What the FTC got wrong in the Google antitrust investigation | Patreon

From 2011-2012, the FTC investigated the possibility of pursuing antitrust action against Google. The FTC decided to close the investigation and not much was publicly known about what happened until Politico released 312 pages of internal FTC memos that from the investigation a decade later. As someone who works in tech, on reading the memos, the most striking thing is how one side, the side that argued to close the investigation, repeatedly displays a lack of basic understanding of tech industry and the memos from directors and other higher-ups don't acknowledge that this at all.

If you don't generally follow what regulators and legislators are saying about tech, seeing the internal c(or any other industry) when these decisions are, apparently, being made with little to no understanding of the industries1.

Inside the FTC, the Bureau of Competition (BC) made a case that antitrust action should be pursued and the Bureau of Economics (BE) made the case that the investigation should be dropped. The BC case is moderately strong. Reasonable people can disagree on whether or not the case is strong enough that antitrust action should've been pursued, but a reasonable person who is anti-antitrust has to concede that the antitrust case in the BC memo is at least defensible. The case against in the BE is not defensible. There are major errors in core parts of the BE memo. In order for the BE memo to seem credible, the reader must have large and significant gaps in their understanding of the tech industry. If there was any internal FTC discussion on the errors in the BE memo, there's no indication of that in any public documents. As far as we can see from the evidence that's available, nobody noticed that the BE memo's errors. The publicly available memos from directors and other higher ups indicate that they gave the BC memo as much or more weight than the BE memo, implying a gap in FTC leadership's understanding of the tech industry.

Brief summary

Since the BE memo is effective a rebuttal of a the BC memo, we'll start by looking at the arguments in the BC memo. The bullet points below summarize the Executive Summary from the BC memo, which roughly summarizes the case made by the BC memo:

In their supplemental memo on mobile, BC staff claim that Google dominates mobile search via exclusivity agreements and that mobile search was rapidly growing at the time. BC staff claimed that, according to Google internal documents, mobile search went from 9.5% to 17.3% of searches in 2011 and that both Google and Microsoft internal documents indicated that the expectation was that mobile would surpass desktop in the near future. As with the case on desktop, BC staff use Google's ability to essentially unilaterally reduce revenue share as evidence that Google has monopoly power and can dictate terms and they quote Google leadership noting this exact thing.

BC staff acknowledge that many of Google's actions have been beneficial to consumers, but balance this against the harms of anticompetitive tactics, saying

the evidence paints a complex portrait of a company working toward an overall goal of maintaining its market share by providing the best user experience, while simultaneously engaging in tactics that resulted in harm to many vertical competitors, and likely helped to entrench Google's monopoly power over search and search advertising

BE staff strongly disagreed with BC staff. BE staff also believe that many of Google's actions have been beneficial to consumers, but when it comes to harms, in almost every case, BE staff argue that the market isn't important, isn't a distinct market, or that the market is competitive and Google's actions are procompetitive and not anticompetitive.

Common errors

At least in the documents provided by Politico, BE staff generally declined to engage with BC staff's arguments and numbers directly. For example, in addition to arguing that Google's agreements and exclusivity (insofar as agreements are exclusive) are procompetitive and foreclosing the possibility of such agreements might have significant negative impacts on the market, they argue that mobile is a small and unimportant market. The BE memo argues that mobile is only 8% of the market and, while it's growing rapidly, is unimportant, as it's only a "small percentage of overall queries and an even smaller percentage of search ad revenues". They also claim that there is robust competition in mobile because, in addition to Apple, there's also BlackBerry and Windows Mobile. Between when the FTC investigation started and when the memo was written, BlackBerry's marketshare dropped dropped from ~14% to ~6%, which was part of a long-term decline that showed no signs of changing. Windows Mobile's drop was less precipitous, from ~6% to ~4%, but in a market with such strong network effects, it's curious that BE staff would argue that these platforms with low and declining marketshare would provide robust competition going forward.

When the authors of the BE memo make a prediction, they seem to have a facility for predicting the opposite of what will happen. To do this, the authors of the BE memo took positions that were opposed to the general consensus at the time. Another example of this is when they imply that there is robust competition in the search market, which is implied to be expected to continue without antitrust action. Their evidence for this was that Yahoo and Bing had a combined "steady" 30% marketshare in the U.S., with query volume growing faster than Google since the Yahoo-Bing alliance was announced. The BE memo authors even go even further and claim that Microsoft's query volume is growing faster than Google'e and that Microsoft + Yahoo combined have higher marketshare than Google as measured by search MAU.

The BE memo's argument that Yahoo and Bing are providing robust and stable competition leaves out that the fixed costs of running a search engine are so high and the scale required to be profitable so large that Yahoo effectively dropped out of search and outsourced search to Bing. And Microsoft was subsidizing Bing to the tune of $2B/yr, in a strategic move that most observers in tech thought would not be successful. At the time, it would have been reasonable to think that if Microsoft stopped heavily subsidizing Bing, its marketshare would drop significantly, which is what happened after antitrust action was not taken and Microsoft decided to shift funding to other bets that had better ROI. Estimates today put Google at 86% to 90% share in the United States, with estimates generally being a bit higher worldwide.

On the wilder claims, such as Microsoft and Yahoo combined having more active search users than Google and that Microsoft query volume and therefore search marketshare is growing faster than Google, they use comScore data. There are a couple of curious things about this.

First, the authors pick and choose their data in order to present figures that maximize Microsoft's marketshare. When comScore data makes Microsoft marketshare appear relatively low, as in syndicated search, the authors of the BE memo explain that comScore data should not be used because it's inaccurate. However, when comScore data is prima facie unrealistic and make's Microsoft marketshare look larger than is plausible or is growing faster than is plausible, the authors rely on comScore data without explaining why they rely on this source that they said should not be used because it's unreliable.

Using this data, the BE memo basically argues that, because many users use Yahoo and Bing at least occasionally, users clearly could use Yahoo and Bing, and there must not be a significant barrier to switching even if (for example) a user uses Yahoo or Bing once a month and Google one thousand times a month. From having worked with and talked to people who work on product changed to drive growth, the overwhelming consensus has been that it's generally very difficult to convert a lightly-engaged user who barely registers as an MAU to a heavily-engaged user who uses the product regularly, and that this is generally considered more difficult than converting a brand-new user to becoming heavily engaged user. Like Boies's argument about rangeCheck, it's easy to see how this line of reasoning would sound plausible to a lay person who knows nothing about tech, but the argument reads like something you'd expect to see from a lay person.

Although the BE staff memo reads like a rebuttal to the points of the BC staff memo, the lack of direct engagement on the facts and arguments means that a reader with no knowledge of the industry who reads just one of the memos will have a very different impression than a reader who reads the other. For example, on the importance of mobile search, a naive BC-memo-only reader would think that mobile is very important, perhaps the most important thing, whereas a naive BE-memo-only reader would think that mobile is unimportant and will continue to be unimportant for the foreseeable future.

Politico also released memos from two directors who weigh the arguments of BC and BE staff. Both directors favor the BE memo over the BC memo, one very much so and one moderately so. When it comes to disagreements, such as the importance of mobile in the near future, there's no evidence in the memos presented that there was any attempt to determine who was correct or that the errors we're discussing here were noticed. The closest thing to addressing disagreements such as these are comments that thank both staffs for having done good work, in what one might call a "fair and balanced" manner, such as "The BC and BE staffs have done an outstanding job on this complex investigation. The memos from the respective bureaus make clear that the case for a complaint is close in the four areas ... ". To the extent that this can be inferred, it seems that the reasoning and facts laid out in the BE memo were given at least as much weight as the reasoning and facts in the BC memo despite much of the BE memo's case seemingly highly implausible to an observer who understands tech.

