Thursday, May 30, 2019

U.S. Internet Access Speeds are Improving Fast

Without any question, the latest Federal Communications Commission report on U.S. broadband is going to be criticized. The data overstates deployment, it will be argued. Some might argue 25 Mbps is not broadband. The thrust of the complaints will be that the U.S. digital divide is either not closing or not closing fast enough. Some conceivably will argue the divide is getting worse.

But coverage and speeds are growing. Sure, there is a gap between rural and urban. There likely always will be some gap. But the gaps are closing.




All methodological shortcomings noted, it is hard to argue that the U.S. speeds are too low, or getting worse. There really is not any evidence for that view. In 2018 alone, U.S. average internet access speed  on fixed networks grew 36 percent, according to Ookla Speedtest.

In 2019, U.S. fixed internet service provider were 96 Mbps downstream, according to Speedtest. Even mobile average speeds were in the 33 Mbps range in 2018.






Why Big Public Spending Projects Often Produce No Significant Economic Benefit

Big public investments including sports stadiums and government-owned broadband often are controversial because the claimed benefits (new economic activity, especially) cannot actually  be shown to exist.

In the case of public funding for sports stadiums, the purported new economic activity is simply shifted from other expenditures in the same community, with no actual net increase in economic activity.

“NFL stadiums do not generate significant local economic growth, and the incremental tax revenue is not sufficient to cover any significant financial contribution by the city,” said Roger Noll, a senior fellow at the Stanford Institute for Economic Policy Research.

“A new sports facility has an extremely small (perhaps even negative) effect on overall economic activity and employment,” an analysis  by the Brookings Institution has found.

In a 2017 poll, 83 percent of the economists surveyed agreed that "providing state and local subsidies to build stadiums for professional sports teams is likely to cost the relevant taxpayers more than any local economic benefits that are generated."

Sports economist Michael Leeds suggests that professional sports have very little economic impact, noting that a baseball team (with 81 regular-season home games per year) "has about the same impact on a community as a midsize department store." Doubtless that small an impact would normally be thought unreasonable for the large amount of public cost.

Leeds suggests that if every professional sports team in Chicago (Cubs, White Sox, Bears, Bulls, and Blackhawks) were to suddenly disappear, the economic impact on Chicago would be a fraction of one percent.

Moreover, most economists highlight an important pitfall when politicians or stadium funding advocates tout the economic impact of stadiums: the failure to include opportunity costs.

The opportunity cost is the value of the next-best alternative when a decision is made; it is what is given up, whether that is in roads, bridges, schools, parks, riverfront improvements or anything else with expected positive economic impact.

Also ignored is the fact that what is spent by consumers at stadiums, and for stadiums, simply displaces spending that would have occurred elsewhere.

If they were not spending on sporting events, they would instead spend on museums, movies, concerts, theater, restaurants, and so on. Because consumers have limited entertainment budgets, dollars spent at a new stadium are simply diverted from other spending.

That might also be true for government-owned broadband networks that compete with private broadband suppliers.

In a new study, The Rewards of Municipal Broadband: An Econometric Analysis of the Labor Market, Phoenix Center Chief Economist Dr. George Ford and Phoenix Center Adjunct Fellow Professor R. Alan Seals (Auburn University) use data obtained from the U.S. Census Bureau’s American Community Survey to quantify the economic impact, if any, of the county-wide government-owned network (GON) in Chattanooga, Tenn. on labor market outcomes.

“Across a variety of empirical models, we find no payoffs in the labor market from the city’s broadband investments,” they conclude. “We find almost no statistically significant effects for a wide range of important labor market variables, with the possible exception of a reduction in labor force participation.”

The study looked at private-sector labor force participation, employment status, wages, information technology employment, self-employment, and business income, “all of which appear unaffected by the GON,” the researchers say.

Though Chattanooga’s Mayor Andy Berke has claimed that the city’s nearly $400-million network was responsible for a decline in unemployment in the city from 7.8 percent to 4.1 percent over the 2012 to 2015 period, over the same post-recession period the nationwide unemployment rate fell from 7.5 percent to 4.7 percent, they note. So it is hard to isolate any impact other than general economic conditions for the decrease.

There are some key caveats. A new Volkswagen factory, planned before the GON was launched, did open at about the same time as the network began operations. “Marginal employment effects in auto manufacturing closely match the plant’s employment levels,” the researchers note.

