A commentary from BCAP President Todd Eachus
To close the digital divide in Pennsylvania, it is critical that funding is targeted to unserved areas based on reliable data.
Unfortunately, The Center for Rural Pennsylvania’s (CRP) report “Pennsylvania Broadband Access: A Speed Test Analysis” makes several fundamental errors resulting in misleading conclusions. It falls far short of the type of reliable information we need to wisely spend taxpayer dollars dedicated to broadband. In short, reliance on data presented in the CRP report will only lead to an inefficient application of broadband expansion funds and hinder efforts to realize the goal of broadband access for all Pennsylvanians.
The fatal methodological flaws of the CRP study explained below have evidently generated false results.
Philadelphia The CRP report states, “Based on these data, no Pennsylvania county had more than 50 percent of tested users fully served by broadband.” This conclusion ignores the reality that measuring Wi-Fi speed only reveals the speed received by customers in the home, which is dictated by the speed tier they purchase based on their individual needs. In-home Wi-Fi testing does not provide a picture of speeds available to customers.
For example, Philadelphia County – which the report shows with only 27% of users with speeds exceeding 100 Mbps – has at least two Gigabit wireline service providers (Xfinity and Fios) and at least two wireless 5G broadband service providers (AT&T and Verizon Wireless) advertising availability throughout the county.
This means that every Philadelphian has several broadband services available to them, including two offering Gigabit speed. This impressive array of competitive options does not mean customers purchase Gigabit service. In fact, Philadelphians purchasing 200 Mbps service and completing a reliable speed test showing they are receiving 200 Mbps have not provided any insight into the other speeds available for purchase, including the availability of Gigabit speed.
Sullivan County The CRP report indicates that less than 2% of Sullivan County residents have access to 100 Mbps. The U.S. Census reports 2,723 households and 157 employer establishments in Sullivan County. Comcast’s network alone passes at least 1,470 homes and businesses in the county. Because Comcast offers gigabit service throughout its footprint in Pennsylvania, simple math shows that Comcast alone serves at least 54% of Sullivan County households with broadband service significantly faster than the 100 Mbps threshold.
Crawford County The CRP report indicates that only 22% of Crawford County homes have access to 100 Mbps service. The U.S. Census reports 45,002 households in Crawford County. Armstrong’s network passes 30,500 (67.8%) of homes and business in the county and a minimum of 500 Mbps service is available to each of them.
PA PUC Telco List In addition to cable broadband providers, Pennsylvanians may also receive broadband service from a variety of telephone company providers. A list of wireline and wireless companies by county compiled by the PA PUC can be found here.
The Measurement Lab (M-Lab) “Network Diagnostic Test” (NDT) is a not a speed test. Experts and M-Lab agree it should not be used to measure broadband speed (adoption) or availability. Attempting to use it as a speed test has well-known deficiencies. Academics have found the M-Lab NDT used in the study is highly unreliable and leads to underestimates of actual access.
“This, in fact, would not be an appropriate use of the data as far too many factors confound such an interpretation.” – MIT’s David Clark & Sara Wedeman “The M-Lab NDT test consistently underestimates access link throughput.” – Princeton University’s Nick Feamster
Industry and think tank observers also have highlighted problems with M-Lab’s NDT. High Tech Forum has explained in detail the technical flaws with M-Lab’s test. Ookla showed how M-Lab speed test data will give a distorted picture of rural connectivity (availability).
The CRP report also confuses broadband subscription speeds with broadband deployment speeds. Even the best speed test will only measure the speed to which the consumer subscribes. However, the CRP study concludes a location is “unserved” – that is, it does not have broadband deployment of at least 25/3 Mbps – if the speed test of the subscriber’s service tier falls below the broadband standard.
Conflating what a consumer subscribes to with what the consumer could subscribe to will misdirect deployment funding to locations that already are served with broadband. The Technology Policy Institute’s Scott Wallsten has explained this common mistake as it relates to a different report in which the same organization that CRP hired for the study (X-lab) was involved.
Equipment Placement/Testing at the Modem
The most reliable method of testing broadband speeds is via a direct wireline connection to the modem consistent with the FCC’s “Measuring Broadband America” program’s methodology. In addition to using unreliable testing mechanisms, such as the M-Lab test, even technically sound in-home Wi-Fi speed test results tend to vary for a number of reasons.
Please visit the Ookla Speedtest website for additional information about how to ensure that Wi-Fi speed tests are as accurate as possible.