Thank you for sharing these analyses of the FCC data. I work in local government and am attempting to summarize the facts around the FCC data for our elected officials and community. I am not an expert on internet service by any means. This national context is really helpful. I have a methodological question about binning the speeds into "best available" categories. I assume this is a boolean test considering both the upload and the download reported. Is that understanding correct? For example, if a provider reported internet speeds at 100 down, 10 up, that would go into the 25/3+ category, since it falls short of 20 up? Same for 1,000/10?
I started to look at data by Provider, to see how well the Tier-1, Tier-2, and MSOs actual deployment numbers track against what they've stated publicly (as well as finding out how much FTTH they've deployed when they don't disclose that to shareholders). The datasets are too big for Excel,although I can get some data out of pivot tables. Unfortunately, I have to put this aside for lack of time.
There's a lot of understanding of the US access networks buried in there for GIS, but for the fact that locations are geocoded only to the census block rather that Lat-Lon or street address. To unlock that data requires a license from CostQuest (!!!), which I doubt is available to researchers. Supposedly at some point, there is supposed to be a click-through license to Lat-Lon.
I wasn't aware that a BSL could be an apartment building, rather than all of the units inside the building. That would make it really difficult to quantify how many households are actually unserved.
And don't get me started on the topic of apartment building owners who won't allow carriers who are at the curb to deploy in their buildings. Grrr.
Congrats for downloading the whole dataset. I have been trying to do that, but my program is still running. I didn't notice some of the data missing that you flagged; for example, I downloaded data for CO cable providers. A few data sets didn't download properly, but otherwise it seems complete to me (at least so far). However, one thing I haven't found is anything that I can download with unserved locations. Where did you find that information? Thanks.
So by unserved locations you are counting locations that show up in the database that have service below 25/3 correct, such as service at 10/1 only? I am wondering how to find locations that don't have service at any speed. That was something that could be estimated before from total households or housing units in a Census block, but that doesn't seem doable now.
Hi Mike,
Thank you for sharing these analyses of the FCC data. I work in local government and am attempting to summarize the facts around the FCC data for our elected officials and community. I am not an expert on internet service by any means. This national context is really helpful. I have a methodological question about binning the speeds into "best available" categories. I assume this is a boolean test considering both the upload and the download reported. Is that understanding correct? For example, if a provider reported internet speeds at 100 down, 10 up, that would go into the 25/3+ category, since it falls short of 20 up? Same for 1,000/10?
Thank you! Looking forward to more updates!
Sara in Oregon
Hi Sara, Yup, that's exactly right. Both of those examples would be "underserved" at 25/3 or better, but not served at 100/20 or better.
Thank you Mike!
Mike,
Thanks for this. Do you know where we can find the underlying dataset that this aggregate spreadsheet was built from. Thanks!
Cheers,
Bryan
it's just all these individual files downloaded from the FCC site and combined together in a database: https://broadbandmap.fcc.gov/data-download
Awesome! Thank you.
Mike,
This is great!
I started to look at data by Provider, to see how well the Tier-1, Tier-2, and MSOs actual deployment numbers track against what they've stated publicly (as well as finding out how much FTTH they've deployed when they don't disclose that to shareholders). The datasets are too big for Excel,although I can get some data out of pivot tables. Unfortunately, I have to put this aside for lack of time.
There's a lot of understanding of the US access networks buried in there for GIS, but for the fact that locations are geocoded only to the census block rather that Lat-Lon or street address. To unlock that data requires a license from CostQuest (!!!), which I doubt is available to researchers. Supposedly at some point, there is supposed to be a click-through license to Lat-Lon.
Regards,
Dan
I wasn't aware that a BSL could be an apartment building, rather than all of the units inside the building. That would make it really difficult to quantify how many households are actually unserved.
And don't get me started on the topic of apartment building owners who won't allow carriers who are at the curb to deploy in their buildings. Grrr.
Exactly correct. See: https://mikeconlow.substack.com/p/fcc-update-on-new-broadband-maps
This is a great update & much appreciated.
Congrats for downloading the whole dataset. I have been trying to do that, but my program is still running. I didn't notice some of the data missing that you flagged; for example, I downloaded data for CO cable providers. A few data sets didn't download properly, but otherwise it seems complete to me (at least so far). However, one thing I haven't found is anything that I can download with unserved locations. Where did you find that information? Thanks.
I tried downloading CO Cable twice and both times it appeared to stop abruptly. Maybe I'll try again.
https://imgur.com/a/qrWIA58
For unserved locations, have to download all the service records for every location and then calculate whether each location is unserved or not...
So by unserved locations you are counting locations that show up in the database that have service below 25/3 correct, such as service at 10/1 only? I am wondering how to find locations that don't have service at any speed. That was something that could be estimated before from total households or housing units in a Census block, but that doesn't seem doable now.