Do you think this overestimate in areas with less tree cover, or faster-growing sunbelt and western states where permits are pulled for developments but no construction has happened?
Asking from Maine, which is very rural, forested, and where not much housing is getting built.
Great question. Here's what I think might be happening, but this may be over-indexing on certain places I've looked at more closely.
I think certain of these counties have been parceled out, but nothing has been permitted or built. Those parcel maps get aggregated into a commercial offering that Cost Quest licenses for the purposes of building the Fabric. They have machine learning that is using satellite imagery and is trying to find a building on all these parcels. And it works very hard to find something, sometimes finding a shed, sometimes assigning a location with even less than that.
It's probably easier said than done, but I think the location assignment should be aware of how many census housing units are in the area, and not try as hard to assign a location if there are way too many locations already. (How many buildings there should be in an area is also a concern).
I might update this post in the future with some of this depending on how much I find...
Did a decent amount of fieldwork in Rocksprings, TX back in the day, never expected to see a mention of that place ever again, that's wild. In regards to Myles' question I think the ideal thing you'd want to see in an area that is overreported is:
1) Heavily parceled up land all marked as 'single family' by the algorithm. Not much public land.
2) Less tree cover as he mentioned
3) A history of mailing addresses at the location. The vast majority of Sunbelt new builds come with cable/fiber internet so they wouldn't change funding. Counterintuitively, a shrinking population might actually benefit more than a rapidly growing population because locations had mailing addresses at some time.
No way to confirm this but I would guess that Western Nebraska, Western Kansas, West Texas, Eastern New Mexico, and Southwestern Wyoming might have the high counts of locations that aren't actually locations. (Just wild speculation, based on no real data whatsoever.) I could see the rust belt also having similar qualities aside from the tree cover.
I’m curious, what is Rocksprings like? After you get out of the town, what is on all these parcels. In some of them there are buildings. Are those ranch outposts? Do they have power? Your real world experience with what’s in these areas would be really helpful
Feb 3, 2023·edited Feb 3, 2023Liked by Mike Conlow
Not a whole heck of a lot. A decent number of the structures the AI might pick up on could be for hunting. Folks put up really high fences and bring in axis deer/exotic animals for city people to drive in and hunt. Farther West it might be picking up on oil and gas infrastructure. Could also just be agriculture structures that folks abandoned.
If you look for the shadow of electric poles near those BSLs and you don't also see the shadow of a transformer there's a good chance it's just a false positive.
Electric co-ops that offer internet probably have the best data of anyone on that stuff. It'd be interesting to see what they would have to say about the quality of the fabric in rural America.
Do you think this overestimate in areas with less tree cover, or faster-growing sunbelt and western states where permits are pulled for developments but no construction has happened?
Asking from Maine, which is very rural, forested, and where not much housing is getting built.
Great question. Here's what I think might be happening, but this may be over-indexing on certain places I've looked at more closely.
I think certain of these counties have been parceled out, but nothing has been permitted or built. Those parcel maps get aggregated into a commercial offering that Cost Quest licenses for the purposes of building the Fabric. They have machine learning that is using satellite imagery and is trying to find a building on all these parcels. And it works very hard to find something, sometimes finding a shed, sometimes assigning a location with even less than that.
It's probably easier said than done, but I think the location assignment should be aware of how many census housing units are in the area, and not try as hard to assign a location if there are way too many locations already. (How many buildings there should be in an area is also a concern).
I might update this post in the future with some of this depending on how much I find...
Did a decent amount of fieldwork in Rocksprings, TX back in the day, never expected to see a mention of that place ever again, that's wild. In regards to Myles' question I think the ideal thing you'd want to see in an area that is overreported is:
1) Heavily parceled up land all marked as 'single family' by the algorithm. Not much public land.
2) Less tree cover as he mentioned
3) A history of mailing addresses at the location. The vast majority of Sunbelt new builds come with cable/fiber internet so they wouldn't change funding. Counterintuitively, a shrinking population might actually benefit more than a rapidly growing population because locations had mailing addresses at some time.
No way to confirm this but I would guess that Western Nebraska, Western Kansas, West Texas, Eastern New Mexico, and Southwestern Wyoming might have the high counts of locations that aren't actually locations. (Just wild speculation, based on no real data whatsoever.) I could see the rust belt also having similar qualities aside from the tree cover.
I’m curious, what is Rocksprings like? After you get out of the town, what is on all these parcels. In some of them there are buildings. Are those ranch outposts? Do they have power? Your real world experience with what’s in these areas would be really helpful
Not a whole heck of a lot. A decent number of the structures the AI might pick up on could be for hunting. Folks put up really high fences and bring in axis deer/exotic animals for city people to drive in and hunt. Farther West it might be picking up on oil and gas infrastructure. Could also just be agriculture structures that folks abandoned.
If you look for the shadow of electric poles near those BSLs and you don't also see the shadow of a transformer there's a good chance it's just a false positive.
Electric co-ops that offer internet probably have the best data of anyone on that stuff. It'd be interesting to see what they would have to say about the quality of the fabric in rural America.
Where is the FCC file that has the number of units? Thanks.
Here you go: https://us-fcc.app.box.com/v/BSLFabricByGeography