Also, since Chattanooga’s system is an overbuild of multiple private providers, “we stress that our findings may not be generalized to areas where broadband services are not available absent the municipal system,” Ford and Seals say.

“Also, our results cannot speak to the benefits of high-speed Internet services generally, since broadband Internet service was and remains available in Chattanooga absent the municipal system,” they say. “Thus, our results indicate only that building a government-owned network in markets where privately provisioned broadband is generally available has no favorable effect on labor market outcomes.”

“The data suggest local governments must look outside the labor market to justify the sizable investments in municipal broadband systems,” the authors say.

Ads or Higher Prices? Most Choose Ads

About 70 percent of Hulu subscribers buy the $5.99-per-month ad-supported plan, some 30 percent taking the $11.99 ad-free version, which gives you some idea of consumer appetite for ad-supported video, when weighed against paying more money for an ad-free experience, at least under conditions when multiple subscriptions are purchased.

And though we almost-never seem to worry about it, when defraying the cost of a video subscription with advertising, consumers are, in part, “the product,” not simply the buyers of the product. With all the apparent present concern for “privacy,” it is worth remembering that “becoming the target of ads” often is the price paid for discounts on products we wish to consume.

Hulu has 28 million accounts and 82 million viewers (an average of 2.9 viewers per account). About 70 percent, or 58 million, are on the ad-supported plan, according to Peter Naylor, senior VP, head of advertising sales.

Hulu’s ad business generated almost $1.5 billion in ad revenue in 2018. Compare that to the $72 per sub (ad version) and $144 per account Hulu earns from subscription fees. If 70 percent of total revenue came from subscription fees that same year, Hulu might be making $5 billion from subscription fees.

By 2020, Hulu might be making as much as 80 percent of its revenue from subscription fees, as the percentage of subscription revenue seems to have been rising steadily over the past couple of years.


Though it is reasonable to suggest that Netflix dominance of the U.S. video streaming market is going to be challenged by Apple, Disney, Warner Media and Hulu, it also is possible to argue that Netflix now is a global brand that some other competitors (Dish, Warner Media) might be hard-pressed to match. Disney and Apple arguably will be stronger globally. Amazon might fall somewhere in between.

Study Finds No Employment Lift from Municipal Broadband

Big public investments including sports stadiums and government-owned broadband often are controversial because the claimed benefits (new economic activity, especially) cannot actually  be shown to exist.

In the case of public funding for sports stadiums, the purported new economic activity is simply shifted from other expenditures in the same community, with no actual net increase in economic activity.

That might also be true for government-owned broadband networks that compete with private broadband suppliers.

In a new study, The Rewards of Municipal Broadband: An Econometric Analysis of the Labor Market, Phoenix Center Chief Economist Dr. George Ford and Phoenix Center Adjunct Fellow Professor R. Alan Seals (Auburn University) use data obtained from the U.S. Census Bureau’s American Community Survey to quantify the economic impact, if any, of the county-wide government-owned network (GON) in Chattanooga, Tenn. on labor market outcomes.

“Across a variety of empirical models, we find no payoffs in the labor market from the city’s broadband investments,” they conclude. “We find almost no statistically significant effects for a wide range of important labor market variables, with the possible exception of a reduction in labor force participation.”

The study looked at private-sector labor force participation, employment status, wages, information technology employment, self-employment, and business income, “all of which appear unaffected by the GON,” the researchers say.

Though Chattanooga’s Mayor Andy Berke has claimed that the city’s nearly $400-million network was responsible for a decline in unemployment in the city from 7.8 percent to 4.1 percent over the 2012 to 2015 period, over the same post-recession period the nationwide unemployment rate fell from 7.5 percent to 4.7 percent, they note. So it is hard to isolate any impact other than general economic conditions for the decrease.

There are some key caveats. A new Volkswagen factory, planned before the GON was launched, did open at about the same time as the network began operations. “Marginal employment effects in auto manufacturing closely match the plant’s employment levels,” the researchers note.

Also, since Chattanooga’s system is an overbuild of multiple private providers, “we stress that our findings may not be generalized to areas where broadband services are not available absent the municipal system,” Ford and Seals say.

