Google’s Search Auctions and Market Power

I have a plan of market division for The Wall Street Journal: I’ll write about antitrust; they will stop writing about it and will instead write about other topics where the local knowledge runs deeper.

Today’s topic, like last week’s, is the proposed Microsoft-Yahoo! search deal. L. Gordon Crovitz addressed this last week in his commentary in a way that I thought was particularly thoughtless (see my post here and Josh Wright’s post as well). Today’s Wall Street Journal editorial at least recognizes that the original deal was Google-Yahoo! though they still fail to connect the dots to see that Microsoft-Yahoo! is only possible because of the prior review by Justice of Google-Yahoo!. They continue to believe that Microsoft-Yahoo! is pro-competitive, as did Crovitz, as do I, but none of us has done a serious antitrust analysis. Monday morning quarterbacking is fun and easy even in antitrust.

So what is my beef with today’s editorial? The Journal has drunk the Google Kool-Aid on pricing in search markets: “search providers like Google and Bing also don’t determine ad prices, which are set through auctions.” This is wrong. From the perspective of the auctioneer, the whole point of auction design is to figure out how to make more money from the auction. I gave a quick seven-minute speech on this as part of as debate hosted by Intelligence Squared in New York in November and you can watch that but let me describe the highlights quickly.

I described three features of Google’s auction design. First, Google had pushed up minimum bid prices over time: you couldn’t pay just a penny when your ad was clicked on, but you had to instead pay some minimum amount designated by Google. That resulted in many organic searches not being matched with any ads, because if you bid less than the set minimum, your ad wouldn’t show at all. It turns out that there is good reason to think the Google makes more money this way. It extracts more from bidders on average even when there is relatively little competition for the available ad slot. As I say in the talk, this is an exercise of market power, one that seems to flow naturally from Google’s dominant position in search. But not illegal: monopolists can legally charge high prices without violating Section 2 of the Sherman act.

Second, an ad auctioneer controls the number of slots available for advertising. Google’s own Hal Varian, its chief economist, has written quite clearly on the revenue consequences to an auctioneer of over- and under-subscribed auctions. The simple version of the idea is that if there are three bidders, an auctioneer will make a lot more money having two slots rather than four slots. With four slots, the auctioneer gets three minimum bids, while with two slots, it gets two maximum bids. Two x max > 3 x min, and of course if it isn’t, the auctioneer can just have four slots. Yes, the search auctioneer does not set price directly but instead powerfully influences price in selecting the number of slots available.

Third, Google has engaged in some subtle bundling. As of November, Google provided ad matching services for its own search engine but also for Ask.com and AOL. You could buy Google alone but you could only buy placement on Ask and  AOL bundled with Google. That kind of bundling can give rise to antitrust issues under both US and EU law.

Market division of course remains per se illegal under US antitrust law, so I don’t want the Wall Street Journal to take me up on my offer. Instead, the WSJ seems to face we what call a buy-or-build decision: what should you build on your own and what should you instead buy in the market? I think the WSJ should exit the building market on antitrust analysis and should buy in the market instead. I await their call.                               

[Disclosure: I consulted for the opposition to the Google-Yahoo! search deal.]

Comments

[michael webster]

I agree with much of this, and would add the following.

1. Google quality score affects your ad placement, and quality score is fairly opaque.

2. On the content network, ad placement is much more opaque.

3. Ads can have high quality score, but Google decides that only the first 10 run on the first page, and nothing runs on the second search page.

4. Google's attitude to trademarks changes constantly. Now limited use of trademarks in adwords is allowed.

This is an excellent anti-trust problem, and I look forward to you following up with a paper called "Auction Design - The Anti Trust Issues."

[Resveratrol]

Well, None of us has done a serious antitrust analysis. Monday morning quarterbacking is fun and easy even in antitrust.

[Katie]

There have been several lawsuits filed against Google the past few months over the use of trademarks in AdWords advertisements, including a class action suit in Texas, so Google's attitude towards trademarks will eventually be defined.