Hacker News Reader

Comment by greggyb | Hacker News Reader

Regarding 2), that seems like an un-justified expectation that 2x in some category should be 2x price.

I can buy a Chevrolet Spark for $14K. Top speed is listed at 110mph. Here's a list of cars that can go 200mph https://www.automobilemag.com/news/all-cars-that-go-200-mph/. Find me a new one for $28K.

I can buy a $90 printer for 27 pages per minute: https://www.amazon.com/Dell-E310DW-Wireless-Monochrome-Print...

I can find a 50 page per minute at >$500, but none less on a quick search.

Here's one of the cheapest modern processors I can find: https://www.amazon.com/Intel-BX80677G3930-Celeron-Desktop-Pr...

Go find me a 5.4ghz processor for $70.

View on HN

I have no idea what point you're trying to make.

People in France or Germany have access to literally 12 different providers. Everywhere. Granted, it looks like the overall speed is below the US's according to the report, but prices are so much lower (my parents probably pay in the order of $10$/mo for their TV/Internet package) and access is way more homogenous.

I suspect that US results are vastly skewed by hyper-connected areas vs huge parts of the country where Internet is total garbage.

Your second point assumes that scaling some attribute of value by a factor of two should be associated with an exact doubling of price.

These were literally the first three things I could think of, but I'm sure we could braistorm many more items where the price of a product or service that is twice as good along some axis costs more than twice as much.

It is therefore not a strong argument to say, "Now, for fun, let's integrate the price dimension to this report. Yes I have better (maybe 1.5 to 2 times as fast) Internet than my parents who are in Europe, but I also pay 3 times what they pay for it. "

No, there was no assumption related to an exact doubling, and it's not reasonable to infer one, and arguing against it by means of analogies to unlike goods is unsound.

Well now we're two people on the internet arguing about what a third meant. This couldn't possibly end badly (:

So I have two questions.

1: I read the original parent comment's second point as indicating that a 2x in performance for >2x price as something that should be seen negatively. Do you disagree with this intepretation? Why?

2: My argument was not clearly stated, so I will first state it here. It is unreasonable to expect that performance/quality and price correlate linearly with a coefficient of 1. I offered examples of other disparate products where performance/quality do not double for an exact double price, and in fact doubling performance is associated with much more than doubling price. With the point that we should not expect double performance = double price, it seems that anyone positing that this should be the case for a specific product bears the burden of proof. The question is this: do you think that I am wrong and that we should expect in general to see double performance for double price?

Expansion on my position that the default should not be an expectation of "double performance for double cost": Diminishing returns are common in many optimization tasks. If we see diminishing returns, we should expect prices to more than double for a doubling of performance in general.

Additional point on pricing of broadband. A cousin commenter observed that most of the broadband in the countries in question is DSL rather than cable. If we see two different technologies in play, one able to double the performance of the other, it seems exceptionally unreasonable to expect such linear scaling of price.

I do not see anything in the parent comment indicating a linear relationship. Making a more general comparison of prices is perfectly sound. If you must make an inference, you should construe the argument in the most favorable terms. Also, your choice of comparisons were of unlike goods, and "argument from analogy" is a weak form of argument.

> Now, for fun, let's integrate the price dimension to this report. Yes I have better (maybe 1.5 to 2 times as fast) Internet than my parents who are in Europe, but I also pay 3 times what they pay for it.

"Yes..., but" is a form of sentence that indicates disagreement. This was in response to a comment indicating that the US has faster internet than most large European nations.

The commenter's first argument was about Norway and Sweden, which were held up as examples of good internet countries.

I do not think it is is unfavorable to interpret the post as saying attempting to diminish the advantage posited in their parent (with the Akamai link). Please let me know if you disagree with this interpretation.

In their argument they said that the speed is 1.5x-2x, but the price is 3x. Again, "yes, but" is typically a form of disagreement. The "yes" portion grants a specific fact, and the "but" is intended to diminish the importance. Let me know if this is too unfavorable.

Since the "but" is intended to diminish the value of the double speed, we must infer that the expectation is that the price should be less than triple for double performance. There's not a lot of wiggle room here. We can bargain on fractions between double price and triple price. Let me know if you'd like to bargain on these fractions.

Additionally, the form of the argument is "performance is double, but price is more than double". It is not such a large stretch as you seem to be implying that the expectation in the sentence is that price and performance are proportional. Let me know if I'm off or misinterpreting you here.

Again, my argument is of two parts.

One, that we should expect in the general case that, due to the widespread phenomenon of diminishing returns, doublings of performance are associated with prices that are much more than double.

Two, that since this is the general case, the burden falls on the one expecting prices and performance to be so linear to show that we are in a special case.

Note: I would make this argument for any value between double and triple price, so feel free to haggle on those fractions.

You'll note I am not make analogies, unless you choose to interpret my quite explicit argument most unfavorably. I am arguing that there is a general case and that the general case predicts what the parent has "yes, but"ed. With this argument I am asking for a justification that we are in a special case. The valid counterarguments seem to me thus:

1. Present good evidence we are in a special case. 2. Argue that the general case is not a superlinear relationship between quality/performance and price.