“Also, our results cannot speak to the benefits of high-speed Internet services generally, since broadband Internet service was and remains available in Chattanooga absent the municipal system,” they say. “Thus, our results indicate only that building a government-owned network in markets where privately provisioned broadband is generally available has no favorable effect on labor market outcomes.”

“The data suggest local governments must look outside the labor market to justify the sizable investments in municipal broadband systems,” the authors say.

Wednesday, May 29, 2019

Frontier Sells 350,000 Lines, 1.7 Million Passings

Frontier Communications is selling the networks and customer accounts of operations in Washington, Oregon, Idaho, and Montana to WaveDivision Capital and  Searchlight Capital Partners for $1.352 billion in cash.

Those networks have a combined 350,000 access lines access lines in service (business and consumer), $619 million of revenue, $46 million of net income and $272 million of adjusted EBITDA.

The deal essentially values each of the active lines at about $3863. By way of comparison, cable TV subscribers in recent years have been valued at about $4,000 to $5,000 each. Charter bought Time Warner for about $5178 per relationship.

Of course, all of Time Warner’s “lines” or relationships were broadband. Perhaps 29 percent of Frontier’s lines are broadband in a way that compares to cable connections.

Across the four states, Frontier’s network passes 1.7 million residential and business locations, of which approximately 500,000 are fiber-to-the-premises capable. As of March 31, 2019, Frontier served approximately 150,000 fiber broadband, 150,000 copper broadband and 35,000 video connections in these states.

Sometimes 4G Spectrum Will Directly be Used by 5G Devices

Dynamic spectrum sharing--the ability to use 4G spectrum to support 5G devices--is a major way 4G networks can be viewed as integral parts of 5G. That applies to 5G and 4G radios and resources on the same mast, or on different masts.

Using spectrum sharing in this way, 4G resources effectively become 5G resources. It is more than using optical backhaul originally built for 4G and then extended to support 5G at the same towers. It is more than having a base of 4G small cells that also can be used to support 5G.


Dynamic spectrum sharing means 4G spectrum can directly be used by 5G devices, from the same or a nearby mast. In that use case, 4G spectrum and bandwidth is directly available for use by a 5G device.

In a larger sense, there are two ways of looking at 4G mobile networks: the precursor to 5G or part of 5G. So 4G coverage, latency and speed might be viewed either as a problem 5G will fix, or part of the way 5G will improve user experience over 4G levels.

If one takes the latter view--that 4G is a part of the 5G experience--then countries with better 4G might well have better 5G as well, in part because dynamic spectrum sharing can be used, perhaps in part because optical backhaul networks are more developed, in part because radio sites can be reused.

On a speed dimension, that includes South Korea, Norway, Canada, the Netherlands, Singapore, Austria, Switzerland, Denmark, Belgium and Japan, among others.

On the coverage dimension,


On the coverage front, South Korea, Japan, Norway, Hungary, the United States, the Netherlands, Taiwan, Hungary, Sweden and India, among others, have 4G coverage that will help them in the 5G area as well. Perhaps the most-surprising fact is that India is among the nations globally with the most-extensive 4G network coverage.



Net Neutrality Starting to Look Like a "Solution" to a Problem that is Going Away

BT fixed network broadband customers now can sign up for a new “Stay Fast Guarantee” that assures consumers their service will be optimized, automatically and remotely, with a quality of service guarantee.

For a few of you who might immediately recognize this, such quality-of-service features are chief among the practices supporters of strong forms of network neutrality always decry. But advancing technology (packet encryption, application requirements, edge computing and much-faster speeds) also undercut the need for strong network neutrality laws, it can be argued.

In other words, best-effort, everyday performance is getting good enough (on mobile and fixed networks) that the “need” for quality of service mechanisms, or even access to higher-speed tiers of service, is largely moot. There is little need for “fast lanes” when “every lane is a fast lane.”

Bluntly, fast, low-latency networks kill the consumer need for QoS-assured tiers of service, as well as killing the service provider market opportunity to sell such tiers of service.

The new BT QoS offer for consumer broadband is among the growing number of reasons why such laws arguably are not needed, in some part because it is becoming impossible for internet access “bad actors” to intentionally speed up or “degrade” a consumer’s connection.

For example, in a market where 5G latency is so low, and typical best-effort speeds so high, what advantage is gained by services that are optimized for latency or speed?

Doing so normally is thought to require use of deep packet inspection, but that becomes quite challenging to impossible when traffic is encrypted, and that is getting to be the case for 80 percent of all traffic, already.

