Key Takeaways

Contents

Key Takeaways 1

Introduction. 2

High-Speed U.S. Broadband Deployment Is Widespread. 2

Deployment of High-Speed Services Outpaces Adoption. 3

Experienced vs. Advertised Speeds in the United States 5

How U.S. Speed Test Results Stack Up. 5

What Is High-Speed, Anyway? 7

Conclusion. 8

Endnotes 9

Introduction

Opponents of the current private-sector-provided broadband system have long engaged in a campaign to convince people that U.S. broadband is deficient. That way, they believe, policymakers will support replacing the current system with government-owned networks. One of their arguments is that U.S. broadband speeds are slow, such as when the Open Technology Institute concluded that “people in the U.S. pay more for slower internet than [do their] European, Canadian, and Asian counterparts.”[1]

Network speeds, maybe more clearly than deployment or adoption rates, can be measured and compared with those of other countries, so this assertion should be easily proven or disproven. In fact, as will be shown ahead, the claim that U.S. speeds are slow is simply wrong. At the same time, there is some room for confusion over how fast fast actually is, especially in the face of continuing technological advancements and enduring disagreement over what baseline speeds should be.

Overall, the question of whether U.S. broadband speeds are slow can be assessed in a few ways. First, how widespread is the infrastructure to support higher-speed broadband? Second, what is the uptake rate of higher-speed services? The point at which high-speed-tier supply outpaces demand can be looked at as the point at which consumers judge additional increases in speed less worthwhile.

Both of these play into the single metric that some would look to when assessing how U.S. speeds stack up: What are average speeds in the United States, and how do they rank both against peer nations and against the requirements of common applications, such as Zoom and Netflix, that have become an integral part of modern life?

High-Speed U.S. Broadband Deployment Is Widespread

The overall speed of a nation’s broadband depends on a few factors working in concert. First is the existence of infrastructure built to support higher speeds. The current Federal Communications Commission (FCC) standard identifies 25 megabits per second (Mbps) download/3 Mbps upload as the minimum for broadband speeds. In some ways, the de facto speed benchmark for all technology deployed from here on out has become 100/20 Mbps regardless of official policy, as new networks built with forthcoming Broadband Equity, Access, and Deployment (BEAD) Program funding must reach that speed benchmark at a minimum.[2] For the most part, existing broadband coverage already supports these higher speeds: As of June 2022, broadband capable of reaching speeds of 100/20 Mbps passed almost 89 percent of U.S. residential units.[3] Infrastructure capable of 250/25 Mbps passed 86 percent of households.[4]

These numbers are both absolutely and comparatively high. To put them in context, Ofcom, the United Kingdom’s version of the FCC, publishes an International Broadband Scorecard that compares deployment statistics from a few key economies. The penultimate edition calculated that, in 2020, U.S. fixed deployment capable of at least 100 Mbps ranked 8th highest out of 18 countries studied.[5] A quarter of U.S. households already were passed by networks at gigabit speeds, on par with South Korea, a country that is generally hailed as a technological frontrunner. The more recent edition, with data as of 2021, has the United States up in the rankings to 6th place in networks reaching 100 Mbps, though only 11 countries are included in this analysis.[6]

Moreover, advancements in mobile technology have moved some mobile offerings into speed tiers that, traditionally, only wireline connections could achieve. The newest generation of mobile technology, 5G, is capable of speeds at or above 100 Mbps. According to the European Telecom Network Operators (ETNO) State of Digital Communications 2022, 5G covered 93.1 percent of the U.S. population in 2021.[7] This is on level with South Korea’s 5G coverage, which slightly led at 93.9 percent, and ahead of Japan’s and the European average, at 81.2 and 62 percent, respectively. U.S. providers are particularly well-positioned to offer the fastest high-speed 5G coverage because of the disproportionate allocation of high-band spectrum to mobile providers in the United States.[8] By September 2022, high-speed 5G already covered 62 percent of Americans, a number that will only rise as more providers come online.[9]

Deployment of High-Speed Services Outpaces Adoption

The speeds available through broadband built out in the United States are so high, in fact, that higher-speed deployment consistently outpaces demand. The truest assessment of what speeds consumers value is how many people choose to subscribe to higher speeds given the availability of the infrastructure. This measure—the adoption rate—is distinct from total uptake because it only measures consumer choice in the areas where the infrastructure to provide those speeds already exists.

