Tuesday, November 13, 2007

Is Google paying you?

Kamal Jain gave a talk today on "atomic economics". Here is my own description of what he meant, though it seemed Kamal disagrees with at least some of my description :)

Think of credit cards. Whenever I pay with one, the store only gets 1-x% of the posted price, i.e. the the credit card company makes x% of the money. That is nice profit for the CC company, but as always, competition is driving all profits towards zero. Since there are multiple banks willing to give me a CC, and they all retain the same percentage from a merchant, they need to compete on the customer side --- they have to fight for the honor of counting me as their customer. So they start giving me points / miles / cash back / etc. Getting 1% of your money back is already quite standard, and customers need more to be impressed. To some extent, this incentivizing actually works (personally, I tend to use my AmEx card due to the better reward system).

Now, Google and Yahoo are making money because of me (merchants pay them when I click on an ad). In principle, this means we are not at an equilibrium: since these companies are competing for customers (equal ad revenue), one of them should start offering me a share of their profits whenever I click, and the other will have to follow suit.

Kamal sees several problems (market inefficiencies) that will prevent this welfare-maximizing scenario. I also see them as problems, but I have a different opinion on the degree of surmontability. Problems:

  • Customer abuse: profit sharing has to be tied to some actual buying event, or otherwise I will have an incentive to spend my day clicking on ads (and not buying anything), driving the merchant out of business. This is a problem for branding ads (e.g. eBay likes to display worthless ads all the time to build its name into collective conscience), and for ads that are separated from the actual purchase (e.g. an ad to a restaurant). But for e-commerce, which seems to be the raison d'etre of search advertising, this detail can be worked out by sharing information between Google and the merchant.

  • Merchant fraud: the merchant marks the price up by $5, adds $5 to the advertising budget (getting a top slot), and tells me that I will get $5 cash-back from Google (which means I don't care their price was $5 higher than the competition). Realistically, this won't happen, for several well-understood reasons. Common sense dictates that Google will only share a percentage of their profit, so I'll only get, say, $2.5 back, which means I care about the price markup. Like credit-card companies, Google and Yahoo have a weapon to enforce this common sense: customer differentiation. This is oligopolistic behavior, but one that is remarkably stable. What happens is that frequent buyers (resp. people with good credit histories) get better cash-bask deals, because they are presumed to have more authority to "negotiate". Then, the merchant gets a mix of people with different cash-back percentages, but who look the same to the merchant.

  • Bounded rationality, a heavily used buzzword, and the really exciting consideration. Unlike traditional advertising, Google and Yahoo have really tiny costs. This means they can still make a ton of money with a razor-thin cut of the merchant's profit. For instance, if Google is paid 0.1% percent of the transaction value, what could they offer me as an incentive? Maybe 0.05% of the price. But that will be irrelevant to me -- the mental cost of remembering that Google gives me more cash back on lingerie gifts vastly outweighs the 0.05% payoff.
Kamal's point is that the last phenomenon is happening more and more often, as the economy moves to an atomic scale, making the relative cost of rationality in each decision increase. He draws a funny parallel to the Heisenberg uncertainty principle, which IMHO only goes as far as "on a small enough scale, things become funky".

The market inefficiency generated by the small scale of Web transactions is great for Web-based companies like Google, since it means they can make huge profits below the radar of competition forces. The sad news (for a scientist) is that it all becomes a social competition, instead of a scientific/economic one. It's much more important if I, the customer, like Google's colors than if I'm getting my fair share of cash back.

On a larger scale, is this a social problem? I guess so, but not a very big one. If our threshold for irrationality is below 1%, the total income of such companies cannot grow above 1% of the money we all spend. But we're already paying the Federal government some 30% of what we make, and quite possibly many of us are happier with the success of Gmail than the hearts-and-minds campaign in Iraq.

PS: Given the topic, the number of jokes and anecdotes in Kamal's talk was too small by an order of magnitude. This reinforces my belief that we shouldn't be doing this kind of thing (i.e. study pure economics, because it is has something to do with networks or computers). If such pursuits will be more than wasted theoretical brain-cycles, they need to impact managers and policy-makers. But for that to happen, we need to give talks at the level of political speeches, and no self-respecting theorist is willing to do that. People in economics (who incidentally serve on policy-making boards every now and then) have already developed the art of communicating to politicians.