I am open to other counterarguments, of course, as I would hate to push you into a path of argumentation which you do not support.

Of course, if you'd like to continue critiquing the form of arguments, rather than their content, it would become more difficult to trust in your good faith. There are so few teachers arrant these days.

The form of the argument is tangential, but the validity of "argument from analogy" is dependent on the strength of the analogy. Few analogies hold up to substantial scrutiny, and the specific examples cited were all fairly distant in kind from the product in question. I certainly don't mean to dwell on that more than might be relevant, but I did think it reasonable to offer this characterization, since you had similarly characterized the OP's form of argument.

Nothing in the OP's comment can be inferred to be a denial of diminishing returns: if there is "wiggle room", then you should give them the benefit of the doubt. Also, there's no inherent reason to expect that the cost of bandwidth would be anything close to linear. The cost to the ISP does not necessarily increase due to the provisioned speed of a modem, and the used bandwidth is in most senses unmetered. Also, for what it's worth, whether the OP feels like they're getting a good deal is more-or-less independent of the actual mathematics at hand.

There is every reason to believe that bandwidth is a special case: the cost to either the customer or the ISP is not linear with either usage or provisioned speed. However, since the parent's argument is not inconsistent with the concept of diminishing returns, it does not seem valid to arrive at that inconsistency by inference, when other more charitable interpretations exist.

P.S. I'm afraid I don't understand your use of the word "arrant".

I guess we'll have to disagree on the interpretation of, "Yes I have better (maybe 1.5 to 2 times as fast) Internet than my parents who are in Europe, but I also pay 3 times what they pay for it." To me, the most charitable interpretation of this is that 3x price is a negative that diminishes the argument made previously that speeds are higher in the US.

For that to be true, then there is an implicit argument that the price should be lower. Either this argument is there, or the parent is simply complaining. I don't see much other possibility.

Again, I am not arguing by analogy but offering a general case, movement against which must be justified.

As was mentioned in a cousin post, and as I mentioned, there are different technologies in play: https://news.ycombinator.com/item?id=18668662

With this, why should one expect that the cost for ~2x speed on cable should have any relation to ~x speed on DSL.

As for your argument on bandwidth, you seem to treat it as unlimited. It is absolutely a scarce resource and often overprovisioned. The network gear of any ISP only supports so much throughput. It is not an uncommon experience on cable to see drops in throughput at peak hours. Price and artificial bandwidth restrictions for individual modems (your package level) are the tools to handle such congestion and scarcity.

Regarding arrant, I mistyped errant, a reference to knights-errant.

No, I am not arguing that bandwidth is unlimited. That would be prima facie impossible. I'm somewhat confused and distressed to read a summation of my argument which implies such an absurdity; it seems objectionable in the context of a calm and reasoned debate. I'm sure that you meant no offense. I'm not incapable of arguing absurdities, but I hope that this is rare, or at least I hope that you may assume so for courtesy's sake.

Bandwidth is not unlimited, but the point seems irrelevant: costs do not scale linearly, in any sense, for any participant. It does not cost the ISP more or less if my modem is provisioned for 10 or 100 Mbps. It may or may not cost them more money if that bandwidth is actually used, but very few individuals have any large impact on their ISP's operating costs. The aggregate utilization does affect this, but the marginal operating cost per customer is much closer to "fixed/negligible" than "linear in proportion to usage". Correspondingly, customers are not billed based on usage (in the normal case).

The implicit argument is certainly that the price should be lower, but they did not argue a linear price relationship. There is no reason to believe their argument would be invalidated if that were the case, but they did not.

When an ISP leases bandwidth to a customer, they must reserve some amount out of their total available. They may reserve the full amount they have leased, or they may reserve some portion of it. Either way, the reserve is finite and proportional to the actual total bandwidth (the proportion may be 1:1, which is the simplest, but the proportion does not matter for this discussion).

The marginal cost of provisioning me 100Mbps is therefore not limited to the cost of the hardware and setup, but is the opportunity cost of reserving bandwidth for my use. They may not lease this reserve to anyone else. This cost is reflected in the price.

Additionally, we can observe that ISPs tend to grow. Even in a fixed area, they will upgrade hardware over time to increase total available bandwidth. This is driven by demand by existing customers for more bandwidth and new customers. This hardware is very expensive to provision. In fact, I expect that going from total bandwidth x to 2x likely costs more than twice as much as x. I would also argue that it likely costs at least thrice as much.

Products of high fixed cost and finite supply will never be sold at marginal cost (or at least not for anything approaching profit). They will approach marginal cost as N increases (where N is the size of the finite supply).

Another way we can think of a customer leasing bandwidth is that they are purchasing a share of that large and expensive network equipment.

I posit that these are valid ways to look at it. The first, because there is not infinite bandwidth, ISPs must price it like the scarce good it is - prices are how we address scarcity. The second, because ISPs must not price their service as if they are never making another network upgrade.

