Antitrust and Google search bias

Lynne Kiesling

For the past year and a half the Federal Trade Commission has been investigating the potential anti-competitive effects of Google’s search-based business model. The European Union has also been pursuing antitrust complaints against Google. The main accusation is Google search bias — Google’s algorithm prioritizes links both to paid advertisers (which are shaded and labeled to indicate the payment) and to affiliated content sites. Google’s competitors complained to the FTC … but they do the same thing! For example, if you do a stock ticker search for, say AAPL (yes, I’m being cheeky), on Google, Bing, and Yahoo, each one will prioritize its own affiliated finance site, before then listing the sites of their competitors, Wall Street Journal, and so on.

We had a panel discussion on this issue at the Northwestern University-Searle Center annual conference on antitrust economics and competition policy last Friday afternoon, with panelists including Stanford’s Susan Athey (who has done some work with Microsoft) and Google’s Hal Varian and Preston McAfee. The discussion was as informed, informative, and lively as you’d expect.

There’s also a panel debate going on right now in DC in on the issue, hosted by Tech Freedom and including friend-of-Knowledge-Problem Geoff Manne, who has written extensive criticisms of the FTC investigation. I think Geoff has an important point when he asks for evidence of consumer harm in comparison to what a likely antitrust remedy would be. If, for example, the FTC required Google to modify its search algorithm to randomize the top results rather than prioritizing their affiliated content sites, then wouldn’t Microsoft and Yahoo have to implement that randomization as well? And if that’s the remedy, does that make consumers better off or worse off?

I think Bob Hahn and Peter Passell get it right when they say, as they did in a post yesterday at their blog regulation 2point0,

Indeed, anybody who’s been paying attention ought to have figured out by now that information technology is simply moving too fast to allow even the most nimble companies to grab the market goodies and lock the door behind them. …

In a hypercompetitive environment like this, where the product mix sometimes changes faster than Lady Gaga’s wardrobe, antitrust regulators would do well to pick and choose their interventions carefully. And to help get them there, academics really need to provide a more careful accounting of the state of competition in IT.

Hear, hear.

What do you think? Do you think you are “locked in” when you perform a search on a specific platform? Do you just click on the top link? Or when presented with search results do you look for specific sources that have a particular reputation or credibility to you?

FTC finds no evidence of illegal activity after investigation of Western New York gasoline prices

Michael Giberson

Apparently you just have to know who to ask. Yesterday, the FTC sent me a copy of the FTC letter to Representative Brian Higgins describing the agency’s investigation of Western New York gasoline prices last fall. (For background see this earlier post and here.)

Be aware that here begins a very long post about gasoline prices in Western New York.

First a few selections from the FTC letter:

Dear Representative Higgins:

You and Senator Charles E. Schumer have requested a public report on the Federal Trade Commission’s investigation into unusually high gasoline prices in Western New York during the fall of 2008. Thank you for bring this important issue to our attention. We share your concern about the impact of high gasoline prices on the day-to-day life of consumers and understand the frustration and hardship that are created when those prices rise significantly above those in surrounding areas without any obvious market explanation, as occurred in this instance. Such situations receive our closest attention.

However, after careful and extensive investigation, FTC staff did not find any evidence of illegal activity in gasoline markets in any of the affected cities. To the contrary, staff found evidence suggesting it is unlikely that illegal conduct caused these price levels, although staff was unable to identify precise reasons why retail gasoline prices in some cities in Western New York and Vermont did not fall as quickly as prices in other Northeast cities. Although we are unable to establish any direct relationship, we do note that prices began to fall soon after you raised public concerns about the elevated prices and both you and Senator Schumer asked us to conduct an investigation.  This letter describes the scope of the investigation and summarizes the findings of Commission staff, subject to the Commission’s obligations not to disclose confidential information.

RE: “we do note that prices began to fall soon after you raised public concerns”

This line, which Rep. Higgins quoted in his press release back in May, continues to strike me as unnecessarily servile in tone — offering the Congressman the flattering innuendo, but not actually claiming post hoc ergo propter hoc.  I think the word I’m looking for is “unctuous.”  But still, … well, I’ll say more on this point in my wrap up remarks.

I. Investigation of Unusual Pricing Activity in Western New York

The Commission’s ongoing Gasoline and Diesel Price Monitoring Project identified retail gasoline prices significantly above predicted values in Western New York cities, and in Burlington, Vermont, during the fall and early winter of 2008.  In response to these observations and to requests from you and Senator Schumer, Commission staff conducted an analysis of retail gasoline prices in Western New York and Burlington, Vermont, to confirm that prices in those markets were unusually high.

