It has been widely taken as “obvious” that a “no blocking” rule for ISPs is a good regulatory policy. Is this really the case? Does it save consumers from harm… or cause harm?

Netflix has reached the point of being well over 30% of Internet traffic at peak time for some ISPs. This places three costs on the ISP and its users:

• Firstly, users must pay for the direct cost of delivery of that traffic. There is a relatively strong coupling to Netflix demand: it’s not spread out over the day, but concentrated. Sizing the network to this “peak of all peaks of demand” is costly.
• Secondly, there is a Quality of Experience (QoE) impact on all other applications, since they now contend with Netflix traffic. Today’s standard approach to restoring (some of) this QoE is to further over-provision the network to lower contention.
• Thirdly, there is a “deadwieight” cost for some quality-sensitive applications and users whose QoE isn’t restored by over-provisioning.

So for someone who isn’t a streaming video user, they are paying a share of the direct costs. They are also, all other things being equal, having a worse experience on their non-Netflix applications, depending on how idle the network is, and what traffic management is in place. You can be pretty darn certain that when Netflix tunes its aggressive “adaptive” algorithms to make their own service work, the impact on rival applications is not a major concern.

So suppose for a moment that a new ISP launched that simply banned Netflix. The traffic volumes of Notflix would be lower, and its QoE higher, again with all other things being equal. Also assume that Notflix did not have significant market power, or a competing IPTV or cable video service, so there were no issues of anti-competitive behaviour.

The consequence would be that customers who dislike Netflix and seek lower prices or better QoE for real-time applications would switch to Notflix. Presuming they act with enlightened self-interest, this would be a net benefit to them.

The remaining customers of existing ISPs would no longer be free-riding on the back of Notflix’s customer base. In the short run, those ISPs would either have to accept reduced margins, or raise prices. In the long run, if the delivery costs of Netflix are sustainable, they will continue to have a profitable business. If those costs are unsustainable, then they will (by definition) go out of business, as they should: users don’t value the service enough to cover its costs.

Is the overall effect a net benefit or harm to society? My guess is it’s a plus, but I can’t prove it, and I’m not an economist. In this (albeit contrived) example, some users are absolutely better off, and others are merely asked to carry the cost of their online activities. The one thing that is clear is that the “harm” demanding preemptive regulatory intervention is far from obvious.

The underlying issue is that the “virtuous circle” of Internet users and applications makes a false assumption: that demand is relatively homogenous. The resulting fallacy is that each and every user and edge provider is fully internalising the costs of the load they offer to the network. Having more applications and users is seen as all benefit, and never has any undesirable cost. That isn’t necessarily the case, as the externalised QoE costs are real.

Blocking a low-value (to enough users) and high-cost application may be a perfectly rational and efficient response. The “net neutrality” regulations creeping across many jurisdictions are built upon flimsy technical models and false logic. Policymakers everywhere are being encouraged to implement rules that they may later come to regret.