More follow-up: Netflix, sewage, and net neutrality

I’ve been hosting a discussion here at Legally Sociable that is turning into a long-running net neutrality debate.

To recap so far:

A.  Alan Roth, a USTelecom Senior Exec VP, wrote an article over at the Hill drawing an interesting analogy between the economics of Netflix/ISPs/home Internet users and the economics of sewage customers/municipal sewer lines/”pooled” sewage processing in the D.C. metro area.

B. I replied with a post questioning the applicability of Alan’s sewage analogy because it didn’t seem to correlate to the Netflix facts (i.e., sewage customers paid just their municipalities directly whereas Netflix customers pay both their ISPs and Netflix directly).

C.  Alan defended his application of the analogy and clarified his position, stating that he was using the analogy of sewage to highlight “the issue of equitable cost-sharing among the users of a network” and noting that “whatever you might want to say about who provides how much ‘value’ and to whom, the fact is that the data bits in question here are largely flowing in one direction, just as the sewage [flows] in one direction” toward the analogical, pooled-cost sewage processing station in D.C.

D.  I replied that we were still at an impasse because

All ISPs are providing here is a connection to the wider Internet (to the “regional front doors”). Retail customers then pay Netflix for the rest because Netflix is providing the rest. On what basis do the ISPs challenge Netflix’s contention that it “should pay only to transport its bits to a regional gateway, after which the costs of delivery to the end point would fall on others”? Doesn’t that precisely reflect how retail customers are being billed?

The only justification I can see for your position is if subsidies are involved-in other words, ISPs are somehow lowering their retail customers’ bills for Internet service because they are paid by content providers. If that’s true, however, that is very different situation from the D.C. sewage situation to which you analogize.

E.  Alan has just replied here.  Please follow the link to get the full text; however, most of Mr. Roth’s response gets reproduced below as I respond to each of his points in turn.


I sense this will probably be the final round of my exchange with Mr. Roth on this issue, so I want to begin by thanking Alan for engaging with me.  I certainly appreciate his creativity in drawing upon his experience as a member of the Board of Directors of the District of Columbia Water and Sewer Authority to help think about the economics of Internet content delivery.

In short, however, Alan and I do still (strongly) disagree.  I will take his latest points one at a time.

1) A small fraction of Internet users are consuming a huge portion of the available bandwidth on any given evening by downloading and/or streaming Netflix video content. See this article — — for confirmation of that phenomenon.

My analysis: Agreed — lots of people use Netflix, and that generates at lot of Internet traffic.  Alan and I don’t disagree on the facts; we disagree on the conclusions to be drawn.

I look at this situation and note that (a) home Internet users are paying their ISPs for Internet access and (b) home Internet users are (separately) paying Netflix for access.  Separate companies are providing separate services (connection and content, respectively), and they are both being paid directly by the people (home Internet users) who are consuming them.

Under these circumstances, I think it is irrelevant that 20% of Internet traffic is consumed by Netflix.  It could be 100%, and the fact would remain that home Internet users are paying ISPs for the connection and Netflix for the content.

In the real world, of course, Netflix isn’t the whole Internet.  100% of Internet activity is caused by 100% of Internet content providers — blogs, YouTube, Hulu, Skype, Netflix, and innumerable other web sites and services.  But so far as I can see, home Internet users pay 100% of their ISP’s bills for their Internet connections, and they pay 100% of their content providers’ bills for the content they enjoy, either directly (as with credit cards in the case of Netflix) or indirectly (as eyeballs for advertisers on ad-supported websites like Hulu).

If (a) ISPs maintain networks that allow home Internet users to connect to the broader Internet and (b) Netflix pays Internet content delivery networks (CDN’s) to deliver traffic right up to the “regional front doors” (where those ISP’s networks begin and the broader Internet ends) and (c) each bills customers accordingly, the fact that Netflix is 20% of Internet traffic is completely irrelevant.  What is relevant is that Alan thinks ISPs should be able to charge twice — even though Alan never really explains why Netflix should also have to pay the ISPs.

2) Enormous amounts of capital investment on the part of both ISPs and backbone providers are required to deliver that Netflix service reliably, while also providing adequate bandwidth for the other 98% of Internet users who are logged on simultaneously for their own various reasons. And as more and more users begin using their broadband connections for streaming video, satisfactory service will require even more investment. And that doesn’t even begin to take into account the huge sums needed to bring broadband to the (mostly very rural) areas of the country that are currently unserved and that presumably would like to enjoy the same online benefits that we city dwellers have.