Staff first analyzed whether average retail price levels in the Buffalo, Rochester, and Jamestown, New York, and Burlington, Vermont, metropolitan areas were higher than would be expected, using their normal relationship with Albany gas prices as a baseline.  Staff analyzed price data for a ten-year period to establish historical differences between average retail prices in these cities and Albany.  This analysis confirmed that average retail gasoline prices in these cities were significantly higher than expected relative to Albany.

I’ll skip the full discussion of the methods of investigation. In brief: Staff looked for, but did not discover, any supply disruptions or other unusual market conditions; they coordinated with the attorneys general of New York and Vermont; they interviewed and obtained documents and data from numerous relevant companies; and purchased data from the Oil Price Information Service.

Through its investigation, staff discovered that no company possessed a monopoly share of any retail gasoline market in Western New York or Vermont, nor was any company large enough to effectively attempt to create a monopoly through illegal means.  Further, staff identified no unfair method of competition that could explain how a company or group of companies could have illegally caused the observed price levels last fall.  Accordingly, staff’s investigation focused on the only remaining plausible theory of illegal behavior … — that companies … might have engaged in collusion.

Collusion in each of the affected cities would have been very difficult because numerous companies set prices at retail gas stations in each city and no single station owner or group of owners controls a large share of the volumes sold in any city….

Collusion across all of the affected cities would have been even more difficult because numerous companies other than those that operate in Buffalo set retail gasoline prices in Rochester and Jamestown. …

Other market factors also would have made collusion very difficult.  For example, as crude oil prices plummeted during the fall, product costs for gasoline retailers throughout the nation fell with unprecedented speed and magnitude.  As wholesale gasoline prices fell substantially on a daily basis, the numerous retail price setters in each affected city would have had to reach agreement on cartel prices on a frequent basis – probably each day if not more frequently.  Having to reach agreement so frequently would have made it very difficult to effectively maintain a collusive scheme throughout the fall of last year.

Nor did market data support the notion that a conspiracy existed to raise prices last fall….

In sum, staff’s investigation yielded no evidence that illegal anticompetitive conduct caused the price levels experienced in Western New York or Vermont last fall.

[Emphasis added. The link above was a footnote in the original letter. The letter continues with a brief discussion of policy options. Probably the less said about them, the better.]

So the “evidence suggesting it is unlikely that illegal conduct caused these price levels” mentioned in the second paragraph of the letter comes down to discovering that many companies are involved in price setting in the affected areas, and that the circumstances make it unlikely in the extreme that these companies could have collusively coordinated prices.  With some confidence we can conclude that the FTC has ruled out most potential supply-side explanations for the relatively higher gasoline prices in Western New York.

Markets are made of supply and demand, of course, and if supply factors are not the cause, then the economist’s attention should tend toward possible demand-side explanations. That the period of interest was one in which wholesale and retail prices were falling “with unprecedented speed and magnitude” is of particular note.

As it happens I have access to a year’s worth of OPIS price data for New York that includes fall and early winter of 2008.  Not the ten years of data that the FTC has, but it does show some of the relevant points.  As of May 1, 2008, prices in Buffalo averaged $3.75, which was about 3 cents below prices in Albany that day. By the end of June, prices in both areas were higher and the price in Buffalo had caught up to the price in Albany.  Mid-July, prices begin to fall, following the wholesale price down.

But, and this is the issue that would catch the Congressman’s eye, prices in Buffalo did not fall as quickly as prices in Albany.  In the OPIS data I have, average retail prices in Buffalo and Albany were both $4.24 on June 30 (some data selection issues behind these averages, email me if you have questions, but the numbers should be pretty good).  Through early July, Buffalo prices were a few pennies higher and Albany a few pennies lower, then prices started to fall “with unprecedented speed and magnitude”, as the FTC put it.  This unprecedented fall can be seen in the following chart created via GasBuddy.com.

Gasbuddy_New_York_prices_2003-2009

Rep. Higgins’ first letter to the FTC requesting an investigation was sent October 22. On that day prices in Buffalo averaged more than 30 cents per gallon higher than prices in Albany, a 35 cent swing in relative prices compared to the May 1 averages.

The question is why, and if the answer is not found on the supply side, then let’s look at the demand side of the market.

Notice in the six years of data charted above that when prices are falling, they tend to fall a little faster in Albany than they do in Buffalo or Rochester. In fact, until 2008 Rochester prices tended to fall much more slowly than prices in Albany or Buffalo.  For Buffalo, and now I’m just eyeballing the chart, the result appears to be a price higher than in Albany by few cents per gallon for as much as a week or so.  Then prices in Buffalo catch up with prices in Albany.  Two things are different in the price fall of 2008: first, with prices falling so fast and so far, it was months – not days – before the slower falling prices in Buffalo caught prices in Albany; and second, this time prices in Rochester tended to fall a little faster than prices in Buffalo.

