LESSWRONG
LW

Comment Permalink

Since the Shapley value of all players also has to sum to the value of the end result, I think the value of each A voter has to be just RB/n. I'm way out of my depth with the combinatorics here, but here's a paper I found that gives a bit more information than the wikipedia page.

SilasBarta12y00

Good point, but I don't think the value of the end result is necessarily equal to RB, for much the same reason that (I suspect) Shapley value would correspond to (something like) "market value" rather than "market value plus consumer surplus". That is, no matter how badly you want you bathroom cleaned, the value of the labor to clean the bathroom is only equal to the market value of that labor, irrespective of how happy I am to have it done.

While voting doesn't directly map onto a market like that, there is a similar sense in which bei... (read more)

See in context

1 How much was creating Google worth?

by JonahS

2nd Jun 2013

1 min read

1

I would greatly appreciate feedback on whether the assumptions in this post, as well as the conceptual framework, are reasonable. I recognize that I don't address positive externalities explicitly. I think that they may be picked up implicitly, but am unsure about this.

Sergey Brin and Larry Page made an outsized contribution to society by speeding the development of high quality online content filtering by creating Google. They became spectacularly wealthy as a result, but plausibly produced far more value than they captured.

One can estimate the value of creating Google by trying to place oneself in the shoes of somebody with average US earnings in 2000 (~$45,000/yr) and thinking about how much one would be willing to pay for Google to have been created, rather than its counterfactual replacement some years later.

Assumption 1: The counterfactual replacement would have been created 3 years later (in expectation).

Assumption 2: Contingent on knowing assumption 1, the mean 2000 US citizen would be willing to pay at least $1/week (about 0.3% of his or her income).

The average person spends about 15% of his or her income on utilities and entertainment combined. In view of this, we’ll assume

Assumption 3: The mean American would not be willing to pay more 5% of his or her income for three years.

Assumption 4: The percentage of income that one would be willing to pay for Google is approximately uniform over the entire population.

Putting these assumptions together, and using the fact that world GDP in 2000 was about $60 trillion, we get a ballpark estimate of the value of Google of 0.6 trillion — 9 trillion.

Since Brin and Page have total net worth about $40 billion, the remarks above suggest that they produced between 15x and 225x as much value as their aggregate earnings.

Thoughts?

Personal Blog

1

New Comment

48 comments, sorted by

top scoring

Click to highlight new comments since: Today at 8:05 AM

[-]CarlShulman12y370

As a consumer surplus estimate for those years, it's too high.

In the year 2000 internet access itself had 7% global penetration, and 31% penetration in the developed world; your estimate seems to assume that people with no internet access were getting substantial benefits out of Google
Search use was less intensive at that time, e.g. because of annoying dial-up connections instead of broadband, the absence of mobile internet, lack of content to search for... see these two pages for data on search traffic, which is up close to two orders of magnitude since 2000
Services like AltaVista were definitely worse, but could still find pages and were improving, so the value contributed is only the difference between the engines
5% of income is way too high to be credible for willingness-to-pay: 20% of American households in 2011 weren't willing to pay for internet access, which includes the marginal search quality of Google among many other features, for a lower cost
Even $1/week might be too much for much of the world's population, and even a nontrivial chunk of the developed world: people who won't pay $10-20/month plus computer costs for internet access itself plausibly wouldn't pay $4/month for marginal Google search quality
On the plus side better search encouraged the development of more websites and internet businesses with various good effects

[-]Eliezer Yudkowsky12y60

AltaVista was bad, Google was awesome and I probably would pay $100/month to make up the difference today if I had no choice but to pay. This is on the order of 2.5% of aftertax, pre-rent income. I wouldn't have been able to afford it when Google first came out.

I usually credit altruistic effects to cohorts of agents with logically correlated decisions, rather than assuming that in any election won by two votes, nobody's vote had a marginal effect; this is a separate big issue but says that I wouldn't just assume the consumer surplus generated by their lives' decisions, vanished after somebody else would've done it anyway - somebody has to be the somebody else and get credit for that.

[-]CarlShulman12y80

You're an infovore.

[-]Eliezer Yudkowsky12y30

I'm also cheap.

[-]JonahS12y10

Thanks Carl!