My analysis: Alan and I also agree that enormous amounts of capital investments are necessary to run the Internet.  Again, however, home Internet users pay ISPs and Netflix pays CDNs, and the two networks meet at the “regional front doors.”  The fact that a lot of money is spent to make such networks possible doesn’t explain why ISPs should be able to charge Netflix for access to its customers any more than it would explain why Netflix should be able to seek reimbursement from ISPs for Netflix’s own CDN bills.  To my mind, what is “fair” is that each business pays its own costs.

To be sure, the economic arrangement between ISPs and Netflix could be different.  For example, ISPs could buy Netflix’s service (content and delivery to the “regional front door”) on a “wholesale” basis and then “retail” it to their customers.  ISPs could do this either as “included” with their Internet service or on an “a la carte” basis.  (Indeed, this is similar to what happens with a service that goes by the name “cable TV”:  “included” services are called “basic cable” and “a la carte” services are sometimes referred to as “premium channels”.)

While Netflix could follow such a model, that’s clearly not what they do.  Netflix has an innovative business model that competes with cable TV (and also the myriad of services vying for consumers’ entertainment dollars).  And so long as home Internet users are paying both their ISPs and Netflix directly, it seems logical to me each should only bill customers for the specific service each is providing.

Alan’s citation of the high capital costs of building the physical Internet is a true fact but an irrelevant argument.  Whether networks are expensive or not, ISPs are paid by their customers to provide connections to the broader Internet, and Netflix delivers content right to the point where the broader Internet starts.  If ISPs feel like the prices they are charging their customers don’t adequately cover their capital costs, they can change those prices.  But it eludes me why Alan thinks the ISPs somehow deserve money from Netflix.  Justifying that sort of double-dipping seems complex at best and unfair at worst.

(As for expanding service into rural areas, Mr. Roth seems to be implying that urban Internet users should be willing to subsidize rural Internet users.  We can debate that point as a separate policy matter, but it seems hardly related to the issue of whether Netflix should be paying the ISPs.)

3) Creating a sound business case for those enormous capital investments requires that someone pay for them. In my view, requiring 98% of end users to subsidize a small handful of mega-users isn’t the right way to generate those capital funds, especially when price is already one obstacle to greater broadband adoption. (About one-third of people who could access broadband now don’t take it!) Admittedly, some form of usage-based pricing might solve that problem — but based on what I’ve seen from my vantage point here in DC, you can bet that howls of protest from that tiny group of bandwidth hogs will drown out any rational discussion of what’s fair to the other 98%.

My analysis: I think Alan and I both agree that Netflix’s CEO was extremely self-serving when he stated that he didn’t want to see ISPs move to metered, per-GB Internet pricing.  It is understandable that Netflix would love people to be able to get cheap access through ISPs (a service Netflix doesn’t provide) to Netflix’s licensed content (a service Netflix does provide).  If Netflix’s customers don’t have to think about a running meter, so much the better.

But let’s not pretend that ISPs don’t also play this self-serving game.  Alan’s employer, USTelecom, is “the nation’s premier trade association representing broadband service providers, manufacturers and suppliers providing advanced applications and entertainment.”  It is in USTelecom’s members’ interest — the very ISPs we have been discussing — to charge as much as possible for Internet connectivity.  Now, there’s nothing wrong with that.  ISPs do incur “enormous capital investments”, and it only makes sense that they want to be compensated.

I think Alan’s deployment of the “fairness” argument at this juncture, however, is quite telling.  As he admits, ISP’s bills could be fair and avoid cross-user subsidies by following true usage-based pricing.  But we all know that this is not what happens.  Perhaps ISPs don’t offer usage-based pricing because market competition will not allow it (which presumably means the market is working).  Or perhaps ISPs like the current system precisely because it allows them to charge $60+/month (or higher in many markets) to individuals who barely use the Internet (which would imply that ISPs are benefiting from the fact that many home Internet users do not have a choice of ISPs).

But either way, let’s not pretend that ISPs — most of which operate under monopoly or duopoly market conditions — are avoiding usage-based pricing on the mere prospect of “howls of protest.”  If ISPs wanted to offer usage based pricing, they would.  But they apparently don’t.  Given this undeniable fact, it’s hard to see how Alan’s position in favor of the ISPs he represents is any less self-serving than Netflix’s position.

Moreover, any unfairness that exists because “98% of end users…subsidize a small handful of mega-users” is completely on the ISP side of the “regional front door”.  If ISPs want to address a perceived unfairness in pricing among their users, they can try charging more for more use.  But it’s hard to see why Netflix — on the other side of that “regional front door” — should pay money to ISPs just because the ISPs own customers don’t all consume exactly the same amount of bandwidth.

4) It isn’t just me who sees Netflix as conducting this unfair “subsidization” campaign. See this analysis too —

My analysis: I note at the outset that Alan’s view of “subsidies” cuts both ways — he doesn’t want to “requir[e] 98% of end users to subsidize a small handful of mega-users” (point #3), but he does want “to take into account the huge sums needed to bring broadband to the (mostly very rural) areas of the country that are currently unserved and that presumably would like to enjoy the same online benefits that we city dwellers have” (point #2).