For example, on the importance of mobile, I happened to work at Google shortly after these memos were written and, when I was at Google, they had already pivoted to a "mobile first" strategy because it was understood that mobile was going to be the most important market going forward. This was also understood at other large tech companies at the time and had been understood going back further than the dates of these memos. Many consumers didn't understand this and redesigns that degraded the desktop experience in order to unify desktop and mobile experiences were a common cause of complaints at the time. But if you looked at the data on this or talked to people at big companies, it was clear that, from a business standpoint, it made sense to focus on mobile and deal with whatever fallout might happen in desktop if that allowed for greater velocity in mobile development.

Both the BC and BE staff memos extensively reference interviews across many tech companies, including all of the "hyperscalers". It's curious that someone could have access to all of these internal documents from these companies as well as interviews and then make the argument that mobile was, at the time, not very important. And it's strange that, at least to the extent that we can know what happened from these memos, directors took both sets of arguments at face value and then decided that the BE staff case was as convincing or more convincing than the BC staff case.

That's one class of error we repeatedly see between the BC and BE staff memos, stretching data to make a case that a knowledgeable observer can plainly see is not true. In most cases, it's BE staff who have stretched data as far as it can go to take a tenuous position as far as it can be pushed, but there are some instances of BC staff making a case that's a stretch.

Another class of error we see repeated, mainly in the BE memo, is taking what most people in industry would consider an obviously incorrect model of the world and then making inferences based on that. An example of this is the discussion on whether or not vertical competitors such as Yelp and TripAdvisor were or would be significantly disadvantaged by actions BC staff allege are anticompetitive. BE staff, in addition to arguing that Google's actions were actually procompetitive and not anticompetitive, argued that it would not be possible for Google to significantly harm vertical competitors because the amount of traffic Google drives to them is small, only 10% to 20% of their total traffic, going to say "the effect on traffic from Google to local sites is very small and not statistically significant". Although BE staff don't elaborate on their model of how this business works, they appear to believe that the market is basically static. If Google removes Yelp from its listings (which they threatened to do if they weren't allowed to integrate Yelp's data into their own vertical product) or downranks Yelp to preference Google's own results, this will, at most, reduce Yelp's traffic by 10% to 20% in the long run because only 10% to 20% of traffic comes from Google.

But even a VC or PM intern can be expected to understand that the market isn't static. What one would expect if Google can persistently take a significant fraction of search traffic away from Yelp and direct it to Google's local offerings instead is that, in the long run, Yelp will end up with very few users and become a shell of what it once was. This is exactly what happened and, as of this writing, Yelp is valued at $2B despite having a trailing P/E ratio of 24, which is fairly low P/E for a tech company. But the P/E ratio is unsurprisingly low because it's not generally believed that Yelp can turn this around due to Google's dominant position in search as well as maps making it very difficult for Yelp to gain or retain users. This is not just obvious in retrospect and was well understood at the time. In fact, I talked to a former colleague at Google who was working on one of a number of local features that leveraged the position that Google had and that Yelp could never reasonably attain; the expected outcome of these features was to cripple Yelp's business. Not only was it understood that this was going to happen, it was also understood that Yelp was not likely to be able to counter this due to Google's ability to leverage its market power from search and maps. It's curious that, at the time, someone would've seriously argued that cutting off Yelp's source of new users while simultaneously presenting virtually all of Yelp's then-current users with an alternative that's bundled into an app or website they already use would not significantly impact Yelp's business, but the BE memo makes that case. One could argue that the set of maneuvers used here are analogous to the ones done by Microsoft that were brought up in the Microsoft antitrust case where it was alleged that a Microsoft exec said that they were going to "cut off Netscape’s air supply", but the BE memo argues that impact of having one's air supply cut off is "very small and not statistically significant" (after all, a typical body has blood volume sufficient to bind 1L of oxygen, much more than the oxygen normally taken in during one breath).

Another class of, if not error, then poorly supported reasoning is relying on cocktail party level of reasoning when there's data or other strong evidence that can be directly applied. This happens throughout the BE memo even though, at other times, when the BC memo has some moderately plausible reasoning, the BE memo's counter is that we should not accept such reasoning and need to look at the data and not just reason about things in the abstract. The BE memo heavily leans on the concept that we must rely on data over reasoning and calls arguments from the BC memo that aren't rooted in rigorous data anecdotal, "beyond speculation", etc., but BE memo only does this in cases where knowledge or reasoning might lead one to conclude that there was some kind of barrier to competition. When the data indicates that Google's behavior creates some kind of barrier in the market, the authors of BE memo ignore all relevant data and instead rely on reasoning over data even when the reasoning is weak and has the character of the Boies argument we referenced earlier. One could argue that the standard of evidence for pursuing an antitrust case should be stronger the standard of evidence for not pursuing one, but if the asymmetry observed here were for that reason, the BE memo could have listed areas where the evidence wasn't strong enough without making its own weak assertions in the face of stronger evidence. An example of this is the discussion of the impact of mobile defaults.

The BE memo argues that defaults are essentially worthless and have little to no impact, saying multiple times that users can switch with just "a few taps", adding that this takes "a few seconds" and that, therefore, "[t]hese are trivial switching costs". The most obvious and direct argument piece of evidence on the impact of defaults is the amount of money Google pays to retain its default status. In a 2023 antitrust action, it was revealed that Google paid Apple $26.3B to retain its default status in 2021. As of this writing, Apple's P/E ratio is 29.53. If we think of this payment as, at the margin, pure profit and having default status is as worthless as indicated by the BE memo, a naive estimate of how much this is worth to Apple is that it can account for something like $776B of Apple's $2.9T market cap. Or, looking at this from Google's standpoint, Google's P/E ratio is 27.49, so Google is willing to give up $722B of its $2.17T market cap. Google is willing to pay this to be the default search for something like 25% to 30% of phones in the world. This calculation is too simplistic, but there's no reasonable adjustment that could give anyone the impression that the value of being the default is as trivial as claimed by the BE memo. For reference, a $776B tech company would be 7th most valuable publicly traded U.S. tech company and the 8th most valuable publicly traded U.S. company (behind Meta/Facebook and Berkshire Hathaway, but ahead of Eli Lilly). Another reference is that YouTube's ad revenue in 2021 was $28.8B. It would be difficult to argue that spending one YouTube worth of revenue, in profit, in order to retain default status makes sense if, in practice, user switching costs are trivial and defaults don't matter. If we look for publicly available numbers close to 2012 instead of 2021, in 2013, TechCrunch reported a rumor that Google was paying Apple $1B/yr for search status and a lawsuit then revealed that Google paid Apple $1B for default search status in 2014. This is not longer after these memos are written and $1B/yr is still a non-trivial amount of money and it belies the BE memo's claim that mobile is unimportant and that defaults don't matter because user switching costs are trivial.

It's curious that, given the heavy emphasis in the BE memo on not trusting plausible reasoning and having to rely on empirical data, that BE staff appeared to make no attempt to find out how much Google was paying for its default status (a memo by a director who agrees with BE staff suggests that someone ought to check on this number, but there's no evidence that this was done and the FTC investigation was dropped shortly afterwards). Given the number of internal documents the FTC was able to obtain, it seems unlikely that the FTC would not have been able to obtain this number from either Apple or Google. But, even if it were the case that the number were unobtainable, it's prima facie implausible that defaults don't matter and switching costs are low in practice. If FTC staff interviewed product-oriented engineers and PMs or looked at the history of products in tech, so in order to make this case, BE staff had to ignore or avoid finding out how much Google was paying for default status, not talk to product-focused engineers, PM, or leadership, and also avoid learning about the tech industry.

