- An independent study assessed the effectiveness of the USD 10 billion Connect America Fund (CAF) program, which seeks to improve Internet access in rural and underserved regions in the USA.
- Over 40% of CAF-funded addresses appear unserved by the corresponding CAF-funded ISP.
- To improve future government-funded Internet access and affordability programs, broadband infrastructure policymaking requires rigorous oversight and empirical evaluation.
The 2021 US Infrastructure Act’s Broadband Equity, Access, and Deployment – or BEAD – Program is the American government’s latest and most significant investment in improving high-speed broadband connectivity and affordability in traditionally underserved areas.
According to the Pew Research Center, 20% of US adults in 2023 didn’t subscribe to a high-speed broadband Internet service at home, down from 30% in 2022. Most of these people live in rural areas. Coincidently, this has made these areas most reliant on smartphones for online access.
Before the BEAD program starts to materialize, there is a lot we can learn from the government’s similar programs, including the Connect America Fund (CAF), which provided USD 10 billion in subsidies to Internet Service Providers (ISPs) in exchange for deploying new infrastructure in these regions. One significant lesson is the need for better government and public visibility into the efficacy of these interventions.
Like similar subsidy programs, CAF relied heavily on self-reported data from the benefitting ISPs. Recently, this data has been questioned for not representing the whole picture of Internet inequity.
At the University of California, Santa Barbara, and Berkeley campuses, our team recently sought to address this problem by designing a more transparent and replicable framework for assessing these policy interventions. We applied this to CAF data to determine whether:
- Addresses certified as served by the ISPs during the CAF program were served.
- The available services met CAF’s minimum service quality requirements of a 10 Mbps download speed and 1 Mbps upload speed.
- Regulated monopolies offered better service quality to users compared to existing unregulated monopolies.
We focused on four ISPs that accounted for more than 50% of the CAF program addresses: AT&T, Frontier, Centurylink, and Consolidated. We used a robust sampling strategy, endeavoring to query at least 10% of the addresses in each census block group covered by CAF in the 15 states we examined. See our paper for further details.
Serviceability and Performance Vary
As shown in Figure 1, we observed varying serviceability rates (the estimated fraction of addresses an ISP actively serves) across our four studied ISPs.
CenturyLink’s highest serviceability rate was encouraging, given that it received the most CAF funding ($1.84 billion) of any ISP. Consolidated Communications (84%) and Frontier (71%) had the second and third-highest serviceability rates. AT&T’s low serviceability rate (32%) was especially concerning, given that it ranked second among all ISPs in terms of total funds received ($1 billion) and first in terms of addresses certified as “served.”
Figure 1—Variability of the estimated fraction of addresses each ISP actively serves. The longer the box, the greater the variability.
We explored program compliance by comparing the speeds and prices of the plans advertised on the ISPs’ websites to the FCC’s CAF program standards that mandate homes get a download speed of at least 10 Mbps.
Overall, our data analysis indicated that 67% of CAF addresses did not receive service from the studied ISPs that comply with the FCC’s program rules (including those addresses where service is unavailable at all). Table 1 shows the available broadband plans per ISP in more detail.
| AT&T (176.53K) | CenturyLink (82.83K) | Consolidated (21.98K) | Frontier (127.86K) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Certified | Advertised | Certified | Advertised | Certified | Advertised | Certified | Advertised | ||||||||
| Mbps | % | Mbps | % | Mbps | % | Mbps | % | Mbps | % | Mbps | % | Mbps | % | Mbps | % |
| 10 | 100 | 0 | 67.660 | 10 | 100 | 0 | 10.024 | 10 | 86.02 | 0 | 10 | 99.98 | 0 | 30.611 | |
| AT&T Internet Air | 5.052 | 0.5 | 0.298 | 25 | 12.87 | 3 | 100 | 0.01 | Frontier Internet | 53.255 | |||||
| 0.768 | 1.153 | 1.5 | 1.996 | 100 | 0.77 | 7 | Unknown Plan | 12.138 | |||||||
| 1 | 0.976 | 3 | 15.036 | 1000 | 0.33 | 10 | 10 | 0 | |||||||
| 3 | 1.786 | 6 | 5.664 | 11-99 | 11-99 | 0 | |||||||||
| 5 | 2.479 | 10 | 32.52 | 100-999 | 100-999 | 0.098 | |||||||||
| 10 | 3.135 | 11-99 | 34.145 | 1000 | 1000 | 3.895 | |||||||||
| 11-99 | 9.628 | 100-999 | 1.78 | ||||||||||||
| 100-999 | 0.359 | 1000 | 0 | ||||||||||||
| 1000 | 7.767 | ||||||||||||||
Table 1—Available broadband plans per ISP showing the percentages that were certified and advertised.
Finally, we compared the service received by addresses covered by the CAF program with nearby monopoly-served neighbors (in the same census block) that are not. The same CAF ISP serves these nearby comparators but without being subject to the CAF program rules.
We found that, in most cases, the plans offered at CAF-subsidized addresses were the same as those provided at other, nearby unregulated addresses. Where there was a difference, ISPs typically offered addresses subject to CAF’s program rules better service than nearby addresses that received unregulated service.
In the cases where the CAF addresses received better plans, there was often a substantial improvement, with a median increase of around 75%. In short, where CAF’s program rules improve consumer service, that improvement is significant.
>40% of CAF-funded Addresses Unserved
These findings are concerning as ISPs have accepted public subsidies on the condition that they offer broadband access that meets specific minimum service standards. Still, these ISPs do not appear to be living up to their end of the bargain.
Over 40 percent of CAF-funded addresses seem to be unserved by the corresponding CAF-funded ISP, thereby contradicting the ISPs’ own certifications.
Our study highlights the need for rigorous oversight and empirical evaluation in broadband infrastructure policymaking. We hope that our conclusions improve both the design of future broadband programs (by, say, considering past compliance with funding conditions for future grants) and the post-hoc verification mechanisms used to evaluate them.
Contributors: Varshika Srinivasavaradhan, Laasya Koduru, Elizabeth Belding, Arpit Gupta, Tejas N. Narechania, Kevin Zhang, and Xuanhe Zhou.
Haarika Manda is a Graduate Student Researcher at the University of California, Santa Barbara.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of the Internet Society.