Berkman Center’s Final Report to the FCC, which combined datasets from the OECD, TeleGeography and PointTopic for its custom pricing analysis.
For next-generation speeds (over 35 Mbps), Canadian retail prices are the second highest out of the 19 countries with NGN offerings - only the United States is worse (Berkman, pp. 73-77).
Another twist to broadband: add low costs to high prices
September 30, 2010 - 11:22am By: David Ellis
The OECD ranks Canada as the 6th most expensive broadband jurisdiction among its 30 member countries, based on average monthly price per advertised Mbit/s (adjusted for purchasing power).
The discrepancies get worse - i.e. Canada gets more expensive - as we go from lower to higher speed tiers. This trend is confirmed in the Berkman Center’s Final Report to the FCC, which combined datasets from the OECD, TeleGeography and PointTopic for its custom pricing analysis. For next-generation speeds (over 35 Mbps), Canadian retail prices are the second highest out of the 19 countries with NGN offerings - only the United States is worse (Berkman, pp. 73-77).
The Berkman report suggests (p. 67) that high prices probably result from “a combined effect of cost and lack of competition that varies by location.” Berkman cites evidence (from the Pew Internet Project) showing that as the number of service providers decline in a given area, retail prices rise significantly. But as Berkman cautions:
“This does not necessarily mean that the price where there are only one or two providers reflects the absence of competition. It may be that the high prices reflect the high costs of providing service in a given area, which in turn results in a lower level of competition as competitors are dissuaded from entering these markets by the high costs of entry. To assume that prices reflect purely higher costs and not the lack of competition would be equally speculative” (p 67).
Why don’t broadband prices fall?
In the US and Canada, high prices didn’t happen all of a sudden, nor are they going away any time soon. Why have prices tended to hold steady or drop only marginally from year to year? MIT’s Technology Review ran a short piece recently that cites a new study by two American economists (Shane Greenstein and Ryan McDevitt). They looked at contracts from 1,500 different DSL and cable service providers, covering the period 2004-2009. What they found was “evidence of only a very small price drop, between 3 and 10 percent, nothing like the rates of price decrease that characterize the rest of the electronic world.”
The piece leans too much on the importance of CPUs in deploying and upgrading broadband networks, since that also involves costly, labor-intensive tasks like trenching - which don’t obey Moore’s Law. Nevertheless, so many “electronic” goods drop rapidly in price after their introduction we treat that as the norm. The Moore’s phenomenon has been paralleled by dramatic drops over the years in hard-drive storage and, to some extent, bandwidth costs - which only adds to the mystery as to why that hasn’t been reflected in broadband pricing.
For Greenstein and McDevitt, the mystery has a lot to do with costs and how far removed retail prices are from any cost-based structure. Their claim is that while deployment may be expensive, maintenance is far cheaper than what might be inferred from prices:
Meanwhile, once companies have installed the lines, their costs are far below prices. “At that point, it becomes pure profit,” Greenstein says. A company might spend around $100 per year to “maintain and service” the connection, but people are paying nearly that amount every other month.
The OECD’s broadband data includes a time series measuring the evolution of speeds and prices in member countries from 2005 to 2008 (4k, “Evolution of a representative broadband subscription over time”). To create this dataset, OECD staff selected two representative suppliers from each country, one for DSL and one for cable. For Canada, the selections were Bell Canada and Shaw respectively. Actual values are shown for each year, then expressed in terms of compound annual growth rate (CAGR):
- Bell (DSL): speed grew 13%, price -1% (2005-08)
- Shaw (cable): speed grew 0%, price -10% (2005-08).
Compare these figures to the US:
- ATT (DSL): speed 0%, price -7%
- Comcast (cable): speed 25%, price -14%.
The Comcast figures reflect a near-doubling of speeds by the company in 2008, with a significant drop in price. Apart from this anomaly, the pattern noted by Greenstein and McDevitt seems to hold, except for one other anomaly that is not explicit - bit caps. The worksheet indicates that across the 30 member nations, only four have bit caps (as of 2008) for both DSL and cable: Australia, Belgium, Canada and New Zealand.
Even if caps are large enough not to affect a lot of subscribers, they work directly against progress towards NGNs, because as networks approach the speeds enjoyed in Japan and Korea - upwards of 100 Mbps - caps will be punitively expensive for subscribers, yet will have no value in supporting traffic management. Nevertheless, this is the role caps have been officially assigned by the CRTC - as “economic” Internet traffic management practices, or ITMPs.
The underlying costs borne by ISPs for bandwidth and other inputs are widely treated as trade secrets. Regulators like the CRTC are obliged under enabling legislation to promote the public interest, in part by ensuring that licensees do not charge unreasonable prices for their services. Canada’s Telecommunications Act provides among other things that “Every rate charged by a Canadian carrier for a telecommunications service shall be just and reasonable.”
Sleuthing out costs
Let’s go back to caps for a moment and an example of why the large ISPs are happy to keep you in the dark. Bell offers a “Performance” tier which at 6 Mbps downlink is the closest option to what I get from my reseller, National Capital Freenet, at 5 Mbps. Not only is this Bell tier more expensive per month than NCF’s. It also comes with a stunning extra feature: a bit cap of 25 GB a month. And equally astonishing, the over-usage charge is... $2 a gigabyte! Meanwhile, NCF has a cap of 200 GB, which comes with an over-usage charge that’s 75% less than Bell’s - 50 cents/GB.