One could make the case that, while defaults are powerful, companies have been able to overcome being non-default, which could lead to a debate on exactly how powerful defaults are. For example, one might argue about the impact of defaults when Google Chrome became the dominant browser and debate how much of it was due to Chrome simply being a better browser than IE, Opera, and Firefox, how much was due to blunders by Microsoft that Google is unlikely to repeat in search, how much was due to things like tricking people into making Chrome default via a bundle deal with badware installers and how much was due to pressuring people into setting Chrome is default via google.com. That's an interesting discussion where a reasonable person with an understanding of the industry could take either side of the debate, unlike the claim that defaults basically don't matter at all and user switching costs are trivial in practice, which is not plausible even without access to the data on how much Google pays Apple and others to retain default status. And as of the 2020 DoJ case against Google, roughly half of Google searches occur via a default search that Google pays for.

Another repeated error, closely related to the one above, is bringing up marketing statements, press releases, or other statements that are generally understood to be exaggerations, and relying on these as if they're meaningful statements of fact. For example, the BE memo states:

Microsoft's public statements are not consistent with statements made to antitrust regulators. Microsoft CEO Steve Ballmer stated in a press release announcing the search agreement with Yahoo: "This agreement with Yahoo! will provide the scale we need to deliver even more rapid advances in relevancy and usefulness. Microsoft and Yahoo! know there's so much more that search could be. This agreement gives us the scale and resources to create the future of search"

This is the kind of marketing pablum that generally accompanies an acquisition or partnership. Because this kind of meaningless statement is common across many industries, one would expect regulators, even ones with no understanding of tech, to recognize this as marketing and not give it as much or more weight as serious evidence.

A few interesting tidbits

Now that we've covered the main classes of errors observed in the memos, we'll look at a tidbits from the memos.

Between the approval of the compulsory process on June 3rd 2011 and the publication of the BC memo dated August 8th 2012, staff received 9.5M pages of documents across 2M docs and said they reviewed "many thousands of these documents", so staff were only able to review a small fraction of the documents.

Prior to the FTC investigation, there were a number of lawsuits related to the same issues, and all were dismissed, some with arguments that would, if they were taken as broad precedent, make it difficult for any litigation to succeed. In SearchKing v. Google, plaintiffs alleged that Google unfairly demoted their results but it was ruled that Google's rankings are constitutionally protected opinion and even malicious manipulation of rankings would not expose Google to liability. In Kinderstart v. Google, part of the ruling was that Google search is not an essential facility for vertical providers (such as Yelp, eBay, and Expedia). Since the memos are ultimately about legal proceedings, there is, of course, extensive discussion of Verizon v. Trinko and Aspen Skiing Co. v. Aspen Highlands Skiing Corp and the implications thereof.

As of the writing of the BC memo, 96% of Google's $38B in revenue was from ads, mostly from search ads. The BC memo makes the case that other forms of advertising, other than social media ads, only have limited potential for growth. That's certainly wrong in retrospect. For example, video ads are a significant market. YouTube's ad revenue was $28.8B in 2021 (a bit more than what Google pays to Apple to retain default search status), Twitch supposedly generated another $2B-$3B in video revenue, and a fair amount of video ad revenue goes directly from sponsors to streamers without passing through YouTube and Twitch, e.g., the #137th largest streamer on Twitch was offered $10M/yr stream online gambling for 30 minutes a day, and he claims that the #42 largest streamer, who he personally knows, was paid $10M/mo from online gambling sponsorships. And this isn't just apparent in retrospect — even at the time, there were strong signs that video would become a major advertising market. It happens that those same signs also showed that Google was likely to dominate the market for video ads, but it's still the case that the specific argument here was overstated.

In general, the BC memo seems to overstate the expected primacy of search ads as well as how distinct a market search ads are, claiming that other online ad spend is not a substitute in any way and, if anything, is a complement. Although one might be able to reasonably argue that search ads are a somewhat distinct market and the elasticity of substitution is low once you start moving a significant amount of your ad spend away from search, the degree to which the BC memo makes this claim is a stretch. Search ads and other ad budgets being complements and not substitutes is a very different position than I've heard from talking to people about how ad spend is allocated in practice. Perhaps one can argue that it makes sense to try to make a strong case here in light of Person V. Google, where Judge Fogel of the Northern District of California criticized the plaintiff's market definition, finding no basis for distinguishing "search advertising market" from the larger market for internet advertising, which likely foreshadows an objection that would be raised in any future litigation. However, as someone who's just trying to understand the facts of the matter at hand and the veracity of the arguments, the argument here seems dubious.

For Google's integrated products like local search and product search (formerly Froogle), the BC memo claims that if Google treated its own properties like other websites, the products wouldn't be ranked and Google artificially placed their own vertical competitors above organic offerings. The webspam team declined to include Froogle results because the results are exactly the kind of thing that Google removes from the index because it's spammy, saying "[o]ur algorithms specifically look for pages like these to either demote or remove from the index". Bill Brougher, product manager for web search said "Generally we like to have the destination pages in the index, not the aggregated pages. So if our local pages are lists of links to other pages, it's more important that we have the other pages in the index". After the webspam team was overruled and the results were inserted, the ads team complained that the less clicked (and implied to be lower quality) results would lead to a loss of $154M/yr. The response to this essentially contained the same content as the BC memo's argument on the importance of scale and why Google's actions to deprive competitors of scale are costly:

We face strong competition and must move quickly. Turning down onebox would hamper progress as follows - Ranking: Losing click data harms ranking; [t]riggering Losing CTR and google.com query distribution data triggering accuracy; [c]omprehensiveness: Losing traffic harms merchant growth and therefore comprehensiveness; [m]erchant cooperation: Losing traffic reduces effort merchants put into offer data, tax, & shipping; PR: Turning off onebox reduces Google's credibility in commerce; [u]ser awareness: Losing shopping-related UI on google.com reduces awareness of Google's shopping features

Normally, CTR is used as a strong signal to rank results, but this would've resulted in a low ranking for Google's own vertical properties, so "Google used occurrence of competing vertical websites to automatically boost the ranking of its own vertical properties above that of competitors" — if a comparison shopping site was relevant, Google would insert Google Product search above any rival, and if a local search site like Yelp or CitySearch was relevant, Google automatically returned Google Local at top of SERP.

Additionally, in order to see content for Google local results, Google took Yelp content and integrated it into Google Places. When Yelp observed this was happening, they objected to this and Google threatened to ban Yelp from traditional Google search results and further threatened to ban any vertical provider that didn't allow its content to be used in Google Places. Marissa Mayer testified that it was, from a technical standpoint, extraordinarily difficult to remove Yelp from Google Places without also removing Yelp from traditional organic search results. But when Yelp sent a cease and desist letter, Google was able to remove Yelp results immediately, seemingly indicating that it was less difficult than claimed. Google then claimed that it was technically infeasible to remove Yelp from Google Places without removing Yelp from the "local merge" interface on SERP. BC staff believe this claim is false as well, and Marissa Mayer later admitted in a hearing that this claim was false and that Google was concerned about the consequences of allowing sites to opt out of Google Places while staying in "local merge". There was also a very similar story with Amazon results and product search. As noted above, the BE memo's counterargument to all of this is that Google traffic is "very small and not statistically significant"

The BC memo claims that the activities above both reduced incentives of companies Yelp, City Search, Amazon, etc., to invest in the area and also reduced the incentives for new companies to form in this area. This seems true. In addition to the evidence presented in the BC memo (which goes beyond what was summarized above), if you just talked to founders looking for an idea or VCs around the time of the FTC investigation, there had already been a real movement away from founding and funding companies like Yelp because it was understood that Google could seriously cripple any similar company in this space by cutting off its air supply.