Anonymous said...

we need to give talks at the level of political speeches, and no self-respecting theorist is willing to do that.

I find it sad that making one's work accessible to the public is looked upon with such disdain among theorists. Maybe this explains the sorry state of theory funding.

MiP said...

Anon, I am all for putting more effort into explaining our results. I actually suspect I'm among the pedantic crowd here.

But convincing a politician/manager to make an important decision based on your theoretical model is not about explaining the result to a non-theory audience. It takes some kind of skill, that I don't recognize in many theory people. The obvious difference is that part of the economics crowd was selected for this skill, while the theory crowd was not.

Anonymous said...

Sounds lame and pointless.

Besides, Google and other search engines are competing for user of search... in search quality, that is, producing relevant search results (including ads). People won't flock to other search engine just because of some 0.03c for non-fraud ad-click.

MiP said...

This is not pointless at all. The question is how much money Google can make without competition cutting into their profits. A lot of political effort is spent thinking how to regulate or not regulate markets.

If you want, an alternative view of the question is: what happens if you start a company which sacrifices search quality a bit (say a few percent), but charges advertisers much less money? This would be the "generic" version of Google. Let me remind you that you are benefitting a lot in life from generic almost-any-product (like aspirin).

bálint said...

I think, Mihai, you cannot compare credit card companies to google like this.

The main activity for google users is to search, and they get that service for free. One gets to see ads, but it is not the main reason to go to search engines. This is why I have to agree to anonymous 2, people won't switch to a worse search engine if they get paid from ad revenue.

And in general there is no reason google should pay back users, google already "pays back" to users by providing excellent free search service.

I think in general, advertising moves away from the old system: "mass advertising, lots of money = people get annoyed by ads." The new scheme is "well-targeted ads, cheaper = happy customers" So it is not the money shaping the future of the ads market, but rather the quality. This is why in my opinion sharing ad revenue on search results is not an issue.

Anonymous said...

If you want, an alternative view of the question is: what happens if you start a company which sacrifices search quality a bit (say a few percent), but charges advertisers much less money.

Ads are given by auctions so this would not effect the status quo at all. Bids will just increase by that factor.

I think in general, advertising moves away from the old system: "mass advertising, lots of money = people get annoyed by ads." The new scheme is "well-targeted ads, cheaper = happy customers" So it is not the money shaping the future of the ads market, but rather the quality. This is why in my opinion sharing ad revenue on search results is not an issue.

If the well-targeted advertising is go good why can't there be well-targeted search. And if there is a well-targeted search why would I ever click on an advertisement.

bálint said...

If the well-targeted advertising is go good why can't there be well-targeted search. And if there is a well-targeted search why would I ever click on an advertisement.

First of all, search engines try to make well targeted search (e.g. results depend on the geographical location of your ip, language settings, you can switch on personalized search...)

And why doesn't this kill ad revenues? I think the answer is because I, as user, don't care if I click on and ad or an organig result, as long as I get what I want. And when I am looking for a product/service, pretty often there are several offers which I would happily take. So the merchants pay the ads so that their link is on the top of the page and probably I will read their offer first. And if they are really a good match for me, I might take it and not look for other offers... Merchants pay to get my attention, and they do this exactly in the moment when I am really open for their offer.

And in a way it is an equilibrium. Merchants have no incentive to trick me and show me irrelevant offers, because anyway I wouldn't purchase... remember, I was looking for one specific thing on the first place. Users happy that they got a good offer, google is happy because he got his cut, and the merchants are happy because they sold something (as long as the system works well and the right offer came to me)

Just as a note. I am not saying that we should mix organic results and ads. But as long as ads are relevant and pagacked the right way they might even be useful.

arnab said...

Hi Mihai, this is off-topic, but I just encountered your blog today. Very good writing and entertaining views. Hope to follow it more and contribute in the future. Best wishes for Monday :)


Anonymous said...

Google is paying me. It gives me a free search service, a free blog service, a free watch-videos service, a free translation service, a free academic index service, a free mail service, etc. Look up the cost of ISI science index to see how much this transfer is worth. Or cost out the price of translation software or extra space in other "free" mail services.