In the year 2000 internet access itself had 7% global penetration, and 31% penetration in the developed world; your estimate seems to assume that people with no internet access were getting substantial benefits out of Google

I was trying to wash out this issue by weighting for income, but maybe the implicit adjustment wasn't large enough. Do you have an idea of what the fraction of world GDP corresponds to that 7% of the population?

Search use was less intensive at that time, e.g. because of annoying dial-up connections instead of broadband, the absence of mobile internet, lack of content to search for... see these two pages for data on search traffic, which is up close to two orders of magnitude since 2000

Very good point.

Services like AltaVista were definitely worse, but could still find pages and were improving, so the value contributed is only the difference between the engines

I had this in mind.

5% of income is way too high to be credible for willingness-to-pay: 20% of American households in 2011 weren't willing to pay for internet access, which includes the marginal search quality of Google among many other features, for a lower cost

Yes, I suspected that my figure was way too large, but wanted to be liberal. Do you have an idea for a suitable anchor for a plausible upper bound? I could replace "5%" by "1%" but that feels too ad hoc....

Even $1/week might be too much for much of the world's population, and even a nontrivial chunk of the developed world: people who won't pay $10-20/month plus computer costs for internet access itself plausibly wouldn't pay $4/month for marginal Google search quality

Here too, I was trying to wash this out by weighting for income, and considering somebody making $45,000/yr.

On the plus side better search encouraged the development of more websites and internet businesses with various good effects

Do you believe that the effects are out of proportion with what people would have been willing to pay to use Google (relative to counterfactual uses of money)?

[-]JonahS12y00

Even $1/week might be too much for much of the world's population, and even a nontrivial chunk of the developed world: people who won't pay $10-20/month plus computer costs for internet access itself plausibly wouldn't pay $4/month for marginal Google search quality

Do you know of data on the distribution of population by amount that they were paying for internet in 2000?

[-]elharo12y270

This commits the fallacy of assigning all of Google's worth to Page and Brin. While they certainly deserve a large share of the credit so too do Craig Silverstein, Jeff Dean, Sanjay Ghemawat, Eric Schmidt, Marissa Mayer, Susan Wojcicki, the original VCs who invested in Google, and at least one hundred thousand other people, probably more.

Anyone know a name for this fallacy? I.e. attributing the entire value of some group or action to a few salient individuals while ignoring the contributions of all the other people involved. This strikes me as a major failure mode in a lot of discussion on LessWrong about effective altruism and life choices. I don't know how to apportion credit and blame to individual people for group actions (and almost all effective actions are group actions). I'm not sure it's even a meaningful question.

[-]saturn12y90

I don't know how to apportion credit and blame to individual people for group actions (and almost all effective actions are group actions). I'm not sure it's even a meaningful question.

Shapley value is one way to answer this question.

[-]SilasBarta12y00

Interesting! I hadn't heard of Shapley Value before.

Regarding voting, let me do a back of the envelope calculation: every voter (that voted the same way) would, by symmetry arguments have contributed equal value. And since Shapley averages over every possible voter subset, and voters would only get credited for those subsets where they are the determining vote (which is proportional to the factorial of the margin of victory I think) then the value each voter (for policy A given two alternatives) contributes is something like:

  RB
----
n MV!

Where n is the number of voters for policy A, RB is relative benefit of policy A compared to B, and MV is the margin of victory. But I think I made a mistake of ome kind somewhere.

[-]saturn12y20

[-]SilasBarta12y00

While voting doesn't directly map onto a market like that, there is a similar sense in which being one of the voters for something that "had no chance of passing" (thus getting the high margin of victory) is worth less -- even per voter -- than voting for something whose fate was less certain.

[-]Eliezer Yudkowsky12y70

It seems to me to be related to the big separate issue of whether, if an election is settled by two votes, nobody's vote had any effect - by the same logic if an election was settled by one vote, everyone who voted for the winning side solely decided the election and get all the credit for it. The altruistic credit due to Page and Brin is not the interval between this universe and the counterfactual universe where they didn't exist, any more than the credit due for an election result is the distance between this universe and the counterfactual universe where you voted the other way.

[-]buybuydandavis12y40

The more general problem is considering something as "the" cause when it is only "a" cause.