We'll defer to the appendix BC memo discussion on the AdWords API restrictions that specifically disallow programmatic porting of campaigns to other platforms, such as Bing. But one interesting bit there is that Google was apparently aware of the legal sensitivity of this matter, so meeting notes and internal documentation on the topic are unusually incomplete. On one meeting, apparently the most informative written record BC staff were able to find consists of a message from Director of PM Richard Holden to SVP of ads Susan Wojicki which reads, "We didn't take notes for obvious reasons hence why I'm not elaborating too much here in email but happy to brief you more verbally".

We'll also defer a detailed discussion of the BC memo comments on Google's exclusive and restrictive syndication agreements to the appendix, except for a couple of funny bits. One is that Google claims they were unaware of the terms and conditions in their standard online service agreements. In particular, the terms and conditions contained a "preferred placement" clause, which a number of parties believe is a de facto exclusivity agreement. When FTC staff questioned Google's VP of search services about this term, the VP claimed they were not aware of this term. Afterwards, Google sent a letter to Barbara Blank of the FTC explaining that they were removing the preferred placement clause in the standard online agreement.

Another funny bit involves Google's market power and how it allowed them to collect an increasingly large share of revenue for themselves and decrease the revenue share their partner received. Only a small number of Google's customers who were impacted by this found this concerning. Those that did find it concerning were some of the largest and most sophisticated customers (such as Amazon and IAC); their concern was that Google's restrictive and exclusive provisions would increase Google's dominance over Bing/Microsoft and allow them to dictate worse terms to customers. Even as Google was executing a systematic strategy to reduce revenue share to customers, which could only be possible due to their dominance of the market, most customers appeared to either not understand the long-term implications of Google's market power in this area or the importance of the internet.

For example, Best Buy didn't find this concerning because Best Buy viewed their website and the web as a way for customers to find presale information before entering a store and Walmart didn't find didn't find this concerning because they viewed the web as an extension to brick and mortar retail. It seems that the same lack of understanding of the importance of the internet which led Walmart and Best Buy to express their lack of concern over Google's dominance here also led to these retailers, which previously had a much stronger position than Amazon, falling greatly behind in both online and overall profit. Walmart later realized its error here and acquired Jet.com for $3.3B in 2016 and also seriously (relative to other retailers) funded programmers to do serious tech work inside Walmart. Since Walmart started taking the internet seriously, it's made a substantial comeback online and has averaged a 30% CAGR in online net sales since 2018, but taking two decades to mount a serious response to Amazon's online presence has put Walmart solidly behind Amazon in online retail despite nearly a decade of serious investment and Best Buy has still not been able to mount an effective response to Amazon after three decades.

The BE memo uses the lack of concern on the part of most customers as evidence that the exclusive and restrictive conditions Google dictated here were not a problem but, in retrospect, it's clear that it was only a lack of understanding of the implications of online business that led customers to be unconcerned here. And when the BE memo refers to the customers who understood the implications here as sophisticated, that's relative to people in lines of business where leadership tended to not understand the internet. While these customers are sophisticated by comparison to a retailer that took two decades to mount a serious response to the threat Amazon poses to their business, if you just talked to people in the tech industry at the time, you wouldn't need to find a particularly sophisticated individual to find someone who understood what was going on. It was generally understood that retail revenue and even moreso, retail profit was going to move online, and you'd have to find someone who was extremely unusually out of the loop to find someone who didn't at least roughly understand the implications here.

There's a lengthy discussion on search and scale in both the BC and BE memos. On this topic, the BE memo seems wrong and the implications of the BC memo are, if not subtle, at least not obvious. Let's start with the BE memo because that one's simpler to discuss, although we'll very briefly discuss the argument in the BC memo in order to frame the discussion in the BE memo. A rough sketch of the argument in the BC memo is that there are multiple markets (search, ads) where scale has a significant impact on product quality. Google's own documents acknowledge this "virtuous cycle" where having more users lets you serve better ads, which gives you better revenue for ads and, likewise in search, having more scale gives you more data which can be used to improve results, which leads to user growth. And for search in particular, the BC memo claims that click data from users is of high importance and that more data allows for better results.

The BE memo claims that this is not really the case. On the importance of click data, the BE memo raises two large objections. First, that this is "contrary to the history of the general search market" and second, that "it is also contrary to the evidence that factors such as the quality of the web crawler and web index; quality of the search algorithm; and the type of content included in the search results [are as important or more important].

Of the first argument, the BE memo elaborates with a case that's roughly "Google used to be smaller than it is today, and the click data at the time was sufficient, therefore being as large as Google used to be means that you have sufficient click data". Independent of knowledge of the tech industry, this seems like a strange line of reasoning. "We now produce a product that's 1/3 as good as our competitor for the same price, but that should be fine because our competitor previously produced a product that's 1/3 as good as their current product when the market was less mature and no one was producing a better product" is generally not going to be a winning move. That's especially true in markets where there's a virtuous cycle between market share and product quality, like in search.

The second argument also seems like a strange argument to make even without knowledge of the tech industry in that it's a classic fallacious argument. It's analogous to saying something like "the BC memo claims that it's important for cars to have a right front tire, but that's contrary to evidence that it's at least as important for a car to have a left front tire and a right rear tire". The argument is even less plausible if you understand tech, especially search. Calling out the quality of the search algorithm as distinct doesn't feel quite right because scale and click data directly feed into algorithm development (and this is discussed at some length in the BE memo — the authors of the BC memo surely had access to the same information and, from their writing, seem to have had access to the argument). And as someone who's worked on search indexing, as much as I'd like to be agree with the BE memo and say that indexing is as important or more important than ranking, I have to admit that indexing is an easier and less important problem than ranking and likewise for crawling vs. ranking. This was generally understood at the time so, given the number of interviews FTC staff did, the authors of the BE memo should've known this as well. Moreover, given the "history of the general search market" which the BE memo refers to, even without talking to engineers, this should've been apparent.

For example, Cuil was famous for building a larger index than Google. While that's not a trivial endeavor, at the time, quite a few people had the expertise to build an index that rivaled Google's index in raw size or whatever other indexing metric you prefer, if given enough funding for a serious infra startup. Cuil and other index-focused attempts failed because having a large index without good search ranking is worth little. While it's technically true that having good ranking with a poor index is also worth little, this is not something we've really seen in practice because ranking is the much harder problem and a company that's competent to build a good search ranker will, as a matter of course, have a good enough index and good enough crawling.

As for the case in the BC memo, I don't know what the implications should be. The BC memo correctly points out that increased scale greatly improves search quality, that the extra data Bing got from the Yahoo greatly increased search quality and increased CTR, that further increased scale should be expected to continue to provide high return, that the costs of creating a competitor to Google are high (Bing was said to be losing $2B/yr at the time and was said to be spending $4.5B/yr "developing its algorithms and building the physical capacity necessary to operate Bing"), and that Google undertook actions that might be deemed anticompetitive which disadvantaged Bing's compared to the counterfactual world where Google did not take those actionts, and they make a similar case for ads. However, despite the strength of the stated BC memo case and the incorrectness of the stated BE memo case, the BE memo's case is correct in spirit, in that there are actions Microsoft could've taken but did not in order to compete much more effectively in search and one could argue that the FTC shouldn't be in the business of rescuing a company from competing ineffectively.

Personally, I don't think it's too interesting to discuss the position of the BC memo vs. the BE memo at length because the positions the BE memo takes seem extremely weak. It's not fair to call it a straw man because it's a real position, and one that carried the day at the FTC, but the decision to take action or not seemed more about philosophy than the arguments in the memos. But we can discuss what else might've been done.