We don’t know for sure what Bell pays its upstream provider for bandwidth, but there are some well-informed guesstimates out there. One analyst who follows the industry numbers very closely, and internationally, is Dave Burstein (he also publishes several newsletters at fastnetnews.com). In the context of his analysis of a 250-gig Comcast cap, Dave writes the following:
“Two carriers report cost per gigabyte of under ten cents and it should be similar at Comcast. Comcast doesn't release a figure, so I'm presenting a range based on 5 cents to 20 cents. Smaller and some rural carriers pay much more” (my emphasis).
Let’s assume the high end of Dave’s range is suitably conservative - 20 cents per gigabyte. So, back of the envelope, it looks like Bell is charging a 1,000% markup on the raw cost of each gig in the cap on this tier - $2 for what costs 20 cents. If we were to take the low end of Burstein’s range, i.e. 5 cents per GB, then a $2 cap would represent a 4,000% markup. These estimates seem to correspond to what we heard from one outraged ISP reseller after the CRTC decision on usage-based billing last May (Telecom Decision CRTC 2010-255):
"The rates are absolutely atrocious. How the hell are we doing above one dollar for extra usage?" said Rocky Gaudrault, president of Chatham, Ont.-based Teksavvy. "It's in the thousands of multiples beyond what the costs are."
Burstein’s analysis applies more widely to ISP costs. They are much lower than prices would indicate and have been falling or holding steady for years. His note from June 24 last is worth some attention. He provides cost estimates from several providers for bandwidth per megabit and these translate into a current average cost for large carriers of about $10. That in turn corresponds to a cost per subscriber for bandwidth of $1 a month. In other words, for provisioning a service like Performance, which costs the subscriber about $50 a month (with modem and without bundle), Bell’s bandwidth cost is about 2% of the retail price.
“This figure is the industry standard and has been roughly stable for 5-8 years. Bandwidth demand has been growing at 30% (AT&T figure) to 40% (Comcast out of date figure) for each user. Switches/routers/transit have been getting cheaper at 25-45% per year driven by Moore's Law. Likely future cost: Almost certainly similar.
“... Result: Costs go down 25-45%, bandwidth demand goes up possibly slower. Cost per month per customer should be flat to down as far out as we can look.”
Burstein uses his cut-to-the-chase cost analysis to make one other crucial point about the access business. Our ISPs charge a much higher price for higher speed tiers, suggesting they are covering higher costs for those speeds. This is not the case, it turns out:
The cost of delivering 40 megabits is almost exactly the same as the cost of delivering 2 megabits once the equipment is in place. The only steep difference is in cost is when additional construction is needed. Regardless of speed, it's one DSLAM port, one modem, and one wire. There's no cost reason to charge significantly different prices. The two or three times difference in price for high and low speeds on Verizon FiOS, for example, comes from weak competition, not ordinary economics.”
Agreeing on what’s “just and reasonable”
The fact that retail broadband prices would seem to be far from cost-based raises some questions about how the CRTC arrives at telecommunications decisions where a price impact on consumers is at issue. I mentioned above the CRTC decision on usage-based billing handed down last May (Telecom Decision CRTC 2010-255) - and alluded earlier to the idea that sanctioning, indeed institutionalizing, bit caps is bad policy, anti-consumer policy.
The implications of 2010-255 extend to both principle and practical details. It’s bad enough the Commission has handed the Bell companies - indeed all of our incumbent ISPs - a new revenue stream. It’s worse still that, in its decision, the Commission explicitly approved tariffs for UBB that include over-usage charges. The Commission would certainly have all the relevant costs at its fingertips. Did it have information indicating that the markup on some caps might range from 1,000% to 4,000%. If it did, does it consider that these markups are “just and reasonable”?
As part of its due diligence, the Commission asks explicitly “Are the Bell companies' proposed rates just and reasonable?” (following para 51). After clawing back some markups that it did not consider just and reasonable, the Commission concludes (para 72) that the rates as adjusted do indeed pass muster.
Should we believe the Commission’s assessment? Well, whether you believe it or you don’t, it’s not for you to know. Because under its rules, “applicants may request that specific types of material or information submitted in an application be treated by the Commission as confidential where the public interest will best be served by doing so.” How do we know the public interest is being served if we have no idea what was submitted by Bell under its cost claims? We don’t.
The explanation lies in footnote 7 at the very end of the decision:
“The markup included in a rate for a service or service component is the difference between the rate and costs, and is typically expressed as a percentage. Markup levels included in the Bell companies' economic ITMP rates and in the cable carriers' TPIA service rates are confidential, since the costs submitted by the Bell companies in this proceeding and by the cable carriers in past proceedings in support of proposed TPIA service rates have been maintained in confidence.”
If it seems that the incumbents are making supranormal profits in residential broadband, thanks in part to extraordinarily low costs, would it not be in the public interest for the CRTC to explain why they believe this not to be the case?
Whether they do or do not, Canada needs much greater transparency in proceedings like this one, especially about costs. Without any indication of actual costs to ISPs, it’s impossible for anyone outside the CRTC(and the ISPs) to judge whether consumers are being harmed by retail pricing, and by Commission-sanctioned measures such as usage-based billing. |