Kamal Jain said...

I just found the article today otherwise I would have participated in it.

Credit card companies charges like 2% commission and they are willing to share 1% in cash-back and sometimes even more. Of course they also provide free service to most credit card holders of enabling a convenient form of transactions. But the competition forced them to do even more on top of that free service. Competitive economy is based on incentivizing companies to offer their best deal. To do one up on a competitor gradually makes the world progress. If the "best" deal incentive is not there then in the long term the industry stagnant. There is no reason why search industry won't stagnant if people do not expect the best deal? People are short sighted in that sense that's why making policies and academic studies enter the picture.

Search advertising cost is not 1% as somebody is suggesting here. It is like 10%, 20% or sometimes even 30%. A company pays like an order of $1 for a click. And not every click results in a purchase. If credit card companies can offer you 1% after charging like 2% from merchants and paying the cost of the free service provided to you, a search engine can do much better. (I had heard that reward cards have 5 times more usage than non-reward cards).

That was not the point of my talk. If that is what Mihai had concluded then I did a poor job in communicating to him. The reason is that I do not myself have full clarity on the issues. I do not pretend I do. It was even a part of the short abstract I distributed.

The point was that search advertising is not advertising in the traditional sense. In search advertising, an economic system forms for a very short duration at atomic scale. The service sold is atomic, and the price is atomic. This atomic system in some sense has the properties of a micro economic system. Some of the properties are those which we would hate in case it was not an atomic system but a micro-economic system. For an example the ability of a search engine to use monopolistic auction. Search ads are sold by auction but the bread at your nearby supermarket is not. Why? Because auction requires monopoly otherwise it is quite an inefficient method of pricing. The question you should ask, how could all the search engines have monopoly? How come an industry with multiple companies can provide monopoly to each company? Oil rich countries are like that but they act as a cartel. But as far as I know, search engines do not act as a cartel.

A system is atomic does not mean we should get suboptimal deal like only free search when there is a potential of doing more. Raise your hands if you do not like rewards from credit cards on top of the free service they provide? If it was only a small number of atoms, we could have ignored the inefficiency in it. But the number of atoms is large. A lot of searches happens everyday. The large number of atoms compensate for the fact that each atom is small. Therefore we all get affected by the problems which happen at atomic level. For an example, on the mobile web people are speculating that search companies not only will offer free search but will also try to subsidize the phone/data-service on it. Why can't the search engines do the same on the landline and help improve the sorry state of broadband penetration in the country? Answer: the mobile web does not have net-neutrality. Don't be confused by the word neutrality.

The mobile web can correct the problem that a search engine in this atomic economic system has created. How? Because in the mobile web there is a complete chain of money from advertisers to a search engine to a phone company to a user. A link is missing in the case of broadband web. An economy always wants to route money from where it is in excess to where it could raise the most social welfare.

We theorists are happy because many of us study monopolistic auctions. Earlier there was no mass market monopoly which a government was not trying to regulate. The applications of monopolistic auctions were like FCC auctions which happen once in a while. (Ebay auctions are rarely monopolistic e.g., in the case of selling a rare piece of art.)

The search engines gave hundreds of million monopolies everyday. We theorists just love that we suddenly became more important! We are researching auctions which can make even more money in a monopolistic setting. But we do not ask, why should we allow a monopoly in the first place? Because if we prevent that, then we should be researching auctions which can make even more *social welfare*. The profits should follow as the social welfare increase.

For an example in the Vickery auction, we compute the social welfare and let the agents get the amount of money equal to the incremental social welfare they create. So in that sense it is not important that many of us use free searches from Google, Yahoo, and Microsoft. What is important is that, if Google was not running a search engine and all our searches were going to Yahoo then how much social welfare we would lose. Personally, I would lose almost nothing. But whatever that loss is, that is the rightful profit claimed by Google or as a matter of fact by any other search engine. Any money on top of that claim belongs to the societies i.e., searchers and advertisers. If that is not happening that means the economic principles are not working as we want them to work. Our trust in economic systems is quite similar to our trust in the pilot of a flight. His life is as much stuck in the safe landing of the plane as ours. Similarly we created or at least we thought we created a system of competitive economics, where each company’s interest in increasing social welfare is as much as in its own profit. Of course sometimes flights crash and we try to find the reason behind it and prevent further crashes. Similarly sometimes our economic designs do not work, e.g., in the case of the search engines. We should research and find ways to correct it.