The FCC’s Fourteenth Broadband Deployment Report measures adoption rates for fixed terrestrial services at certain speed tiers using 2019 data. For the 25/3 Mbps speed tier, which coincides with the FCC’s current speed benchmark, a total of 69.4 percent of households in areas that have been passed by broadband at that speed benchmark subscribe to that broadband.[10] Bandwidth level and proportion of subscribers have an inverse relationship: By the 50/5 Mbps benchmark, 64.8 percent of households, given the opportunity, choose to subscribe, and by the 100/10 Mbps benchmark, the adoption rate falls to 50.9 percent.[11]

Every given speed tier already sees diminishing adoption rates, but somewhere past the 100 Mbps benchmark is a clear inflection point after which consumers no longer signal the value of higher speeds. Of the percentage of households passed by networks capable of 250/25 Mbps, only 9.1 percent of them choose to adopt that speed level, according to the 2019 data.[12]

These numbers tell us something important: The majority of consumers, even given the option of higher speeds, choose a lower speed tier as one that is sufficient for their everyday needs. Therefore, by this definition, U.S. speeds are more than fast enough, as the majority of consumers elect to subscribe to a plan lower than what is available.

However, the last few years have seen a marked trajectory toward higher speeds. The FCC’s Measuring Broadband America (MBA) report, an expansive study of broadband speeds in the United States, calculates a 2020 weighted average advertised download speed of 193.9 Mbps, a 33 percent increase from the year before.[13] The FCC notes that a large percentage of consumers migrated to a higher speed tier between 2019 and 2020, either because they increased their subscription or because the provider increased the speeds of an existing plan.[14] OpenVault’s Broadband Insights Q3 2022 provides even more-recent data and finds that the majority of subscriptions now exist within the 200–400 Mbps speed tier, which nearly doubled to 54.8 percent of subscribers within a year.[15]

Because OpenVault reports percentage of subscribers rather than adoption rate, we cannot distinguish instances of consumers choosing to purchase high speed services from those that simply buy what is available. In general, upward shifts in subscribers’ speed tiers can represent either a change in public opinion as to what speeds are necessary or a pattern of providers increasing the speed allowances of popular existing services. But by every account, high-speed plans are widely available across the United States. Their availability is also consistently increasing, as providers tend to boost speeds as technology and consumer demand allow. Comcast, for example, recently announced that it is raising speeds for many of its existing plans, some by as much as 200 Mbps.[16]

Moreover, these high-speed plans are available at reasonable prices. The International Telecommunication Union publishes pricing data that benchmarks international fixed broadband prices against an affordability target set at 2 percent of Gross National Income per capita (GNIpc). As of 2021, U.S. fixed broadband cost 1 percent of GNIpc, slightly cheaper than the average for high-income countries in the ITU dataset.[17] The highest-quality mobile option in the dataset is even cheaper, at 0.7 percent of GNIpc.[18]

The true pricing situation in the United States is also likely better than those numbers might suggest, since pricing comparisons should account for network quality and cost factors, both of which play a role in what prices the market sets. In the 2020 Communications Marketplace Report, the FCC constructed a pricing index that compared international broadband prices while considering quality factors such as network speed and availability of content, alongside cost factors such as demographic and geographical differences among countries. Once these factors are taken into account, the United States ranks 2nd least expensive out of 26 countries.[19] A recently released Oxford Economics study looks at the cost of mobile plans as a percentage of average household disposable income and similarly finds that U.S. offerings rank favorably against those of its international peers. Even the highest-quality offering—with unlimited voice, SMS, and data—ranks 4th cheapest out of 15 countries studied, at 0.6 percent of average disposable income.[20]

Experienced vs. Advertised Speeds in the United States

The reality of providers offering “best effort” plans with speeds advertised as “up to” a certain number is that consumers might sometimes experience bandwidth capacity that’s lower than the number attached to their plan. However, in the United States, broadband plans have historically achieved—and continue to achieve—close to their advertised speeds.

The FCC’s MBA report is the most in-depth official study of the actual speeds of major participating U.S. providers, and the most recent iteration reports data up until 2020. The report measures speeds from 10 Internet service providers (ISPs), some of which deploy multiple broadband technologies, for a total of 12 ISP and technology combinations.[21] Though the report’s findings are limited to participating providers, they include the provision of such major services as Verizon’s fiber service and Charter’s and Comcast’s cable. Among participating providers and technologies, advertised speeds across plans are weighted by the number of subscribers to find the most popular speed tiers. Overall, the report finds that American broadband consumers largely get what they pay for: Of the 12 combinations, 9 had a median that met or exceeded the advertised download speed during peak Internet use hours, and only 1 achieved below 90 percent of that speed.[22]