What might've happened

What happened after the FTC declined to pursue antitrust action was that Microsoft effectively defunded Bing as a serious bet, taking resources that could've gone to continuing to fund a very expensive fight against Google, and moving them to other bets that it deemed to be higher ROI. The big bets Microsoft pursued were Azure, Office, and HoloLens (and arguably Xbox). Hololens was a pie-in-the-sky bet, but Azure and Office were lines of business where Microsoft could, instead of fighting an uphill battle where their competitor can use its dominance in related markets to push around competitors, Microsoft could fight downhill battles where they can use their dominance in related markets to push around competitors, resulting in a much higher return per dollar invested. As someone who worked on Bing and thought that BIng had the potential to seriously compete with Google given sustained, unprofitable, heavy investment, I find that disappointing but also likely the correct business decision. If you look at any particular submarket, like Teams vs. Slack, the Microsoft product doesn't need to be nearly as good as the competing product to take over the market, which is the opposite of the case in search, where Google's ability to push competitors around means that Bing would have to be much better than Google to attain marketshare parity.

Based on their public statements, Biden's DoJ Antitrust AAG appointee, Jonathan Kanter, would argue for pursuing antitrust action under the circumstances, as would Biden's FTC commissioner and chair appointee Lina Khan. Prior to her appointment as FTC commissioner and chair, Khan was probably best known for writing Amazon's Antitrust Paradox, which has been influential as well as controversial. Obama appointees, who more frequently agreed with the kind of reasoning from the BE memo, would have argued against antitrust action and the investigation under discussion was stopped on their watch. More broadly, they argued against the philosophy driving Kanter and Khan. Obama's FTC Commissioner appointee, GMU economist and legal scholar Josh Wright actually wrote a rebuttal titled "Requiem for a Paradox: The Dubious Rise and Inevitable Fall of Hipster Antitrust", a scathing critique of Khan's position.

If, in 2012, the FTC and DoJ were run by Biden appointees instead of Obama appointees, what difference would that have made? We can only speculate, but one possibility would be that they would've taken action and then lost, as happened with the recent cases against Meta and Microsoft which seem like they would not have been undertaken under an Obama FTC and DoJ. Under Biden appointees, there's been much more vigorous use of the laws that are on the books, the Sherman Act, the Clayton Act, the FTC Act, the Robinson–Patman Act, as well as "smaller" antitrust laws, but the opinion of the courts hasn't changed under Biden and this has led to a number of unsuccessful antitrust cases in tech. Both the BE and BC memos dedicate significant space to whether or not a particular line of reasoning will hold up in court. Biden's appointees are much less concerned with this than previous appointees and multiple people in the DoJ and the FTC are on the record saying things like "it is our duty to enforce the law", meaning that when they see violations of the antitrust laws that were put into place by elected officials, it's their job to pursue these violations even if courts may not agree with the law.

Another possibility is that there would've been some action, but the action would've been in line with most corporate penalties we see. Something like a small fine that costs the company an insignificant fraction of marginal profit they made from their actions, or some kind of consent decree (basically a cease and desist), where the company will be required to stop doing specific actions while keeping their marketshare, keeping the main thing they wanted to gain, a massive advantage in a market dominated by network effects. Perhaps there will be a few more meetings where "[w]e didn't take notes for obvious reasons" to work around the new limitations and business as usual will continue. Given the specific allegations in the FTC memos and the attitudes of the courts at the time, my guess is that something like this second set of possibilities would've been the most likely outcome had the FTC proceeded with their antitrust investigation instead of dropping it, some kind of nominal victory that makes little to no difference in practice. Given how long it takes for these cases to play out, it's overwhelmingly likely that Microsoft would've already scaled back its investment in Bing and moved Bing from a subsidized bet it was trying to grow to a profitable business it wanted to keep by the time any decision was made. There are a number of cases that were brought by other countries which had remedies that were in line with what we might've expected if the FTC investigation continued. On Google using market power in mobile to push software Google wants to nearly all Android phones, an EU and was nominally successful but made little to no difference in practice. Cristina Caffara of the Centre for Economic Policy Research characterized this as

Europe has failed to drive change on the ground. Why? Because we told them, don't do it again, bad dog, don't do it again. But in fact, they all went and said 'ok, ok', and then went out, ran back from the back door and did it again, because they're smarter than the regulator, right? And that's what happens.

So, on the tying case, in Android, the issue was, don't tie again so they say, "ok, we don't tie". Now we got a new system. If you want Google Play Store, you pay $100. But if you want to put search in every entry point, you get a discount of $100 ... the remedy failed, and everyone else says, "oh, that's a nice way to think about it, very clever"

Another pair of related cases are Yandex's Russian case on mobile search defaults and a later EU consent decree. In 2015, Yandex brought a suit about mobile default status on Android in Russia, which was settled by adding a "choice screen" which has users pick their search engine without preferencing a default. This immediately caused Yandex to start gaining marketshare on Google and Yandex eventually surpassed Google in marketshare in Russia according to statcounter. In 2018, the EU required a similar choice screen in Europe, which didn't make much of a difference, except maybe sort of in the Czech republic. There are a number of differences between the situation in Russia and in the EU. One, arguably the most important, is that when Yandex brought the case against Google in Russia, Yandex was still fairly competitive, with marketshare in the high 30% range. At the time of the EU decision in 2018, Bing was the #2 search engine in Europe, with about 3.6% marketshare. Giving consumers a choice when one search engine completely dominates the market can be expected to have fairly little impact. One argument the BE memo heavily relies on is the idea that, if we intervene in any way, that could have bad effects down the line, so we should be very careful and probably not do anything, just in case. But in these winner-take-most markets with such strong network effects, there's a relatively small window in which you can cheaply intervene. Perhaps, and this is highly speculative, if the FTC required a choice screen in 2012, Bing would've continued to invest enough to at least maintain its marketshare against Google.

For verticals, in shopping, the EU required some changes to how Google presents results in 2017. This appears to have had little to no impact, being both perhaps 5-10 years too late and also a trivial change that wouldn't have made much difference even if enacted a decade earlier. The 2017 ruling came out of a case that started in 2010, and in the 7 years it took to take action, Google managed to outcompete its vertical competitors, making them barely relevant at best.

Another place we could look is at the Microsoft antitrust trial. That's a long story, at least as long as this document, but to very briefly summarize, in 1990, the FTC started an investigation over Microsoft's allegedly anticompetitive conduct. A vote to continue the investigation ended up in a 2-2 tie, causing the investigation to be closed. The DoJ then did its own investigation, which led to a consent decree that was generally considered to not be too effective. There was then a 1998 suit by the DoJ about Microsoft's use of monopoly power in the browser market, which initially led to a decision to break Microsoft up. But, on appeal, the breakup was overturned, which led to a settlement in 2002. A major component of the 1998 case was about browser bundling and Microsoft's attack on Netscape. By the time the case was settled, in 2002, Netscape was effectively dead. The parts of the settlements having to do with interoperability were widely regarded as ineffective at the time, not only because Netscape was dead, but because they weren't going to be generally useful. A number of economists took the same position as the BE memo, that no intervention should've happened at the time and that any intervention is dangerous and could lead to a fettering of innovation. Nobel Prize winning economist Milton Friedman wrote a Cato Policy Forum essay titled "The Business Community's Suicidal Impulse", predicting that tech companies calling for antitrust action against Microsoft were committing suicide, and that a critical threshold had been passed and that this would lead to the bureaucratization of Silicon Valley

When I started in this business, as a believer in competition, I was a great supporter of antitrust laws; I thought enforcing them was one of the few desirable things that the government could do to promote more competition. But as I watched what actually happened, I saw that, instead of promoting competition, antitrust laws tended to do exactly the opposite, because they tended, like so many government activities, to be taken over by the people they were supposed to regulate and control. And so over time I have gradually come to the conclusion that antitrust laws do far more harm than good and that we would be better off if we didn’t have them at all, if we could get rid of them. But we do have them.