But quite opposite, we are finding ways to extract as much profit in a monopolistic auction from the societies as possible instead of the other way around. Does not it disturb you? It surely does to me.

The fact is that a monopoly is a monopoly even if it is at an atomic scale. The competition should break it. I want the search engines who are not yet winning in this game to break this monopoly. Unfortunately I do not see an answer. If it were a microeconomics system, like credit card, then the existing solutions would have worked. But this atomic economic system is lot more puzzling.

If you think you have a solution, please do publish a paper. If there was an easy solution, believe me the competitive economic systems would have been working. The non-dominating players are always looking ways to pass the social welfare back to societies. If they do not have ways, like undercutting the dominating player then that’s not a good position for the societies.

PS: Sorry for the long comment. I wrote a comment but when I pressed publish, either the blogger site went down or my connection broke. So I have to re-write the whole comment.

bálint said...

I think that your comparison of credit card companies with search engines is not appropriate, Kamal. As I have written before, I think there is a conceptual difference between credit card companies and search engines. As a credit card customer I get a payment method and I pay for this service. As a search engine customer I get a free service to search, and if I want I can check out ads, but I never have to give anything to the search engine. If I think it is value for me, I click on some ads, but if I don't want to, I don't give anything back to the search engine (don't click on ads).

On the surface it is true that the ad auctions are small monopolies. But if you look more closely, the merchant buying the ads can freely choose which monopoly (search engine and keyword) he wants to pay for, so these monopolies are indirectly connected by those who pay for the ads. Thus they aren't real monopolies.

And you say the mobile web fixes problems with the landline data services. Well, mobile operators I think are special case by themselves because of the high entry-barrier market, and they don't work as real free market. If mobile data communication infrastructure would be as cheap as the landline, we wouldn't have only a couple of operators and we wouldn't have to sign a 2 year contract when we buy a new handset. In my view it is the other way around. Mobile operators keep strong control of the customers and because of the limited choice for users, the operators can afford to charge them maybe higher than the real value of the service. By "subsidizing" service (if it ever happens) with ads they are not being nice to the customers, they just adapt to the new conditions of the market.

Maybe I didn't understand you, but I don't see real reasons to say that the search engines' economic systems are not the right ones, and they harm the society for the profit of the search engines.

Kamal Jain said...

Balint, I do not know how credit card works in your country. But in the USA most people do not pay for the credit card.

People get the credit card and it comes with the free method of peyment. Credit cards are free for the users and of course if you pay on time you get a free loan for a month. Merchants pay for the credit card. They pay like 2% plus slightly more.

It takes 2% to serve some users and less than 2% to some other. For an example, a user who is a credit risk is expensive. But a user who is not a credit risk but spends thousands of dollars every month is cheap to serve. Many costs in the credit card business scales with the number of users and not with the amount of dollars. For an example, the cost of sending bill. So a big spender is really cheap to serve.

So there is a heterogeneity on one side of this two sided market. Since some customers are cheaper, credit card companies try to attract these customers. In doing so they give back the benefit of being cheaper to serve, to the cheaper customers. For an example 1% reward. I think come credit card companies even pay 2% rewards to big spenders.

Just like credit card companies provides free service, search engines provide free service too. Neither credit cards nor search engines have "rewards" in the core of their service. Reward credit cards is an outcome of competition in the market. Earlier, credit card companies used to charge a fee from users. Then the fee become zero. Now it has become negative (i.e., rewards).

By your logic, credit cards should not also give the reward back to users. You are free to return your reward check back but I love to deposit it in the bank.

The good thing is that the competition has forced even the best of the credit companies to offer the rewards. In the united states, every major bank has reward cards. So a user is not losing on the quality of the transaction service either.

