can be seen here. As explained the the background section there are 7 baseline values for 7 days of week. So we should not have a weekly pattern with around 20% increase in the mobility. This essentially mean some workers are going to workplace once a week.
• If you can think of any interesting reason to explain this, please comment below.

Residential Mobility: covers places of residence. We should expect to see some increase here.

• As expected we do see increase in the residential area.
• Interestingly, the worst affected state (Maharashtra) is one of the states showing maximum increase in mobility around residential places. Perhaps people are staying home a lot but still interacting within the neighborhood.

4. Summary

• The lockdowns were able to restrict the mobility significantly (more than 65 to 70%) in Retail and Recreations category and (more than 50%) workplaces and transit stations in the worst affected states.
• When it comes to Grocery and Pharmacy places there is significant variation based on phase of lockdown or the states. Same is the case with park, public plazas and gardens. These areas could use more uniform policies across the state borders at least for these worst affected ones.

5. More about the data

More about Google's mobility data here.

Raw mobility dataset for India looks like this:

6. Explore on your own

Want to play with the interactive visualizations seen above? All you need to do is 3 clicks.

1. Click here (to launch the notebook in binder, takes 30 seconds)
2. Click "Cell"
3. Click "Run All"
4. Scroll to the end and start exploring