The FCC also assesses networks using an 80/80 consistent speed measure, which reports the ratio of experienced to advertised speeds for at least 80 percent of users, during 80 percent of peak hours. This measures the consistency of plans’ achieved speeds. Overall, the included ISPs, particularly those using cable and fiber technologies, were consistent almost across the board, suggesting that U.S. speeds are largely reliable even during peak usage hours during which there is the most traffic to contend with.[23]

How U.S. Speed Test Results Stack Up

All these factors taken together suggest that U.S. speeds are consistently fast. And to put these numbers in context, the United States tends to perform well in international speed tests. These datasets aggregate individual results that measure the speeds of participating consumers’ connections at the time they run the test. Because of this, overall speed test results capture something about both the universe of available consumer speed tier choices and the percentage of advertised speeds providers generally achieve.

Speed tests are generally not the best way of assessing a nation’s average speeds because the population that participates in them is not truly random. Many of the users taking speed tests are likely experiencing slow speeds and running a test to diagnose that. The results also incorporate, and cannot be disaggregated from, factors separate from a network’s native speed, such as computer processing power and in-home Wi-Fi; thus, speed test results are always a lower bound and likely underestimate the speeds an ISP delivers to a subscriber’s home. But for all these inherent flaws, speed tests remain one of the best ways to compare international speeds.

The Organization for Economic Co-operation and Development (OECD) publishes and ranks international average speeds based on major speed test platforms, including Ookla and Measurement Lab (M-Lab).[24] The Ookla data is from January 2021, and includes every OECD country save Iceland. Switzerland ranks first, at an average experienced fixed download speed of 188.9 Mbps.[25] Denmark is in second place, at 185.4 Mbps. At 173.7 Mbps download speed, the United States ranks fifth out of OECD, significantly ahead of peer nations such as Japan, Germany, and the United Kingdom.[26] Even South Korea, which generally heads the technological curve, actually ranks below the United States in this dataset at 171.3 Mbps.

OECD also published M-Lab data from July 2019 to June 2020.[27] Every OECD country has data available for this metric, ranking from 118 Mbps download (Luxembourg) all the way down to 8.8 Mbps (Colombia). The United States measures up in ninth place at 71.3 Mbps.[28]

Though its rank is similar to that of the Ookla data, overall, the M-Lab statistics are biased toward a lower range of experienced speeds in every country. Differences in the results can be largely attributed to differences in methodologies: Ookla runs its tests through multiple streams of data, so it essentially measures the link capacity of a network or its total ability to carry traffic on a certain path.[29] Ookla tests are also performed on servers that are generally hosted within, and therefore only really measure the performance of, a single network. In contrast, M-Lab test servers are housed in data centers, so they are more likely to capture performance across various networks and interconnection points.[30] They are also conducted using a single stream of data, which is more vulnerable to one-off issues, such as packet loss, that impact reliability. One interpretation of these differences is Ookla’s methodology is measuring something more akin to “best-case scenario” speeds, while M-Lab’s is more aligned with individual user experience and more able to identify performance issues. On the flip side, many such performance issues may be on the user side, or caused by infrequent network blips that aren’t consistent enough to play a significant role in overall average speeds.

Data directly from Ookla’s Speedtest Global Index offers average speeds of both mobile and fixed connections. The November 2022 data ranks the United States sixth overall in fixed broadband, at a median download speed of 189.5 Mbps.[31] (Recall that this is over seven times the amount set by the FCC as the baseline for broadband download speeds.) Chile, China, Singapore, Thailand, and Hong Kong are the only economies that rank above the United States’, with median download speeds ranging from 216.5 Mbps (Chile) to 194.4 (Hong Kong).[32] U.S. median upload speeds in the same month breached the 20 Mbps threshold at 22.5 Mbps, again roughly seven times the upload benchmark set by the FCC.

In addition, Ookla offers average latencies: U.S. latency as of the November 2022 data is 14 ms.[33] Latency is perhaps the least suited to any type of ranking system, because once a provider achieves average latency in the low double digits, there is very little to be gained by pitting one country against another because of a difference of a few milliseconds (ms). The difference between the U.S. average and that of the top-ranking country’s (Chile) latency of 6 ms is not enough to make a significant difference in the user experience. In fact, multiple sources recommend that online gamers—possibly the most latency-sensitive users today—aim for a latency at or below 100 ms.[34] Given that, results in the low double digits all appear to be interchangeably quick.