Note that just like credit card companies have heterogeniety of users, search engines have the same. Many users click ads and do purchases. Some users, like an academic researcher, just search and do not click ads. Many users click ads. Some do not even know that the ads are ads. Search engines do not label ads as ads. They label them "sponsored links, sites, results or some such thing". But they never label an ad as an ad. (some of them do label ads as ads on their publisher network pages).

In any case, I agree that search engine situation is different than credit card situation. That was the whole point of my talk. It is the atomicity of the economic system.

I am not sure if you have attended any of my talk on the subject. Neither I nor Mihai has expressed the content of my presentation in this blog. So we can't discuss on technicalities. An important part of the talk was modeling the search engine business.

Anonymous said...

Kamal, you keep saying that credit cards provide a free service, or provide a service that merchants pay for, not customers. You're buying (wholesale!) the idea that the only cost of a credit card is the high interest rate, which can easily be avoided by paying your bill every month. In response to the 2% fees merchants simply raise their prices accordingly ***which affects every customer***. One of the key parts of the merchant-creditcard contract is that they do *not* pass that 2% fee onto the credit card purchases. If merchants could do that they would, and kill any illusions you had about a "free" service.

Kamal Jain said...

Anonymous@13, who says the advertisement costs are not passed to you?

In a competetive economy you pay the cost of the second least cost efficient supplier. It includes all kind of costs.

Anyway, you need to read the discussion from the top. Balint claim that search is free (to customers) therefore a search engine should give only a good enough deal to consumers and not the best under the competetive system. In credit card cashback he said, credit cards are not free to consumers.

To me either both are free. Because a user does not pay directly. Or both are pricey because the merchant adds up the cost to whatever the merchant is selling.

Take your pick and model the situation.

The big difference you will find is in the details. One is an atomic economic system and the other is a microeconomic system.

One is actually selling the information (so competetive equilibrium theory does not hold which assume perfect information), the other reaches the competetive equilibrium because the information is assumed to be there.

Anonymous said...

The big difference you will find is in the details. One is an atomic economic system and the other is a microeconomic system.

The difference is in the details, but the most relevant detail here is contractual obligation. It doesn't make sense to say merchants or consumers act in certain way because it is rational if they are contractually obligated to do otherwise. As you point out, all kinds of costs can be passed along to the customer such as advertising, the electric bill, etc. The difference here is that merchants *do not want to* pass the cost of advertising to specific customers, whereas they are contractually obligated not to pass the cost of credit cards to specific customers.

There are several other situations where an absurd (but stable) situation could be created by contractual obligation. For instance, cigarettes have a much higher tax rate than everything else sold in a 7-eleven. The cigarette companies could insist that they be sold for the nominal price and let 7-eleven push the tax burden on to non-cigarette buying customers. Absurd? Yes! Here's another, which is less absurd. Clothing stores sell in malls and online and the cost of the former is significantly larger due to high rents in the malls. They do occasionally give better deals online, which pisses off the malls for obvious reasons. The malls would prefer to have a clause in the lease that says the clothing store cannot maintain two price structures.

The reason the situations above don't actually happen is only structural. The contractual structure of credit cards has many of the properties of a protection racket. This isn't to say it is a scam. The mafia DID provide protection and other services and maintained stability. However they were not efficient.

Kamal Jain said...

In today's world the cost of accepting credit card may be cheaper than the cost of handling cash. Similarly if a check writer is holding the line for two extra minutes that's also a cost.

That's irrelevant. The relevant part is that certain credit cards gives back the "extra" price charged by the merchants to society via rebates to consumers. They are not doing through altruism. They are doing because there is a price competition in the market which brings the best deal to users.

There is no price competition in the search engine market. Search engine is among the cheapest form of marketing (in terms of social cost), but the society do not get to enjoy the cheapness of this marketing vehicle. Ideally, search engines try to get the true utility of selling an item as bid. True utility is the maximum profit! Or it is the amount of money a merchants was paying to inefficient sales channel. So the search engines cost the same to societies as other inefficient channels. As the world have progresses, should not the society get the monetary benefit of the progress?

Why the society can't claim its fair share in the Search engine but could claim its fair share in other goods/services?

The answer lies in atomic economics! The search engine is actually not a singular example of an atomic economics. I guess I mentioned about 10 of those in my presentation at MIT.