[-]IlyaShpitser12y20

http://www.cs.cornell.edu/home/halpern/papers/blame.pdf (etc by Halpern)

[-]elharo12y00

Yes, that's a very good analogy; and perhaps one that's a little easier to get a handle on mathematically since it's very well specified. The logic of the problems seems exactly the same.

[-]Decius12y00

Conservation of credit/blame?

If we give Brin and Page full credit for everything Google did for three years, then we must give zero credit to everyone else for those outcomes.

[-]ozziegooen11y00

Not exactly. Counterfactual credit on individuals isn't additive. For instance, we can say that almost every piece of the car is counterfactually (for the removal of the piece) essential (worth the entire value of the car).

That said, in this case, one would have to wonder what these employees would have done otherwise. I imagine the value of the founders for Google counterfactually, is somewhat equal to the value of Google minus the value of what the used-by-Google resources (money and lots of talent), had they been used elsewhere.

[-]Decius11y00

I'll buy your car, minus a windshield wiper blade, for a dollar. I'll also sell you a windshield wiper blade from mine for half of the bluebook value.

It certainly doesn't generalize to physical objects composed of smaller objects; and the standard assumption is that almost all talent has non-exceptional results.

[-]JonahS12y00

Craig Silverstein, Jeff Dean, Sanjay Ghemawat, Eric Schmidt, Marissa Mayer, Susan Wojcicki, the original VCs who invested in Google, and at least one hundred thousand other people, probably more.

Working at Google diverted these people's skilled labor from other worthwhile endeavors. I agree that these people being involved contributed by choosing Google as opposed to choosing something else. But I don't think that the effect size is very large. Google's market cap is about 200 billion, which is only 5x Larry Page and Sergey Brin's wealth. Much of this valuation comes from Google having a dominant position rather than from Google's contribution relative to the counterfactual. I would be interested in making an adjustment to account for the issue that you mention, but I doubt that it would be greater than 2x.

[-]ThisSpaceAvailable12y00

Working at Google diverted these people's skilled labor from other worthwhile endeavors.

Huh? If someone writes a software program worth $100,000, then they have contributed $100,000 to the economy. If the person could have written a program worth $80,000, that doesn't mean that the person has only contributed $20,000 to the economy. Your comment makes sense if evaluating the value of Google to the economy, but not when evaluating the value of the employees to Google or the economy.

[-]JonahS12y00

Ok, so I intended to ask the question "How much value did Brin and Page create by founding Google?" rather than the question "How much value did the early Google staff collectively create in contributing to Google's growth?" When I get a chance, I'll edit the post to this effect.

[-]buybuydandavis12y00

Google's market cap is about 200 billion

Don't mistake Google's market cap for the value people have gotten out of Google. Shareholders only get a fraction of the value delivered by their companies.

[-]ThisSpaceAvailable12y30

Sometimes, that "fraction" is greater than 1 (not that I'm saying that it is in the case of Google). Rent seeking, regulatory capture, arbitrage, and tournament theory are all cases where people make more money than their contribution to the economy.

[-]JonahS12y-10

Indeed, this was the premise of my post :-). I was using the ratio of the founders' earnings to market cap as a proxy for the ratio of social value produced by the founders to social value produced by Google.

[-]David_Gerard12y-20

They created a hell of a lot of value with just the original pagerank algorithm. I remember just how bloody useless search engines were in 1998. Google was the first one that was any good whatsoever.

[-]wedrifid12y100

Since Brin and Page have total net worth about $40 billion, the remarks above suggest that they produced between 15x and 225x as much value as their aggregate earnings.

It would be more accurate to say that value of 15X to 225X their aggregate earnings was created via a path that happened to include Brin and Page at a key point. It would be an error to, for example, suggest that this means if we went back in time and... used some crude intervention to prevent Brin and Page from contributing then we could expect a result of an earth in 2013 that contains 9 trillion less in value.

One can estimate the value of creating Google by trying to place oneself in the shoes of somebody with average US earnings in 2000 (~$45,000/yr) and thinking about how much one would be willing to pay for Google to have been created, rather than its counterfactual replacement some years later.

I suspect that this estimate of value creation while interesting is rather likely to be used with misleading connotations. For example using it in a naive manner would result in 'altruistic' decision making that is grossly miscalibrated with respect to actually increasing expected value.