Though speed test results are a helpful comparison among countries, some of the limitations laid out previously mean they generally provide more of a lower bound than an average speed for a country. Ookla’s U.S. speed test result, if taken as a national average speed, would give an impression of lower speeds in the United States than others have calculated. In fact, OpenVault, which provides market data on broadband consumption patterns, finds that the average U.S. household has a broadband download speed of 347.8 Mbps, and an upload speed of 23.5 Mbps.[35] But whatever the true average speed, U.S. speeds are doing comparatively very well, ranking in the top 10 countries in both speed tests.

What Is High-Speed, Anyway?

One final way to evaluate the existence of high-speed connections is to compare them against basic requirements for everyday online use. Today, speed requirements for the most commonly used applications make up an extremely small, and diminishing, percentage of average speeds. Basic online functions such as web browsing probably eat up a maximum of 1 Mbps; instant messaging requires even less than that.[36] Even intensive use of high-bandwidth services, such as holding multiple simultaneous Zoom calls on the same connection, requires bandwidth in the low double digits—Zoom’s HD video, for example, requires a connection of 3 Mbps down/3.8 Mbps up.[37] Google Meet’s HD video option requires an upload speed of at least 3.2 Mbps and a download speed up to 4 Mbps, depending on the number of participants.[38] To stream a movie in the highest quality available, 4K/Ultra HD, Netflix recommends a download speed of at least 15 Mbps.[39]

Calculating bandwidth speeds to support multiple simultaneous uses has become more important as Internet reliance has intensified, and the average household’s bandwidth needs have increased. Many multiperson households are now more likely to have members either working from home, attending school, or both, and therefore the need for a home connection to support multiple video calls has intensified. Important to note here is that bandwidth usage is additive: Concurrent streaming of video on, say, Netflix (15 Mbps) while participating in a group Zoom call (3 Mbps) generally uses a total of 18 Mbps down.

Overall, the FCC has calculated that the average student or telecommuter requires a downstream bandwidth connection somewhere in the range of 5–25 Mbps.[40] The FCC also estimates the bandwidth range needed for different usage tiers, each of which considers factors such as the number of devices and users and the type of activity the connection needs to support.[41] Of the household types and usage patterns the FCC lays out, only a few require connections of over 25 Mbps download: generally, households with multiple connected devices intended for multiple, often intensive, uses. The FCC categorizes one user on one device whose Internet habits are moderately intensive—including basic functions such as email and browsing concurrent with a bandwidth-heavy application such as video conferencing or HD streaming—as a user in need of basic service, which it defines as between 3 and 8 Mbps down.[42]

However, given the demands placed on Internet connections today, it’s probably fair to assume that the real average use per person is closer to 15 Mbps. This accommodates higher-usage tasks such as the streaming of some high-quality video and leaves some room for the increasingly common occurrence of multiple devices per person (though even if somebody has multiple devices, it seems unlikely they would need to use them all, simultaneously, to their full capacity). If, per Census data, the average U.S. household has 2.5 people, the amount of bandwidth actually used by an average household maxes out around 40 Mbps.[43]

By definition, this 40 Mbps criteria would be above that needed by households with smaller-than-average household size; on the flip side, it would be lower than what larger or more intensive-use households would need. Using 40 Mbps as the rough average bandwidth requirement in the United States, we found that both Ookla’s and M-Lab’s speed test results significantly outpace this average, even with additional bandwidth left unused as a buffer to head off throttling or accommodate traffic increases. The infrastructure to support these speeds covers the vast majority of U.S. households.[44]

These back-of-the-napkin calculations highlight the discrepancy between U.S. average speeds and the theoretical bandwidth requirements for most everyday applications. It appears that U.S. broadband is more than fast enough for common modern-day use.

Conclusion

U.S. high-speed deployment is very strong. In fact, the FCC could officially shift the broadband benchmark all the way up to 100/20 Mbps and the significant majority of U.S. households would still have coverage. U.S. high-speed networks are widespread and competitive with peer nations’ at every speed tier, and actual speeds experienced by U.S. customers are consistent with those advertised and among the fastest in the world. Consumer-side data supports this: Adoption rates drop off long before availability declines.

Because so many factors play into average speeds, there is no one perfect metric to make international comparisons. But from every angle, U.S. broadband speeds are strong and in no way support the narrative advanced by those seeking to establish a government-owned broadband system. Policymakers should reject their policy advice. Since speed is not an area of weakness, piling resources there will waste funds that otherwise could have been directed toward actually closing the digital divide.

Endnotes