Under the circumstances, given that we do have antitrust laws, is it really in the self-interest of Silicon Valley to set the government on Microsoft? ... you will rue the day when you called in the government. From now on the computer industry, which has been very fortunate in that it has been relatively free of government intrusion, will experience a continuous increase in government regulation. Antitrust very quickly becomes regulation. Here again is a case that seems to me to illustrate the suicidal impulse of the business community.

In retrospect, we can see that this wasn't correct and, if anything, was the opposite of correct. On the idea that even attempting antirust action against Microsoft would lead to an inevitable increase in government intervention, we saw the opposite, a two-decade long period of relatively light regulation and antitrust activity. And in terms of the impacts on innovation, although the case against Microsoft was too little and too late to save Netscape, Google's success appears to be causally linked to the antitrust trial. At one point, in the early days of Google, when Google had no market power and Microsoft effectively controlled how people access the internet, Microsoft internally discussed proposals aimed at killing Google. One proposal involved redirecting users who tried to navigate to Google to Bing (at the time, called MSN Search, and of course this was before Chrome existed and IE dominated the browser market). Another idea was to put up a big scary warning that warned users that Google was dangerous, much like the malware warnings browsers have today. Gene Burrus, a lawyer for Microsoft at the time, stated that Microsoft chose not to attempt to stop users from navigating to google.com due to concerns about further antitrust action after they'd been through nearly a decade of serious antitrust scrutiny. People at both Google and Microsoft who were interviewed about this both believe that Microsoft would've killed Google had they done this so, in retrospect, we can see that Milton Friedman was wrong about the impacts of the Microsoft antitrust investigations and that one can make the case that it's only because of the antitrust investigations that web 1.0 companies like Google and Facebook were able to survive, let alone flourish.

Another possibility is that a significant antitrust action would've been undertaken, been successful, and been successful quickly enough to matter. It's possible that, by itself, a remedy wouldn't have changed the equation for Bing vs. Google, but if a reasonable remedy was found and enacted, it still could've been in time to keep Yelp and other vertical sites as serious concerns and maybe even spur more vertical startups. And in the hypothetical universe where people with the same philosophy as Biden's appointees were running the FTC and the DoJ, we might've also seen antitrust action against Microsoft in markets where they can leverage their dominance in adjacent markets, making Bing a more appealing area for continued heavy investment. Perhaps that would've resulted in Bing being competitive with Google and the aforementioned concerns that "sophisticated customers" like Amazon and IAC had may not have come to pass. With antitrust against Microsoft and other large companies that can use their dominance to push competitors around, perhaps Slack would still be an independent product and we'd see more startups in enterprise tools (a number of commenters believe that Slack was basically forced into being acquired because it's too difficult to compete with Teams given Microsoft's dominance in related markets). And Slack continuing to exist and innovate is small potatoes — the larger hypothetical impact would be all of the new startups and products that would be created that no one even bothers to attempt because they're concerned that a behemoth with an integrated bundle like Microsoft would crush their standalone product. If you add up all of these, if not best-case, at least very-good-case outcomes for antitrust advocates, one could argue that consumers and businesses would be better off. But, realistically, it's hard to see how this very-good-case set of outcomes could have come to pass.

Coming back to the FTC memo, if we think about what it would take to put together a set of antitrust actions that actually fosters real competition, that seems extraordinarily difficult. A number of the more straightforward and plausible sounding solutions are off the table for political reasons, due to legal precedent, or due to arguments like the Boies argument we referenced or some of the arguments in the BE memo that are clearly incorrect, but appear to be convincing to very important people.

For the solutions that seem to be on the table, weighing the harms caused by them is non-trivial. For example, let's say the FTC mandated a mobile and desktop choice screen in 2012. This would've killed Mozilla in fairly short order unless Mozilla completely changed its business model because Mozilla basically relies on payments from Google for default status to survive. We've seen with Opera that even when you have a superior browser that introduces features that other browsers later copy, which has better performance than other browsers, etc., you can't really compete with free browsers when you have a paid browser. So then we would've quickly been down to IE/Edge and Chrome. And in terms of browser engines, just Chrome after not too long as Edge is now running Chrome under the hood. Maybe we can come up with another remedy that allows for browser competition as well, but the BE memo isn't wrong to note that antitrust remedies can cause other harms.

Another example which highlights the difficulty of crafting a politically suitable remedy are the restrictions the Bundeskartellamt imposed against Facebook, which have to do with user privacy and use of data (for personalization, ranking, general ML training, etc.), which is considered an antitrust issue in Germany. Michal Gal, Professor and Director of the Forum on Law and Markets at the University of Haifa pointed out that, of course Facebook, in response to the rulings, is careful to only limit its use of data if Facebook detects that you're German. If the concern is that ML models are trained on user data, this doesn't do much to impair Facebook's capability. Hypothetically, if Germany had a tech scene that was competitive with American tech and German companies were concerned about a similar ruling being leveled against them, this would be disadvantageous to nascent German companies that initially focus on the German market before expanding internationally. For Germany, this is only a theoretical concern as, other than SAP, no German company has even approached the size and scope of large American tech companies. But when looking at American remedies and American regulation, this isn't a theoretical concern, and some lawmakers will want to weigh the protection of American consumers against the drag imposed on American firms when compared to Korean, Chinese, and other foreign firms that can grow in local markets with fewer privacy concerns before expanding to international markets. This concern, if taken seriously, could be used to argue against nearly any pro-antitrust action argument.

What can we do going forward?

This document is already long enough, so we'll defer a detailed discussion of policy specifics for another time, but in terms of high-level actions, one thing that seems like it would be helpful is to have tech people intimately involved in crafting remedies and regulation as well as during investigations2. From the directors memos on the 2011-2021 FTC investigation that are publicly available, it would appear this was not done because the arguments from the BE memos that wouldn't pass the sniff test for a tech person appear to have been taken seriously. Another example is the one EU remedy that Cristina Caffara noted was immediately worked around by Google, in a way that many people in tech would find to be a delightful "hack".

There's a long history of this kind of "hacking the system" being lauded in tech going back to before anyone called it "tech" and it was just physics and electrical engineering. To pick a more recent example, one of the reasons Sam Altman become President of Y Combinator, which eventually led to him becoming CEO of Open AI was that Paul Graham admired his ability to hack systems; in his 2010 essay on founders, under the section titled "Naughtiness", Paul wrote:

Though the most successful founders are usually good people, they tend to have a piratical gleam in their eye. They're not Goody Two-Shoes type good. Morally, they care about getting the big questions right, but not about observing proprieties. That's why I'd use the word naughty rather than evil. They delight in breaking rules, but not rules that matter. This quality may be redundant though; it may be implied by imagination.

Sam Altman of Loopt is one of the most successful alumni, so we asked him what question we could put on the Y Combinator application that would help us discover more people like him. He said to ask about a time when they'd hacked something to their advantage—hacked in the sense of beating the system, not breaking into computers. It has become one of the questions we pay most attention to when judging applications.

Or, to pick one of countless examples from Google, in order to reduce travel costs at Google, Google engineers implemented a system where they computed some kind of baseline "expected cost for flights, and then gave people a credit for taking flights that came in under the baseline costs that could be used to upgrade future flights and travel accommodations. This was a nice experience for employees compared to what stodgier companies were doing in terms of expense limits and Google engineers were proud of creating a system that made things better for everyone, which was one kind of hacking the system. The next level of hacking the system was when some employees optimized their flights and even set up trips to locations that were highly optimizable (many engineers would consider this a fun challenge, a variant of classic dynamic programming problems that are given in interviews, etc.), allowing them to upgrade to first class flights and the nicest hotels.