[-]JonahS12y-20

It would be more accurate to say that value of 15X to 225X their aggregate earnings was created via a path that happened to include Brin and Page at a key point. It would be an error to, for example, suggest that this means if we went back in time and... used some crude intervention to prevent Brin and Page from contributing then we could expect a result of an earth in 2013 that contains 9 trillion less in value.

See my response to elharo. I agree that $9 trillion is too high.

I suspect that this estimate of value creation while interesting is rather likely to be used with misleading connotations. For example using it in a naive manner would result in 'altruistic' decision making that is grossly miscalibrated with respect to actually increasing expected value.

Do you think that I misused it in the present context? In giving my estimates I was attempting to strip out the altruistic component.

[-]wedrifid12y10

Do you think that I misused it in the present context?

Not yet. However to the extent that this post is a prelude to future analysis I recommend caution.

[-]JonahS12y00

I agree.

[-]ESRogs12y-10

The counterfactual replacement would have been created 3 years later (in expectation).

Interesting, my reaction was that it would have taken less time. So just based on the the contribution of the search engine, I'd rate the value of Google lower, but this estimate doesn't seem to take into account the long term value of creating the kind of company that Google became -- an engineer-led haven for top talent, with the motto of "Don't be Evil."

I think that's probably more significant than a 3 year jump in search quality.

[-]JonahS12y10

Thanks. My post is just a very crude initial calculation — I'll be doing more later.

[-]Shmi12y-10

The counterfactual replacement would have been created 3 years later (in expectation).

I doubt it. There were search engines before and after, none ever came close. Google and now Amazon have been creating far more wealth than a simple calculation may suggest.

[-]Luke_A_Somers12y40

When I switched to Google, the improvement over Yahoo felt modest (for the searches I was making). Barely enough to change my bookmarks.

[-]David_Gerard12y30

I found the difference remarkable and spectacular.

[-]gjm12y00

Did the two of you switch at different times? (If so, that might give some information about how long it would have taken for someone else to improve as much as Google did -- though they might have improved slower with less competitive pressure.)

[-]David_Gerard12y00

I started using Google in early 1999, when it still had "Stanford search" on the front.

[-]ThisSpaceAvailable12y20

At one point, Yahoo was actually using Google to create its searchable index.

[-]gjm12y00

When did you switch? (See my reply to David_Gerard for why I'm asking.)

[-]Luke_A_Somers12y00

Hmm. I'm pretty sure it was late 1999.

I think it's more likely that my searches were either really easy or nearly impossible, so the difference in performance didn't stand out.

[-]gjm12y00

Oh well. Another beautiful hypothesis slain by an ugly fact. So it goes.

[-]Locaha12y30

Do you think no one else could invent PageRank?

[-]SilasBarta12y70

I'm not sure myself, but the general PageRank algorithm (the method of adjacency matrix eigenvectors for determining critical nodes) had been around since the 60s, but hadn't been properly mapped to the site ranking problem until them (which is not a trivial problem at self, at least at the stage of recognizing the relevance).

[-]gwern12y20

Supposedly PageRank isn't even very important anymore, and hasn't been for a long time.

[-]JonahS12y00

There were search engines before and after, none ever came close.

But "after" Google developed dominant position and everyone started using it, and working for them. How long do you think that the lag would have been?

[-]DanielLC12y-10

rather than its counterfactual replacement some years later.

Why later? Perhaps its replacement would have been Yahoo!, created earlier.

[-]JonahS12y30

Yes, it could have been Yahoo! — what I meant is "counterfactual replacement of the same quality." It would have taken a while for an existing search engine to reach the same level of quality.

[-]sanddbox12y-20

Obviously the actual number itself is completely arbitrary, although I think you did a pretty good job estimating and giving a relatively realistic range. The full impact of Google, of course, can't really be quantified; it's impacted the world culturally, technically, socially, economically, etc. When you think about it, things that we understand qualitatively but not quantitatively are usually massively complex.

[-]JonahS12y10

You're right.

My gut inclination is to to say "there are lots of impacts that we can't quantify, but there would also have been lots of impacts that we couldn't quantify in the case of the counterfactual, and I don't imagine that they would have been systematically better in one case or the other, so I'll ignore such impacts" but this feels like a dodge. Do you have any thoughts there?

I do think that the "how much would people have been willing to pay for it?" metric picks up on more qualitative impacts than might initially meet the eye.

Moderation Log