Another concern we heard from other researchers using Safegraph data is that park visits are likely undercounted because Safegraph is missing many park locations. Do you think this is true (especially with regards to other locations in the data)?

Questions & Help
1
  1. Another concern we heard from other researchers using Safegraph data is that park visits are likely undercounted because Safegraph is missing many park locations. Do you think this is true (especially with regards to other locations in the data).
2

It is true that our park coverage is not as good as our consumer POI business coverage.

we have good coverage of national parks, but our coverage for small city parks is something we are actively trying to improve.

cc: @Bryan_SafeGraph FYI

3

@Christos_Nicolaides_University_of_Cyprus_MIT as @Ryan_Fox_Squire_SafeGraph mentioned, this is true when comparing against our consumer POI coverage.

That said, we now have 100k+ park POIs in the U.S. and Canada. The attributes tied to these records are fairly reliable, but the geometry is sub par since parks require a more unique polygon instead of a standard building footprint. We plan to have parks along with a few other unique type POIs drawn over the next few months so that the polygons allow us to pick up better foot traffic insights to these locations.

4

Ok very good to know. As a ballpark, do you think this means that park visits and density tend to be over-counted or under-counted?

5

It would be hard to say for sure without a more statistical approach, but I would estimate that park visits are under counted more than they are over counted. This is because the existing park polygons that need to be drawn are usually too small rather than too large.

6

I

7

I’m late to the game here, but I am having trouble estimating true visitation to state and local parks using either the macro or micro normalization approach. Some park counts seem very close to the numbers we have based on pyrobox auto counters, but some are 6x fewer than reported by park managers. I would expect the numbers to be off a bit, but I’m afraid we will lose credibility very quickly if we present numbers that far from their manual park count estimations. Is there any way to determine which types of parks (i.e. large state parks versus small ball parks) are more reliable? Am I just misunderstanding the process-- are the home-cbg and visitor_panel (i.e. micro and macro methods) supposed to give true visitation counts, or is it supposed to be an order of magnitude off? Thanks for any guidance.