When I've talked about this with people in management in traditional industries, they've frequently been horrified and can't believe that these employees weren't censured or even fired for cheating the system. But when I was at Google, people generally found this to be admirable, as it exemplified the hacker spirit.

We can see, from the history of antitrust in tech going back at least two decades, that courts, regulators, and legislators have not been prepared for the vigor, speed, and delight with which tech companies hack the system.

And there's precedent for bringing in tech folks to work on the other side of the table. For example, this was done in the big Microsoft antitrust case. But there are incentive issues that make this difficult at every level that stem from, among other things, the sheer amount of money that tech companies are willing to pay out. If I think about tech folks I know who are very good at the kind of hacking the system described here, the ones who want to be employed at big companies frequently make seven figures (or more) annually, a sum not likely to be rivaled by an individual consulting contract with the DoJ or FTC. If we look at the example of Microsoft again, the tech group that was involved was managed by Ron Schnell, who was taking a break from working after his third exit, but people like that are relatively few and far between. Of course there are people who don't want to work at big companies for a variety of reasons, often moral reasons or a dislike of big company corporate politics, but most people I know who fit that description haven't spent enough time at big companies to really understand the mechanics of how big companies operate and are the wrong people for this job even if they're great engineers and great hackers.

At an antitrust conference a while back, a speaker noted that the mixing and collaboration between the legal and economics communities was a great boon for antitrust work. Notably absent from the speech as well as the conference were practitioners from industry. The conference had the feel of an academic conference, so you might see CS academics at the conference some day, but even if that were to happen, many of the policy-level discussions are ones that are outside the area of interest of CS academics. For example, one of the arguments from the BE memo that we noted as implausible was the way they used MAU to basically argue that switching costs were low. That's something outside the area of research of almost every CS academic, so even if the conference were to expand and bring in folks who work closely with tech, the natural attendees would still not be the right people to weigh in on the topic when it comes to the plausibility of nitty gritty details.

Besides the aforementioned impact on policy discussions, the lack of collaboration with tech folks also meant that, when people spoke about the motives of actors, they would often make assumptions that were unwarranted. On one specific example of what someone might call a hack of the system, the speaker described an exec's reaction (high-fives, etc.), and inferred a contempt for lawmakers and the law that was not in evidence. It's possible the exec in question does, in fact, have a contempt and disdain for lawmakers and the law, but that celebration is exactly what you might've seen after someone at Google figured out how to get upgraded to first class "for free" on almost all their flights by hacking the system at Google, which wouldn't indicate contempt or disdain at all.

Coming back to the incentive problem, it goes beyond getting people who understand tech on the other side of the table in antitrust discussions. If you ask Capitol Hill staffers who were around at the time, the general belief is that the primary factor that scuttled the FTC investigation was Google's lobbying, and of course Google and other large tech companies spend more on lobbying than entities that are interested in increased antitrust scrutiny.

And in the civil service, if we look at the lead of the BC investigation and the first author on the BC memo, they're now Director and Associate General Counsel of Competition and Regulatory Affairs at Facebook. I don't know them, so I can't speak to their motivations, but if I were offered as much money as I expect they make to work on antitrust and other regulatory issues at Facebook, I'd probably take the offer. Even putting aside the pay, if I was a strong believer in the goals of increased antitrust enforcement, that would still be a very compelling offer. Working for the FTC, maybe you lead another investigation where you write a memo that's much stronger than the opposition memo, which doesn't matter when a big tech company pours more lobbying money into D.C. and the investigation is closed. Or maybe your investigation leads to an outcome like the EU investigation that led to a "choice screen" that was too little and far too late. Or maybe it leads to something like the Android Play Store untying case where, seven years after the investigation was started, an enterprising Google employee figures out a "hack" that makes the consent decree useless in about five minutes. At least inside Facebook, you can nudge the company towards what you think is right and have some impact on how Facebook treats consumers and competitors.

Looking at it from the standpoint of people in tech (as opposed to people working in antitrust), in my extended social circles, it's common to hear people say "I'd never work at company X for moral reasons". That's a fine position to take but, almost everyone I know who does this ends up working at a much smaller company that has almost no impact on the world. If you want to take a moral stand, you're more likely to make a difference by working from the inside or finding a smaller direct competitor and helping it become more successful.

Thanks to Laurence Tratt, Yossi Kreinin, Justin Hong, kouhai@treehouse.systems, Sophia Wisdom, @cursv@ioc.exchange, and Misha Yagudin for comments/corrections/discussion

Appendix: non-statements

This is analogous to the "non-goals" section of a technical design doc, but weaker, in that a non-goal in a design doc is often a positive statement that implies something that couldn't be implied from reading the doc, whereas the non-goal statements themselves don't add any informatio

BC Staff Memo

By "Barbara R. Blank, Gustav P. Chiarello, Melissa Westman-Cherry, Matthew Accornero, Jennifer Nagle, Anticompetitive Practices Division; James Rhilinger, Healthcare Division; James Frost, Office of Policy and Coordination; Priya B. Viswanath, Office of the Director; Stuart Hirschfeld, Danica Noble, Northwest Region; Thomas Dahdouh, Western Region-San Francisco, Attorneys; Daniel Gross, Robert Hilliard, Catherine McNally, Cristobal Ramon, Sarah Sajewski, Brian Stone, Honors Paralegals; Stephanie Langley, Investigator"

Dated August 8, 2012

Executive Summary







1. Google

2. General search competitors

a. Microsoft
b. Yahoo

3. Major Vertical Competition


2. Online Advertising

3. Syndicated Search and Search Advertising

1. Search Query Volume

2. Advertising Volume

3. Scale Curve


1. Google's Preferencing of Google Vertical Properties Within Its Search Engine Results Page ("SERP")

a. Overview of Changes to Google's SERP
b. Google's Development and Introduction of Vertical Properties
c. The Evolution of Display of Google's Vertical Properties on the SERP
d. Google's Preferential Display of Google Vertical Properties on the SERP
e. Google's Demotion of Competing Vertical Websites
f. Effects of Google's SERP Changes on Vertical Rivals

2. Google's "Scraping" of Rivals' Vertical Content

a. The "Local" Story
b. The "Shopping" Story
c. Effects of Google's "Scraping" on Vertical Rivals

3. Google's API Restrictions

a. Overview of the AdWords Platform
b. The Restrictive Conditions
c. Effects of the Restrictive Conditions
i. Effects on Advertisers and Search Engine Marketers ("SEMs")
ii. Effects on Competitors
d. Internal Google Discussions Regarding the Restrictions

4. Google's Exclusive and Restrictive Syndication Agreements

a. Publishers and Market Structure
b. Development of the Market for Search Syndication

c. Specifics of Google's Syndication Agreements

d. Effects of Exclusivity and Preferred Placement
i. Common Publisher Responses
ii. Publishers' Views of Exclusivity Provisions
iii.Effects on Competitors


1. Relevant Markets and Market Shares

b. Search Advertising
c. Syndicated Search and Search Advertising ("Search Intermediation")

2. Substantial Barriers to Entry Exist

a. Technology and Specialization
b. Substantial Upfront Investment
c. Scale Effects
d. Reputation, Brand Loyalty, and the "Halo Effect"
e. Exclusive and Restrictive Agreements -