Again, I am willing to haggle on the cost of 2x performance. I would make all of these arguments for any value under 3x price, as I mentioned above, which is clearly stated in the original parent. There is no need to stick to a strict linear interpretation, and I have said as much more than once now.

For residential service, there is no reason to expect a 1:1 reservation. This would be extremely wasteful. The exact number is not relevant, but residential bandwidth is oversold as a matter of course. Talking about capex and aggregate costs also seems to be drifting from the main theme here: whatever it costs to double the network's capacity is going to be pretty circumstantial, and not directly relevant to the topic of consumer pricing. <i>Ceteris paribus</i>, the marginal cost per individual is still negligible, and the price per unit of provisioned bandwidth only loosely coupled to cost. ISPs charge extra for higher-tier packages because having only one tier of service is leaving money on the table, and because they can.

You're right. The reservation ratio doesn't matter - we agree on this and have both stated it's irrelevant.

We're talking about average internet performance in a nation. This entire thread was kicked off by a comment about Akamai indicating the US as a whole has faster broadband than e.g. Germany.

The cost to wire up the US with an average speed of 2x that of Germany is absolutely an influencing factor in the cost of bandwidth to a consumer. We are talking about performance, not widgets. To double the performance of a nation's network from p to 2p, it is not unreasonable to expect that the cost is greater than double or greater than 3x the cost of performance p.

But, if you want to proceed from an analysis of marginal cost, we can do that as well. ISPs do not fit the model you are using.

Marginal cost for something like an ISP is not a single number. There may be 10,000 consumers with a marginal cost near 0. Then customer 10,001 may have a marginal cost that is "install a new backhaul to the network backbone and extra 10G or 100G routing equipment to serve your neighborhood". Obviously customer 10,001 doesn't get stuck with a bill of $100Ks to be added to the network.

You are attempting to pigeonhole an ISP into a model that assumes a single, flat fixed cost and (effectively) unlimited production of a good at a low marginal cost. That is not the industry we are in.

The fixed costs of an ISP are a stepwise function, not a constant. You cannot model an ISP effectively with an assumption of constant fixed costs.

All other things being equal, the marginal cost of a given customer is <cost of new network equipment or major network upgrade> / N, where N is the number of customers that can be served before the ISP has to install new routing hardware, or in the unhappy path lay new wire.

Now, it is obvious that I have simplified things above. ISPs do not build infra for N customers, then wait for customer N+1 to build out the infra for the next N customers. ISPs invest in their infra in long planning cycles. They cannot effectively build out in this manner in response to a single customer.

Thus, for years at a time, the supply of bandwidth is effectively fixed. With a fixed supply of bandwidth, they can only lease so much. We have agreed that the reservation amount per bandwidth leased is not relevant. The model you want to look at has a vertical supply curve.

Finally, the theory that says consumer price should be the producer's marginal cost is predicated on the assumption that barriers to entry are low. This is not at all the case in ISPs. There is lots of expensive hardware, and there is a significant amount of technical expertise required. Thus, we cannot apply the standard econ 101 model, because we have violated its assumptions.

France and Germany have lots of cheap options because they rely heavily on DSL: https://www.oecd.org/sti/broadband/1.2.OECD-FixedMobileBB-20... (see the "Fixed Broadband" tab).

70-75% of France and German broadband connections are DSL, versus 22% in the U.S. The copper networks are mostly depreciated, and copper-loop unbundling has created a lot of competition. But that's also why internet in those countries is so slow. According to Akamai, just 18% of French broadband connections are above 15 mbps, versus 48% of U.S. connections. Copper is also a complete dead-end, technology wise.

Cable in those countries is not super cheap. Kabel Deutschland's gigabit service has a non-promo price of 69.99 euro ($80): https://translate.google.com/translate?hl=en&sl=de&u=https:/.... Comcast's non-promo pricing is $105, which is a bit more expensive in nominal terms, but cheaper as a percentage of net adjusted disposable income (which is 33% higher in the U.S. than in Germany).

You also need to account for differences in labor costs. Broadband isn't like an iPhone, where it's the same product made in China whether you buy it in the U.S. or in Germany. It's a service, like a hotel stay or restaurant meal. Labor to build, maintain, operate, and support the network is a huge part of the cost of broadband, and skilled labor is significantly more expensive in the U.S. than in Germany or France. That's one of the things that makes pricing comparisons between the U.S. and say Romania completely specious. You can buy a 2-ride subway ticket in Bucharest for $1.22--that doesn't tell you anything about what's a reasonable price for a subway ticket in D.C. The $10/month cost for gigabit fiber in Romania is equivalent to $80/month in the U.S.

What taxes would someone expect on the $105/month Comcast service?

Since all applicable taxes are included in the €70 Kabel Deutschland price.

Depends on where you live.

My total bill for internet service is approximately $110. Of that $110 total, approximately $1.50 is tax/fee.

Reply on news.ycombinator.com