It’s perfectly understandable that Alan wants subsidies that will help USTelecom’s ISP clients and not subsidies that don’t.  But it’s a little much for him to be shocked when Netflix also wants a subsidy.  And whatever rhetorical games ISPs and Netflix are playing, the fact remains that ISPs are in the driver’s seat on Internet connection pricing.  If they think the current arrangement is unfair because “unlimited access” means some people use way more, they can change their pricing structure (see #3, above).  But again, there’s no discernible reason to pull Netflix, a third party, into this issue.

As far as the Digital Society article is concerned, I’ll let my readers look at it for themselves.  From what I can tell, this is the most relevant excerpt:

Netflix and their CDN partners want’s [sic] the government (the FCC in particular) to declare this peering negotiation as a Net Neutrality violation and force broadband providers to give away thousands of Gbps of broadband capacity for free.

I direct readers’ attention to the comment by Jeremy Stench at the bottom of the page, which is more fully explained in his linked blog post over at Packet Life:

If Comcast [an ISP] were charging Level 3 [one of Netflix’s CDN’s] for transit service [i.e., passing data through the Comcast network on the way to some other final destination], this would be business as usual, not even worth commenting on. But this situation is markedly different as Comcast is demanding payment for traffic terminating on its own network. The traffic in question ultimately must traverse Comcast’s infrastructure, regardless of who the immediate peer is. Comcast effectively is charging two parties for a single service: it wants Level 3 to pay for sending data, and its own subscribers to pay for receiving that same data.

In an ideal economy, Comcast should be able to charge what it likes. And Level 3 should be able to decline. And Comcast subscribers should be able to switch to another provider if they want to watch Netflix. Sadly, this is not the situation of Internet access in the US. For millions of residents, Comcast is their only choice for broadband Internet access. Terminating peering with Comcast would mean an inability to deliver Netflix to those millions of potential customers, and Comcast knows it. [emphasis added]

This is well said (and illustrative of the monopoly/duopoly pricing power of many broadband ISPs discussed at #3, above).

5) When Netflix has been paying $600-700 million annually to the US Postal Service to deliver DVDs through the mail, one would think they would see some value to sharing in the costs of building out that expensive broadband infrastructure by agreeing to share in a fair proportion of the costs their own service is imposing. But instead, they seem to want to offload all those costs onto the 98% of consumers who AREN’T using their services, as the same time as they work on getting out of the mail delivery business — thereby saving themselves hundreds of millions of dollars each year.

My analysis: Of all Alan’s points, this seems the flimsiest.  Other than the subsidy issues already discussed above, Alan’s argument reduces to the mere assertion that (a) because Netflix used to pay the USPS a lot of money to mail DVD’s (“$600-700 million annually”), (b) Netflix should have to pay ISPs a lot of money to deliver those movies digitally (rather than the ~$35 million they actually pay to CDN’s like Level3, according to the Digital Society article).

First, this is misleading.  Netflix may only pay $35 million to its CDN’s to deliver content to those “regional front doors”, but Netflix’s 20 million+ customers pay their ISPs hundreds of millions (if not billions) of dollars for Internet service (which, in practice, means a connection to those “regional front doors”).

Second, why a company’s previous cost structure is relevant to its current business model escapes me.  Netflix’s prior, significant expenses in using the postal service to send DVD’s by mail seems as irrelevant as it would be for FedEx to argue that L. L. Bean should pay more to ship its packages than because L. L. Bean was founded in 1912 (when shipping was more difficult and expensive) and was founded in 1994 (when shipping was relatively cheaper).  It’s hard to see the logic in that.

[6] Joel, it seems you and I should be able to agree on the end result we both want — a strong, robust, and high-capacity broadband infrastructure that serves everyone’s needs. But instead of working to figure out how best to get there, we find ourselves arguing about a subsidiary question that reminds me of that famous dialogue from an old Marx Brothers film, where Groucho finds himself negotiating with a beautiful woman over how they might spend the night together:

Groucho: Would you sleep with me for a million dollars?

Pretty lady (laughing): Of course I would!

Groucho: Would you sleep with me for $10?

Pretty lady: Certainly not! What do you take me for?

Groucho: We’ve already established that. Now we’re just haggling over the price.

My analysis: Yes, of course, price is all Alan and I have been arguing about.  It would be disingenuous to suggest otherwise.  Alan argues that Netflix should pay some (unspecified) amount to ISPs in addition to what ISP’s customers pay; I have been arguing that ISPs should not be able to double-dip.

At this point, I will let our respective arguments speak for themselves.