That is not to deny some utility for the BT QoS guarantee. In the U.S. fixed network market, consumers own their in-home wiring. So when that network malfunctions, it is the consumer who pays the cost of the repairs. Such in-home wiring might not malfunction or degrade very often, but it does happen.

That might be especially true for wiring that is on the exterior of a home or building.

When BT customers sign up to a new BT broadband plan or extend their existing contract, they will be given a guarantee of speed based on the estimated capability of the line. If it’s believed a broadband customer could get a faster line speed, BT will remotely optimize broadband performance without the customer having to do a thing, or will dispatch a technician, the company says.

If BT has not managed to get a customer’s broadband speeds back to where they should be after 30 days of a fault being identified, customers will be eligible to receive £20 back, up to four times a year, BT says.

BT will also ensure customer broadband speeds are being monitored and optimized remotely 24 hours a day, every day.

Some will complain that the Stay Fast offer violates net neutrality principles. Others might argue it should not be illegal to sell a quality-of-service-assured access service. In principle, BT’s Stay Fast offer only offers consistency of service, not a fast lane.

Beyond that, it might be argued that such offers offer less value than might once have been the case. Faster fixed network speeds and low-latency, high-bandwidth 5G, plus packet encryption, all combine to reduce the potential value of QoS-assured services.

When problems are resolved, it makes little sense to continue trying to “fix” them. It is beginning to look as though strong forms of network neutrality are proposed solutions to problems that are going away. Time to move on?

SD-WAN Interest Seemingly Jumps in 1 Year

About 46 percent of information technology executives surveyed by Cato Networks indicated that they had, or were considering, deploying SD-WAN in 12 months, up from about 25 percent with such intentions in 2018.

Another 33 percent of respondents are considering SD-WAN but have no current plans to deploy the technology.

The primary motivations for considering SD-WAN include:
  • Improved Internet access (46 percent)
  • bandwidth (39 percent)
  • improved last-mile availability (38 percent)
  • excessive WAN related costs (37 percent).

As argued by Cato Networks, the advantages of SD-WAN over MPLS include:
  • Secure, direct Internet access from branch offices
  • Predictable, responsive global application performance without
  • Lower cost than MPLS
  • Easy and affordable optimized cloud access
  • Optimized and secure mobile experience worldwide

But connectivity providers (telcos) still seem to face mixed reviews from enterprise users.
Respondents said  telcos provide “average service level.”

“There is nothing about our telco which makes us like or dislike them, they are just ‘as expected,’” Cato Networks summarizes.

On a numerical ranking, telcos got ratings of 54 out of 100, when respondents were asked if they thought network service pricing was fair.

On other measures telcos arguably scored better. Respondents gave high marks for the overall
experience with cloud providers (3.71 for cloud application providers, 3.70 for cloud datacenter providers), while global telcos scored 3.24.

Cato Networks calls that performance by cloud service suppliers “high,” while telcos scored “lowest.” It is not possible for me to determine how significant those differences are, as Cato did not provide the range of possible scores. I would guess the range was 1 to 5. In that case, though telcos scored lower, the difference between cloud providers and telcos, while clear, might not be a gulf.

Tuesday, May 28, 2019

Revenue Per Square Mile: Population Density Really Matters

The relationship between population density and network investment is clear in this table illustrating revenue per square mile for a fixed network supplying internet access at $50 per month.

Area
Population
(people/mile2)
Household Income
(median, 2017)
Broadband Adoption
Revenue
per mile2
Manhattan, New York
69,467.5
$85,071
81.9%
$347,337.50
City of Chicago
11,841.8
$55,295
73.5%
$59,209.00
Santa Clara County, CA
6,327.3
$106,761
89.4%
$31,636.50
City of Palo Alto, CA
2,696.5
$147,537
91.3%
$13,482.50
Carbon County, PA
171.1
$51,236
74.4%
$855.50
United States
87.4
$60,336
78.1%
$437.00
Campbell County, WY
9.6
$80,178
86.4%
$48.00
Loving County, TX
0.1
$ 80,938
64.5%
$0.50

"Tokens" are the New "FLOPS," "MIPS" or "Gbps"

Modern computing has some virtually-universal reference metrics. For Gemini 1.5 and other large language models, tokens are a basic measure...