FCC broadband data is biased

The real number of homes with service is probably 93%

Beginning on May 16th, the $42 billion set aside to build out broadband in the 2021 bipartisan infrastructure bill will begin to be dispersed. But there is a growing concern that the money will not be allocated to the areas that need it most. Experts, policymakers, and even agency commissioners agree that the maps created by the FCC, on which funding will be based, are biased. They overstate how many Americans have reliable Internet connections. 

Bad maps mean bad funding outcomes. 

At first glance, biased maps might seem to be a minor issue. However, this minor issue compounds significantly under the decision-making requirements of the infrastructure bill. The bill has a waterfall construction. First, the truly unserved areas will be funded. Then, whatever is left will be allocated to areas that have broadband according to a set of rules. Overstating broadband deployment means that the first tranche of money won’t go to all areas in need.

Although all 50 states are vying for these funds, Georgia alone has released data on broadband deployment. As expected, Georgia’s data does not match the FCC’s data. For reasons described below, Georgia’s data is likely more accurate than the FCC’s data. I have taken Georgia’s dataset and applied a workhorse economic model to estimate which counties are still without broadband in the 49 other states. 

After correcting for two known biases in the data, my model suggests that 92.9 percent of all US homes have broadband, but this number might be as low as 89.85 percent. According to the FCC, 97.5 percent of the population has access to broadband. This 5 to 8 percent difference isn’t negligible, almost 11 million homes are without broadband. And the difference is larger in some areas. The maps below compare differences at the county level. 

None of the maps in this analysis offer the last word on this topic. Rather, they are intended to offer guidance for policymakers and a best guess of where gaps still exist.  

The bias in Form 477 data

All Internet service providers are required to file Form 477 with the FCC twice a year. On this form, providers indicate where they offer Internet service at speeds exceeding 200 kbps in at least one direction (download or upload). This data is compiled and can be downloaded in its raw format on the FCC's Open Data site.

Under the FCC’s current definition, one house in a census block needs to be served for the entire block to be considered completely served in the official numbers. In small blocks in urban regions, this bias doesn't matter all that much, but larger blocks in more rural areas probably mean more bias. This definition has the effect of overcounting the number of served homes and thus undercounting the number of households truly in need. 

Forty states have been working with the National Telecommunications and Information Administration to produce the National Broadband Availability Map (NBAM). Most states have chosen to survey residents or set up a portal to submit requests for service. 

Georgia and Vermont are unique because they have released datasets that effectively count every address as either connected or not connected. But only Georgia has released data that is meant to be used by researchers.  

Thus, the State of Georgia offers an estimate of the true number of served and unserved locations. Importantly, its data collection methods differ from the FCC. It doesn't ask providers if they are offering service in the block as the FCC does. Rather, the state collected map data via a commercial address system and then went to each address and found the number and kind of providers for Internet service. 

Thus the Georgia dataset, with its geographic and demographic differences, stands as a true estimate for broadband. 

Georgia data allows for a new estimate to be calculated

The raw data from The Georgia Broadband Deployment Initiative can be found here.

For each census block, the State of Georgia provides the number of served and unserved locations. This provides one estimate of the number of the unconnected. According to this data, there are 5,283,882 locations, of which 4,825,081 are served, equaling 91.3 percent served. 

The FCC’s own estimates suggest that about 96.0% of people in Georgia have broadband. This leaves a 5 percent gap in the population that may not have broadband access, but are being overlooked in the FCC’s data.

I’ve used the Georgia dataset to calculate new estimates of broadband access in the United States. In the language of economics, we want conditional means. This blog post explains the model in detail.

The map below charts out three measures: the State of Georgia’s estimate of broadband connections, the FCC’s estimate of broadband connections, and the difference between the two.

Other broadband estimates

George Ford, Chief Economist of the Phoenix Center for Advanced Legal and Economic Public Policy Studies, has written extensively on this subject. Using Georgia’s data, Ford found that "there are 14 million unserved locations in the US, though about 5 million of these were addressed in the recent Rural Digital Opportunity Fund (“RDOF”) auction, leaving approximately 9.1 million unserved locations." The maps I offer in this blog post update Ford’s model using more recent datasets from the FCC and the 2020 Census. 

In his earlier work using 2019 data, Ford estimated an overstatement of 5.6 percentage points. More recently, Busby, Tanberk, and Cooper (2021) estimate an overstatement of 11.2 percentage points. 

We know from Ford (2021) that the FCC bias is correlated with the census block size as well as the number of homes in the block. But it is also safe to assume that the census block bias is probably also connected to the age, income, and education of a region. As explained in the full write-up, two models were tested. From here, national data was used to calculate a new estimate of homes with broadband. 

Which areas still lack service?

According to FCC estimates, 97.5% of the population has access to broadband at a service level with 25 Mbps upload and 3 Mbps download. I tested two models, and the real number is probably closer to 92.9% of all homes. 

This first estimate corrects for the size of a region and the number of homes. The model predicts 132,830,990 homes out of a total of 143,620,264 were served for a difference of 10,789,274 homes. 

My second model controls for age, income, and education as well as geographic size and the number of homes. It suggests that the number of homes with broadband might be as low as 89.85%. This model predicts that only 129,043,167 out of 143,620,264 homes have broadband, a difference of 14,577,097. 

It should be noted that the FCC calculates its estimates using population counts. The analysis here uses total housing units, which represent the total number of premises that need to be connected. The map below takes the difference between the FCC estimates and our full model. While the difference is between a housing count and a population count, it highlights those regions where our estimate picks up overcounting by the FCC. 

Estimating broadband service levels helps to narrow the search for funding. The last map in the series aims to do that. It includes expected connections from the Rural Digital Opportunity Fund auction (RDOF). This fund was created to bring “high speed fixed broadband service to rural homes and small businesses that lack it.” 

At the fund’s close in late 2020, $9.2 billion in support was given out by the FCC to 180 different groups to bring service to over 5.2 million unserved homes and businesses. Assuming that the RDOF auction results hold, the total number of places without broadband stands between 9,377,097 and 5,589,274.  

In a letter to the National Telecommunications and Information Administration (NTIA), Senator John Thune reminded the agency that the infrastructure bill “included provisions requiring the use of accurate maps, preventing overbuilding, and prohibiting rate regulation and other unnecessary and burdensome requirements on providers receiving funds.” 

Senator Thune is right. We need to “ensure that this money is effectively and efficiently spent.” But the agencies should also be engaging with the public by providing as much data as possible. 

Georgia has been a leader in providing open data. The federal government needs to do the same. The NTIA needs to follow the lead of Georgia and release a machine-readable set of data to the public. 

A lot of money is about to go out. It is in the public interest to understand where it is headed.

CGO scholars and fellows frequently comment on a variety of topics for the popular press. The views expressed therein are those of the authors and do not necessarily reflect the views of the Center for Growth and Opportunity or the views of Utah State University.