Is there a reliable/existing method to match survey respondents based on IP address to the records in Social Distancing Metrics? The raw records exist at census block group level and are available from early 2019.
We overlooked how accurate lat-lon are reported from Qualtrics. With an existing survey, the only “accurate” location records are the IP addresses. My current plan is to use a GeoIP2 service that return (lat-lon, R) , where R denotes a radius in kilometers that the real address shall fall in the circle centered at lat-lon with X%. Then, I’ll pull for a “tile-cover” using the census block groups that shall intersect with the circle.
Before I set out to perform the tile-covering exercise, I’d like to ask if there is an existing method to match Social Distancing Metrics data based on IP addresses. At the end of the day, we plan to use the changes in local mobility (average over a fixed window in 2020 vs avg over the same window in 2019) as an instrument variable.
Hi @Linfeng_Li, I think I understand your proposed methodology, but I am a bit in the weeds on what you are asking - are you asking if SafeGraph can provide IP addresses to records in SDM (i.e. data not already in the SDM)?
Hi Jack, thanks for chiming in. Here are two things I had in mind:
To my understanding, SDM has and will only have data available at the census block group level? I wish I am wrong about this, and that SafeGraph can match a few thousands IP addresses from the survey respondents directly. So, my first question is whether there is a channel for directly matching SDM records with known IP addresses.
Secondly, I was asking if what I proposed has been implemented already. I believe people familiar with the matters can improve the accuracy of matching IP addresses to collections of CBG by a huge margin. I was asking if I can avoid inventing my diesel engine and get on a Tesla lol.
@Linfeng_Li, first off, thank you for that diesel v Tesla analogy - it’s great haha
unfortunately, your understanding on the limitation of the SDM data is correct as we see via the SDM release document:
> It is important to note that we take privacy very seriously. In aggregating this data, we employed differential privacy and anonymization protocols to protect and ensure user privacy at all times. We report our insights only at the Census Block Group level.
As for whether or not people have implemented IP matching, I have not seen anything on it. I recently worked with Ookla data and used GeoPandas to match that data to SafeGraph traffic – you might can get some benefit from that (check out the columns and see if you can build a crosswalk) – I am sorry I can’t be of more help here
LINK to Ookla
Thanks a lot for confirming the first assumption. I’ll update here should I manage to pull the threads together.
Great! Looking forward to it!