1. Google's Preferencing of Google Vertical Properties Within Its SERP

a. Google's Product Design Impedes Vertical Competitors
b. Google's SERP Changes Have Resulted In Anticompetitive Effects
c. Google's Justifications for the Conduct

2. Google's "Scraping" of Rivals' Vertical Content

a. Google's "Scraping" Constitutes a Conditional Refusal to Deal or Unfair Method Of Competition
b. Google's "Scraping" Has Resulted In Anticompetitive Effects
c. Google's "Scraping" Is Not Justified By Efficiencies

3. Google's API Restrictions

a. The Restrictive Conditions Are Unreasonable
b. The Restrictive Conditions Have Resulted In Anticompetitive Effects
c. The Restrictive Conditions Are Not Justified By Efficiencies

4. Google's Exclusive and Restrictive Syndication Agreements

a. Google's Agreements Foreclose a Substantial Portion of the Relevant Market
b. Google's Agreements Have Resulted In Anticompetitive Effects
c. Google's Agreements Are Not Justified By Efficiencies


A. Scraping

B. API Restrictions

C. Exclusive and Restrictive Syndication Agreements



FTC BE staff memo

"Bureau of Economics

August 8, 2012

From: Christopher Adams and John Yun, Economists"

Executive Summary

1. Does Google possess monopoly power in the relevant antitrust market?

Theory 1: The preferencing theory

2.1 Overview

2.2 Analysis

2.3 Empirical evidence

2.4 Documentary evidence

2.5 Summary of the preferencing theory

Theory 2: Exclusionary practices in search distribution

3.1 Overview

3.2 Are the various Google distribution agreements in fact exclusionary?

3.3 Are rival search engines being excluded from the market?

3.4 Does Microsoft have sufficient scale to be competitive?

3.5 Theory based on raising rivals' costs

[I stopped taking detailed notes at this point because taking notes that are legible to other people (as opposed to just for myself) takes about an order of magnitude longer, and I didn't think that there was much of interest here. I generally find comments of the form "I stopped reading at X" to be quite poor, in that people making such comments generally seem to pick some trivial thing that's unimportant and then declare and entire document to be worthless based on that. This pattern is also common when it comes to engineers, institutions, sports players, etc. and I generally find it counterproductive in those cases as well. However, in this case, there isn't really a single, non-representative, issue. The majority of the reasoning seems not just wrong, but highly disconnected from the on-the-ground situation. More notes indicating that the authors are making further misleading or incorrect arguments in the same style don't seem very useful. I did read the rest of the document and I also continue to summarize a few bits, below. I don't want to call them "highlights" because that would imply that I pulled out particularly interesting or compelling or incorrect bits and it's more of a smattering of miscellaneous parts with no particular theme]

Other memos

[for these, I continued writing high-level summaries, not detailed summaries]

  1. By analogy to a case that many people in tech are familiar with, consider this exchange between Oracle counsel David Boies and Judge William Alsup on the rangeCheck function, which checks if a range is a valid array access or not given the length of an array and throws an exception if the access is out of range:

    • Boies: [argument that Google copied the rangeCheck function in order to accelerate development]
    • Alsup: All right. I have — I was not good — I couldn't have told you the first thing about Java before this trial. But, I have done and still do a lot of programming myself in other languages. I have written blocks of code like rangeCheck a hundred times or more. I could do it. You could do it. It is so simple. The idea that somebody copied that in order to get to market faster, when it would be just as fast to write it out, it was an accident that that thing got in there. There was no way that you could say that that was speeding them along to the marketplace. That is not a good argument.
    • Boies: Your Honor
    • Alsup: [cutting off Boies] You're one of the best lawyers in America. How can you even make that argument? You know, maybe the answer is because you are so good it sounds legit. But it is not legit. That is not a good argument.
    • Boies: Your Honor, let me approach it this way, first, okay. I want to come back to rangeCheck. All right.
    • Alsup: RangeCheck. All it does is it makes sure that the numbers you're inputting are within a range. And if they're not, they give it some kind of exceptional treatment. It is so — that witness, when he said a high school student would do this, is absolutely right.
    • Boies: He didn't say a high school student would do it in an hour, all right.
    • Alsup: Less than — in five minutes, Mr. Boies.

    Boies previously brought up this function as a non-trivial piece of work and then argues that, in their haste, a Google engineer copied this function from Oracle. As Alsup points out, the function is trivial, so trivial that it wouldn't be worth looking it up to copy and that even a high school student could easily produce the function from scratch. Boies then objects that, sure, maybe a high school student could write the function, but it might take an hour or more and Alsup correctly responds that an hour is implausible and that it might take five minutes.

    Although nearly anyone who could pass a high school programming class would find Boeis's argument not just wrong but absurd3, more like a joke than something that someone might say seriously, it seems reasonable for Boies to make the argument because people presiding over these decisions in court, in regulatory agencies, and in the legislature, sometimes demonstrate a lack of basic understanding of tech. Since my background is in tech and not law or economics, I have no doubt that this analysis will miss some basics about law and economics in the same way that most analyses I've read seem miss basics about tech, but since there's been extensive commentary on this case from people with strong law and economics backgrounds, I don't see a need to cover those issues in depth here because anyone who's interested can read another analysis instead of or in addition to this one.

  2. Although this document is focused on tech, the lack of hands-on industry-expertise in regulatory bodies, legislation, and the courts, appears to cause problems in other industries as well. An example that's relatively well known due to a NY Times article that was turned into a movie is DuPont's involvement in the popularization of PFAS and, in particular, PFOA. Scientists at 3M and DuPont had evidence of the harms of PFAS going back at least to the 60s, and possibly even as far back as the 50s. Given the severe harms that PFOA caused to people who were exposed to it in significant concentrations, it would've been difficult to set up a production process for PFOA without seeing the harm it caused, but this knowledge, which must've been apparent to senior scientists and decision makers in 3M and DuPont, wasn't understood by regulatory agencies for almost four decades after it was apparent to chemical companies.

    By the way, the NY Times article is titled "The Lawyer Who Became DuPont’s Worst Nightmare" and it describes how DuPont made $1B/yr in profit for years while hiding the harms of PFOA, which was used in the manufacturing process for Teflon. This lawyer brought cases against DuPont that were settled for hundreds of millions of dollars; according to the article and movie, the litigation didn't even cost DuPont a single year's worth of PFOA profit. Also, DuPont manage to drag out the litigation for many years, continuing to reap the profit from PFOA. Now that enough evidence has mounted against PFOA, Teflon is now manufactured using PFO2OA or FRD-903, which are newer and have a less well understood safety profile than PFOA. Perhaps the article could be titled "The Lawyer Who Became DuPont's Largest Mild Annoyance".

  3. In the media, I've sometimes seen this framed as a conflict between tech vs. non-tech folks, but we can see analogous comments from people outside of tech. For example, in a panel discussion with Yale SOM professor Fiona Scott Morton and DoJ Antitrust Principal Deputy AAG Doha Mekki, Scott Morton noted that the judge presiding over the Sprint/T-mobile merger proceedings, a case she was an expert witness for, had comically wrong misunderstandings about the market, and that it's common for decisions to be made which are disconnected from "market realities". Mekki seconded this sentiment, saying "what's so fascinating about some of the bad opinions that Fiona identified, and there are many, there's AT&T Time Warner, Sabre Farelogix, T-mobile Sprint, they're everywhere, there's Amex, you know ..."

    If you're seeing this or the other footnote in mouseover text and/or tied to a broken link, this is an issue with Hugo. At this point, I've spent more than an entire blog post's worth of effort working around Hugo breakage and am trying to avoid spending more time working around issues in a tool that makes breaking changes at a high rate. If you have a suggestion to fix this, I'll try it, otherwise I'll try to fix it when I switch away from Hugo.