Asymmetric price adjustment – prices falling more slowly than they rise (also called the “rockets and feathers” phenomena) – have been extensively studied in gasoline markets. Work by Matthew Lewis has put forward some interesting consumer-search based explanations of the phenomena. (See this paper and related papers available here.)  It might be the case that consumers in Albany tend to acquire more price information before buying gasoline when prices are falling, at least relative to consumers in Rochester and Buffalo, and that could explain why prices tend to fall faster in Albany.  Maybe consumers in Buffalo were just happy to get lower prices, relatively speaking, and not overly focused on searching out the lowest possible among the lower prices before buying gasoline.

Interestingly, this argument brings me back to the FTC’s flattering innuendo. If the explanation for the relatively slow fall of prices in Buffalo was a lack of consumer attention to finding the lowest possible among lower prices, then the Congressman’s efforts to put the issue back in the news just might have spurred a little more consumer search behavior, which then served to push prices down a little faster. I’m not convinced, but wouldn’t claim “no effect,” either, without more study.

Also interesting is the right-most edge of the chart above.  Again using GasBuddy.com, the chart below shows average prices for Albany, Rochester, and Buffalo just for the most recent three months. Prices peaked in the state around June 20, and have been falling since.

Gasbuddy_New_York_prices_Three-months-in-2009

Notice that prices are again falling faster in Albany than in Buffalo or Rochester.  Alert the media!

Congressman to Western New York gasoline retailers: We will be watching you

Michael Giberson

From the Buffalo, New York, BusinessFirst:

In a letter sent May 13 by FTC Chairman Jon Leibowitz to Higgins, the agency said after a careful and extensive investigation, regulators could not find any evidence of illegal activity in gasoline markets in any of the affected cities. The agency monitored prices in Buffalo, Jamestown, Rochester and Burlington, Vt.

“To the contrary, staff found evidence suggesting that it is unlikely that illegal conduct caused those price levels, although staff was unable to identify precise reasons why retail gas prices in Western New York did not fall as quickly as prices in other Northeast cities,” Leibowitz wrote.

What the agency did note was that after Higgins released an Oil Price Information Service (OPIS) report on Dec. 4, 2008 citing Jamestown and the Buffalo-Niagara regions among the top 5 most “profitable” for gasoline retailers, the prices for unleaded gas decreased from an average of $2.25 to $1.85 by the end of 2009.

Does this last paragraph suggest that a servile and pandering attitude on the behalf of the FTC toward the congressman? I don’t find the FTC letter online at either the FTC’s website or that of the Congressman, so I’m just raising the question based on the news article.

The congressman’s press release suggests, surprise!, that he is quite willing to encourage the view that his actions had something to do with prices falling back in line with state averages:

“Western New York consumers were getting ripped off and we sounded the alarm, which caused WNY gas prices to fall in line with state averages, again proving that when we stand up for ourselves we can get things done,” Higgins added.  After Congressman Higgins publicly released an Oil Price Information Service (OPIS) report naming Jamestown & the Buffalo-Niagara regions among the top 5 most “Profitable” for gasoline retailers on December 4th, 2008, the prices of unleaded gas decreased from an average to $2.25 in Buffalo to $1.85 by the year’s end.

To the congressman, absence of evidence is no barrier to action:

“While we might not have proof of illegal activity or a clear definition of why our prices were so high, what is clear is retailers were acting in bad faith trough some type of implicit collusion and retailers and consumers should know that we were watching then and are watching now and will continue to work to make sure this doesn’t happen again,” said Higgins.

A quick trip to the price charts at http://www.buffalogasprices.com tells some of the story (start here at the default Buffalo price chart for one month, then add two more upstate New York cities to the chart – Rochester, Albany, or Syracuse are options – and expand to at least 18 months).  Up until the 2008 mid-July gasoline price peak, retail prices in Buffalo tracked prices elsewhere in upstate New York pretty closely. When prices started falling, they fell more slowly in Buffalo than in the rest of the state.  By the time retail prices hit bottom at the end of the year, prices in Buffalo were back in line with prices elsewhere in New York, and since the beginning of the year prices in Buffalo have moved in line with other gasoline prices in the state.

So, yes, prices did fall more slowly in Buffalo. Presumably, a good economic study could uncover likely causes. Those causes may not turn out to be politically useful to local politicians. In any case, of the congressman’s agenda intended to “prevent price disparity in [the Western New York] region” — raise public awareness, push for passage of a federal price gouging bill, push for passage of a bill to prevent excess speculation in the oil market, and invest in renewable energy — only the first is likely to have any real impact on local “price disparity.”