A Look at Google Mobility Data

I haven’t looked at the Google mobility data for a while but the latest update takes us through to the period of the Cairns 3 day lockdown. Mobility data is relative to a baseline for the 5 week period from January 3 to February 6 2020. Data by category since May 2021:

Google Mobility 110821 1

When it comes to lockdown effects it’s important to note a difference in methodology for the relatively stable residential category:

The Residential category shows a change in duration—the other categories measure a change in total visitors. Because people already spend much of the day at places of residence (even on workdays), the capacity for change isn’t so large.

You shouldn’t compare the change in Residential with other categories because they have different units of measurement.

To look more closely at the residential category going back to the start of the series in February 2020:

Google Mobility 110821 2

Another quirk in the daily Google data is that the baseline relates to an equivalent day of the week. Consequently all the previous spikes on this chart relate to public holiday periods which are not relevant to the current lockdown spike. So the 3 day lockdown is actually well above previously comparable levels in early 2020 where the spikes represent Easter days. Could do an average to smooth this out but these days are really just bad data and should probably look at just stripping them out.

The other category which stands out at the bottom is transit stations:

Google Mobility 110821 3

This is represented elsewhere including the Google PDF reports as “public transport”. This is a bit opaque but the data notes indicate the category includes “subway station, sea port, taxi stand, highway rest stop, car rental agency”. Which makes me wonder where activity around the airport fits in as the patterns here seem to be reflective of that. The composition of this category could be very different for different regions and cities.

The workplace category is also interesting here:

Google Mobility 110821 4

Again the large downward spikes represent public holidays which is just bad data. There are also smaller periodic depressed periods which correlate with school holidays. What is striking though is how closely this workplace mobility resembles the ABS payrolls data:

Google Mobility 110821 5

Note: Workplace graph here is daily with a 28 day average. Payrolls is weekly with a 4 week average.

There is a dashboard presentation here which is quite good to play with to compare countries and regions without the gargantuan Google csv download files.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: